A power load-side resource dispatching system

Abstract

This application provides a power load-side resource scheduling system, comprising: a load aggregation module generating a first load curve based on historical operating data of a supercomputing center and dividing it into stages to obtain a second load curve, and performing calculations on it to obtain a predicted allocation curve; a task scheduling module allocating the execution time of the tasks to be calculated according to the priority of the tasks to be calculated and the second load curve to obtain a first scheduling scheme, and sending it to a load adjustment module; the load adjustment module calculating the real-time total power of the supercomputing center according to the first scheduling scheme, and adjusting the operating power corresponding to different modules of the supercomputing center according to the real-time total power and the predicted allocation curve to obtain a second scheduling scheme, controlling the operation of the supercomputing center, thus solving the problem in the prior art of lacking a scheduling method that adjusts the execution time of projects based on project priority, grid load, and computing resource requirements to avoid the impact of peak load periods on computing tasks.

Description

Technical Field

[0001] This invention relates to the field of power load dispatching technology, and more specifically, to a dispatching system for power load-side resources. Background Technology

[0002] In today's era of rapid digital development, supercomputing centers, as major electricity consumers on the power grid's load side, play a crucial role. However, supercomputing centers face complex power management challenges. Typically, supercomputing centers run numerous computing projects, which vary significantly in priority, resource requirements, and execution time. For example, some urgent scientific computing projects may have the highest priority and require priority resource allocation, while some routine data processing projects have relatively lower priority. Furthermore, the resource requirements of different projects vary; some projects may require substantial computing power, while others rely more heavily on storage resources.

[0003] Because of these differences, load scheduling and management of supercomputing centers are particularly critical. Ineffective load scheduling can lead to grid load imbalances, affecting the stable operation of the entire grid. For example, during peak electricity consumption periods, excessive load on the supercomputing center could exacerbate the grid burden and even trigger power outages. Therefore, how to achieve effective power load scheduling to match the grid load has become a crucial problem that urgently needs to be solved. Summary of the Invention

[0004] The main objective of this application is to provide a scheduling system for power load-side resources, so as to at least solve the problem in the prior art of lacking a scheduling method that adjusts the project execution time based on project priority, grid load, and computing resource requirements to avoid the impact of peak load periods on computing tasks.

[0005] To achieve the above objectives, according to one aspect of this application, a power load-side resource scheduling system is provided. The scheduling system includes a load aggregation module, a task scheduling module, and a load adjustment module, which are communicatively connected. Specifically: the load aggregation module generates a first load curve based on historical operating data of a supercomputing center, divides the first load curve into stages based on load values ​​and preset thresholds to obtain a second load curve, performs calculations based on the second load curve to obtain a predicted allocation curve, sends the second load curve to the task scheduling module, and sends the predicted allocation curve to the load adjustment module. The first load curve is a curve representing the change of power grid load over time. The task scheduling module allocates the execution time of the tasks to be calculated based on their priority and the second load curve to obtain a first scheduling scheme, and sends the first scheduling scheme to the load adjustment module. The load adjustment module calculates the real-time total power of the supercomputing center based on the first scheduling scheme, adjusts the operating power corresponding to different modules of the supercomputing center based on the real-time total power and the predicted allocation curve to obtain a second scheduling scheme, and controls the operation of the supercomputing center according to the second scheduling scheme.

[0006] Optionally, the load aggregation module includes a load curve generation module, a load curve division module, and a scheduling curve generation module, wherein: the load curve generation module is used to predict based on historical power grid operation data, generate a curve showing the change of the power grid load over time within a first preset time period, obtain the first load curve, and send the first load curve to the load curve division module, where the length of the first preset time period is a preset duration, and the start time of the first preset time period is the current time; the load curve division module is used to traverse the first load curve, determine the curve in the first load curve whose corresponding power grid load is less than or equal to a first threshold as a low-valley stage, determine the curve in the first load curve whose corresponding power grid load is greater than the first threshold and less than a second threshold as a balanced stage, and determine the curve in the first load curve whose corresponding power grid load is greater than or equal to the second threshold as a peak stage, obtain the second load curve, and send the second load curve to the scheduling curve generation module and the task scheduling module; the scheduling curve generation module is used to fit a predicted scheduling curve based on a first preset coefficient, a second preset coefficient, and the second load curve, and send the predicted scheduling curve to the load adjustment module.

[0007] Optionally, the task scheduling module includes a priority determination module and a task matching module, wherein: the priority determination module is used to traverse the required computing speed and security score of the tasks to be computed and divide the tasks to be computed into a first priority, a second priority, and a third priority, wherein the first priority is higher than the third priority, and the third priority is higher than the second priority; the task matching module is used to allocate the tasks to be computed with the first priority to the off-peak period for execution, allocate the tasks to be computed with the second priority to the peak period for execution, and allocate the tasks to be computed with the third priority to the balanced period for execution, thereby obtaining the first scheduling scheme.

[0008] Optionally, the load adjustment module includes a power consumption calculation module, a safety power consumption adjustment module, and a computing power consumption adjustment module, wherein: the power consumption calculation module is used to determine the safety power consumption of the safety components, the computing power consumption of the computing components, and other power consumption of other power-consuming components in the supercomputing center according to the first scheduling scheme; calculate the heat dissipation power consumption of the heat dissipation components of the supercomputing center according to the safety power consumption, the computing power consumption, and the other power consumption; and calculate the real-time total power according to the safety power consumption, the computing power consumption, the heat dissipation power consumption, and the other power consumption; the safety power consumption adjustment module is used to adjust the safety power consumption according to the real-time total power and the predicted allocation curve when the task to be computed is of the first priority; and the computing power consumption adjustment module is used to adjust the computing power consumption according to the real-time total power and the predicted allocation curve.

[0009] Optionally, the predicted scheduling curve is obtained by fitting the first preset coefficient, the second preset coefficient, and the second load curve, including: calculating the product of the second load curve and the first preset coefficient to obtain a third load curve; and summing the third load curve and the second preset coefficient to obtain the predicted scheduling curve.

[0010] Optionally, the computational speed and security score of the tasks to be computed are traversed to divide the tasks into a first priority, a second priority, and a third priority, including: determining the tasks to be computed with a computational speed greater than a third threshold and a security score greater than a fourth threshold as the first priority; determining the tasks to be computed with a computational speed less than a fifth threshold and a security score less than a sixth threshold as the second priority, wherein the third threshold is greater than the fifth threshold and the fourth threshold is greater than the sixth threshold; and determining the tasks to be computed other than the first priority and the second priority as the third priority.

[0011] Optionally, the heat dissipation power consumption of the supercomputing center's heat dissipation components is calculated based on the safe power consumption, the computing power consumption, and the other power consumption. The real-time total power is calculated based on the safe power consumption, the computing power consumption, the heat dissipation power consumption, and the other power consumption, including: calculating the sum of the safe power consumption, the computing power consumption, and the other power consumption to obtain the component power consumption; calculating the product of the component power consumption and a third preset coefficient to obtain the heat dissipation power consumption; and summing the safe power consumption, the computing power consumption, the heat dissipation power consumption, and the other power consumption to obtain the real-time total power.

[0012] Optionally, if the task to be calculated is of the first priority, the safe power consumption is adjusted according to the real-time total power and the predicted allocation curve, including: determining the current scheduling value according to the predicted allocation curve; and increasing the safe power consumption if the current scheduling value is greater than the real-time total power.

[0013] Optionally, if the task to be computed is of the first priority, the safe power consumption is adjusted according to the real-time total power and the predicted allocation curve, including: determining the current scheduling value according to the predicted allocation curve; and increasing the computing power consumption if the current scheduling value is greater than the real-time total power.

[0014] According to another aspect of this application, an electrical grid is provided, comprising any of the systems described herein.

[0015] This application provides a power load-side resource scheduling system. The system includes a load aggregation module, a task scheduling module, and a load adjustment module, all interconnected. Specifically: the load aggregation module generates a first load curve based on historical operating data from a supercomputing center, divides the first load curve into stages based on load values ​​and preset thresholds to obtain a second load curve, calculates a predicted allocation curve based on the second load curve, sends the second load curve to the task scheduling module, and sends the predicted allocation curve to the load adjustment module. The first load curve is a curve representing the change in grid load over time. The task scheduling module allocates the execution time of the tasks to be calculated based on their priority and the second load curve to obtain a first scheduling scheme, and sends the first scheduling scheme to the load adjustment module. The load adjustment module calculates the real-time total power of the supercomputing center based on the first scheduling scheme, adjusts the operating power of different modules of the supercomputing center based on the real-time total power and the predicted allocation curve to obtain a second scheduling scheme, and controls the operation of the supercomputing center according to the second scheduling scheme. This application divides time according to the power grid load and classifies computing tasks based on project priority. It then matches the classified computing tasks with the divided time periods to achieve project computing adapted to the power grid load. This solves the problem in the prior art of lacking a scheduling method that adjusts project execution time based on project priority, power grid load, and computing resource requirements to avoid peak load periods affecting computing tasks. Attached Figure Description

[0016] Figure 1 A structural block diagram of a power load-side resource scheduling system according to an embodiment of this application is shown;

[0017] Figure 2 A structural block diagram of a load aggregation module provided according to an embodiment of this application is shown;

[0018] Figure 3 This illustrates a power grid load time period division diagram of a load curve division module provided according to an embodiment of this application;

[0019] Figure 4 A structural block diagram of a task scheduling module provided according to an embodiment of this application is shown;

[0020] Figure 5 A structural block diagram of a load regulation module provided according to an embodiment of this application is shown;

[0021] Figure 6 The diagram illustrates a scheduling curve generation module that generates a predicted scheduling curve according to an embodiment of this application. Detailed Implementation

[0022] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0023] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0024] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0025] As described in the background section, existing technologies typically involve a large number of computational projects that need to be run. These projects may have different priorities, resource requirements, and execution times. Therefore, effective load scheduling is required to manage these projects and adapt them to the power grid load. To address the lack of a scheduling method in existing technologies that adjusts project execution times based on project priority, power grid load, and computational resource requirements to avoid impacting computational tasks during peak load periods, embodiments of this application provide a power load-side resource scheduling system.

[0026] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0027] Figure 1 This is a structural block diagram of a power load-side resource scheduling system according to an embodiment of this application. Figure 2 As shown, the scheduling system includes: a load aggregation module, a task scheduling module, and a load adjustment module. The load aggregation module, the task scheduling module, and the load adjustment module are communicatively connected, wherein:

[0028] The load aggregation module is used to generate a first load curve based on the historical operating data of the supercomputing center, and divide the first load curve into stages according to the load value and preset threshold to obtain a second load curve. The module then performs calculations based on the second load curve to obtain a predicted allocation curve, sends the second load curve to the task scheduling module, and sends the predicted allocation curve to the load adjustment module. The first load curve is a curve of the power grid load changing over time.

[0029] Specifically, the load aggregation module obtains power data of the grid load changing over time within a preset time period from the supercomputing center and plots it as a curve of the grid load changing over time, i.e., the first load curve. A preset threshold is set, and the first load curve is divided into different grid load stages according to the power values ​​of the grid load at different times in the first load curve to obtain the second load curve. The second load curve is then sent to the task scheduling module. The second load curve is negatively calculated, and a dispatch coefficient is added to it to obtain the dispatch curves of different stages of grid load. These curves are then sent to the load adjustment module.

[0030] The task scheduling module is used to allocate the execution time of the task to be calculated according to the priority of the task to be calculated and the second load curve to obtain a first scheduling scheme, and send the first scheduling scheme to the load adjustment module.

[0031] Specifically, in the task scheduling module mentioned above, execution priorities are assigned to tasks according to the different types of tasks to be calculated, and execution scheduling schemes for different tasks to be calculated are comprehensively allocated according to the computational priorities of different tasks to be calculated and the power grid load time period corresponding to the second load curve mentioned above.

[0032] The load adjustment module is used to calculate the real-time total power of the supercomputing center according to the first scheduling scheme, and adjust the operating power of different modules of the supercomputing center according to the real-time total power and the predicted allocation curve to obtain a second scheduling scheme, and control the operation of the supercomputing center according to the second scheduling scheme.

[0033] Specifically, the load adjustment module compares the real-time total power with the predicted allocation curve and performs real-time adjustments. When the scheduling value is greater than the total power before allocation, the load adjustment module sends a computing power boosting instruction to the supercomputing center.

[0034] According to the technical solution of this application, in the aforementioned power load-side resource dispatching system, the dispatching system includes a load aggregation module, a task scheduling module, and a load adjustment module. The load aggregation module, the task scheduling module, and the load adjustment module are communicatively connected. Specifically: the load aggregation module generates a first load curve based on historical operating data of the supercomputing center, divides the first load curve into stages based on load values ​​and preset thresholds to obtain a second load curve, calculates a predicted allocation curve based on the second load curve, sends the second load curve to the task scheduling module, and sends the predicted allocation curve to the load adjustment module. The first load curve is a curve showing the change of power grid load over time. The task scheduling module allocates the execution time of the task to be calculated based on the priority of the task and the second load curve to obtain a first dispatching scheme, and sends the first dispatching scheme to the load adjustment module. The load adjustment module calculates the real-time total power of the supercomputing center based on the first dispatching scheme, adjusts the operating power corresponding to different modules of the supercomputing center based on the real-time total power and the predicted allocation curve to obtain a second dispatching scheme, and controls the operation of the supercomputing center according to the second dispatching scheme. This application divides time according to the power grid load and classifies computing tasks based on project priority. It then matches the classified computing tasks with the divided time periods to achieve project computing adapted to the power grid load. This solves the problem in the prior art of lacking a scheduling method that adjusts project execution time based on project priority, power grid load, and computing resource requirements to avoid peak load periods affecting computing tasks.

[0035] In one embodiment of this application, such as Figure 3 As shown, the above load aggregation module includes a load curve generation module, a load curve partitioning module, and a scheduling curve generation module, wherein:

[0036] The load curve generation module is used to predict based on the historical operation data of the power grid, generate a curve of the power grid load changing with time within a first preset period, obtain the first load curve, and send the first load curve to the load curve division module. The duration corresponding to the first preset period is the preset duration, and the start time of the first preset period is the current time.

[0037] Specifically, the load curve generation module first obtains historical operating data of the power grid load changing over time within a preset time period, generates a load curve, namely the first load curve, based on the data of the power grid load changing over time within the first preset time period, and then sends the generated first load curve to the load curve division module.

[0038] The load curve division module is used to traverse the first load curve, determine the curve in the first load curve whose corresponding grid load is less than or equal to the first threshold as the valley stage, determine the curve in the first load curve whose corresponding grid load is greater than the first threshold and less than the second threshold as the balance stage, and determine the curve in the first load curve whose corresponding grid load is greater than or equal to the second threshold as the peak stage, thereby obtaining the second load curve, and sending the second load curve to the scheduling curve generation module and the task scheduling module.

[0039] Specifically, such as Figure 4 As shown, the load curve division module divides the power grid load period into a low-valley phase (less than or equal to a first threshold), a balanced phase (greater than the first threshold but less than a second threshold), and a peak phase (greater than or equal to the second threshold) based on the power grid load threshold at different times within a preset time period, thereby obtaining the second load curve. The first threshold is 800, and the second threshold is 1500. Then, the second load curve is sent to the scheduling curve generation module and the task scheduling module.

[0040] The aforementioned scheduling curve generation module is used to fit the aforementioned predicted scheduling curve based on the first preset coefficient, the second preset coefficient, and the aforementioned second load curve, and then send the aforementioned predicted scheduling curve to the aforementioned load adjustment module.

[0041] Specifically, the dispatch curve generation module first takes a negative value for the second load curve, i.e., the first preset coefficient, and then adds the allocation coefficient, i.e. the second preset coefficient, to obtain the predicted dispatch curve.

[0042] In one embodiment of this application, such as Figure 5 As shown, the task scheduling module includes a priority determination module and a task matching module, wherein:

[0043] The priority determination module described above is used to iterate through the required computing speed and security scores of the tasks to be computed and divide the tasks to be computed into first priority, second priority and third priority, where the first priority is higher than the third priority and the third priority is higher than the second priority.

[0044] Specifically, the above-mentioned computing tasks can be divided into security tasks, general tasks, and restricted tasks. Security tasks need to be processed first and require high security and fast computing speed, so they are classified as first priority. General tasks have a medium priority and can have their computing speed appropriately reduced, so they are classified as third priority. Restricted tasks have a low priority and can be terminated if necessary, so they are classified as second priority.

[0045] The task matching module is used to assign the tasks to be calculated with the first priority to the low-priority phase for execution, assign the tasks to be calculated with the second priority to the peak phase for execution, and assign the tasks to be calculated with the third priority to the balanced phase for execution, thereby obtaining the first scheduling scheme.

[0046] Specifically, the different task types mentioned above are assigned to different power grid load phases for execution according to priority. Among them, the guarantee type tasks are suitable for execution during the off-peak electricity consumption phase, that is, the above-mentioned first priority tasks are assigned to the above-mentioned off-peak electricity consumption phase for execution. The ordinary type tasks are suitable for execution during the balanced period, that is, the above-mentioned third priority tasks are assigned to the above-mentioned balanced period for execution. The restricted type tasks are suitable for execution during the off-peak electricity consumption phase, that is, the above-mentioned second priority tasks are assigned to the above-mentioned off-peak electricity consumption phase for execution.

[0047] In one embodiment of this application, such as Figure 6 As shown, the load regulation module includes a power consumption calculation module, a safety power consumption adjustment module, and a computing power power consumption adjustment module, wherein:

[0048] The aforementioned power consumption calculation module is used to determine the safe power consumption of the security component, the computing power consumption of the computing component, and other power consumption of other power-consuming components in the supercomputing center according to the aforementioned first scheduling scheme; calculate the heat dissipation power consumption of the heat dissipation component of the supercomputing center according to the aforementioned safe power consumption, the aforementioned computing power consumption, and the aforementioned other power consumption; and calculate the aforementioned real-time total power according to the aforementioned safe power consumption, the aforementioned computing power consumption, the aforementioned heat dissipation power consumption, and the aforementioned other power consumption.

[0049] Specifically, the formula for calculating the real-time total power by the power consumption calculation module is: Total power = safe power consumption + computing power power consumption + other power consumption + heat dissipation power consumption, heat dissipation power consumption = (safe power consumption + computing power power consumption + other power consumption) * heat dissipation coefficient, where the heat dissipation coefficient is 0.1.

[0050] The aforementioned safe power consumption adjustment module is used to adjust the safe power consumption based on the real-time total power and the predicted allocation curve when the task to be calculated is of the first priority.

[0051] Specifically, the aforementioned safe power consumption adjustment module is used to adjust the safe power consumption of protection tasks in real time. When the power consumption during the off-peak electricity period is greater than the total power before adjustment, the aforementioned safe power consumption adjustment module sends an instruction to the aforementioned supercomputing center to increase the aforementioned safe power consumption.

[0052] The aforementioned computing power and power consumption adjustment module is used to adjust the aforementioned computing power and power consumption based on the aforementioned real-time total power and the aforementioned predicted allocation curve.

[0053] Specifically, the aforementioned computing power and power consumption adjustment module is used to adjust the computing power and power consumption in real time. When the power consumption of the predicted allocation curve is greater than the total power before adjustment, the aforementioned computing power and power consumption adjustment module sends an instruction to the aforementioned supercomputing center to increase the computing power.

[0054] In one embodiment of this application, the predicted scheduling curve is obtained by fitting a first preset coefficient, a second preset coefficient, and the aforementioned second load curve, including:

[0055] The third load curve is obtained by multiplying the second load curve mentioned above with the first preset coefficient mentioned above.

[0056] Specifically, the first preset coefficient is negative, and the second load curve is multiplied by the first preset coefficient to obtain the third load curve with a negative power value.

[0057] The predicted scheduling curve is obtained by summing the third load curve and the second preset coefficient.

[0058] Specifically, the second preset coefficient is a positive constant. The third load curve and the second preset coefficient are summed to obtain the predicted dispatch curve with a positive power value.

[0059] The above-mentioned computational tasks can be divided into security tasks, normal tasks, and restricted tasks. Security tasks need to be processed first and require high security and fast computation speed, so they are classified as first priority. Normal tasks have a medium priority and can have their computation speed appropriately reduced, so they are classified as third priority. Restricted tasks have a low priority and can be terminated if necessary, so they are classified as second priority.

[0060] In one embodiment of this application, the computational speed and security scores of the above-mentioned tasks to be computed are traversed to divide the tasks into a first priority, a second priority, and a third priority, including:

[0061] The above-mentioned tasks to be calculated with a corresponding calculation speed greater than the third threshold and a requirement safety score greater than the fourth threshold are identified as the first priority.

[0062] Specifically, the computation tasks that require both high computation speed and high security are designated as the first priority.

[0063] The above-mentioned tasks to be calculated that correspond to a calculation speed less than the fifth threshold and a requirement security score less than the sixth threshold are determined as the second priority, the third threshold is greater than the fifth threshold, and the fourth threshold is greater than the sixth threshold.

[0064] Specifically, the above-mentioned tasks that can be computed at a significantly reduced speed, can be terminated if necessary, and have low security requirements are identified as the second priority.

[0065] The tasks to be calculated, other than the first and second priorities mentioned above, are designated as the third priority.

[0066] Specifically, the above-mentioned computation tasks that require moderate computation speed, but can reduce computation speed and have moderate security are determined as the third priority.

[0067] In one embodiment of this application, the heat dissipation power consumption of the supercomputing center heat dissipation component is calculated based on the aforementioned security power consumption, computing power power consumption, and other power consumption. The real-time total power consumption is calculated based on the aforementioned security power consumption, computing power power consumption, heat dissipation power consumption, and other power consumption, including:

[0068] The power consumption of the component is obtained by summing the above-mentioned safe power consumption, computing power power consumption, and other power consumption.

[0069] Specifically, the power consumption of the aforementioned components refers to the power consumption generated by the system components of the scheduling system during operation. Component power consumption = security power consumption + computing power power consumption + other power consumption.

[0070] The above heat dissipation power consumption is obtained by multiplying the power consumption of the above components by the third preset coefficient.

[0071] Specifically, the third preset coefficient is 0.1, and the heat dissipation power consumption = (security power consumption + computing power power consumption + other power consumption) * 0.1.

[0072] The above-mentioned safe power consumption, computing power consumption, heat dissipation power consumption, and other power consumption are summed to obtain the above-mentioned real-time total power.

[0073] Specifically, real-time total power = safe power consumption + computing power power consumption + other power consumption + heat dissipation power consumption.

[0074] In one embodiment of this application, when the task to be calculated is of the first priority, adjusting the safe power consumption based on the real-time total power and the predicted allocation curve includes:

[0075] The current scheduling value is determined based on the above-mentioned predicted allocation curve;

[0076] Specifically, the load aggregation module calculates the predicted allocation curve based on the second load curve and sends the predicted allocation curve to the load adjustment module. The load adjustment module then determines the current scheduling value based on the predicted allocation curve.

[0077] If the current scheduling value is greater than the real-time total power, increase the safe power consumption.

[0078] Specifically, the power consumption of the first priority task to be computed is controlled. When the current scheduling value is greater than the real-time total power, an instruction to increase the power consumption of the security system is sent to the supercomputing center.

[0079] In one embodiment of this application, when the task to be computed is of the first priority, adjusting the safe power consumption based on the real-time total power and the predicted allocation curve includes:

[0080] The current scheduling value is determined based on the predicted allocation curve;

[0081] Specifically, the load aggregation module calculates the predicted allocation curve based on the second load curve and sends the predicted allocation curve to the load adjustment module. The load adjustment module then determines the current scheduling value based on the predicted allocation curve.

[0082] If the current scheduling value is greater than the real-time total power, the computing power consumption is increased.

[0083] Specifically, the computing power and power consumption of the tasks to be computed with the first priority are controlled. When the current scheduling value is greater than the real-time total power, an instruction to increase computing power is sent to the supercomputing center.

[0084] This application provides a power grid, including any one of the systems described above for implementing a power load-side resource dispatching system.

[0085] Specifically, a power load-side resource scheduling system includes: a load aggregation module, a task scheduling module, and a load adjustment module, wherein the load aggregation module, the task scheduling module, and the load adjustment module are communicatively connected, wherein:

[0086] The load aggregation module is used to generate a first load curve based on the historical operating data of the supercomputing center, and divide the first load curve into stages according to the load value and preset threshold to obtain a second load curve. The module then performs calculations based on the second load curve to obtain a predicted allocation curve, sends the second load curve to the task scheduling module, and sends the predicted allocation curve to the load adjustment module. The first load curve is a curve of the power grid load changing over time.

[0087] Specifically, the load aggregation module obtains power data of the grid load changing over time within a preset time period from the supercomputing center and plots it as a curve of the grid load changing over time, i.e., the first load curve. A preset threshold is set, and the first load curve is divided into different grid load stages according to the power values ​​of the grid load at different times in the first load curve to obtain the second load curve. The second load curve is then sent to the task scheduling module. The second load curve is negatively calculated, and a dispatch coefficient is added to it to obtain the dispatch curves of different stages of grid load. These curves are then sent to the load adjustment module.

[0088] The task scheduling module is used to allocate the execution time of the task to be calculated according to the priority of the task to be calculated and the second load curve to obtain a first scheduling scheme, and send the first scheduling scheme to the load adjustment module.

[0089] Specifically, in the task scheduling module mentioned above, execution priorities are assigned to tasks according to the different types of tasks to be calculated, and execution scheduling schemes for different tasks to be calculated are comprehensively allocated according to the computational priorities of different tasks to be calculated and the power grid load time period corresponding to the second load curve mentioned above.

[0090] The load adjustment module is used to calculate the real-time total power of the supercomputing center according to the first scheduling scheme, and adjust the operating power of different modules of the supercomputing center according to the real-time total power and the predicted allocation curve to obtain a second scheduling scheme, and control the operation of the supercomputing center according to the second scheduling scheme.

[0091] Specifically, the load adjustment module compares the real-time total power with the predicted allocation curve and performs real-time adjustments. When the scheduling value is greater than the total power before allocation, the load adjustment module sends a computing power boosting instruction to the supercomputing center.

[0092] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0093] As can be seen from the above description, the embodiments of this application achieve the following technical effects:

[0094] 1) In a power load-side resource scheduling system according to this application, the scheduling system includes a load aggregation module, a task scheduling module, and a load adjustment module. The load aggregation module, the task scheduling module, and the load adjustment module are communicatively connected. The load aggregation module generates a first load curve based on historical operating data of a supercomputing center and divides the first load curve into stages based on load values ​​and preset thresholds to obtain a second load curve. It then performs calculations based on the second load curve to obtain a predicted allocation curve, sends the second load curve to the task scheduling module, and sends the predicted allocation curve to the load adjustment module. The first load curve is a curve showing the change of power grid load over time. The task scheduling module allocates the execution time of the task to be calculated based on the priority of the task and the second load curve to obtain a first scheduling scheme, and sends the first scheduling scheme to the load adjustment module. The load adjustment module calculates the real-time total power of the supercomputing center based on the first scheduling scheme, adjusts the operating power corresponding to different modules of the supercomputing center based on the real-time total power and the predicted allocation curve to obtain a second scheduling scheme, and controls the operation of the supercomputing center according to the second scheduling scheme. This application divides time according to the power grid load and classifies computing tasks based on project priority. It then matches the classified computing tasks with the divided time periods to achieve project computing adapted to the power grid load. This solves the problem in the prior art of lacking a scheduling method that adjusts project execution time based on project priority, power grid load, and computing resource requirements to avoid peak load periods affecting computing tasks.

[0095] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A power load-side resource dispatching system, characterized in that, The scheduling system includes a load aggregation module, a task scheduling module, and a load adjustment module. The load aggregation module, the task scheduling module, and the load adjustment module are communicatively connected, wherein: The load aggregation module is used to generate a first load curve based on the historical operating data of the supercomputing center, and divide the first load curve into stages according to the load value and a preset threshold to obtain a second load curve. It then performs calculations based on the second load curve to obtain a predicted allocation curve, sends the second load curve to the task scheduling module, and sends the predicted allocation curve to the load adjustment module. The first load curve is a curve of the power grid load changing over time. The task scheduling module is used to allocate the execution time of the task to be calculated according to the priority of the task to be calculated and the second load curve, obtain a first scheduling scheme, and send the first scheduling scheme to the load adjustment module; The load adjustment module is used to calculate the real-time total power of the supercomputing center according to the first scheduling scheme, and adjust the operating power of different modules of the supercomputing center according to the real-time total power and the predicted allocation curve to obtain a second scheduling scheme, and control the operation of the supercomputing center according to the second scheduling scheme. The load aggregation module includes a scheduling curve generation module, which is used to fit a predicted scheduling curve based on a first preset coefficient, a second preset coefficient, and a second load curve, and then send the predicted scheduling curve to the load adjustment module. The predicted scheduling curve is obtained by fitting the first preset coefficient, the second preset coefficient, and the second load curve, including: Calculate the product of the second load curve and the first preset coefficient to obtain the third load curve, wherein the first preset coefficient is a negative value; The predicted scheduling curve is obtained by summing the third load curve and the second preset coefficient, wherein the second preset coefficient is a positive constant used to make the predicted scheduling curve positive.

2. The system according to claim 1, characterized in that, The load aggregation module includes a load curve generation module, a load curve partitioning module, and a scheduling curve generation module, wherein: The load curve generation module is used to predict based on the historical operation data of the power grid, generate a curve of the power grid load changing with time within a first preset time period, obtain the first load curve, and send the first load curve to the load curve division module. The duration corresponding to the first preset time period is a preset duration, and the start time of the first preset time period is the current time. The load curve segmentation module is used to traverse the first load curve, determine the curve in the first load curve whose corresponding grid load is less than or equal to a first threshold as a low-valley phase, determine the curve in the first load curve whose corresponding grid load is greater than the first threshold and less than a second threshold as a balanced phase, and determine the curve in the first load curve whose corresponding grid load is greater than or equal to the second threshold as a peak phase, thereby obtaining the second load curve, and sending the second load curve to the scheduling curve generation module and the task scheduling module.

3. The system according to claim 2, characterized in that, The task scheduling module includes a priority determination module and a task matching module, wherein: The priority determination module is used to iterate through the required computing speed and security score of the task to be computed and divide the task to be computed into a first priority, a second priority and a third priority, wherein the first priority is higher than the third priority and the third priority is higher than the second priority. The task matching module is used to allocate the task to be calculated with the first priority to the low-priority phase for execution, allocate the task to be calculated with the second priority to the peak phase for execution, and allocate the task to be calculated with the third priority to the balanced phase for execution, thereby obtaining the first scheduling scheme.

4. The system according to claim 3, characterized in that, The load regulation module includes a power consumption calculation module, a safety power consumption adjustment module, and a computing power power consumption adjustment module, wherein: The power consumption calculation module is used to determine the safe power consumption of the security component, the computing power consumption of the computing component, and other power consumption of other power-consuming components in the supercomputing center according to the first scheduling scheme; calculate the heat dissipation power consumption of the heat dissipation component of the supercomputing center according to the safe power consumption, the computing power consumption, and the other power consumption; and calculate the real-time total power according to the safe power consumption, the computing power consumption, the heat dissipation power consumption, and the other power consumption. The safe power consumption adjustment module is used to adjust the safe power consumption according to the real-time total power and the predicted allocation curve when the task to be calculated is of the first priority. The computing power and power consumption adjustment module is used to adjust the computing power and power consumption according to the real-time total power and the predicted allocation curve.

5. The system according to claim 3, characterized in that, The computational speed and security scores of the tasks to be computed are used to classify the tasks into three priorities: first priority, second priority, and third priority. The task to be computed that has a corresponding computing speed greater than the third threshold and a security requirement score greater than the fourth threshold is determined as the first priority. The task to be computed that corresponds to a computing speed less than the fifth threshold and a requirement security score less than the sixth threshold is determined as the second priority, wherein the third threshold is greater than the fifth threshold and the fourth threshold is greater than the sixth threshold; The tasks to be computed, other than those of the first and second priorities, are designated as the third priority.

6. The system according to claim 4, characterized in that, The heat dissipation power consumption of the supercomputing center's heat dissipation components is calculated based on the safe power consumption, the computing power power consumption, and the other power consumption. The real-time total power is also calculated based on the safe power consumption, the computing power power consumption, the heat dissipation power consumption, and the other power consumption, including: The component power consumption is obtained by summing the security power consumption, the computing power power consumption, and the other power consumptions. The heat dissipation power consumption is obtained by multiplying the power consumption of the component by the third preset coefficient. The real-time total power is obtained by summing the security power consumption, the computing power consumption, the heat dissipation power consumption, and the other power consumption.

7. The system according to claim 4, characterized in that, When the task to be calculated is of the first priority, the safe power consumption is adjusted according to the real-time total power and the predicted allocation curve, including: The current scheduling value is determined based on the predicted allocation curve; If the current scheduling value is greater than the real-time total power, the safe power consumption is increased.

8. The system according to claim 4, characterized in that, When the task to be calculated is of the first priority, the safe power consumption is adjusted according to the real-time total power and the predicted allocation curve, including: The current scheduling value is determined based on the predicted allocation curve; If the current scheduling value is greater than the real-time total power, the computing power consumption is increased.

9. A power grid, characterized in that, Includes the system described in any one of claims 1 to 6.

Images