Management of power distribution into and out of data center

Abstract

Management of power distribution into and out of a data center is provided. Systems and methods for providing consistent and reliable power to data centers executing high power and variable high performance computing (HPC) and artificial intelligence (AI) workloads are disclosed. By installing a three-phase current sensor at or near the data center or along the grid before power reaches a power conversion system (PCS), the power conversion system is configured to respond faster to changes in rapid and inconsistent rising and falling and variable workloads. A controller within the PCS is then configured to customize an amount of power sent to the data center from the converted power from the grid and from the battery energy storage system.

Description

[0001] Cross-reference to related applications

[0002] This U.S. non-provisional patent application claims the benefit and priority of U.S. Provisional Patent Application No. 63 / 719,332, filed November 12, 2024, the contents of which are incorporated herein by reference in their entirety. Technical Field

[0003] This disclosure relates to systems for stabilizing power for data centers performing large-scale and variable workloads. Background Technology

[0004] The rapid development of advanced computing systems driven by high-performance computing (HPC), artificial intelligence (AI), and other advanced computing technologies has transformed power demand patterns within data centers hosting computing equipment capable of performing these HPC and AI workloads. The amount of energy required to run concurrent and more complex operations causes power demands in such data centers to fluctuate rapidly, for example, hundreds of times per minute, thus creating “pulsating energy” demands. Power providers such as the public power grid can indeed determine the size of the total power required for such types of operations, but they cannot provide the pulsed energy needed for these new loads. This situation can destabilize the grid, potentially leading to power quality issues or even other more serious problems, such as data center providers deploying large-scale HPC and AI workloads being “kicked off” from the grid. Summary of the Invention

[0005] The disclosed embodiments include a system for managing high-power and variable workloads at a data center. The system includes: a three-phase current sensor configured to provide current measurements to a power conversion system, wherein the three-phase current sensor is located near the data center or along the power grid supplying power to the data center; and a power conversion system for the data center, including a controller, wherein: the power conversion system is configured to control the rate of power supplied to the data center from the power grid and from a battery energy storage system coupled to the power conversion system; and, in order to control the rate of power, the controller is configured to: receive an indication of a current difference between a current measured at the data center and a current measured by the power grid, at least in part based on the current measurements provided by the three-phase current sensor; and adjust one or both of the rate of power from the power grid and the rate of power from the battery energy storage system, at least in part based on the indication of the difference, to reduce the difference to approximately zero.

[0006] Another aspect of the disclosed embodiments includes a method for managing power to a data center. The method includes: receiving a three-phase voltage from a power grid; receiving a first three-phase current from a power conversion system coupled to a battery energy storage system; receiving a second three-phase current from the data center; applying a phase-locked loop to measure the frequency and phase angle of the three-phase voltage from the power grid relative to a given phase; performing a first Park-Clarke transformation on the phase angles of the first three-phase current and the three-phase voltage to output a first equivalent phase current; performing a second Park-Clarke transformation on the phase angles of the second three-phase current and the three-phase voltage to output a second equivalent phase current; comparing the first and second equivalent phase currents with an average equivalent phase current from the data center to output the difference between the three equivalent phase currents; and outputting a pulse-width modulation command to the power conversion system to generate a fourth equivalent phase current that reduces the difference to approximately zero. Attached Figure Description

[0007] The present disclosure can be best understood from the following detailed description when read in conjunction with the accompanying drawings. It should be emphasized that, by convention, the various features in the drawings are not drawn to scale. Rather, for clarity, the dimensions of the various features may be arbitrarily enlarged or reduced.

[0008] Figure 1 A controller within a power conversion system (PCS) according to some embodiments is shown, the controller being configured to manage power entering and leaving a data center from the grid and based at least in part on three-phase current sensors located near the data center side or grid side of the overall system.

[0009] Figure 2 Visual indications of voltage and current measurement results according to some embodiments are shown, which are used in applications such as... Figure 1 The system shown manages the power entering and leaving the data center.

[0010] Figure 3 A feedback control loop according to some embodiments is shown, which corresponds to, for example, Figure 1 The diagram shows the overall system for managing the power entering and leaving the data center.

[0011] Figure 4 (A) illustrates an example of D-phase current measured at a data center according to some implementations, the measurement starting at a moment corresponding to the start of AI workloads at the data center and throughout the subsequent time period when power is managed to flow in and out of the data center.

[0012] Figure 4(B) shows an example of a D-phase current measured on the grid side according to some implementations, the measurement starting at a moment corresponding to the start of AI workloads at the data center and throughout subsequent time periods when power is managed to enter and exit the data center.

[0013] Figure 4 (C) shows an example of D-phase current measured at the PCS according to some implementations, the measurement starting at a moment corresponding to the start of AI workloads at the data center and throughout the subsequent time period when power is managed to enter and exit the data center.

[0014] Figure 5 This is a flowchart illustrating a control algorithm executed by a controller within a PCS according to some implementations.

[0015] Figure 6A A simulation of the current versus time relationship over the duration of an example AI workload executed at a data center, according to some implementation methods, is shown.

[0016] Figure 6B A simulation of the relationship between current and time measured at a PCS during the duration of an example AI workload performed at a data center, according to some implementations, is shown.

[0017] Figure 6C The diagram illustrates the relationship between input current and time over the duration of an example AI workload executed at a data center, according to some implementation methods.

[0018] Figure 6D The battery voltage and power measured at BESS during the duration of an example AI workload performed at a data center, according to some implementations, are shown.

[0019] Figure 6E The battery current measured at BESS during the duration of an example AI workload performed at a data center, according to some implementations, is shown.

[0020] Figure 6F The diagram illustrates the current versus time relationship of the AI ​​load current at the data center, the current on the grid side, and the current measured at the PCS during the duration of an example AI workload executed at a data center, according to some implementations.

[0021] Figure 7Another example of a controller within a power conversion system (PCS) according to some implementations is shown, the controller being configured to manage power entering and exiting server racks at the data center from a low-voltage AC distribution unit and based at least in part on three-phase current sensors located near the server racks or low-voltage AC distribution unit of the overall system. Detailed Implementation

[0022] Reference will now be made in detail to the exemplary embodiments shown in the accompanying drawings, in which the same reference numerals always denote the same elements. In this respect, the exemplary embodiments may take different forms and should not be construed as limited to the description set forth herein.

[0023] It should be understood that, when used in this specification, the terms “include,” “including,” “comprise,” and / or “comprising” specify the presence of the stated feature, integer, step, operation, element, and / or component, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0024] It will be further understood that although the terms “first,” “second,” “third,” etc., may be used herein to describe various elements, components, regions, layers, and / or sections, these elements, components, regions, layers, and / or sections may not be limited by these terms. These terms are used only to distinguish one element, component, region, layer, or section from another element, component, region, layer, or section.

[0025] Furthermore, it will be understood that terms such as “optimized,” “best,” and “approximate” as used herein may be used to mean producing or achieving the most efficient or best possible performance. However, those skilled in the art will recognize upon reading this document that optimal performance is not always achievable. Therefore, these terms may also encompass producing or achieving the best possible or most efficient performance, or performance that is practical in a given situation, or performance that is better than performance that could be achieved using other settings or parameters, or performance that is better than not performing the actions, calculations, etc., described herein.

[0026] As used herein, the term "and / or" includes any and all combinations of one or more of the listed items. When preceding a list of elements, expressions such as "at least one" modify the entire list of elements and not the individual elements of the list.

[0027] Various terms are used to refer to specific system components. Different companies may use different names to refer to components—this document does not intend to distinguish between components with different names but the same function.

[0028] Matters of these exemplary embodiments that will be obvious to those skilled in the art to which these exemplary embodiments pertain will not be described in detail herein.

[0029] It is understood that the exemplary embodiments described herein are to be considered in a descriptive sense only and not for limiting purposes. The description of features or aspects within each exemplary embodiment may be considered applicable to other similar features or aspects in other exemplary embodiments.

[0030] As mentioned above, computer servers, computing devices, processors, and support equipment used for HPC and AI workloads typically require high power and variable power. Meeting the high power and variable power requirements of data centers is challenging for both the power grid company and the data center provider.

[0031] This disclosure overcomes this challenge by incorporating measurements from an additional three-phase current sensor into a control algorithm executed by a controller within the PCS. By implementing a three-phase current sensor that provides periodic current measurements from the grid side or data center side to the controller within the PCS, the PCS is configured to be more sensitive to the rapid and highly variable power demands of HPC and AI workloads performed within the data center. By incorporating the measurements from the additional three-phase current sensor into a control algorithm managed by the controller within the PCS, power consumption at the data center is better suited to HPC and AI workloads. For example, the systems and methods described herein consider peak shaving (also known as load shedding), frequency stabilization, and the ability to sell excess power back to the grid.

[0032] Figure 1 A controller within a PCS according to some embodiments is shown, which is configured to manage power entering and leaving the data center from the power grid and is based at least in part on three-phase current sensors located near the data center side or the power grid side of the overall system.

[0033] exist Figure 1 In the overall system 100 shown, power grid 102 is configured to supply power to data center 106. Since the input power from power grid 102 is typically high-voltage AC power, such as 34.5 kV, ... Figure 1 As shown, PCS 108 can therefore be located near data center 106 and can be configured to convert high-voltage AC power to lower-voltage AC power, and then to DC power. Thus, Battery Energy Storage System (BESS) 110 is coupled to grid 102 via PCS 108, where PCS 108 manages the power supplied by the grid.

[0034] To improve the response time of PCS 108 and BESS 110 in meeting the high power and variable demands of data center 106, a three-phase current sensor 104 can be located on the data center side, such as... Figure 1 As shown in the specific embodiments illustrated, it may be located on the grid side. In some embodiments, the three-phase current sensor 104 is configured to measure the rate of electricity at the sensor location and provide those measurements to the controller of the PCS 108. Via Figure 3 The controller 316 in the diagram shows a schematic of the controller of PCS 108.

[0035] As further described below, current measurements from three-phase current sensor 104 and another current sensor within PCS 108 can be provided as input to a control algorithm that is then configured to rapidly detect changes in workload at data center 106 and subsequently respond to those changes by supplying or drawing power from the grid. In some implementations, this detection and response method enables a constant or near-constant power delivery from grid 102 to data center 106, even when a given AI workload may fluctuate between 30% and 100% per second.

[0036] The overall system 100 is also configured to, for example, track the AC voltage from the power grid 102 using a phase-locked loop (PLL) to transform the AC voltage and current into an equivalent DC signal using a Parker-Clark transformation. Such a transformation from AC to an equivalent DC signal can be written as follows: ,in It can also be written as:

[0037]

[0038] The angle theta(θ) in the above equation can be derived from the PLL and is equivalent to the instantaneous angle of the grid voltage.

[0039] Figure 2 Visual indications of voltage and current measurement results according to some embodiments are shown, which are used in applications such as... Figure 1 The system shown manages the power entering and leaving the data center.

[0040] The overall system 200 can be similar to the overall system 100 described above. For the sake of this discussion, and especially for the purpose of discussing the following... Figure 3The feedback control loop shown in the figure includes arrows indicating "three-phase grid voltage", "three-phase PCS current", "BESS voltage", "BESS current", and "three-phase AI data center current". It should be understood that, according to some embodiments, the grid voltage of grid 102 can be measured within PCS 108, but it is shown in the figure for ease of discussion.

[0041] Figure 3 A feedback control loop according to some embodiments is shown, which corresponds to, for example, Figure 1 The diagram shows the overall system for managing the power entering and leaving the data center.

[0042] In some implementations, the components of the feedback control loop 300 may correspond to Figure 1 and Figure 2 The hardware components shown and Figure 2 The sensor signals are displayed intuitively. For example, the mains voltage 302 refers to... Figure 2 The arrow marked "Three-phase grid voltage" refers to the three-phase voltage of the input AC power from grid 102. Similarly, PCS current 306 refers to... Figure 2 The arrow marked "Three-phase PCS current" indicates that the AI ​​data center current 308 refers to... Figure 2 The arrow in the middle is marked "Three-phase AI data center current".

[0043] like Figure 3 As shown, the grid voltage 302 (also referred to as the three-phase ABC voltage in the figure) is supplied to PLL 304, which quickly and accurately measures the phase angle and frequency of the grid voltage relative to a given phase, also referred to as the grid voltage phase angle in the figure. In some embodiments, phase A is chosen as the reference, but any phase can be similarly used as the reference phase. In other embodiments, multiple phases can be used as reference phases.

[0044] The PCS current 306 (also referred to as the three-phase ABC current in the figure) is then fed as input to the Park-Clark transformer 310 along with the grid voltage phase angle. The output of the Park-Clark transformer 310 is the transformation to the equivalent D-phase current.

[0045] The AI ​​data center current 308 (also referred to as the three-phase ABC current in the figure) is then fed as input to the Park-Clark transformer 312 along with the grid voltage phase angle. The output of the Park-Clark transformer 312 is the transformation to the equivalent D-phase current.

[0046] According to some implementations, when an instantaneous grid voltage angle is applied, both D-phase currents become DC signals. Furthermore, for the sake of discussion herein, it can be assumed that the voltage and current are balanced and free of harmonic components, and therefore the Q-phase and D-phase values ​​are zero. However, it should be understood that the control feedback loop 300 will be implemented in a manner that includes additional feedback loops for those phases that will accordingly adjust them to specified values.

[0047] Continuing with the diagram of control feedback loop 300, at the step shown as comparison (Σ) 314, the D-phase current of the PCS and AI data center is compared with another signal labeled "Grid Target D-phase Current" in the diagram. The "Grid Target D-phase Current" signal can be defined as the average value of the AI ​​data center's D-phase current plus or minus a certain value, which ensures that the battery within the BESS 110 does not fully discharge or become overcharged.

[0048] The output of Comparison (Σ) 314 is the difference or "error" between the individual signals. This error is then provided to Controller 316, such as a Proportional-Integral-Derivative (PID) controller, which is then configured to generate a Pulse Width Modulation (PWM) command, which is then provided to PCS 318. PCS 318 then generates a current that reduces the error to zero.

[0049] The current in phase D of the PCS is compared with the current in phase D of the AI ​​data center, generating a difference (error). This error is sent to controller 316, such as a proportional-integral-derivative (PID) controller. Controller 316 generates a PWM command to PCS 318, which generates a current that reduces the error to zero.

[0050] Figure 4 (A) Figure 4 (B) and Figure 4 (C) illustrates examples of D-phase currents measured at data center 106, on the grid side, and at PCS 108, respectively, according to some embodiments. This measurement begins at a time corresponding to the start of the AI ​​workload at data center 106 and continues throughout subsequent periods when power to and from grid 102 is managed to flow to and from data center 106. This should be understood within the context of the operation of controller 316. Figure 4 (A) Figure 4 (B) and Figure 4 (C).

[0051] We can assume Figure 4 (A) Figure 4 (B) and Figure 4Each of the three D-phase currents shown in (C) starts from zero at time t = 0. Time t = 0 can also refer to the moment when the HPC or AI workload at data center 106 gradually rises to a certain value. Then, the grid current jumps to the same value, as... Figure 4 As shown in (B), this is to support the power requirements of new HPC or AI workloads. Then, controller 316 measures... Figure 3 The difference or "error" shown in the figure is then used to adjust its current level to draw power from the BESS 110 until it reaches the same value as the HPC or AI workload.

[0052] Then, at time t = t1, the HPC or AI workload drops to zero, and therefore the current from BESS 110 now flows back to grid 102, as... Figure 4 The negative grid current is indicated by shown in (B). Then, controller 316 quickly jumps to a negative value approximately equal to the average of the last AI load pulse. The grid current reverses direction, as... Figure 4 As shown in (B), it stabilizes at a positive value, which is approximately the average value of the last AI load pulse.

[0053] At time t = t2, the AI ​​workload then rises, and the controller 316 then quickly reverses the current direction, causing the BESS 110 to provide the peak of the AI ​​workload, while the grid 102 continues to provide the average value.

[0054] This process can then be repeated, as shown at time t = t3, such that the grid 102 provides average power to the data center 106, while the BESS 110 provides peak power and draws average current from the grid to recharge the battery within the BESS 110. Figure 4 In (B), the incremental change at time t = t3 and thereafter indicates the response time of controller 316.

[0055] Figure 5 This is a flowchart illustrating a control algorithm executed by a controller within PCS 108 according to some implementations.

[0056] The above information can be used. Figures 1 to 4 The hardware components and feedback control loop architecture described in (C) are used to implement the processing of 500. Figure 5As shown, the target D-phase current of the grid in box 514 is obtained through the following operations: First, the average AI current is calculated as indicated in box 502. Next, in box 504, the BESS state of charge (SOC) is calculated. Then, as indicated in diamond 506, if the battery is about to discharge, the controller increases the average grid current by a certain amount (+K), as indicated in box 508; if the battery is about to overcharge, the controller decreases the average grid current by a certain amount (-K), as indicated in box 510.

[0057] In other implementations, the feedback controller can slowly adjust the average target current over a period of time to balance the battery's state of charge (SOC).

[0058] In some implementations, the target D-phase current of the power grid can therefore be defined as an average value plus an increment (±K). The calculation 512 for generating the average D-phase current can be written as follows:

[0059]

[0060] Among them, T min It is a time when it is at a low level, and T max It is the peak time of the current HPC or AI workload currently being executed in the data center.

[0061] In other implementations, alternative methods such as analog or digital filtering can be used to derive the average AI D phase current.

[0062] Figure 6A A simulation of the current versus time relationship over the duration of an example AI workload executed at a data center, according to some implementation methods, is shown.

[0063] Figure 6B A simulation of the relationship between current and time measured at a PCS during the duration of an example AI workload performed at a data center, according to some implementations, is shown.

[0064] Figure 6C The diagram illustrates the relationship between input current and time over the duration of an example AI workload executed at a data center, according to some implementation methods.

[0065] Figure 6D The battery voltage and power measured at BESS during the duration of an example AI workload performed at a data center, according to some implementations, are shown.

[0066] Figure 6E The battery current measured at BESS during the duration of an example AI workload performed at a data center, according to some implementations, is shown.

[0067] Figure 6F The diagram illustrates the current versus time relationship of the AI ​​load current at the data center, the current on the grid side, and the current measured at the PCS during the duration of an example AI workload executed at a data center, according to some implementations.

[0068] Figure 7 Another example of a controller within PCS 708 according to some embodiments is shown, the controller being configured to manage power entering and exiting server rack 706 at the data center from a low-voltage AC distribution unit in the data center and based at least in part on a three-phase current sensor 704 located near the server rack 706 or the low-voltage AC distribution unit in the overall system 700. Via Figure 3 The controller 316 in the diagram shows a schematic of the controller of the PCS 708.

[0069] In such an implementation, a similar current measurement scheme using a three-phase current sensor 704 located near a low-voltage AC distribution unit in a server rack 706 or data center can be provided to a similar feedback control loop, such as... Figure 3 As shown in the diagram, the electrical positions of PCS 708 and energy storage device 710 can be modified accordingly.

[0070] Furthermore, the energy storage device 710 of the overall system 700 can refer to any energy storage device compatible with low-voltage AC distribution, such as electrolytic capacitors, supercapacitors, lithium-ion batteries, etc.

[0071] Although exemplary embodiments have been described with reference to the accompanying drawings, those skilled in the art will understand that various changes in form and detail may be made therein without departing from the spirit and scope defined by the appended claims.

Claims

1. A system for managing power to a data center, the system comprising: A three-phase current sensor, configured to provide current measurement results to a power conversion system, wherein the three-phase current sensor is located near the data center; and The power conversion system for the data center includes a controller, wherein: The power conversion system is configured to control the rated power supplied to the data center from the power grid and from a battery energy storage system coupled to the power conversion system; and In order to control the rated power, the controller is configured to: The indication of the current difference between the current measured at the data center and the current measured by the power grid is received, at least in part, based on the current measurement results provided by the three-phase current sensor; and The indication that at least part of the rated power from the power grid and the rated power from the battery energy storage system is adjusted is used to reduce the difference to near zero.

2. The system according to claim 1, wherein, The controller is a proportional-integral-derivative controller.

3. The system according to claim 1, wherein, The controller is also configured to calculate the state of charge of the battery energy storage system.

4. The system according to claim 3, wherein, The controller is also configured to: Based on the calculated state of charge, it is determined that the battery energy storage system is about to discharge; as well as Increase the rated power from the battery energy storage system to rebalance the state of charge of the battery energy storage system.

5. The system according to claim 3, wherein, The controller is also configured to: Based on the calculated state of charge, it is determined that the battery energy storage system is about to become overcharged; as well as The rated power from the battery energy storage system is reduced to rebalance the state of charge of the battery energy storage system.

6. The system according to claim 1, wherein, The battery energy storage system includes electrolytic capacitors compatible with low-voltage AC distribution.

7. The system according to claim 1, wherein, The battery energy storage system includes a supercapacitor compatible with low-voltage AC distribution.

8. The system according to claim 1, wherein, The battery energy storage system includes lithium-ion batteries compatible with low-voltage AC distribution.

9. A system for managing power to a data center, the system comprising: A three-phase current sensor, configured to provide current measurements to a power conversion system, wherein the three-phase current sensor is located along the power grid supplying power to the data center; and The power conversion system for the data center includes a controller, wherein: The power conversion system is configured to control the rated power supplied to the data center from the power grid and from a battery energy storage system coupled to the power conversion system; and In order to control the rated power, the controller is configured to: Based at least in part on the current measurement results provided by the three-phase current sensors, an indication of the current difference between the current measured at the data center and the current measured by the power grid is received; and The indication that at least part of the rated power from the power grid and the rated power from the battery energy storage system is adjusted is used to reduce the difference to near zero.

10. The system according to claim 9, wherein, The controller is a proportional-integral-derivative controller.

11. The system according to claim 9, wherein, The controller is also configured to calculate the state of charge of the battery energy storage system.

12. The system according to claim 11, wherein, The controller is also configured to: Based on the calculated state of charge, it is determined that the battery energy storage system is about to discharge; as well as Increase the rated power from the battery energy storage system to rebalance the state of charge of the battery energy storage system.

13. The system according to claim 11, wherein, The controller is also configured to: Based on the calculated state of charge, it is determined that the battery energy storage system is about to become overcharged; as well as The rated power from the battery energy storage system is reduced to rebalance the state of charge of the battery energy storage system.

14. The system according to claim 9, wherein, The battery energy storage system includes electrolytic capacitors compatible with low-voltage AC distribution.

15. The system according to claim 9, wherein, The battery energy storage system includes a supercapacitor compatible with low-voltage AC distribution.

16. The system according to claim 9, wherein, The battery energy storage system includes lithium-ion batteries compatible with low-voltage AC distribution.

17. A method for managing power to a data center, the method comprising: Receive three-phase voltage from the power grid; The power conversion system coupled to the battery energy storage system receives the first three-phase current; Receive second and third phase current from the data center; A phase-locked loop is used to measure the phase angle and frequency of the three-phase voltage from the power grid relative to a given phase; A first Park-Clark transformation is performed on the phase angles of the first three-phase current and the three-phase voltage to output a first equivalent phase current; A second Park-Clark transformation is performed on the phase angles of the second three-phase current and the three-phase voltage to output a second equivalent phase current; The first equivalent phase current and the second equivalent phase current are compared with the average equivalent phase current from the data center to output the difference between the three equivalent phase currents; as well as A pulse width modulation command is output to the power conversion system to generate a fourth equivalent phase current that reduces the difference to near zero.

18. The method of claim 17, further comprising calculating the state of charge of the battery energy storage system.

19. The method of claim 17, further comprising: Based on the calculated state of charge, it is determined that the battery energy storage system is about to discharge; as well as Increase the rated power from the battery energy storage system to rebalance the state of charge of the battery energy storage system.

20. The method of claim 17, further comprising: Based on the calculated state of charge, it is determined that the battery energy storage system is about to become overcharged; as well as The rated power from the battery energy storage system is reduced to rebalance the state of charge of the battery energy storage system.

Images