Managing queue distribution between critical datacenter and flexible datacenter

[0001] This specification relates to a system for controlling the use of “behind-the-meter” power.

[0002] The price for power distributed through regional and national electric power grids is composed of Generation, Administration, and Transmission & Distribution (“T&D”) costs. T&D costs are a significant portion of the overall price paid by consumers for electricity. T&D costs include capital costs (land, equipment, substations, wire, etc.), electrical transmission losses, and operation and maintenance costs. Electrical power is typically generated at local stations (e.g., coal, natural gas, nuclear, and renewable sources) in the Medium Voltage class of 2.4 kVAC to 69 kVAC before being converted in an AC-AC step up transformer to High Voltage at 115 kVAC or above. T&D costs are accrued at the point the generated power leaves the local station and is converted to High Voltage electricity for transmission onto the grid.

[0003] Local station operators are paid a variable market price for the amount of power leaving the local station and entering the grid. However, grid stability requires that a balance exist between the amount of power entering the grid and the amount of power used from the grid. Grid stability and congestion is the responsibility of the grid operator and grid operators take steps, including curtailment, to reduce power supply from local stations when necessary. Frequently, the market price paid for generated power will be decreased in order to disincentivize local stations from generating power. In some cases, the market price will go negative, resulting in a cost to local station operators who continue to supply power onto a grid. Grid operators may sometimes explicitly direct a local station operator to reduce or stop the amount of power the local station is supplying to the grid.

[0004] Power market fluctuations, power system conditions such as power factor fluctuation or local station startup and testing, and operational directives resulting in reduced or discontinued generation all can have disparate effects on renewal energy generators and can occur multiple times in a day and last for indeterminate periods of time. Curtailment, in particular, is particularly problematic.

[0005] According to the National Renewable Energy Laboratory's Technical Report TP-6A20-60983 (March 2014):

• • [0006] [C]urtailment [is] a reduction in the output of a generator from what it could otherwise produce given available resources (e.g., wind or sunlight), typically on an involuntary basis. Curtailments can result when operators or utilities command wind and solar generators to reduce output to minimize transmission congestion or otherwise manage the system or achieve the optimal mix of resources. Curtailment of wind and solar resources typically occurs because of transmission congestion or lack of transmission access, but it can also occur for reasons such as excess generation during low load periods that could cause baseload generators to reach minimum generation thresholds, because of voltage or interconnection issues, or to maintain frequency requirements, particularly for small, isolated grids. Curtailment is one among many tools to maintain system energy balance, which can also include grid capacity, hydropower and thermal generation, demand response, storage, and institutional changes. Deciding which method to use is primarily a matter of economics and operational practice.
  • [0007] “Curtailment” today does not necessarily mean what it did in the early 2000 s. Two sea changes in the electric sector have shaped curtailment practices since that time: the utility-scale deployment of wind power, which has no fuel cost, and the evolution of wholesale power markets. These simultaneous changes have led to new operational challenges but have also expanded the array of market-based tools for addressing them.
  • [0008] Practices vary significantly by region and market design. In places with centrally-organized wholesale power markets and experience with wind power, manual wind energy curtailment processes are increasingly being replaced by transparent offer-based market mechanisms that base dispatch on economics. Market protocols that dispatch generation based on economics can also result in renewable energy plants generating less than what they could potentially produce with available wind or sunlight. This is often referred to by grid operators by other terms, such as “downward dispatch.” In places served primarily by vertically integrated utilities, power purchase agreements (PPAs) between the utility and the wind developer increasingly contain financial provisions for curtailment contingencies.
  • [0009] Some reductions in output are determined by how a wind operator values dispatch versus non-dispatch. Other curtailments of wind are determined by the grid operator in response to potential reliability events. Still other curtailments result from overdevelopment of wind power in transmission-constrained areas.
  • [0010] Dispatch below maximum output (curtailment) can be more of an issue for wind and solar generators than it is for fossil generation units because of differences in their cost structures. The economics of wind and solar generation depend on the ability to generate electricity whenever there is sufficient sunlight or wind to power their facilities.
  • [0011] Because wind and solar generators have substantial capital costs but no fuel costs (i.e., minimal variable costs), maximizing output improves their ability to recover capital costs. In contrast, fossil generators have higher variable costs, such as fuel costs. Avoiding these costs can, depending on the economics of a specific generator, to some degree reduce the financial impact of curtailment, especially if the generator's capital costs are included in a utility's rate base.

[0012] Curtailment may result in available energy being wasted (which may not be true to the same extent for fossil generation units which can simply reduce the amount of fuel that is being used). With wind generation, in particular, it may also take some time for a wind farm to become fully operational following curtailment. As such, until the time that the wind farm is fully operational, the wind farm may not be operating with optimum efficiency and/or may not be able to provide power to the grid.

[0013] In an example, a system is described. The system includes a flexible datacenter comprising: a behind-the-meter power input system, a first power distribution system, a datacenter control system, and a first plurality of computing systems powered by the behind-the-meter power input system via the first power distribution system. The flexible datacenter control system is configured to modulate power delivery to the plurality of computing systems based on one or more monitored power system conditions or an operational directive. The system further includes a critical datacenter comprising: a power input system, a second power distribution system, a critical datacenter control system, and a second plurality of computing systems powered by the power input system via the second power distribution system. The system further includes a queue system configured to organize a plurality of computational operations and a first communication link connecting the flexible datacenter, the critical datacenter, and the queue system. The system also includes a routing control system configured to (i) identify, using the queue system, a computational operation to be performed, (ii) determine whether to route the computational operation to the flexible datacenter, and (iii) based on a determination to route the computational operation to the flexible datacenter, cause the computational operation to be sent to the flexible datacenter via the first communication link.

[0014] In another example, a system is described. The system includes a plurality of flexible datacenters, each flexible datacenter comprising: a behind-the-meter power input system, a first power distribution system, a datacenter control system, and a first plurality of computing systems powered by the behind-the-meter power input system. The flexible datacenter control system is configured to modulate power delivery to the plurality of computing systems based on one or more monitored power system conditions or an operational directive. The system further includes a critical datacenter comprising: a power input system, a second power distribution system, a critical datacenter control system, and a second plurality of computing systems powered by the power input system via the second power distribution system. The system also includes a queue system configured to organize a plurality of computational operations and a first communication link connecting the plurality of flexible datacenter, the critical datacenter, and the queue system. The system further includes a routing control system configured to (i) identify, using the queue system, a computational operation to be performed, (ii) determine whether to route the computational operation to a flexible datacenter in the plurality of flexible datacenters, (iii) based on a determination to route the computational operation to a flexible datacenter in the plurality of flexible datacenters, determine a specific flexible datacenter in the plurality of flexible datacenters to route the computational operation to, and (iv) cause the computational operation to be sent to the specific flexible datacenter via the first communication link.

[0015] Other aspects of the present invention will be apparent from the following description and claims.

[0030] One or more embodiments of the present invention are described in detail with reference to the accompanying figures. For consistency, like elements in the various figures are denoted by like reference numerals. In the following detailed description of the present invention, specific details are set forth in order to provide a thorough understanding of the present invention. In other instances, well-known features to one having ordinary skill in the art are not described to avoid obscuring the description of the present invention.

[0031] The embodiments provided herein relate to providing an electrical load “behind the meter” at local stations such that generated power can be directed to the behind-the-meter load instead of onto the grid, typically for intermittent periods of time. “Behind-the-meter” power includes power that is received from a power generation system (for instance, but not limited to, a wind or solar power generation system) prior to the power undergoing step-up transformation to High Voltage class AC power for transmission to the grid. Behind-the-meter power may therefore include power drawn directly from an intermittent grid-scale power generation system (e.g. a wind farm or a solar array) and not from the grid.

[0032] The embodiments herein provide an advantage when, for example, the power system conditions exhibit excess local power generation at a local station level, excess local power generation that a grid cannot receive, local power generation that is subject to economic curtailment, local power generation that is subject to reliability curtailment, local power generation that is subject to power factor correction, low local power generation, start up local power generation situations, transient local power generation situations, conditions where the cost for power is economically viable (e.g., low cost for power), or testing local power generation situations where there is an economic advantage to using local behind-the-meter power generation. This is not least because the excess power can be utilized by the behind-the-meter electrical load rather than going to waste. In addition, by providing an electrical load behind-the-meter rather than connected to the grid, electrical transmission losses resulting from transmission of power through the grid can be reduced. In addition, any degradation in the power generation systems which may result from curtailment may be reduced.

[0033] Preferably, controlled computing systems that consume electrical power through computational operations can provide a behind-the-meter electrical load that can be granularly ramped up and down quickly under the supervision of control systems that monitor power system conditions and direct the power state and/or computational activity of the computing systems. In one embodiment, the computing systems preferably receive all their power for computational operations from a behind-the-meter power source. In another embodiment, the computing systems may additionally include a connection to grid power for supervisory and communication systems or other ancillary needs. In yet another embodiment, the computing systems can be configured to switch between behind-the-meter power and grid power under the direction of a control system.

[0034] Among other benefits, a computing system load with controlled granular ramping allows a local station to avoid negative power market pricing and to respond quickly to grid directives. Local stations may include a station capable of controlling power direction and supply and may be referred to as substations or station controls. For instance, a local station may control access to power from the power grid.

[0035] Various computing systems can provide granular behind-the-meter ramping. Preferably the computing systems perform computational tasks that are immune to, or not substantially hindered by, frequent interruptions or slow-downs in processing as the computing systems ramp up and down. In one embodiment, control systems can activate or de-activate one or more computing systems in an array of similar or identical computing systems sited behind the meter. For example, one or more blockchain miners, or groups of blockchain miners, in an array may be turned on or off. In another embodiment, control systems can direct time-insensitive computational tasks to computational hardware, such as CPUs and GPUs, sited behind the meter, while other hardware is sited in front of the meter and possibly remote from the behind-the-meter hardware. Any parallel computing processes, such as Monte Carlo simulations, batch processing of financial transactions, graphics rendering, and oil and gas field simulation models are all good candidates for such interruptible computational operations.

[0036] A typical datacenter provides computational resources to support computational operations. Particularly, one or more enterprises may assign computational operations to the typical datacenter with expectations that the typical datacenter reliably provides resources to support the computational operations, such as processing abilities, networking, and/or storage. The computational operations assigned to a typical datacenter may vary in their requirements. Some computational operations may require low-latency processing, or are extremely time sensitive, or require a high degree of support and reliability from the datacenter. Other computational operations are not time sensitive and can be batch processed over time, or can be distributed across multiple computational systems with interruptible parallel processing, or can be run on specialized hardware for more efficient processing. Therefore, there can be an economic advantage to sending computational operations to different types of datacenters that have different costs for different types of computational operations. According to embodiments disclosed here, a system of one or more high-compute-cost critical datacenters and one or more low-compute-cost flexible datacenters provides such an economic advantage.

[0037] A critical datacenter may have a similar configuration to a typical datacenter. Due to the need to reliably provide computing resources to support critical operations, a critical datacenter is preferably connected to a reliable power source, such as the power grid with multiple redundant power supply systems. The power grid will offer a constant power supply that the critical datacenter uses to meet the needs of assigned computational operations. However, the grid power that enables the critical datacenter to provide the required computational resources is a very significant expense.

[0038] In addition, it might also be difficult to estimate future costs associated with utilizing the critical datacenter for critical computational operations. The cost for power from the power grid can fluctuate in price depending on various factors, including the location of the critical datacenter using the power, the overall demand for the power, weather conditions, fuel costs endured by suppliers of the power to the power grid, and time of use, among others.

[0039] Example embodiments presented herein aim to reduce the cost associated with using a critical database to perform computational operations. In particular, some examples involve using one or more flexible datacenters to offload computational operations from a critical datacenter. A flexible datacenter may also initially assume and support computational operations rather than a critical datacenter supporting the computational operations. As described below with regards to FIG. 2, a flexible datacenter may use behind-the-meter power in order to provide processing abilities and other computing resources. By using behind-the-meter power from renewable energy sources (e.g., wind farm 600, solar farm 700) and other behind-the-meter power sources, a flexible datacenter can provide computing resources at very low costs, significantly below the costs incurred to power a critical datacenter. As such, one or more flexible datacenters may assist a critical datacenter in efficiently handling computational operations assigned to the critical datacenter by one or more enterprises. The addition of one or more flexible datacenters can also increase the quantity of computational resources available to perform and support computational operations.

[0040] In addition, some example implementations presented herein involve the use of queue system. A queue system is a data model that can organize computational operations for subsequent access and distribution to available datacenters, including critical datacenters and flexible datacenters. In various embodiments, one or more control systems may support and maintain the queue system. For example, the remote master control system 420, the local station control system 410, or the datacenter control system 220 may support operations of queue system. By using the queue system, computational operations can be organized and distributed efficiently to datacenters that have the capabilities and availabilities to handle the computational operations.

[0041] The structure and operation of the queue system may vary within examples. Particularly, the queue system may include one or more queues that arrange the computational operations according to a variety of factors. Example factors include parameters of each computational operation, deadlines for completing each computational operation, time of submission of each computational operation, computing resources required for each computational operation, and price obtained to perform each computational operation, among others. As such, the queue system may organize and maintain the computational operations waiting for performance by the critical datacenter or the flexible datacenter.

[0042] Some examples may involve using a centralized queue system maintained centrally by a control system. For instance, the remote master control system 420 may support and maintain the queue system as a centralized queue. The remote master control system 420 may receive new computational operations requests from enterprises or other sources. The remote master control system 420 may then place each new computational operation request in the centralized queue system for access and distribution to one or more datacenters. In some examples, the remote master control system 420 may manage the distribution of computational operations to available datacenters. The distribution may involve factoring the availability of the datacenters, the type of datacenters (e.g., critical datacenter or flexible datacenter), and the cost for a datacenter to perform the computational operation, among others. In other examples, the data control systems at each datacenter may access and obtain computational operations from the centralized queue system.

[0043] Some example embodiments may involve using a decentralized queue system. Particularly, the queue system may be distributed across multiple control systems. For example, a first queue subsystem of the decentralized queue system may receive and organize computational operations for a first set of datacenters to support and a second queue subsystem of the decentralized queue system may receive and organize computational operations for a second set of datacenters to support. The decentralized queue system may require less computational resources to support and maintain. Further, by dividing computational operations into multiple queue subsystems, sets of datacenters may be similarly divided to quickly and efficiently address computational operations within each subsystem.

[0044] In various embodiments, the access to computational operations may differ. In some examples, the data control system or computing system managing the queue system may control access to computational operations assigned to the queue system. Particularly, the remote master control system 420 or another control system may communicate and distribute computational operations to critical datacenters and flexible datacenters. In other examples, the control system at each datacenter may access and assume responsibility for computational operations placed within the queue system. This way, the datacenter control system 220 may manage computational operations performed at the flexible datacenter 200 based on capabilities and availability of the computing systems 100 at the flexible datacenter 200.

[0045] In some examples, a critical datacenter may offload some or all of a set of computational operations to the queue system. The critical datacenter may also release some or all of the set of computational operations directly to flexible datacenter. Particularly, when conditions signal that use of a flexible datacenter is economically viable (i.e., at the same or decreased costs relative to using power from the power grid at the critical datacenter), a flexible datacenter may assume some or even all of one or more sets of computational operations from the critical datacenter.

[0046] In some instances, a flexible datacenter may assume less critical computational operations from the queue system or directly from a critical datacenter. This way, the critical datacenter may offload less critical computational operations directly or indirectly to a flexible datacenter to support and manage. In such a configuration, the critical datacenter may continue to support critical operations assigned to the critical datacenter by one or more enterprises while offloading less critical operations directly or indirectly to one or more flexible datacenters. As a result, the critical datacenter may ensure that the critical operations remain supported by computational resources powered by grid power.

[0047] In other examples, a flexible datacenter may assume critical operations, augmenting the resources provided by the critical datacenter. Particularly, situations can arise where the flexible datacenter can operate at a lower cost than the critical datacenter. For instance, one or more behind-the-meter power sources (e.g., wind farm 600, solar farm 700) may enable the flexible datacenter to operate at a lower cost than the critical datacenter. As a result, using the flexible datacenter instead of the critical datacenter can lower the costs required to support assigned computing operations. If the situation changes such that the flexible datacenter is no longer less costly than the critical datacenter, the critical datacenter can reassume the computing operations from the flexible datacenter.

[0048] As shown herein, by having one or more flexible datacenters powered by one or more behind-the-meter power sources available, computing operations can be managed in a dynamic manner between the critical datacenter and the flexible datacenters. The dynamic management can lower costs and, in some cases, decrease the time needed to complete time-sensitive computing operations submitted to the critical datacenter by an enterprise.

[0049] In some embodiments, one or more flexible datacenters may perform computing processes obtained through an auction process. The one or more flexible datacenters may use behind-the-meter power to acquire and perform computational operations made available via the auction process. For example, an auction process may be used to connect companies or entities requesting computational operations to be supported and performed at one or more datacenters with datacenters capable of handling the computational operations. Particularly, the auction process may involve datacenters placing bids in a competition for the various computational operations available in the auction process. For instance, the datacenter that bids to perform a computational operation at the lowest cost may win and receive the right to enter into a contract to perform the computational for the priced bid or subsequently agreed upon. As such, flexible datacenters may compete and receive the right to perform computational operations by bidding prices based on using low cost power, such as behind-the-meter power. A datacenter control system of a flexible datacenter may monitor available computational operations in multiple auctions simultaneously to determine when to bid for computational operations based on the cost of power available and competing bids.

[0050] FIG. 1 shows a computing system 100 in accordance with one or more embodiments of the present invention. Computing system 100 may include one or more central processing units (singular “CPU” or plural “CPUs”) 105, host bridge 110, input/output (“IO”) bridge 115, graphics processing units (singular “GPU” or plural “GPUs”) 125, and/or application-specific integrated circuits (singular “ASIC or plural “ASICs”) (not shown) disposed on one or more printed circuit boards (not shown) that are configured to perform computational operations. Each of the one or more CPUs 105, GPUs 125, or ASICs (not shown) may be a single-core (not independently illustrated) device or a multi-core (not independently illustrated) device. Multi-core devices typically include a plurality of cores (not shown) disposed on the same physical die (not shown) or a plurality of cores (not shown) disposed on multiple die (not shown) that are collectively disposed within the same mechanical package (not shown).

[0051] CPU 105 may be a general purpose computational device typically configured to execute software instructions. CPU 105 may include an interface 108 to host bridge 110, an interface 118 to system memory 120, and an interface 123 to one or more IO devices, such as, for example, one or more GPUs 125. GPU 125 may serve as a specialized computational device typically configured to perform graphics functions related to frame buffer manipulation. However, one of ordinary skill in the art will recognize that GPU 125 may be used to perform non-graphics related functions that are computationally intensive. In certain embodiments, GPU 125 may interface 123 directly with CPU 125 (and interface 118 with system memory 120 through CPU 105). In other embodiments, GPU 125 may interface 121 with host bridge 110 (and interface 116 or 118 with system memory 120 through host bridge 110 or CPU 105 depending on the application or design). In still other embodiments, GPU 125 may interface 133 with 10 bridge 115 (and interface 116 or 118 with system memory 120 through host bridge 110 or CPU 105 depending on the application or design). The functionality of GPU 125 may be integrated, in whole or in part, with CPU 105.

[0052] Host bridge 110 may be an interface device configured to interface between the one or more computational devices and IO bridge 115 and, in some embodiments, system memory 120. Host bridge 110 may include an interface 108 to CPU 105, an interface 113 to IO bridge 115, for embodiments where CPU 105 does not include an interface 118 to system memory 120, an interface 116 to system memory 120, and for embodiments where CPU 105 does not include an integrated GPU 125 or an interface 123 to GPU 125, an interface 121 to GPU 125. The functionality of host bridge 110 may be integrated, in whole or in part, with CPU 105. IO bridge 115 may be an interface device configured to interface between the one or more computational devices and various IO devices (e.g., 140, 145) and IO expansion, or add-on, devices (not independently illustrated). IO bridge 115 may include an interface 113 to host bridge 110, one or more interfaces 133 to one or more IO expansion devices 135, an interface 138 to keyboard 140, an interface 143 to mouse 145, an interface 148 to one or more local storage devices 150, and an interface 153 to one or more network interface devices 155. The functionality of IO bridge 115 may be integrated, in whole or in part, with CPU 105 or host bridge 110. Each local storage device 150, if any, may be a solid-state memory device, a solid-state memory device array, a hard disk drive, a hard disk drive array, or any other non-transitory computer readable medium. Network interface device 155 may provide one or more network interfaces including any network protocol suitable to facilitate networked communications.

[0053] Computing system 100 may include one or more network-attached storage devices 160 in addition to, or instead of, one or more local storage devices 150. Each network-attached storage device 160, if any, may be a solid-state memory device, a solid-state memory device array, a hard disk drive, a hard disk drive array, or any other non-transitory computer readable medium. Network-attached storage device 160 may or may not be collocated with computing system 100 and may be accessible to computing system 100 via one or more network interfaces provided by one or more network interface devices 155.

[0054] One of ordinary skill in the art will recognize that computing system 100 may be a conventional computing system or an application-specific computing system. In certain embodiments, an application-specific computing system may include one or more ASICs (not shown) that are configured to perform one or more functions, such as distributed computing processes or hashing, in a more efficient manner. The one or more ASICs (not shown) may interface directly with CPU 105, host bridge 110, or GPU 125 or interface through IO bridge 115. Alternatively, in other embodiments, an application-specific computing system may be reduced to only those components necessary to perform a desired function in an effort to reduce one or more of chip count, printed circuit board footprint, thermal design power, and power consumption. The one or more ASICs (not shown) may be used instead of one or more of CPU 105, host bridge 110, TO bridge 115, or GPU 125. In such systems, the one or more ASICs may incorporate sufficient functionality to perform certain network and computational functions in a minimal footprint with substantially fewer component devices.

[0055] As such, one of ordinary skill in the art will recognize that CPU 105, host bridge 110, IO bridge 115, GPU 125, or ASIC (not shown) or a subset, superset, or combination of functions or features thereof, may be integrated, distributed, or excluded, in whole or in part, based on an application, design, or form factor in accordance with one or more embodiments of the present invention. Thus, the description of computing system 100 is merely exemplary and not intended to limit the type, kind, or configuration of component devices that constitute a computing system 100 suitable for performing computing operations in accordance with one or more embodiments of the present invention.

[0056] One of ordinary skill in the art will recognize that computing system 100 may be a stand-alone, laptop, desktop, server, blade, or rack mountable system and may vary based on an application or design.

[0057] FIG. 2 shows a flexible datacenter 200 in accordance with one or more embodiments of the present invention. Flexible datacenter 200 may include a mobile container 205, a behind-the-meter power input system 210, a power distribution system 215, a climate control system (e.g., 250, 260, 270, 280, and/or 290), a datacenter control system 220, and a plurality of computing systems 100 disposed in one or more racks 240. Datacenter control system 220 may be a computing system (e.g., 100 of FIG. 1) configured to dynamically modulate power delivery to one or more computing systems 100 disposed within flexible datacenter 200 based on behind-the-meter power availability or an operational directive from a local station control system (not shown), a remote master control system (not shown), or a grid operator (not shown).

[0058] In certain embodiments, mobile container 205 may be a storage trailer disposed on wheels and configured for rapid deployment. In other embodiments, mobile container 205 may be a storage container (not shown) configured for placement on the ground and potentially stacked in a vertical or horizontal manner (not shown). In still other embodiments, mobile container 205 may be an inflatable container, a floating container, or any other type or kind of container suitable for housing a mobile datacenter 200. And in still other embodiments, flexible datacenter 200 might not include a mobile container. For example, flexible datacenter 200 may be situated within a building or another type of stationary environment.

[0059] Flexible datacenter 200 may be rapidly deployed on site near a source of unutilized behind-the-meter power generation. Behind-the-meter power input system 210 may be configured to input power to flexible datacenter 200. Behind-the-meter power input system 210 may include a first input (not independently illustrated) configured to receive three-phase behind-the-meter alternating current (“AC”) voltage. In certain embodiments, behind-the-meter power input system 210 may include a supervisory AC-to-AC step-down transformer (not shown) configured to step down three-phase behind-the-meter AC voltage to single-phase supervisory nominal AC voltage or a second input (not independently illustrated) configured to receive single-phase supervisory nominal AC voltage from the local station (not shown) or a metered source (not shown). Behind-the-meter power input system 210 may provide single-phase supervisory nominal AC voltage to datacenter control system 220, which may remain powered at almost all times to control the operation of flexible datacenter 200. The first input (not independently illustrated) or a third input (not independently illustrated) of behind-the-meter power input system 210 may direct three-phase behind-the-meter AC voltage to an operational AC-to-AC step-down transformer (not shown) configured to controllably step down three-phase behind-the-meter AC voltage to three-phase nominal AC voltage. Datacenter control system 220 may controllably enable or disable generation or provision of three-phase nominal AC voltage by the operational AC-to-AC step-down transformer (not shown).

[0060] Behind-the-meter power input system 210 may provide three phases of three-phase nominal AC voltage to power distribution system 215. Power distribution system 215 may controllably provide a single phase of three-phase nominal AC voltage to each computing system 100 or group 240 of computing systems 100 disposed within flexible datacenter 200. Datacenter control system 220 may controllably select which phase of three-phase nominal AC voltage that power distribution system 215 provides to each computing system 100 or group 240 of computing systems 100. In this way, datacenter control system 220 may modulate power delivery by either ramping-up flexible datacenter 200 to fully operational status, ramping-down flexible datacenter 200 to offline status (where only datacenter control system 220 remains powered), reducing power consumption by withdrawing power delivery from, or reducing power to, one or more computing systems 100 or groups 240 of computing systems 100, or modulating a power factor correction factor for the local station by controllably adjusting which phases of three-phase nominal AC voltage are used by one or more computing systems 100 or groups 240 of computing systems 100. In some embodiments, flexible datacenter 200 may receive DC power to power computing systems 100.

[0061] Flexible datacenter 200 may include a climate control system (e.g., 250, 260, 270, 280, 290) configured to maintain the plurality of computing systems 100 within their operational temperature range. In certain embodiments, the climate control system may include an air intake 250, an evaporative cooling system 270, a fan 280, and an air outtake 260. In other embodiments, the climate control system may include an air intake 250, an air conditioner or refrigerant cooling system 290, and an air outtake 260. In still other embodiments, the climate control system may include a computer room air conditioner system (not shown), a computer room air handler system (not shown), or an immersive cooling system (not shown). One of ordinary skill in the art will recognize that any suitable heat extraction system (not shown) configured to maintain the operation of the plurality of computing systems 100 within their operational temperature range may be used in accordance with one or more embodiments of the present invention.

[0062] Flexible datacenter 200 may include a battery system (not shown) configured to convert three-phase nominal AC voltage to nominal DC voltage and store power in a plurality of storage cells. The battery system (not shown) may include a DC-to-AC inverter configured to convert nominal DC voltage to three-phase nominal AC voltage for flexible datacenter 200 use. Alternatively, the battery system (not shown) may include a DC-to-AC inverter configured to convert nominal DC voltage to single-phase nominal AC voltage to power datacenter control system 220.

[0063] One of ordinary skill in the art will recognize that a voltage level of three-phase behind-the-meter AC voltage may vary based on an application or design and the type or kind of local power generation. As such, a type, kind, or configuration of the operational AC-to-AC step down transformer (not shown) may vary based on the application or design. In addition, the frequency and voltage level of three-phase nominal AC voltage, single-phase nominal AC voltage, and nominal DC voltage may vary based on the application or design in accordance with one or more embodiments of the present invention.

[0064] FIG. 3 shows a three-phase power distribution of a flexible datacenter 200 in accordance with one or more embodiments of the present invention. Flexible datacenter 200 may include a plurality of racks 240, each of which may include one or more computing systems 100 disposed therein. As discussed above, the behind-the-meter power input system (210 of FIG. 2) may provide three phases of three-phase nominal AC voltage to the power distribution system (215 of FIG. 2). The power distribution system (215 of FIG. 2) may controllably provide a single phase of three-phase nominal AC voltage to each computing system 100 or group 240 of computing systems 100 disposed within flexible datacenter 200. For example, a flexible datacenter 200 may include eighteen racks 240, each of which may include eighteen computing systems 100. The power distribution system (215 of FIG. 2) may control which phase of three-phase nominal AC voltage is provided to one or more computing systems 100, a rack 240 of computing systems 100, or a group (e.g., 310, 320, or 330) of racks 240 of computing systems 100.

[0065] In the figure, for purposes of illustration only, eighteen racks 240 are divided into a first group of six racks 310, a second group of six racks 320, and a third group of six racks 330, where each rack contains eighteen computing systems 100. The power distribution system (215 of FIG. 2) may, for example, provide a first phase of three-phase nominal AC voltage to the first group of six racks 310, a second phase of three-phase nominal AC voltage to the second group of six racks 320, and a third phase of three-phase nominal AC voltage to the third group of six racks 330. If the flexible datacenter (200 of FIG. 2) receives an operational directive from the local station (not shown) to provide power factor correction, the datacenter control system (220 of FIG. 2) may direct the power distribution system (215 of FIG. 2) to adjust which phase or phases of three-phase nominal AC voltage are used to provide the power factor correction required by the local station (not shown) or grid operator (not shown). One of ordinary skill in the art will recognize that, in addition to the power distribution, the load may be varied by adjusting the number of computing systems 100 operatively powered. As such, the flexible datacenter (200 of FIG. 2) may be configured to act as a capacitive or inductive load to provide the appropriate reactance necessary to achieve the power factor correction required by the local station (not shown).

[0066] FIG. 4 shows a control distribution scheme 400 of a flexible datacenter 200 in accordance with one or more embodiments of the present invention. Datacenter control system 220 may independently, or cooperatively with one or more of local station control system 410, remote master control system 420, and grid operator 440, modulate power delivery to flexible datacenter 200. Specifically, power delivery may be dynamically adjusted based on conditions or operational directives.

[0067] Local station control system 410 may be a computing system (e.g., 100 of FIG. 1) that is configured to control various aspects of the local station (not independently illustrated) that generates power and sometimes generates unutilized behind-the-meter power. Local station control system 410 may communicate with remote master control system 420 over a networked connection 430 and with datacenter control system 220 over a networked or hardwired connection 415. Remote master control system 420 may be a computing system (e.g., 100 of FIG. 1) that is located offsite, but connected via a network connection 425 to datacenter control system 220, that is configured to provide supervisory or override control of flexible datacenter 200 or a fleet (not shown) of flexible datacenters 200. Grid operator 440 may be a computing system (e.g., 100 of FIG. 1) that is configured to control various aspects of the grid (not independently illustrated) that receives power from the local station (not independently illustrated). Grid operator 440 may communicate with local station control system 440 over a networked or hardwired connection 445.

[0068] Datacenter control system 220 may monitor unutilized behind-the-meter power availability at the local station (not independently illustrated) and determine when a datacenter ramp-up condition is met. Unutilized behind-the-meter power availability may include one or more of excess local power generation, excess local power generation that the grid cannot accept, local power generation that is subject to economic curtailment, local power generation that is subject to reliability curtailment, local power generation that is subject to power factor correction, conditions where the cost for power is economically viable (e.g., low cost for power), situations where local power generation is prohibitively low, start up situations, transient situations, or testing situations where there is an economic advantage to using locally generated behind-the-meter power generation, specifically power available at little to no cost and with no associated transmission or distribution losses or costs.

[0069] The datacenter ramp-up condition may be met if there is sufficient behind-the-meter power availability and there is no operational directive from local station control system 410, remote master control system 420, or grid operator 440 to go offline or reduce power. As such, datacenter control system 220 may enable 435 behind-the-meter power input system 210 to provide three-phase nominal AC voltage to the power distribution system (215 of FIG. 2) to power the plurality of computing systems (100 of FIG. 2) or a subset thereof. Datacenter control system 220 may optionally direct one or more computing systems (100 of FIG. 2) to perform predetermined computational operations (e.g., distributed computing processes). For example, if the one or more computing systems (100 of FIG. 2) are configured to perform blockchain hashing operations, datacenter control system 220 may direct them to perform blockchain hashing operations for a specific blockchain application, such as, for example, Bitcoin, Litecoin, or Ethereum. Alternatively, one or more computing systems (100 of FIG. 2) may be configured to independently receive a computational directive from a network connection (not shown) to a peer-to-peer blockchain network (not shown) such as, for example, a network for a specific blockchain application, to perform predetermined computational operations.

[0070] Remote master control system 420 may specify to datacenter control system 220 what sufficient behind-the-meter power availability constitutes, or datacenter control system 220 may be programmed with a predetermined preference or criteria on which to make the determination independently. For example, in certain circumstances, sufficient behind-the-meter power availability may be less than that required to fully power the entire flexible datacenter 200. In such circumstances, datacenter control system 220 may provide power to only a subset of computing systems (100 of FIG. 2), or operate the plurality of computing systems (100 of FIG. 2) in a lower power mode, that is within the sufficient, but less than full, range of power that is available.

[0071] While flexible datacenter 200 is online and operational, a datacenter ramp-down condition may be met when there is insufficient, or anticipated to be insufficient, behind-the-meter power availability or there is an operational directive from local station control system 410, remote master control system 420, or grid operator 440. Datacenter control system 220 may monitor and determine when there is insufficient, or anticipated to be insufficient, behind-the-meter power availability. As noted above, sufficiency may be specified by remote master control system 420 or datacenter control system 220 may be programmed with a predetermined preference or criteria on which to make the determination independently. An operational directive may be based on current dispatchability, forward looking forecasts for when unutilized behind-the-meter power is, or is expected to be, available, economic considerations, reliability considerations, operational considerations, or the discretion of the local station 410, remote master control 420, or grid operator 440. For example, local station control system 410, remote master control system 420, or grid operator 440 may issue an operational directive to flexible datacenter 200 to go offline and power down. When the datacenter ramp-down condition is met, datacenter control system 220 may disable power delivery to the plurality of computing systems (100 of FIG. 2). Datacenter control system 220 may disable 435 behind-the-meter power input system 210 from providing three-phase nominal AC voltage to the power distribution system (215 of FIG. 2) to power down the plurality of computing systems (100 of FIG. 2), while datacenter control system 220 remains powered and is capable of rebooting flexible datacenter 200 when unutilized behind-the-meter power becomes available again.

[0072] While flexible datacenter 200 is online and operational, changed conditions or an operational directive may cause datacenter control system 220 to modulate power consumption by flexible datacenter 200. Datacenter control system 220 may determine, or local station control system 410, remote master control system 420, or grid operator 440 may communicate, that a change in local conditions may result in less power generation, availability, or economic feasibility, than would be necessary to fully power flexible datacenter 200. In such situations, datacenter control system 220 may take steps to reduce or stop power consumption by flexible datacenter 200 (other than that required to maintain operation of datacenter control system 220). Alternatively, local station control system 410, remote master control system 420, or grid operator 440, may issue an operational directive to reduce power consumption for any reason, the cause of which may be unknown. In response, datacenter control system 220 may dynamically reduce or withdraw power delivery to one or more computing systems (100 of FIG. 2) to meet the dictate. Datacenter control system 220 may controllably provide three-phase nominal AC voltage to a smaller subset of computing systems (100 of FIG. 2) to reduce power consumption. Datacenter control system 220 may dynamically reduce the power consumption of one or more computing systems (100 of FIG. 2) by reducing their operating frequency or forcing them into a lower power mode through a network directive.

[0073] One of ordinary skill in the art will recognize that datacenter control system 220 may be configured to have a number of different configurations, such as a number or type or kind of computing systems (100 of FIG. 2) that may be powered, and in what operating mode, that correspond to a number of different ranges of sufficient and available unutilized behind-the-meter power availability. As such, datacenter control system 220 may modulate power delivery over a variety of ranges of sufficient and available unutilized behind-the-meter power availability.

[0074] FIG. 5 shows a control distribution of a fleet 500 of flexible datacenters 200 in accordance with one or more embodiments of the present invention. The control distribution of a flexible datacenter 200 shown and described with respect to FIG. 4 may be extended to a fleet 500 of flexible datacenters 200. For example, a first local station (not independently illustrated), such as, for example, a wind farm (not shown), may include a first plurality 510 of flexible datacenters 200a through 200d, which may be collocated or distributed across the local station (not shown). A second local station (not independently illustrated), such as, for example, another wind farm or a solar farm (not shown), may include a second plurality 520 of flexible datacenters 200e through 200h, which may be collocated or distributed across the local station (not shown). One of ordinary skill in the art will recognize that the number of flexible datacenters 200 deployed at a given station and the number of stations within the fleet may vary based on an application or design in accordance with one or more embodiments of the present invention.

[0075] Remote master control system 420 may provide supervisory control over fleet 500 of flexible datacenters 200 in a similar manner to that shown and described with respect to FIG. 4, with the added flexibility to make high level decisions with respect to fleet 500 that may be counterintuitive to a given station. Remote master control system 420 may make decisions regarding the issuance of operational directives to a given local station based on, for example, the status of each local station where flexible datacenters 200 are deployed, the workload distributed across fleet 500, and the expected computational demand required for the expected workload. In addition, remote master control system 420 may shift workloads from a first plurality 510 of flexible datacenters 200 to a second plurality 520 of flexible datacenters 200 for any reason, including, for example, a loss of unutilized behind-the-meter power availability at one local station and the availability of unutilized behind-the-meter power at another local station.

[0076] FIG. 6 shows a flexible datacenter 200 powered by one or more wind turbines 610 in accordance with one or more embodiments of the present invention. A wind farm 600 typically includes a plurality of wind turbines 610, each of which intermittently generates a wind-generated AC voltage. The wind-generated AC voltage may vary based on a type, kind, or configuration of farm 600, turbine 610, and incident wind speed. The wind-generated AC voltage is typically input into a turbine AC-to-AC step-up transformer (not shown) that is disposed within the nacelle (not independently illustrated) or at the base of the mast (not independently illustrated) of turbine 610. The turbine AC-to-AC step up transformer (not shown) outputs three-phase wind-generated AC voltage 620. Three-phase wind-generated AC voltage 620 produced by the plurality of wind turbines 610 is collected 625 and provided 630 to another AC-to-AC step-up transformer 640 that steps up three-phase wind-generated AC voltage 620 to three-phase grid AC voltage 650 suitable for delivery to grid 660. Three-phase grid AC voltage 650 may be stepped down with an AC-to-AC step-down transformer 670 configured to produce three-phase local station AC voltage 680 provided to local station 690. One of ordinary skill in the art will recognize that the actual voltage levels may vary based on the type, kind, or number of wind turbines 610, the configuration or design of wind farm 600, and grid 660 that it feeds into.

[0077] The output side of AC-to-AC step-up transformer 640 that connects to grid 660 may be metered and is typically subject to transmission and distribution costs. In contrast, power consumed on the input side of AC-to-AC step-up transformer 640 may be considered behind-the-meter and is typically not subject to transmission and distribution costs. As such, one or more flexible datacenters 200 may be powered by three-phase wind-generated AC voltage 620. Specifically, in wind farm 600 applications, the three-phase behind-the-meter AC voltage used to power flexible datacenter 200 may be three-phase wind-generated AC voltage 620. As such, flexible datacenter 200 may reside behind-the-meter, avoid transmission and distribution costs, and may be dynamically powered when unutilized behind-the-meter power is available.

[0078] Unutilized behind-the-meter power availability may occur when there is excess local power generation. In high wind conditions, wind farm 600 may generate more power than, for example, AC-to-AC step-up transformer 640 is rated for. In such situations, wind farm 600 may have to take steps to protect its equipment from damage, which may include taking one or more turbines 610 offline or shunting their voltage to dummy loads or ground. Advantageously, one or more flexible datacenters 200 may be used to consume power on the input side of AC-to-AC step-up transformer 640, thereby allowing wind farm 600 to operate equipment within operating ranges while flexible datacenter 200 receives behind-the-meter power without transmission or distribution costs. The local station control system (not independently illustrated) of local station 690 may issue an operational directive to the one or more flexible datacenters 200 or to the remote master control system (420 of FIG. 4) to ramp-up to the desired power consumption level. When the operational directive requires the cooperative action of multiple flexible datacenters 200, the remote mater control system (420 of FIG. 4) may determine how to power each individual flexible datacenter 200 in accordance with the operational directive or provide an override to each flexible datacenter 200.

[0079] Another example of unutilized behind-the-meter power availability is when grid 660 cannot, for whatever reason, take the power being produced by wind farm 600. In such situations, wind farm 600 may have to take one or more turbines 610 offline or shunt their voltage to dummy loads or ground. Advantageously, one or more flexible datacenters 200 may be used to consume power on the input side of AC-to-AC step-up transformer 640, thereby allowing wind farm 600 to either produce power to grid 660 at a lower level or shut down transformer 640 entirely while flexible datacenter 200 receives behind-the-meter power without transmission or distribution costs. The local station control system (not independently illustrated) of local station 690 or the grid operator (not independently illustrated) of grid 660 may issue an operational directive to the one or more flexible datacenters 200 or to the remote master control system (420 of FIG. 4) to ramp-up to the desired power consumption level. When the operational directive requires the cooperative action of multiple flexible datacenters 200, the remote master control system (420 of FIG. 4) may determine how to power each individual flexible datacenter 200 in accordance with the operational directive or provide an override to each flexible datacenter 200.

[0080] Another example of unutilized behind-the-meter power availability is when wind farm 600 is selling power to grid 660 at a negative price that is offset by a production tax credit. In certain circumstances, the value of the production tax credit may exceed the price wind farm 600 would have to pay to grid 660 to offload their generated power. Advantageously, one or more flexible datacenters 200 may be used to consume power behind-the-meter, thereby allowing wind farm 600 to produce and obtain the production tax credit, but sell less power to grid 660 at the negative price. The local station control system (not independently illustrated) of local station 690 may issue an operational directive to the one or more flexible datacenters 200 or to the remote master control system (420 of FIG. 4) to ramp-up to the desired power consumption level. When the operational directive requires the cooperative action of multiple flexible datacenter 200, the remote master control system (420 of FIG. 4) may determine how to power each individual flexible datacenter 200 in accordance with the operational directive or provide an override to each flexible datacenter 200.

[0081] Another example of unutilized behind-the-meter power availability is when wind farm 600 is selling power to grid 660 at a negative price because grid 660 is oversupplied or is instructed to stand down and stop producing altogether. The grid operator (not independently illustrated) may select certain power generation stations to go offline and stop producing power to grid 660. Advantageously, one or more flexible datacenters 200 may be used to consume power behind-the-meter, thereby allowing wind farm 600 to stop producing power to grid 660, but making productive use of the power generated behind-the-meter without transmission or distribution costs. The local station control system (not independently illustrated) of the local station 690 or the grid operator (not independently illustrated) of grid 660 may issue an operational directive to the one or more flexible datacenters 200 or to the remote master control system (420 of FIG. 4) to ramp-up to the desired power consumption level. When the operational directive requires the cooperative action of multiple flexible datacenters 200, the remote master control system (420 of FIG. 4) may determine how to power each individual flexible datacenter 200 in accordance with the operational directive or provide an override to each flexible datacenter 200.

[0082] Another example of unutilized behind-the-meter power availability is when wind farm 600 is producing power to grid 660 that is unstable, out of phase, or at the wrong frequency, or grid 660 is already unstable, out of phase, or at the wrong frequency for whatever reason. The grid operator (not independently illustrated) may select certain power generation stations to go offline and stop producing power to grid 660. Advantageously, one or more flexible datacenters 200 may be used to consume power behind-the-meter, thereby allowing wind farm 600 to stop producing power to grid 660, but make productive use of the power generated behind-the-meter without transmission or distribution costs. The local station control system (not independently illustrated) of local station 690 may issue an operational directive to the one or more flexible datacenters 200 or to the remote master control system (420 of FIG. 4) to ramp-up to the desired power consumption level. When the operational directive requires the cooperative action of multiple flexible datacenters 200, the remote master control system (420 of FIG. 4) may determine how to power each individual flexible datacenter 200 in accordance with the operational directive or provide an override to each flexible datacenter 200.

[0083] Further examples of unutilized behind-the-meter power availability is when wind farm 600 experiences low wind conditions that make it not economically feasible to power up certain components, such as, for example, the local station (not independently illustrated), but there may be sufficient behind-the-meter power availability to power one or more flexible datacenters 200. Similarly, unutilized behind-the-meter power availability may occur when wind farm 600 is starting up, or testing, one or more turbines 610. Turbines 610 are frequently offline for installation, maintenance, and service and must be tested prior to coming online as part of the array. One or more flexible datacenters 200 may be powered by one or more turbines 610 that are offline from farm 600. The above-noted examples of when unutilized behind-the-meter power is available are merely exemplary and are not intended to limit the scope of what one of ordinary skill in the art would recognize as unutilized behind-the-meter power availability. Unutilized behind-the-meter power availability may occur anytime there is power available and accessible behind-the-meter that is not subject to transmission and distribution costs and there is an economic advantage to using it.

[0084] One of ordinary skill in the art will recognize that wind farm 600 and wind turbine 610 may vary based on an application or design in accordance with one or more embodiments of the present invention.

[0085] FIG. 7 shows a flexible datacenter 200 powered by one or more solar panels 710 in accordance with one or more embodiments of the present invention. A solar farm 700 typically includes a plurality of solar panels 710, each of which intermittently generates a solar-generated DC voltage 720. Solar-generated DC voltage 720 may vary based on a type, kind, or configuration of farm 700, panel 710, and incident sunlight. Solar-generated DC voltage 720 produced by the plurality of solar panels 710 is collected 725 and provided 730 to a DC-to-AC inverter 740 that converts solar-generated DC voltage into three-phase solar-generated AC voltage 750. Three-phase solar-generated AC voltage 750 is provided to an AC-to-AC step-up transformer 760 that steps up three-phase solar-generated AC voltage to three-phase grid AC voltage 790. Three-phase grid AC voltage 790 may be stepped down with an AC-to-AC step-down transformer 785 configured to produce three-phase local station AC voltage 777 provided to local station 775. One of ordinary skill in the art will recognize that the actual voltage levels may vary based on the type, kind, or number of solar panels 710, the configuration or design of solar farm 700, and grid 790 that it feeds into. In some embodiments, the solar farm 700 may provide DC power directly to flexible datacenter 200 without a conversion to AC via the DC-to-AC inverter 740.

[0086] The output side of AC-to-AC step-up transformer 760 that connects to grid 790 may be metered and is typically subject to transmission and distribution costs. In contrast, power consumed on the input side of AC-to-AC step-up transformer 760 may be considered behind-the-meter and is typically not subject to transmission and distribution costs. As such, one or more flexible datacenters 200 may be powered by three-phase solar-generated AC voltage 750. Specifically, in solar farm 700 applications, the three-phase behind-the-meter AC voltage used to power flexible datacenter 200 may be three-phase solar-generated AC voltage 750. As such, flexible datacenter 200 may reside behind-the-meter, avoid transmission and distribution costs, and may be dynamically powered when unutilized behind-the-meter power is available.

[0087] Unutilized behind-the-meter power availability may occur when there is excess local power generation. In high incident sunlight situations, solar farm 700 may generate more power than, for example, AC-to-AC step-up transformer 760 is rated for. In such situations, solar farm 700 may have to take steps to protect its equipment from damage, which may include taking one or more panels 710 offline or shunting their voltage to dummy loads or ground. Advantageously, one or more flexible datacenters 200 may be used to consume power on the input side of AC-to-AC step-up transformer 760, thereby allowing solar farm 700 to operate equipment within operating ranges while flexible datacenter 200 receives behind-the-meter power without transmission or distribution costs. The local station control system (not independently illustrated) of local station 775 may issue an operational directive to the one or more flexible datacenters 200 or to the remote master control system (420 of FIG. 4) to ramp-up to the desired power consumption level. When the operational directive requires the cooperative action of multiple flexible datacenters 200, the remote mater control system (420 of FIG. 4) may determine how to power each individual flexible datacenter 200 in accordance with the operational directive or provide an override to each flexible datacenter 200.

[0088] Another example of unutilized behind-the-meter power availability is when grid 790 cannot, for whatever reason, take the power being produced by solar farm 700. In such situations, solar farm 700 may have to take one or more panels 710 offline or shunt their voltage to dummy loads or ground. Advantageously, one or more flexible datacenters 200 may be used to consume power on the input side of AC-to-AC step-up transformer 760, thereby allowing solar farm 700 to either produce power to grid 790 at a lower level or shut down transformer 760 entirely while flexible datacenter 200 receives behind-the-meter power without transmission or distribution costs. The local station control system (not independently illustrated) of local station 775 or the grid operator (not independently illustrated) of grid 790 may issue an operational directive to the one or more flexible datacenters 200 or to the remote master control system (420 of FIG. 4) to ramp-up to the desired power consumption level. When the operational directive requires the cooperative action of multiple flexible datacenters 200, the remote master control system (420 of FIG. 4) may determine how to power each individual flexible datacenter 200 in accordance with the operational directive or provide an override to each flexible datacenter 200.

[0089] Another example of unutilized behind-the-meter power availability is when solar farm 700 is selling power to grid 790 at a negative price that is offset by a production tax credit. In certain circumstances, the value of the production tax credit may exceed the price solar farm 700 would have to pay to grid 790 to offload their generated power. Advantageously, one or more flexible datacenters 200 may be used to consume power behind-the-meter, thereby allowing solar farm 700 to produce and obtain the production tax credit, but sell less power to grid 790 at the negative price. The local station control system (not independently illustrated) of local station 775 may issue an operational directive to the one or more flexible datacenters 200 or to the remote master control system (420 of FIG. 4) to ramp-up to the desired power consumption level. When the operational directive requires the cooperative action of multiple flexible datacenter 200, the remote master control system (420 of FIG. 4) may determine how to power each individual flexible datacenter 200 in accordance with the operational directive or provide an override to each flexible datacenter 200.

[0090] Another example of unutilized behind-the-meter power availability is when solar farm 700 is selling power to grid 790 at a negative price because grid 790 is oversupplied or is instructed to stand down and stop producing altogether. The grid operator (not independently illustrated) may select certain power generation stations to go offline and stop producing power to grid 790. Advantageously, one or more flexible datacenters 200 may be used to consume power behind-the-meter, thereby allowing solar farm 700 to stop producing power to grid 790, but making productive use of the power generated behind-the-meter without transmission or distribution costs. The local station control system (not independently illustrated) of the local station 775 or the grid operator (not independently illustrated) of grid 790 may issue an operational directive to the one or more flexible datacenters 200 or to the remote master control system (420 of FIG. 4) to ramp-up to the desired power consumption level. When the operational directive requires the cooperative action of multiple flexible datacenters 200, the remote master control system (420 of FIG. 4) may determine how to power each individual flexible datacenter 200 in accordance with the operational directive or provide an override to each flexible datacenter 200.

[0091] Another example of unutilized behind-the-meter power availability is when solar farm 700 is producing power to grid 790 that is unstable, out of phase, or at the wrong frequency, or grid 790 is already unstable, out of phase, or at the wrong frequency for whatever reason. The grid operator (not independently illustrated) may select certain power generation stations to go offline and stop producing power to grid 790. Advantageously, one or more flexible datacenters 200 may be used to consume power behind-the-meter, thereby allowing solar farm 700 to stop producing power to grid 790, but make productive use of the power generated behind-the-meter without transmission or distribution costs. The local station control system (not independently illustrated) of local station 775 may issue an operational directive to the one or more flexible datacenters 200 or to the remote master control system (420 of FIG. 4) to ramp-up to the desired power consumption level. When the operational directive requires the cooperative action of multiple flexible datacenters 200, the remote master control system (420 of FIG. 4) may determine how to power each individual flexible datacenter 200 in accordance with the operational directive or provide an override to each flexible datacenter 200.

[0092] Further examples of unutilized behind-the-meter power availability is when solar farm 700 experiences intermittent cloud cover such that it is not economically feasible to power up certain components, such as, for example local station 775, but there may be sufficient behind-the-meter power availability to power one or more flexible datacenters 200. Similarly, unutilized behind-the-meter power availability may occur when solar farm 700 is starting up, or testing, one or more panels 710. Panels 710 are frequently offline for installation, maintenance, and service and must be tested prior to coming online as part of the array. One or more flexible datacenters 200 may be powered by one or more panels 710 that are offline from farm 700. The above-noted examples of when unutilized behind-the-meter power is available are merely exemplary and are not intended to limit the scope of what one of ordinary skill in the art would recognize as unutilized behind-the-meter power availability. Behind-the-meter power availability may occur anytime there is power available and accessible behind-the-meter that is not subject to transmission and distribution costs and there is an economic advantage to using it.

[0093] One of ordinary skill in the art will recognize that solar farm 700 and solar panel 710 may vary based on an application or design in accordance with one or more embodiments of the present invention.

[0094] FIG. 8 shows a flexible datacenter 200 powered by flare gas 800 in accordance with one or more embodiments of the present invention. Flare gas 800 is combustible gas produced as a product or by-product of petroleum refineries, chemical plants, natural gas processing plants, oil and gas drilling rigs, and oil and gas production facilities. Flare gas 800 is typically burned off through a flare stack (not shown) or vented into the air. In one or more embodiments of the present invention, flare gas 800 may be diverted 812 to a gas-powered generator that produces three-phase gas-generated AC voltage 822. This power may be considered behind-the-meter and is not subject to transmission and distribution costs. As such, one or more flexible datacenters 200 may be powered by three-phase gas-generated AC voltage. Specifically, the three-phase behind-the-meter AC voltage used to power flexible datacenter 200 may be three-phase gas-generated AC voltage 822. Accordingly, flexible datacenter 200 may reside behind-the-meter, avoid transmission and distribution costs, and may be dynamically powered when unutilized behind-the-meter power is available.

[0095] FIG. 9A shows a method of dynamic power delivery to a flexible datacenter (200 of FIG. 2) using behind-the-meter power 900 in accordance with one or more embodiments of the present invention. In step 910, the datacenter control system (220 of FIG. 4), or the remote master control system (420 of FIG. 4), may monitor behind-the-meter power availability. In certain embodiments, monitoring may include receiving information or an operational directive from the local station control system (410 of FIG. 4) or the grid operator (440 of FIG. 4) corresponding to behind-the-meter power availability.

[0096] In step 920, the datacenter control system (220 of FIG. 4), or the remote master control system (420 of FIG. 4), may determine when a datacenter ramp-up condition is met. In certain embodiments, the datacenter ramp-up condition may be met when there is sufficient behind-the-meter power availability and there is no operational directive from the local station to go offline or reduce power. In step 930, the datacenter control system (220 of FIG. 4) may enable behind-the-meter power delivery to one or more computing systems (100 of FIG. 2). In step 940, once ramped-up, the datacenter control system (220 of FIG. 4) or the remote master control system (420 of FIG. 4) may direct one or more computing systems (100 of FIG. 2) to perform predetermined computational operations. In certain embodiments, the predetermined computational operations may include the execution of one or more distributed computing processes, parallel processes, and/or hashing functions, among other types of processes.

[0097] While operational, the datacenter control system (220 of FIG. 4), or the remote master control system (420 of FIG. 4), may receive an operational directive to modulate power consumption. In certain embodiments, the operational directive may be a directive to reduce power consumption. In such embodiments, the datacenter control system (220 of FIG. 4) or the remote master control system (420 of FIG. 4) may dynamically reduce power delivery to one or more computing systems (100 of FIG. 2) or dynamically reduce power consumption of one or more computing systems. In other embodiments, the operational directive may be a directive to provide a power factor correction factor. In such embodiments, the datacenter control system (220 of FIG. 4) or the remote master control system (420 of FIG. 4) may dynamically adjust power delivery to one or more computing systems (100 of FIG. 2) to achieve a desired power factor correction factor. In still other embodiments, the operational directive may be a directive to go offline or power down. In such embodiments, the datacenter control system (220 of FIG. 4) may disable power delivery to one or more computing systems (100 of FIG. 2).

[0098] As such, FIG. 9B shows a method of dynamic power delivery to a flexible datacenter (200 of FIG. 2) using behind-the-meter power 950 in accordance with one or more embodiments of the present invention. In step 960, the datacenter control system (220 of FIG. 4), or the remote master control system (420 of FIG. 4), may monitor behind-the-meter power availability. In certain embodiments, monitoring may include receiving information or an operational directive from the local station control system (410 of FIG. 4) or the grid operator (440 of FIG. 4) corresponding to behind-the-meter power availability.

[0099] In step 970, the datacenter control system (220 of FIG. 4), or the remote master control system (420 of FIG. 4), may determine when a datacenter ramp-down condition is met. In certain embodiments, the datacenter ramp-down condition may be met when there is insufficient behind-the-meter power availability or anticipated to be insufficient behind-the-meter power availability or there is an operational directive from the local station to go offline or reduce power. In step 980, the datacenter control system (220 of FIG. 4) may disable behind-the-meter power delivery to one or more computing systems (100 of FIG. 2). In step 990, once ramped-down, the datacenter control system (220 of FIG. 4) remains powered and in communication with the remote master control system (420 of FIG. 4) so that it may dynamically power the flexible datacenter (200 of FIG. 2) when conditions change.

[0100] One of ordinary skill in the art will recognize that a datacenter control system (220 of FIG. 4) may dynamically modulate power delivery to one or more computing systems (100 of FIG. 2) of a flexible datacenter (200 of FIG. 2) based on behind-the-meter power availability or an operational directive. The flexible datacenter (200 of FIG. 2) may transition between a fully powered down state (while the datacenter control system remains powered), a fully powered up state, and various intermediate states in between. In addition, flexible datacenter (200 of FIG. 2) may have a blackout state, where all power consumption, including that of the datacenter control system (220 of FIG. 4) is halted. However, once the flexible datacenter (200 of FIG. 2) enters the blackout state, it will have to be manually rebooted to restore power to datacenter control system (220 of FIG. 4). Local station conditions or operational directives may cause flexible datacenter (200 of FIG. 2) to ramp-up, reduce power consumption, change power factor, or ramp-down.

[0101] FIG. 10 illustrates a system for managing queue distribution among a critical datacenter and behind-the-meter flexible datacenters in accordance with one or more embodiments of the present invention. The system 1000 includes a flexible datacenter 200, a critical datacenter 1004, a communication link 1006, a queue system 1008, a communication link 1010, communication links 1012a, 1012b, and a remote master control system 420. The system 1000 represents an example configuration scheme for a system that can distribute computing operations using a queue system 1008 between the critical datacenter 1004 and one or more flexible datacenters (e.g., the flexible datacenter 200). In other examples, the system 1000 may include more or fewer components in other potential configurations. For instance, the system 1000 may not include the queue system 1008 or may include routes that bypass the queue system 1008.

[0102] The system 1000 may be configured to manage computational operations requested to be performed by enterprises or other entities. Computational operations may include various tasks that can be performed or generally supported by one or more computing systems within a datacenter. The parameters of each set of computational operations submitted by an enterprise may differ. For instance, the amount of computational resources (e.g., number of computing systems), the degree of difficulty, the duration and degree of support required, etc., may vary for each set of computational operations. As such, the system 1000 may use the queue system 1008 to organize incoming computational operations requests to enable efficient distribution to the flexible datacenter 200 and the critical datacenter 1004. The system 1000 may use the queue system 1008 to organize sets of computational operations thereby increasing the speed of distribution and performance of the different computational operations requested by various enterprises as compared to a system without a queue system 1008.

[0103] In some examples, the queue system 1008 may enable the system 1000 to efficiently utilize the flexible datacenter 200 to perform some sets of computational operations in a manner that can reduce costs or time required to complete the sets. In particular, one or more components within the system 1000, such as the control systems 220, 420, or 1022, may be configured to identify situations that may arise where using the flexible datacenter 200 can reduce costs or increase productivity of the system 1000, as compared to using the critical datacenter 1004 for computational operations. For example, a component within the system 1000, such as the control systems 220, 420, or 1022, may identify when using behind-the-meter power to power the computing systems 100 within the flexible datacenter 200 is at a lower cost compared to using the computing systems 1020 within the critical datacenter 1004, which are powered by grid power. Additionally, a component in the system 1000, control systems 220, 420, or 1022, may be configured to determine situations when offloading computational operations from the critical datacenter 1004 indirectly (i.e., via the queue system 1008) or directly (i.e., bypassing the queue system 1008) to the flexible datacenter 200 can increase the performance allotted to the computational operations requested by an enterprise (e.g., reduce the time required to complete time-sensitive computational operations).

[0104] Within system 1000, the flexible datacenter 200 may represent one or more flexible datacenters capable of offering computational processing and other computing resources using behind-the-meter power from behind-the-meter sources, such as illustrated in FIGS. 6, 7, and 8. As shown in FIG. 10, the flexible datacenter 200 may include a behind-the-meter power input system 215 that is connected to a behind-the-meter power source, a power distribution system 215, computing systems 100, and a datacenter control system 220, and may take the form of a mobile container or another configuration. The flexible datacenter 200 may additionally be connected to other power sources, such as other behind-the-meter power sources, the power grid, and/or an energy storage system. Additionally, the flexible datacenter 200 may include other components not shown in FIG. 10, such as a climate control system.

[0105] The location of the flexible datacenter 200 relative to the critical datacenter 1004 can vary within embodiments. In some examples, the flexible datacenter 200 may be collocated with critical datacenter 1004. For instance, one or more flexible datacenters 200 may be positioned in the same general location as the critical datacenter 1004 or even share a building with the critical datacenter 1004. In other examples, the flexible datacenter 200 and the critical datacenter 1004 are not collocated. Particularly, one or more flexible datacenters 200 within the system 1000 can have a different location from the critical datacenter 1004. In further examples, one or more flexible datacenters 200 can share a location with the critical datacenter 1004 while other flexible datacenters 200 can have a location away from the critical datacenter 1004.

[0106] In order to provide computing resources to perform or support computational operations, the flexible datacenter 200 may be deployed near or otherwise connected to one or more sources of behind-the-meter power generation. For instance, one or more flexible datacenters 200 may be connected behind-the-meter to the wind farm 600, the solar farm 700, and/or other potentially intermittent power generation sources. As such, the behind-the-meter power input system 210 may be configured to receive behind-the-meter power from one or more sources and input the power to the flexible datacenter 200. For example, the behind-the-meter power input system 210 may provide three-phase nominal AC voltage to the power distribution system 215. The power distribution system 215 may controllably provide a single phase of three-phase nominal AC voltage to one or more computing systems 100 of flexible datacenter 200. For instance, power distribution system 215 may distribute power to the computing systems 100 individually or according to groups of computing systems. The computing systems 100 may then use the power received from the behind-the-meter sources to provide processing/computing abilities, networking, storage, and other resources. In some examples, the computing systems 100 may include one or more ASIC computing systems, GPU computing systems, and/or CPU computing systems.

[0107] In some examples, power received at the flexible datacenter 200 may actively switch between different behind-the-meter sources. For example, the flexible datacenter 200 may actively switch from receiving power from either or both the wind farm 600 and the solar farm 700 (or other types of sources). A control system associated with the flexible datacenter 200 (e.g., the datacenter control system 220) or associated with the system 1000 (e.g., remote master control system 420) generally may monitor various input signals, such as, but not limited to, the price for power, availability of power, computing analysis, and order from an operator, etc., to determine which sources to receive power from at a given time. In some situations, the control system may determine that no source is currently a viable option for supplying power to the flexible datacenter 200. Other sources of behind-the-meter power or grid power can also be used to power the flexible datacenter 200 within examples. For example, the flexible datacenter 200 may receive grid power from the local station where it is cited.

[0108] In some examples, the datacenter control system 220 may monitor activity of the computing systems 100 within the flexible datacenter 200 and use the activity to determine when to obtain computational operations from the queue system 1008. The datacenter control system 220 may analyze various factors prior to requesting or accessing a set of computational operations or an indication of the computational operations for the computing systems 100 to perform. The various factors may include power availability at the flexible datacenter 200, availability of the computing systems 100, type of computational operations available, estimated cost to perform the computational operations at the flexible datacenter 200, cost for power, cost for power relative to cost for grid power, and instructions from other components within the system 1000, among others. The datacenter control system 220 may analyze one or more of the factors when determining whether to obtain a new set of computational operations for the computing systems 100 to perform. In such a configuration, the datacenter control system 220 manages the activity of the flexible datacenter 200, including determining when to acquire new sets of computational operations when capacity among the computing systems 100 permit.

[0109] In other examples, a component (e.g., the remote master control system 420) within the system 1000 may assign or distribute one or more sets of computational operations organized by the queue system 1008 to the flexible datacenter 200. For example, the remote master control system 420 may manage the queue system 1008, including the distribution of computational operations organized by the queue system 1008 to the flexible datacenter 1002 and the critical datacenter 1004.

[0110] The system 1000 also includes the critical datacenter 1004, which represents one or more datacenters assigned to provide computational resources to fulfill critical operations. Particularly, the critical datacenter 1004 may receive one or more assignments to support computational operations from an enterprise. In some examples, the critical datacenter 1004 may receive sets of computational operations directly from the enterprise or the remote master control system 420. The critical datacenter 1004 may also receive sets of computational operations organized by the queue system 1008. As such, to warrant that critical operations are supported, the critical datacenter 1004 is preferably connected to a power grid to ensure that reliable (i.e., non-intermittent) power is available.

[0111] The critical datacenter 1004 may include a power input system 1016, a power distribution system 1018, a critical datacenter control system 1022, and computing systems 1020. The power input system 1016 may be configured to receive power from a power grid and distribute the power to the computing systems 1020 via the power distribution system 1018. In some embodiments, the critical datacenter control system 1022 can manage the assignment and support of computational operations received from enterprises, including the distribution of computational operations among the flexible datacenter 200 and the critical datacenter 1004. This is further described below with respect to remote master control system 420, and management operations described with respect to remote master control system 420 may alternatively or additionally be handled by critical datacenter control system 1022.

[0112] Similar to the flexible datacenter, the critical datacenter 1004 may access and obtain sets of computational operations organized by the queue system 1008. The critical datacenter control system 1022 may monitor activity of the computing systems 1020 and obtain computational operations to perform from the queue system 1008. The critical datacenter control system 1022 may analyze various factors prior to requesting or accessing a set of computational operations or an indication of the computational operations for the computing systems 1020 to perform. Various factors may include power availability at the critical datacenter 1004, power availability at the flexible datacenter 200, availability of the computing systems 1020, type of computational operations available, cost for power from the grid, estimated cost for the critical datacenter 1004 to perform the set computational operations, and instructions from other components within the system 1000, among others. In other examples, a component (e.g., the remote master control system 420) within the system 1000 may assign or distribute one or more sets of computational operations organized by the queue system 1008 to the critical datacenter 1004.

[0113] The communication link 1006 represents one or more links that may serve to connect the flexible datacenter 200, the critical datacenter 1004, and other components within the system 1000 (e.g., the remote master control system 420, the queue system 1008—connections not shown). In particular, the communication link 1006 may enable direct or indirect communication between the flexible datacenter 200 and the critical datacenter 1004. The type of communication link 1006 may depend on the locations of the flexible datacenter 200 and the critical datacenter 1004. Within embodiments, different types of communication links can be used, including but not limited to WAN connectivity, cloud-based connectivity, and wired and wireless communication links.

[0114] The queue system 1008 represents an abstract data type capable of organizing computational operation requests received from enterprises. As each request for computational operations are received, the queue system 1008 may organize the request in some manner for subsequent distribution to a datacenter.

[0115] Different types of queues can make up the queue system 1008 within embodiments. The queue system 1008 may be a centralized queue that organizes all requests for computational operations. As a centralized queue, all incoming requests for computational operations may be organized by the centralized queue.

[0116] In other examples, the queue system 1008 may be distributed consisting of multiple queue subsystems. In the distributed configuration, the queue system 1008 may use multiple queue subsystems to organize different sets of computational operations. Each queue subsystem may be used to organize computational operations based on various factors, such as according to deadlines for completing each set of computational operations, locations of enterprises submitting the computational operations, economic value associated with the completion of computational operations, and quantity of computing resources required to perform each set of computational operations. For instance, a first queue subsystem may organize sets of non-intensive computational operations and a second queue subsystem may organize sets of intensive computational operations.

[0117] Within the system 1000, the queue system 1008 is shown connected to the remote master control system 420 via the communication link 1010. In addition, the queue system 1008 is also shown connected to the flexible datacenter via the communication 1012a and to the critical datacenter 1004 via the communication link 1012b. The communication links 1010, 1012a, 1012b may be similar to the communication link 1006 and can be various types of communication links within examples.

[0118] The organizational design of the queue system 1008 may vary within examples. In some examples, the queue system 1008 may organize indications (e.g., tags, pointers to) to sets of computational operations requested by various enterprises. The queue system 1008 may operate as a First-In-First-Out (FIFO) data structure. In a FIFO data structure, the first element added to the queue will be the first one to be removed. As such, the queue system 1008 may include one or more queues that operate using the FIFO data structure.

[0119] In some examples, one or more queues within the queue system 1008 may use other configurations of queues, including rules to rank or organize queues in a particular manner that can prioritize some sets of computational operations over others. The rules may include one or more of an estimated cost and/or revenue to perform each set of computational operations, an importance assigned to each set of computational operations, and deadlines for initiating or completing each set of computational operations, among others.

[0120] The queue system 1008 may include a computing system configured to organize and maintain queues within the queue system 1008. In another example, one or more other components of the system 1000 may maintain and support queues within the queue system 1008. For instance, the remote master control system 420 may maintain and support the queue system 1008. In other examples, multiple components may maintain and support the queue system 1008 in a distributed manner, such as a blockchain configuration.

[0121] In some examples, the queue system 1008 may include queue subsystems located at each datacenter. This way, each datacenter (e.g., via a datacenter control system) may organize computational operations obtained at the datacenter until computing systems are able to start executing the computational operations.

[0122] The remote master control system 420 represents a component within the system 1000 that, in some embodiments, can manage the assignment and support of computational operations received from enterprises, including the distribution of computational operations among the flexible datacenter 200 and the critical datacenter 1004. As shown in FIG. 10, the remote master control system 420 may connect to the flexible datacenter 200 via communication link 425 and the critical datacenter 1004 via communication link 1002. Alternatively or additionally, remote master control system 420 may connect to flexible datacenter 200 and critical datacenter 1004 via communication link 1006 (not shown) or alternative communication links.

[0123] In some embodiments, remote master control system 420 may serve as an intermediary that facilitates all communication between flexible datacenter 200 and critical datacenter 1004. Particularly, critical datacenter 1004 or flexible datacenter 200 might need to transmit communications to remote master control system 420 in order to communicate with the other datacenter. As also shown, the remote master control system 420 may connect to the queue system 1008 via the communication link 1010. Computational operations may be distributed between the queue system 1008 and the remote master control system 420 via the communication link 1010.

[0124] The remote master control system 420 may assist with management of operations assigned to one or both of the flexible datacenter 200 and the critical datacenter 1004. For instance, the remote master control system 420 may be configured to monitor input signals from behind-the-meter sources in order to identify situations where utilizing the flexible datacenter 200 can reduce costs or increase efficiency of the system 1000. For instance, the remote master control system 420 may determine when flexible datacenter 200 could use power from one or more behind-the-meter power sources to advantageously supplement the computing resources offered by the critical datacenter 1004.

[0125] In addition, the remote master control system 420 may manage the queue system 1008, including providing resources to support queues within the queue system 1008. The remote master control system 420 may also manage the distribution of computational resources from the queue system 1008 to the flexible datacenter 200 and the critical datacenter 1004. The remote master control system 420 may communicate with the datacenter control system 220 in the flexible datacenter 200 via the communication link 425 and the critical datacenter control system 1022 in the critical datacenter 1004 via the communication link 1012b. The communication may include checking whether either datacenter can receive sets of operations. This way, the remote master control system 420 may distribute sets of computational operations organized by the queue system 1008, efficiently increasing the rate which computational operations are completed.

[0126] As an example, the remote master control system 420 (or another component within the system 1000) may determine when power from a behind-the-meter source is being sold at a negative price back to the grid. As another example, the remote master control system 420 may monitor power system conditions and issue operational directives to the flexible datacenter 200. Operational directives may include, but are not limited to, a local station directive, a remote master control directive, a grid directive, a dispatchability directive, a forecast directive, a workload directive based on actual behind-the-meter power availability or projected behind-the-meter power availability. Power system conditions, which may additionally or alternatively be monitored by one or more of the control systems 220, 420, and/or 1020 may include, but are not limited to, excess local power generation at a local station level, excess local power generation that a grid cannot receive, local power generation subject to economic curtailment, local power generation subject to reliability curtailment, local power generation subject to power factor correction, low local power generation, start up local power generation situations, transient local power generation situations, or testing local power generation situations where there is an economic advantage to using local behind-the-meter power generation. As another example, remote master control system 420 (or critical datacenter control system 1022) may monitor the types of computational operations requested of the critical datacenter 1004 and make determinations alone or in conjunction with other control systems, power system conditions, and/or operational directives to decide when or how to offload computational operations to a flexible datacenter 200.

[0127] As a result, the remote master control system 420 may offload some or all of the computational operations assigned to the critical datacenter 1004 to the flexible datacenter 200. This way, flexible datacenter 200 can reduce overall computational costs by using the behind-the-meter power to provide computational resources to assist critical datacenter 1004. The remote master control system 420 may use the queue system 1008 to temporarily store and organize the offloaded computational operations until a flexible datacenter (e.g., the flexible datacenter 200) is available to perform them. The flexible datacenter 200 consumes behind-the-meter power without transmission or distribution costs, which lowers the costs associated with performing computational operations originally assigned to critical datacenter 1004.

[0128] In further examples, remote master control system 420 may identify other situations that may benefit from using one or more flexible datacenters (e.g., flexible datacenter 200) to supplement or replace computational resources provided by critical datacenter 1004.

[0129] In some examples, remote master control system 420 may facilitate communication among components within system 1000 using communication links 425, 1002, 1006, 1010, 1012a, and/or 1012b. The communications may include computation requests from components within system 1000. In one embodiment, the remote master control system 420 may identify a computational operation to be performed at a critical datacenter 1004. The computational operation may be identified by querying the critical datacenter 1004 or by receiving a request from the critical datacenter 1004. Information regarding active or requested computational operations at the critical datacenter 1004 may be considered as part of the identification process. The communications may also include a variety of other information, such as an indication of a current workload at the critical datacenter 1004, a current status of operation at critical datacenter 1004 (e.g., a report indicating current capacity available and power consumption at critical datacenter 1004). Upon receiving the information, the remote master control system 420 may determine whether to route the computational operations to the flexible datacenter 200.

[0130] The determination process may involve considering various factors, including power availability and associated costs from the power grid and behind-the-meter sources, availability of flexible datacenter 200, and type and deadlines associated with assigned computational operations, among others. In some situations, remote master control system 420 may then send the computational operation to flexible datacenter 200 (e.g., via communication link 1006). In these situations, remote master control system 420 may determine that utilizing the flexible datacenter 200 could enhance the operation of system 1000 overall (i.e. improving profitability or timely performance). Particularly, using the flexible datacenter 200 may reduce costs and increase efficiency of system 1000. The flexible datacenter 200 may also help reduce the amount of unutilized or under-utilized power being produced by one or more behind-the-meter sources.

[0131] In some examples, the remote master control system 420 may reassign computational operations from critical datacenter 1004 over to the flexible datacenter 200 for the flexible datacenter 200 to support or complete. For instance, the remote master control system 420 may determine that using the flexible datacenter 200 is more cost efficient that only using critical datacenter 1004. As such, the remote master control system 420 may facilitate a direct transfer of responsibility for the computational operations from the critical datacenter 1004 to the flexible datacenter 200. Alternatively, the remote master control system 420 may use the queue system 1008 to facilitate an indirect transfer of computational operations from the critical datacenter 1004 to the flexible datacenter 200. Particularly, the remote master control system 420 may transfer the offloaded computational operations from the critical datacenter into the queue system 1008 until a flexible datacenter 200 is able to perform the computational operations. The flexible datacenter 200 may access and obtain the offloaded computational operations from the queue system 1008 or may be assigned the computational operations from the queue system 1008 by the remote master control system 420 or another component within the system 1000.

[0132] In further examples, the remote master control system 420 may determine that the flexible datacenter 200 is available to support and provide computing resources to new computational operations received from an enterprise. This way, the remote master control system 420 may route the new computational operations directly to the flexible datacenter 200 or indirectly via use of the queue system 1008 without impacting the workload on the critical datacenter 1004.

[0133] When determining whether to route a computational operation to the flexible datacenter 200, the remote master control system 420 may be configured to consider different factors, such as the availability of the flexible datacenter 200 and availability of behind-the-meter power. In some situations, the remote master control system 420 or another component within the system 1000 (e.g., datacenter control system 220) may determine that the flexible datacenter 200 might not have enough computing systems 100 available to satisfy the computational operation. As a result, the remote master control system 420 may refrain from sending the computational operation to flexible datacenter 200. The remote master control system 420 may then transmit an indication that the flexible datacenter 200 is unavailable back to the critical datacenter 1004. In some examples, the remote master control system 420 may utilize the queue system 1008 to organize the computational requests until a flexible datacenter or the critical datacenter 1004 is available.

[0134] In some examples, the remote master control system 420 may further analyze the workloads of other flexible datacenters to identify a flexible datacenter that is capable of handling the computational operation. Upon identifying an available flexible datacenter, the remote master control system 420 may transmit the computational operation to that flexible datacenter instead. In further examples, the remote master control system 420 may divide operations associated with one or more identified computational operation among multiple flexible datacenters.

[0135] In some examples, the remote master control system 420 may determine whether to route a computational operation to the flexible datacenter 200 based on the availability of between-the-meter power for the flexible datacenter 200. Additionally or alternatively, the remote master control system 420, the flexible datacenter control system 220, or another computing device may monitor one or more other power system operation conditions to make the determination. The remote master control system 420 may also determine whether a datacenter ramp-up condition is met when determining whether to route a computational operation to the flexible datacenter 200. For instance, the remote master control system 420 may check whether the flexible datacenter 200 is ramped-up to a fully online status, ramped-down to a fully offline status, or in another state (e.g., acting as a load balancer). As such, the remote master control system 420 may determine whether to route a computation request to the flexible datacenter 200 based on the status of the flexible datacenter 200.

[0136] As previously discussed, the system 1000 may include a flexible datacenter control system 220, which may be configured to modulate power delivery to computing systems 100 of flexible datacenter 200. For example, the flexible datacenter control system 220 may modulate power delivery to the computing systems 100 based on a threshold level of unutilized behind-the-meter power availability or some other monitored power system condition. In some instances, the flexible datacenter control system 220 may be configured to modulate power delivery to computing systems 100 by selectively enabling or disabling a subset of computing systems 100.

[0137] The flexible datacenter control system 220 may alternatively or additionally be configured to modulate power delivery to the computing systems 100 based on an operational directive. For instance, the flexible datacenter control system 220 or another system may receive an operational directive from a user interface to modulate the power delivery to computing systems 100. As discussed above, the operational directive may be a local station directive, a remote master control directive, a grid directive, a dispatchability directive, or a forecast directive. In some instances, the operational directive may also include a workload directive based on a threshold level actual behind-the-meter power availability or a threshold level of projected behind-the-meter power availability.

[0138] FIG. 11 illustrates a system for managing queue distribution among a critical datacenter and a plurality of behind-the-meter flexible datacenters in accordance with one or more embodiments of the present invention. The system 1100 is similar to the schemes illustrated in FIG. 5, with the addition of the critical datacenter 1004, the queue subsystem 1008a, the queue subsystem 1008b, and communication links 1002, 1006a, 1006b, 1014a, and 1014b. Local stations 410a and 410b, and other control paths not required for illustrative purposes, are removed for clarity. Components and aspects illustrated and/or described in FIG. 10 that are similar or the same as components or aspects illustrated and/or described in FIG. 5 should be considered to have the same characteristics as previously illustrated and/or described.

[0139] The system 1100 may operate similarly to the system 1000 shown in FIG. 10. Similarly labeled components in FIG. 11 may have the same characteristics and/or capabilities as described with respect to FIG. 10. The system 1100, however, is shown configured with the queue system 1008 arranged in a distributed queue subsystem format (e.g., a queue subsystem 1008a and a queue subsystem 1008b). Although queue subsystems 1008a, 1008b are shown physically separate, the queue subsystems 1008a, 1008b may be collocated and/or supported by the same component of the system 1100.

[0140] The system 1100 includes the queue subsystem 1008a connected to the remote master control system 420 via the communication link 1014a and to flexible datacenters 200a, 200b, 200c, 200d and the critical datacenter 1004 via the communication link 1006a. In such a configuration, the remote master control system 420, the flexible datacenters 200a, 200b, 200c, 200d and the critical datacenter 1004 may communicate with the queue subsystem 1008a. The communication may involve obtaining computational operations organized by the queue subsystem 1008a.

[0141] In some examples, the remote master control system 420 may distribute computational operations organized by the queue subsystem 1008a to the flexible datacenters 200a, 200b, 200c, 200d and the critical datacenter 1004. In some examples, the remote master control system 420 may distribute computational operations organized by the queue subsystem 1008a to the flexible datacenters 200e, 200f, 200g, 200h, which are shown not directly connected to the queue subsystem 1008a.

[0142] In some embodiments, the flexible datacenters 200a, 200b, 200c, 200d and the critical datacenter 1004 may communicate directly with the queue subsystem 1008a to obtain computational operations to perform. This way, control systems at each datacenter may monitor and balance the workload supported by their computing systems.

[0143] The system 1100 also includes the queue subsystem 1008b connected to the remote master control system 420 via the communication link 1014b and to flexible datacenters 200e, 200f, 200g, 200h and the critical datacenter 1004 via the communication link 1006b. In such a configuration, the remote master control system 420, the flexible datacenters 200e, 200f, 200g, 200h and the critical datacenter 1004 may communicate with the queue subsystem 1008b. The communication may involve obtaining computational operations organized by the queue subsystem 1008b.

[0144] In some examples, the remote master control system 420 may distribute computational operations organized by the queue subsystem 1008b to the flexible datacenters 200e, 200f, 200g, 200h and the critical datacenter 1004. In some examples, the remote master control system 420 may distribute computational operations organized by the queue subsystem 1008b to the flexible datacenters 200a, 200b, 200c, 200d, which are shown not directly connected to the queue subsystem 1008b.

[0145] In some embodiments, the flexible datacenters 200e, 200f, 200g, 200h and the critical datacenter 1004 may communicate directly with the queue subsystem 1008b to obtain computational operations to perform. This way, control systems at each datacenter may monitor and balance the workload supported by their computing systems.

[0146] In some embodiments, the remote master control system 420 may be configured to determine whether to route a computational operation to a particular flexible datacenter (e.g., flexible datacenter 200a) from among multiple flexible datacenters. The determination process may involve initially determining whether to route the computational operation to a flexible datacenter and then further selecting a specific flexible datacenter to route the computational operation to. The remote master control system 420 or another component (e.g., one or more flexible datacenter control systems 220) may be configured to determine a cost of execution of the computing instructions by computing systems at the specific flexible datacenter. Particularly, each flexible datacenter may be capable of providing computing resources at different costs based on various factors, such as the locations of the flexible datacenters 200 and the availability of behind-the-meter power to each flexible datacenter (i.e., the flexible datacenters 200 may connect to different behind-the-meter power sources). As such, the remote master control system 420 may be configured to consider the different factors to select a specific flexible datacenter to use to fulfill a computational operation. In some examples, the remote master control system 420 may use the queue subsystems 1008a, 1008b to temporarily organize sets of computational operations until a flexible datacenter or the critical datacenter 1004 is available to perform each set.

[0147] FIG. 12 illustrates a method for managing queue distribution between a critical datacenter and a flexible datacenter in accordance with one or more embodiments of the present invention. The method serves an example and may include other steps within other examples. At step 1202, the method involves identifying, using a queue system, a computational operation to be performed. For instance, a component within the system 1000 may identify a computational operation to be performed by analyzing the queue system 1008. The component may be the remote master control system 420, the datacenter control system 220, the critical datacenter control system 1022, and/or a computing system at the queue system 1008 in some examples.

[0148] Identifying the computational operation can include examining various types of information, such as a request for processing, networking, or storage capabilities or a request to offload some work from the critical datacenter. In some instances, the computational operation may be identified in association with an incoming computational operation request received from an outside enterprise. In some examples, the computational operation may be identified based on the organization of the queue system 1008. For instance, the computational operation may be the next operation to be selected based on a FIFO format of the queue system 1008.

[0149] At step 1204, the method involves determining whether to route the computational operation to a flexible datacenter. Different components may be configured to determine whether to route the computational operation to a flexible datacenter. For example, remote master control 420 or critical datacenter control system 1022 within system 1000 may be configured to determine whether to route the computational operation to flexible datacenter 1002. In other examples, a flexible datacenter control system 220 may determine whether to route the computational operation to flexible datacenter 1002. For instance, the flexible datacenter control system 220 may determine whether the computing systems 100 have the availability to perform one or more computational operations within the queue system 1008. In further examples, other components can perform the determination step.

[0150] Determining whether to route the computational operation to a flexible datacenter, such as flexible datacenter 200, can involve considering various factors, such as a cost of execution to provide computing resources at the flexible datacenter relative to the cost of providing computing resources at the critical datacenter. The determination may also factor the availability of the flexible datacenter as well as the cost and availability of unutilized behind-the-meter power from one or more behind-the-meter sources. Other factors can be considered within examples, such as monitored power system conditions and operational directives.

[0151] At step 1206, the method involves causing the computational operation to the flexible datacenter via a communication link, such as links 1006, 425, 1002, 1010, 1012a, and/or 1012b, based on a determination to route the computational operation to the flexible datacenter. Sending the computational operation may enable computing systems of the flexible datacenter to provide computing resources to fulfill the request.

[0152] In some examples, remote master control 420, critical datacenter control system 1022, or another component within system 1000 may determine that the identified computational operation should be routed to the critical datacenter 1004. The determination may be based on various factors, such as a cost of execution to provide computing resources at the flexible datacenter relative to the cost of providing computing resources at the critical datacenter. The determination may also factor the availabilities of the critical datacenter 1004 and the flexible datacenter 200 as well as the cost and availability of unutilized behind-the-meter power from one or more behind-the-meter sources. Other factors may be considered. As such, one or more components may route the computational operation to the critical datacenter 1004 to enable the computing systems 1020 to fulfill the computational request.

[0153] FIG. 13 illustrates a method for managing queue distribution between a critical datacenter and a plurality of flexible datacenter in accordance with one or more embodiments of the present invention. The method serves an example and may include other steps within other examples. The method of FIG. 13 is similar to the method of FIG. 12, and steps, components, and aspects illustrated and/or described in FIG. 13 that are similar to or the same as components or aspects illustrated and/or described in FIG. 12 should be considered to have the same characteristics as previously illustrated and/or described.

[0154] At step 1302, the method involves identifying, using a queue system, a computational operation to be performed. The computational operation may be performed at a critical datacenter, one or more flexible datacenters, or a combination of datacenters. In some examples, the queue system may be distributed into multiple queue subsystems with each queue subsystem organizing a set of computational operations according to rules associated with that queue subsystem. For example, the queue system may include a first queue subsystem and a second queue subsystem where each queue subsystem organizes computational operations accessibly potentially only by a subset of the plurality of flexible datacenters. For instance, a first set of flexible datacenters of the plurality of flexible may be configured to obtain computational operations from the first queue subsystem and a second set of flexible datacenters of the plurality of flexible datacenters are configured to obtain computational operations from the second queue subsystem.

[0155] At step 1304, the method involves determining whether to route the computational operation to a flexible datacenter in a plurality of flexible datacenters. However, in particular, multiple flexible datacenters may be available to receive the computational operation. As such, a computing system, such as remote master control system 420 or critical datacenter control system 1022, may determine whether to route the computational operation to a flexible datacenter out of the multiple available.

[0156] In some examples, the determination may be made by one or more datacenter control systems associated with the plurality of flexible datacenters. Each datacenter control system may determine whether or not its computing systems could currently handle the computational operation.

[0157] At step 1306, the method involves, based on a determination to route the computational operation to a flexible datacenter in the plurality of flexible datacenters, determining a specific flexible datacenter in the plurality of flexible datacenters to route the computational operation to. The computing system may select a specific datacenter based on cost, availability, source of unutilized behind-the-meter power, or other factors. For example, the computing system may compare the cost associated with sending the computational operation to different flexible datacenters.

[0158] In some examples, a flexible datacenter or a critical datacenter may access and obtain the computational operation from the queue system. For example, a flexible datacenter from the plurality of flexible datacenters may obtain the computational operation upon determining that its computing systems are capable of supporting the computational operation (e.g., power is available, enough computing systems are free to operate on the computational operation).

[0159] At step 1308, the method involves causing the computational operation to be sent to the specific flexible datacenter via the communication link. Various components within the system may enable the computational operation to reach the specific flexible datacenter.

[0160] In further examples, the method described above may involve dividing the computational operation among multiple flexible datacenters.

[0161] Advantages of one or more embodiments of the present invention may include one or more of the following:

[0162] One or more embodiments of the present invention provides a green solution to two prominent problems: the exponential increase in power required for growing blockchain operations and the unutilized and typically wasted energy generated from renewable energy sources.

[0163] One or more embodiments of the present invention allows for the rapid deployment of mobile datacenters to local stations. The mobile datacenters may be deployed on site, near the source of power generation, and receive unutilized behind-the-meter power when it is available.

[0164] One or more embodiments of the present invention provide the use of a queue system to organize computational operations and enable efficient distribution of the computational operations to datacenters.

[0165] One or more embodiments of the present invention enable datacenters to access and obtain computational operations organized by a queue system.

[0166] One or more embodiments of the present invention allows for the power delivery to the datacenter to be modulated based on conditions or an operational directive received from the local station or the grid operator.

[0167] One or more embodiments of the present invention may dynamically adjust power consumption by ramping-up, ramping-down, or adjusting the power consumption of one or more computing systems within the flexible datacenter.

[0168] One or more embodiments of the present invention may be powered by behind-the-meter power that is free from transmission and distribution costs. As such, the flexible datacenter may perform computational operations, such as distributed computing processes, with little to no energy cost.

[0169] One or more embodiments of the present invention provides a number of benefits to the hosting local station. The local station may use the flexible datacenter to adjust a load, provide a power factor correction, to offload power, or operate in a manner that invokes a production tax credit and/or generates incremental revenue.

[0170] One or more embodiments of the present invention allows for continued shunting of behind-the-meter power into a storage solution when a flexible datacenter cannot fully utilize excess generated behind-the-meter power.

[0171] One or more embodiments of the present invention allows for continued use of stored behind-the-meter power when a flexible datacenter can be operational but there is not an excess of generated behind-the-meter power.

[0172] It will also be recognized by the skilled worker that, in addition to improved efficiencies in controlling power delivery from intermittent generation sources, such as wind farms and solar panel arrays, to regulated power grids, the invention provides more economically efficient control and stability of such power grids in the implementation of the technical features as set forth herein.

[0173] While the present invention has been described with respect to the above-noted embodiments, those skilled in the art, having the benefit of this disclosure, will recognize that other embodiments may be devised that are within the scope of the invention as disclosed herein. Accordingly, the scope of the invention should be limited only by the appended claims.