A computing power and power cooperative scheduling method and system

Abstract

This application discloses a method and system for collaborative scheduling of computing power and electricity, comprising: acquiring real-time power supply surplus and supply-demand fluctuation trend data of the power grid within a preset future time period; acquiring computing power resource surplus, mixed load characteristics, task constraints, and real-time electricity demand of the computing power center; collecting electricity data from the energy storage system of the computing power center in real time; inputting the above data as multi-dimensional data into a large model; performing data fusion processing on the multi-dimensional data to construct an integrated dataset of electricity supply, computing power demand, task constraints, and energy consumption loss characteristics, and performing deep correlation feature mining based on the integrated dataset; intelligently generating a collaborative scheduling strategy for computing power and electricity based on the multi-dimensional data and deep correlation features; and performing collaborative scheduling of computing power and electricity based on the collaborative scheduling strategy for computing power and electricity. This method combines real-time supply and demand data with future forecast data to perform collaborative scheduling of computing power and electricity, thereby optimizing the allocation of computing power and electricity resources.

Description

Technical Field

[0001] This invention relates to the field of resource scheduling technology, and in particular to a method and system for coordinated scheduling of computing power and electricity. Background Technology

[0002] With the rapid development of the digital economy and artificial intelligence industry, the computing power industry has become the core carrier of new productivity. At the same time, it has also brought about a series of practical problems such as the surge in electricity consumption for computing power, the imbalance of energy consumption structure, and the mismatch between computing power load and new energy output.

[0003] The use of two independent systems for scheduling and operation of the power grid and computing centers, with inconsistent protocols, data incompatibility, and uncoordinated commands, makes this separation of systems unsuitable for the demands of modern industrial development. Therefore, promoting the coordinated development of computing infrastructure and the new power system to achieve integrated computing and power dispatching has become an urgent issue for the industry.

[0004] In existing technologies, when coordinating computing power and power scheduling, a fixed-rule scheduling architecture is often used, which leads to problems of lag and rigidity in computing-power coordination, and makes it impossible to achieve optimal allocation of computing and power resources.

[0005] Therefore, there is an urgent need to provide a computing power and power coordinated scheduling scheme that can solve at least one of the above technical problems. Summary of the Invention

[0006] To address the above technical issues, embodiments of the present invention provide a computing power and power coordinated scheduling method and system, which can combine real-time supply and demand data with future forecast data to perform computing power coordinated scheduling and optimize the allocation of computing and power resources.

[0007] On one hand, embodiments of the present invention provide a method for coordinated scheduling of computing power and power. The method is used in a computing-power coordinated system, which includes: a computing power center, a power grid, and a coordinated scheduling system for computing power and power. The coordinated scheduling system is connected to both the computing power center and the power grid. A large-scale model is deployed within the coordinated scheduling system. An energy storage system is deployed in the computing power center. The computing power center receives power from the power grid or the energy storage system. The method includes: To obtain real-time power supply capacity of the power grid and supply and demand fluctuation trend data within a preset future time period; Obtain the computing power resource reserves, mixed load characteristics, task constraints, and real-time power demand of the computing power center; Real-time collection of power data from the energy storage system in the computing center; The real-time power supply margin of the power grid and the supply and demand fluctuation trend data within a future preset time period, the computing power resource margin of the computing center, the characteristics of mixed loads, task constraints and real-time power demand, and the power data of the energy storage system are input into the large model as multi-dimensional data. The large model performs data fusion processing on the received multidimensional data to construct an integrated dataset of power supply, computing power demand, task constraints, and energy consumption loss characteristics. Based on the integrated dataset, deep correlation feature mining is performed. The deep correlation features are used to characterize the time-series fluctuation pattern of power output, the power output fluctuation trend within a future preset time period, the time period of power supply surplus and the corresponding power surplus data, and the time period of power supply shortage and the corresponding power shortage data. The large model intelligently generates a computing and power collaborative scheduling strategy based on the received multidimensional data and the mined deep correlation features. The computing power and power coordinated scheduling system performs coordinated scheduling of computing power and power based on the aforementioned computing power and power coordinated scheduling strategy.

[0008] Furthermore, the method also includes: Collect real-time power output data released by the power grid. The power output data includes at least: power output data of various new energy sources and traditional thermal power output data, power generation data and power output forecast data. Based on the power output data, the real-time power supply margin of the power grid and the supply and demand fluctuation trend data within a future preset time period are calculated. The system collects full-domain operational data from the computing center. This full-domain operational data includes: computing resource status data, tenant task operation data, and computing load and energy consumption data. The computing resource status data includes: the number of online heterogeneous computing resources, idle computing slices, remaining video memory, available cores, and reclaimable redundant resources. The tenant task operation data includes: task type, priority, tenant credit rating, latency threshold, operation progress, and off-peak attribute. The task type includes real-time task type and offline task type. The computing load and energy consumption data includes: computing utilization rate, video memory occupancy rate, device instantaneous power consumption, and total cluster load. Based on the online quantity, idle computing power slices, remaining video memory, available cores, and recyclable redundant resources of various heterogeneous computing power resources in the computing power resource status data, the remaining amount of heterogeneous computing power resources is calculated. Based on the task types in the tenant task execution data, mixed load characteristics are extracted; Based on the priority, tenant credit rating, latency threshold, running progress, and off-peak attribute in the tenant task running data, task constraints are determined. Among them, based on the task constraints, tasks are divided into tasks that can be deloaded, tasks that must be guaranteed, and tasks that can be moved. Based on the computing power utilization rate, video memory occupancy rate, instantaneous power consumption of devices, and total cluster load in the computing power load and energy consumption data, the real-time power demand is calculated.

[0009] Furthermore, the method also includes: Obtain electricity price data for the same period in previous years; Based on the power output data and the electricity price data for the same period in previous years, a routine charging and discharging plan for the energy storage system is determined. The routine charging and discharging plan includes at least the following: charging the energy storage system through the power grid during a first time period when the power output data is greater than or equal to a first set value and the electricity price is lower than a second set value; and discharging the energy storage system to supply power to the computing center during a second time period when the power output data is less than a third set value and the electricity price is higher than a fourth set value. The first set value is greater than the third set value, and the second set value is less than the fourth set value.

[0010] Furthermore, the method also includes: When the time reaches a preset time point before the first and second time periods of the routine charging and discharging plan, the real-time electricity price data corresponding to the preset time point is obtained; The real-time electricity price data is compared with the average electricity price in the first time period of the same period in previous years and the average electricity price in the second time period of the same period in previous years, respectively. If the electricity price difference is less than the preset electricity price difference, the routine charging and discharging plan will be executed after the time reaches the time point corresponding to the routine charging and discharging plan. If the electricity price difference is greater than or equal to the preset electricity price difference, the routine charging and discharging plan is adjusted based on the real-time electricity price data, and the energy storage system is charged and discharged based on the adjusted routine charging and discharging plan.

[0011] Furthermore, the energy storage system includes a voltage stabilizing unit, and the method further includes: Real-time monitoring of voltage fluctuations in power grid output data; If the voltage fluctuation reaches the preset fluctuation threshold, and the computing center is currently powered by the power grid, then the remaining power of the energy storage system is detected. If the remaining power of the energy storage system is greater than or equal to the first power threshold, the connection between the power grid and the computing center is cut off, and the energy storage system is switched to supply power to the computing center. If the remaining power of the energy storage system is less than the first power threshold but greater than the second power threshold, the grid will be connected to the energy storage system so that the power output from the grid can be stabilized by the voltage stabilization unit in the energy storage system to supply power to the computing center. The second power threshold is a low power warning value or a discharge protection threshold, and the second power threshold is less than the first power threshold.

[0012] Furthermore, the computing center is also equipped with voltage stabilization equipment, and the method further includes: If the remaining power of the energy storage system is less than the second power threshold, the power supply connection between the energy storage system and the computing center is disconnected, the voltage stabilizing device is activated, and the voltage stabilizing device is connected between the power grid and the computing center so that the power output from the power grid is stabilized by the voltage stabilizing device before being supplied to the computing center; and The energy storage system is connected to the power grid so that the power grid can charge the energy storage system. During the charging process, the voltage of the power output from the power grid is regulated by a voltage stabilizing unit in the energy storage system.

[0013] Furthermore, the method also includes: Real-time acquisition of abnormal data, including: abnormal power load data, historical power outage records, power grid fault data or power equipment maintenance notices, temporary power outage notices, and major event notices where the power load reaches a set value. Abnormal data is uploaded to a large model, which then makes predictions and outputs the estimated time when the power grid will experience a power outage. During normal power grid operation, the energy storage system is charged and a special reserve identifier is set for it. This identifier indicates that the energy storage system will only be activated when a preset emergency condition is met.

[0014] Furthermore, if it is detected that the energy storage system is set with a dedicated storage identifier and the energy storage system is in the process of discharging, the discharge operation will be stopped. If the time reaches the routine discharge time of the energy storage system, the discharge operation will not be performed until the preset emergency conditions are met, at which point the energy storage system will be activated to power the computing center. If the power grid is in a power outage state after the energy storage system is activated, the load operation strategy is calculated through a large model based on the power data of the energy storage system, the power demand of each load in all the mixed loads of the computing center, and the priority of each load. The mixed loads include rigid loads and adjustable flexible loads, and the priority of rigid loads is higher than that of adjustable flexible loads. The priority of different loads in the adjustable flexible loads is positively correlated with the real-time requirements of the corresponding loads and negatively correlated with the energy consumption of the corresponding loads. The operating status of each load is controlled based on the load operation strategy. The load operation strategy can ensure the normal operation of rigid loads. Among the flexible adjustable loads, each load is determined to operate normally or stop operating according to priority. The lower the priority of the flexible adjustable load, the more likely it is to stop operating. This is so that the power grid ends the power outage and restores power supply before the energy storage system's power data is exhausted.

[0015] On the other hand, embodiments of the present invention provide a computing power and power coordinated dispatch system, which is connected to a computing power center and a power grid, respectively. The computing power and power coordinated dispatch system internally deploys a large-scale model, and the computing power center deploys an energy storage system. The computing power center receives power from the power grid or the energy storage system. The computing power and power coordinated dispatch system includes: The first acquisition module is used to acquire the real-time power supply margin of the power grid and the supply and demand fluctuation trend data within a future preset time period; The second acquisition module is used to acquire the computing power resource reserves, mixed load characteristics, task constraints and real-time power demand of the computing power center. The data acquisition module is used to collect power data from the energy storage system of the computing center in real time. The input module is used to input the real-time power supply margin of the power grid and the supply and demand fluctuation trend data within a preset time period, the computing power resource margin of the computing center, the characteristics of mixed loads, task constraints and real-time power demand, and the power data of the energy storage system as multi-dimensional data into the large model. The large model is used to perform data fusion processing on the received multidimensional data, construct an integrated dataset of power supply, computing power demand, task constraints and energy consumption loss characteristics, and perform deep correlation feature mining based on the integrated dataset. The deep correlation features are used to characterize the time-series fluctuation pattern of power output, the fluctuation trend of power output in the future preset time period, the time period of power supply surplus and the corresponding power surplus data, and the time period of power supply shortage and the corresponding power shortage data. The large model also intelligently generates a computing and power collaborative scheduling strategy based on the received multidimensional data and the mined deep correlation features. The scheduling module performs coordinated scheduling of computing power and electricity based on a computing-electricity coordinated scheduling strategy.

[0016] Thirdly, embodiments of the present invention provide an electronic device, including: a memory and a processor, wherein the memory and the processor are connected; Memory, used to store computer programs; A processor is used to invoke a computer program stored in memory to perform any of the above methods.

[0017] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when run by a computer, performs any of the methods described above.

[0018] This invention provides a method and system for collaborative scheduling of computing power and power. The method is used in a computing-power collaborative system, which includes a computing power center, a power grid, and a collaborative scheduling system for computing power and power. The collaborative scheduling system is connected to both the computing power center and the power grid. A large-scale model is deployed within the collaborative scheduling system. An energy storage system is deployed in the computing power center, which receives power from the power grid or the energy storage system. The method includes: acquiring the real-time power supply surplus of the power grid and supply-demand fluctuation trend data within a preset future time period; acquiring the computing power resource surplus, mixed load characteristics, task constraints, and real-time power demand of the computing power center; collecting real-time power data from the energy storage system of the computing power center; and combining the real-time power supply surplus of the power grid and the supply-demand fluctuation trend data within a preset future time period with the data from the computing power center. The system inputs multidimensional data, including computing power resource surplus, mixed load characteristics, task constraints, real-time electricity demand, and electricity data from the energy storage system, into a large model. The large model performs data fusion processing on the received multidimensional data to construct an integrated dataset of power supply, computing power demand, task constraints, and energy consumption characteristics. Based on this integrated dataset, it performs deep correlation feature mining. These deep correlation features characterize the temporal fluctuation patterns of power output, the trend of power output fluctuations within a preset future time period, periods of surplus power supply and corresponding surplus electricity data, and periods of power shortage and corresponding shortage electricity data. Based on the received multidimensional data and the mined deep correlation features, the large model intelligently generates a computing-power collaborative scheduling strategy. The computing power and power collaborative scheduling system then performs collaborative scheduling of computing power and power based on this strategy. This method and system, based on real-time collected data and future predicted data, and after deep correlation feature mining by the large model, intelligently generates a computing-power collaborative scheduling strategy. This effectively cooperates with the power grid to peak and valley shaving, smooth the power load curve, improve the supply and demand balance of power resources, and achieve optimal allocation of computing and power resources. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 A flowchart of a computing power and power coordinated scheduling method provided in this application embodiment; Figure 2 A block diagram of a computing power and power coordinated scheduling system provided in this application embodiment; Figure 3 This is a block diagram of a computing and power coordination system provided in an embodiment of this application. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with specific embodiments and accompanying drawings. It should be particularly noted that the embodiments described in this application are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0022] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this application should have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms "first," "second," and similar terms used in the embodiments of this application do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word covers the element or object listed after the word and its equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are only used to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[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 skilled in the art without creative effort are within the scope of protection of the present application.

[0024] One embodiment of this patent application provides a method for coordinated scheduling of computing power and electricity, such as... Figure 1 As shown, the method is used in a computer-computer cooperative system, refer to Figure 3 The computing and power coordination system 30 shown includes: a computing center 301, a power grid 302, and a computing and power coordinated dispatching system 303. The computing and power coordinated dispatching system 303 is connected to the computing center 301 and the power grid 302 respectively. A large model 3031 is deployed inside the computing and power coordinated dispatching system 303. An energy storage system 3011 is deployed in the computing center 301. The computing center 301 is powered by the power grid 302 or the energy storage system 3011. The method includes steps 101-107: Step 101: Obtain the real-time power supply capacity of the power grid and the supply and demand fluctuation trend data within a preset future time period; Step 102: Obtain the computing power resource reserves, mixed load characteristics, task constraints, and real-time power demand of the computing power center; Step 103: Collect power data from the energy storage system in the computing center in real time; Step 104: Input the real-time power supply margin of the power grid and the supply and demand fluctuation trend data within the future preset time period, the computing power resource margin of the computing center, the characteristics of mixed loads, task constraints and real-time power demand, and the power data of the energy storage system as multi-dimensional data into the large model. Step 105: The large model performs data fusion processing on the received multidimensional data to construct an integrated dataset of power supply, computing power demand, task constraints, and energy consumption loss features. Based on the integrated dataset, deep correlation feature mining is performed. The deep correlation features are used to characterize the power output time-series fluctuation pattern, the power output fluctuation trend in the future preset time period, the power supply surplus period and the corresponding power surplus data, and the power supply shortage period and the corresponding power shortage data. Step 106: Based on the received multidimensional data and the mined deep correlation features, the large model intelligently generates a computing and power collaborative scheduling strategy. Step 107: The computing power and power coordinated scheduling system performs coordinated scheduling of computing power and power based on the computing power and power coordinated scheduling strategy.

[0025] In this embodiment, multi-dimensional data from the power grid, computing center, and energy storage center are collected and input into a large model for data fusion processing. This constructs an integrated dataset of power supply, computing demand, task constraints, and energy consumption characteristics. Deep correlation feature mining is then performed to intelligently generate a computing-power collaborative scheduling strategy. Implementing computing-power collaborative scheduling can effectively cooperate with the power grid to smooth peak shaving and valley filling, improve the supply and demand balance of power resources, and achieve optimal resource allocation.

[0026] The multidimensional data obtained in steps 101 and 102 of the above embodiments can be achieved through the following steps: Collect real-time power output data released by the power grid. The power output data includes at least: power output data of various new energy sources and traditional thermal power output data, power generation data and power output forecast data. Among them, the power output data of various new energy sources mainly include: photovoltaic power output data, wind power output data, hydropower output data, etc.

[0027] Based on the power output data, the real-time power supply margin of the power grid and the supply and demand fluctuation trend data within a preset time period in step 101 are calculated.

[0028] The power grid refers to the entire power transmission and distribution system that transmits electricity generated by power plants to power users. It mainly includes three parts: transmission network, distribution network, and substation facilities. In a broad sense, the power grid here can include public power grids, as well as regional, industrial park, or enterprise self-use power grids (not connected to the public grid). There are no specific restrictions here. Any system that can provide relatively stable power output and supply and requires power dispatch can be considered as the power grid here.

[0029] Those skilled in the art will readily understand that the connection between the power grid and the computing center, and the computing power and power coordinated dispatch system, can be determined as needed, and is not limited to direct or indirect physical connections, communication connections, electrical connections, etc. Depending on the different power grid voltages, isolation measures can also be taken, and transmission and transformation equipment can be added, etc. The embodiments of this application do not make specific limitations.

[0030] The system collects full-domain operational data from the computing center. This full-domain operational data includes: computing resource status data, tenant task operation data, and computing load and energy consumption data. The computing resource status data includes: the number of online heterogeneous computing resources, idle computing slices, remaining video memory, available cores, and reclaimable redundant resources. The tenant task operation data includes: task type, priority, tenant credit rating, latency threshold, operation progress, and off-peak attribute. The task type includes real-time task type and offline task type. The computing load and energy consumption data includes: computing utilization rate, video memory occupancy rate, device instantaneous power consumption, and total cluster load. Based on the online number of various heterogeneous computing resources (such as GPU, NPU, FPGA, etc.), idle computing slices, remaining video memory, available cores, and recyclable redundant resources in the computing resource status data, the heterogeneous computing resource reserve in step 102 is calculated. Based on the task type in the tenant task execution data, the mixed load characteristics in step 102 are extracted; Based on the priority, tenant credit rating, latency threshold, running progress, and off-peak attribute in the tenant task running data, the task constraints in step 102 are determined, wherein the tasks are divided into tasks that can be deloaded, tasks that must be guaranteed, and tasks that can be moved based on the task constraints. Based on the computing power utilization rate, memory occupancy rate, instantaneous power consumption of the device, and total cluster load in the computing power load and energy consumption data, the real-time power demand in step 102 is calculated.

[0031] In another embodiment, the method further includes: Obtain electricity price data for the same period in previous years; Based on the power output data and the electricity price data for the same period in previous years, a routine charging and discharging plan for the energy storage system is determined. The routine charging and discharging plan includes at least the following: charging the energy storage system through the power grid during a first time period when the power output data is greater than or equal to a first set value and the electricity price is lower than a second set value; and discharging the energy storage system to supply power to the computing center during a second time period when the power output data is less than a third set value and the electricity price is higher than a fourth set value. The first set value is greater than the third set value, and the second set value is less than the fourth set value.

[0032] In this embodiment, a routine charging and discharging plan is determined for the energy storage system using electricity price data from the same period in previous years. This allows the energy storage system to be charged during peak power output periods and when electricity prices are low, and conversely, to be discharged to supply power to the computing center during off-peak power output periods and when electricity prices are high. This effectively helps to achieve peak shaving and valley filling of the power grid, smooth the power grid load curve, and effectively reduce electricity costs while meeting the power supply needs of the computing center.

[0033] Specifically, when the time reaches a preset time point before the first and second time periods of the routine charging and discharging plan, the real-time electricity price data corresponding to the preset time point is obtained; The real-time electricity price data is compared with the average electricity price in the first time period of the same period in previous years and the average electricity price in the second time period of the same period in previous years, respectively. If the electricity price difference is less than the preset electricity price difference, the routine charging and discharging plan will be executed after the time reaches the time point corresponding to the routine charging and discharging plan. If the electricity price difference is greater than or equal to the preset electricity price difference, the routine charging and discharging plan is adjusted based on the real-time electricity price data, and the energy storage system is charged and discharged based on the adjusted routine charging and discharging plan.

[0034] Because the electricity market uses real-time spot pricing, routine charging and discharging plans based on historical electricity price data may not be economically optimal. Therefore, at a preset time point before the scheduled charging and discharging period (e.g., 15 minutes, 30 minutes, or 1 hour in advance), since the electricity price at this time is close to the real-time price during the scheduled charging and discharging period, the price at this time is compared with the average price during the same period in previous years. This helps determine if the price difference is too large. If the price difference is small, the energy storage system is charged and discharged according to the set routine charging and discharging plan. If the price difference is too large, it means that executing the charging and discharging according to the original routine charging and discharging plan will not be economical. The routine charging and discharging plan needs to be adjusted based on the real-time electricity price before execution, thereby ensuring that the adjusted charging and discharging plan is more economical and saves electricity costs.

[0035] To address the impact of grid voltage fluctuations on computing center equipment, a voltage stabilization unit is incorporated into the energy storage system. The method further includes: Real-time monitoring of voltage fluctuations in power grid output data; If the voltage fluctuation reaches the preset fluctuation threshold, and the computing center is currently powered by the power grid, then the remaining power of the energy storage system is detected. If the remaining power of the energy storage system is greater than or equal to the first power threshold, the connection between the power grid and the computing center is cut off, and the energy storage system is switched to supply power to the computing center. If the remaining power of the energy storage system is less than the first power threshold but greater than the second power threshold, the grid will be connected to the energy storage system so that the power output from the grid can be stabilized by the voltage stabilization unit in the energy storage system to supply power to the computing center. The second power threshold is a low power warning value or a discharge protection threshold, and the second power threshold is less than the first power threshold.

[0036] In this embodiment, when the grid voltage fluctuates significantly, if the remaining power of the energy storage system is large (greater than or equal to the first power threshold), for example, if the first power threshold is 50% or 60% and the remaining power is 70% or 80%, then the system is directly switched to supply power to the energy storage system. When the remaining power of the energy storage system is low (less than the first power threshold and greater than the second power threshold), for example, if the second power threshold is 20% or 15% and the remaining power is 30%, then it is considered that there is a risk of insufficient power if the energy storage system is used to supply power alone. Therefore, the grid needs to be connected to the energy storage system so that the voltage stabilization unit in the energy storage system can stabilize the power output from the grid before supplying power to the computing center.

[0037] As another solution to address the impact of power grid voltage fluctuations on computing center equipment, the computing center is also equipped with voltage stabilization devices. The method further includes: If the remaining power of the energy storage system is less than the second power threshold, the power supply connection between the energy storage system and the computing center is disconnected, the voltage stabilizing device is activated, and the voltage stabilizing device is connected between the power grid and the computing center so that the power output from the power grid is stabilized by the voltage stabilizing device before being supplied to the computing center; and The energy storage system is connected to the power grid to charge the energy storage system through the power grid. During the charging process, the voltage of the power output from the power grid is regulated by a voltage stabilizing unit in the energy storage system.

[0038] In this embodiment, if the grid voltage fluctuates significantly and the remaining power of the energy storage system is low (less than a second power threshold, which is a low power warning value or a discharge protection threshold), for example, 12% or 10% of the remaining power, then the energy storage system is considered to be too low and needs to be replenished immediately. Therefore, the power supply connection between the energy storage system and the computing center needs to be disconnected, and the energy storage system needs to be connected to the grid to charge the energy storage system (at this time, the energy storage system does not supply power to the computing center so that it can be replenished as soon as possible). In this case, a separate voltage regulator deployed at the computing center needs to be activated to stabilize the grid voltage before supplying power to the computing center. By using a separately set voltage regulator, it can be activated in an emergency when the remaining power of the energy storage system is too low, thereby solving the grid voltage fluctuation problem without affecting the charging of the energy storage system, thus effectively improving the emergency power reserve capacity.

[0039] In one embodiment, the method further includes: Real-time acquisition of abnormal data, including: abnormal power load data, historical power outage records, power grid fault data or power equipment maintenance notices, temporary power outage notices, and major event notices where the power load reaches a set value. Abnormal data is uploaded to a large model, which then makes predictions and outputs the estimated time when the power grid will experience a power outage. During normal power grid operation, the energy storage system is charged and a special reserve identifier is set for it. This identifier indicates that the energy storage system will only be activated when a preset emergency condition is met.

[0040] In this embodiment, considering that unforeseen events such as sudden grid failures or excessive power consumption during major events may affect the normal power supply, the predictive reasoning capabilities of a large model are introduced to analyze and reason about abnormal data and notifications obtained in real time from the internet or relevant departments, outputting predicted power outage information to facilitate the development of emergency plans in advance. As an important part of the emergency plan, the energy storage system needs to be fully charged when power is available normally, and a dedicated reserve identifier should be set up for its use as an emergency power reserve.

[0041] If the energy storage system is detected to have a dedicated storage identifier and is in the process of discharging, the discharge operation will be stopped. If the time reaches the routine discharge time of the energy storage system, the discharge operation will not be performed until the preset emergency conditions are met, at which point the energy storage system will be activated to power the computing center. If the power grid is in a power outage state after the energy storage system is activated, the load operation strategy is calculated through a large model based on the power data of the energy storage system, the power demand of each load in all the mixed loads of the computing center, and the priority of each load. The mixed loads include rigid loads and adjustable flexible loads, and the priority of rigid loads is higher than that of adjustable flexible loads. The priority of different loads in the adjustable flexible loads is positively correlated with the real-time requirements of the corresponding loads and negatively correlated with the energy consumption of the corresponding loads. The operating status of each load is controlled based on the load operation strategy. The load operation strategy can ensure the normal operation of rigid loads. Among the flexible adjustable loads, each load is determined to operate normally or stop operating according to priority. The lower the priority of the flexible adjustable load, the more likely it is to stop operating. This is so that the power grid ends the power outage and restores power supply before the energy storage system's power data is exhausted.

[0042] In this embodiment, for energy storage systems that are marked with a special reserve identifier, they need to be activated only when preset emergency conditions are met. Therefore, routine discharge plans are not executed, and ongoing discharge operations are immediately stopped to ensure that the energy storage system can truly serve as a reserve power source and play its maximum role in an emergency.

[0043] After the energy storage system is activated under preset emergency conditions, since the power grid is still in a power outage state, to ensure that the energy stored in the system has sufficient power until the power grid resumes normal operation, the optimal load operation strategy is calculated using the reasoning and computational capabilities of a large-scale model. This strategy ensures the normal operation of rigid loads, while for flexible adjustable loads, each load is prioritized for normal or stopped operation, with lower-priority flexible adjustable loads being stopped first. This ensures that the power grid has ended the power outage and resumed power supply before the energy storage system's energy data is completely consumed. To ensure power supply security, a certain time margin can be set as needed. For example, the energy storage system's energy will not be completely consumed until 15 or 30 minutes after the power grid resumes power supply, thus preventing power outage intervals and preventing irreversible negative impacts on equipment performance from the complete depletion of the energy storage system's energy.

[0044] This invention provides a method for coordinated scheduling of computing power and power. The method is used in a computing-power coordinated system, which includes a computing center, a power grid, and a coordinated scheduling system for computing power and power. The coordinated scheduling system is connected to both the computing center and the power grid. A large-scale model is deployed within the coordinated scheduling system. An energy storage system is deployed in the computing center. The computing center receives power from the power grid or the energy storage system. The method includes: acquiring the real-time power supply surplus of the power grid and supply-demand fluctuation trend data within a preset future time period; acquiring the computing power resource surplus, mixed load characteristics, task constraints, and real-time power demand of the computing center; collecting real-time power data from the energy storage system of the computing center; and combining the real-time power supply surplus of the power grid and the supply-demand fluctuation trend data within a preset future time period with the computing power resource surplus of the computing center. Resource surplus, mixed load characteristics, task constraints, real-time electricity demand, and electricity data from the energy storage system are input as multi-dimensional data into a large model. The large model performs data fusion processing on the received multi-dimensional data to construct an integrated dataset of power supply, computing power demand, task constraints, and energy consumption loss characteristics. Based on this integrated dataset, deep correlation feature mining is performed. The deep correlation features are used to characterize the temporal fluctuation pattern of power output, the power output fluctuation trend within a preset future time period, the power supply surplus period and corresponding electricity surplus data, and the power supply shortage period and corresponding shortage electricity data. Based on the received multi-dimensional data and the mined deep correlation features, the large model intelligently generates a computing and power coordinated scheduling strategy. The computing power and power coordinated scheduling system performs coordinated scheduling of computing power and power based on the computing and power coordinated scheduling strategy. This method, based on real-time collected data and future predicted data, and after deep correlation feature mining by the large model, intelligently generates a computing and power coordinated scheduling strategy, which can effectively cooperate with the power grid to peak shaving and valley filling, smooth the power load curve, improve the supply and demand balance of power resources, and achieve optimal resource allocation.

[0045] On the other hand, embodiments of the present invention provide a computing power and power coordinated scheduling system, such as Figure 2 As shown, the computing power and power coordinated dispatch system is connected to the computing power center and the power grid, respectively. A large-scale model is deployed within the computing power and power coordinated dispatch system, and an energy storage system is deployed in the computing power center. The computing power center receives power from the power grid or the energy storage system. The computing power and power coordinated dispatch system includes: The first acquisition module 201 is used to acquire the real-time power supply margin of the power grid and the supply and demand fluctuation trend data within a future preset time period; The second acquisition module 202 is used to acquire the computing power resource surplus, mixed load characteristics, task constraints and real-time power demand of the computing power center. The acquisition module 203 is used to acquire power data from the energy storage system of the computing center in real time. Input module 204 is used to input the real-time power supply margin of the power grid and the supply and demand fluctuation trend data within a future preset time period, the computing power resource margin of the computing center, the characteristics of mixed loads, task constraints and real-time power demand, and the power data of the energy storage system as multi-dimensional data into the large model. The large model 205 is used to perform data fusion processing on the received multidimensional data, construct an integrated dataset of power supply, computing power demand, task constraints and energy consumption loss characteristics, and perform deep correlation feature mining based on the integrated dataset. The deep correlation features are used to characterize the power output time-series fluctuation pattern, the power output fluctuation trend in the future preset time period, the power supply surplus period and the corresponding power surplus data, and the power supply shortage period and the corresponding shortage power data. The large model 205 also intelligently generates a computing and power collaborative scheduling strategy based on the received multidimensional data and the mined deep correlation features. The scheduling module 206 performs coordinated scheduling of computing power and electricity based on the computing-electricity coordinated scheduling strategy.

[0046] Thirdly, embodiments of the present invention provide an electronic device, including: a memory and a processor, wherein the memory and the processor are connected; Memory, used to store computer programs; A processor is used to invoke a computer program stored in memory to perform any of the above methods.

[0047] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when run by a computer, performs any of the methods described above.

[0048] The computer-readable media in this application embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0049] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to execute the methods of any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0050] It should be noted that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for systems, devices, and media, since they are basically similar to the method embodiments, the descriptions are relatively simple, and relevant parts can be referred to the descriptions in the method embodiments. The methods, systems, and media described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components indicated as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the solution in this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0051] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for coordinated scheduling of computing power and electricity, characterized in that, The method is used in a computing-power collaborative system, which includes: a computing center, a power grid, and a computing and power collaborative scheduling system. The computing and power collaborative scheduling system is connected to the computing center and the power grid, respectively. A large-scale model is deployed within the computing and power collaborative scheduling system. An energy storage system is deployed in the computing center, which receives power from the power grid or the energy storage system. The method includes: To obtain real-time power supply capacity of the power grid and supply and demand fluctuation trend data within a preset future time period; Obtain the computing power resource reserves, mixed load characteristics, task constraints, and real-time power demand of the computing power center; Real-time collection of power data from the energy storage system in the computing center; The real-time power supply margin of the power grid and the supply and demand fluctuation trend data within a future preset time period, the computing power resource margin of the computing center, the characteristics of mixed loads, task constraints and real-time power demand, and the power data of the energy storage system are input into the large model as multi-dimensional data. The large model performs data fusion processing on the received multidimensional data to construct an integrated dataset of power supply, computing power demand, task constraints, and energy consumption loss characteristics. Based on the integrated dataset, deep correlation feature mining is performed. The deep correlation features are used to characterize the time-series fluctuation pattern of power output, the power output fluctuation trend within a future preset time period, the time period of power supply surplus and the corresponding power surplus data, and the time period of power supply shortage and the corresponding power shortage data. The large model intelligently generates a computing and power collaborative scheduling strategy based on the received multidimensional data and the mined deep correlation features. The computing power and power coordinated scheduling system performs coordinated scheduling of computing power and power based on the aforementioned computing power and power coordinated scheduling strategy.

2. The method according to claim 1, characterized in that, The method further includes: Collect real-time power output data released by the power grid. The power output data includes at least: power output data of various new energy sources and traditional thermal power output data, power generation data and power output forecast data. Based on the power output data, the real-time power supply margin of the power grid and the supply and demand fluctuation trend data within a future preset time period are calculated. The system collects full-domain operational data from the computing center. This full-domain operational data includes: computing resource status data, tenant task operation data, and computing load and energy consumption data. The computing resource status data includes: the number of online heterogeneous computing resources, idle computing slices, remaining video memory, available cores, and reclaimable redundant resources. The tenant task operation data includes: task type, priority, tenant credit rating, latency threshold, operation progress, and off-peak attribute. The task type includes real-time task type and offline task type. The computing load and energy consumption data includes: computing utilization rate, video memory occupancy rate, device instantaneous power consumption, and total cluster load. Based on the online quantity, idle computing power slices, remaining video memory, available cores, and recyclable redundant resources of various heterogeneous computing power resources in the computing power resource status data, the remaining amount of heterogeneous computing power resources is calculated. Based on the task types in the tenant task execution data, mixed load characteristics are extracted; Based on the priority, tenant credit rating, latency threshold, running progress, and off-peak attribute in the tenant task running data, task constraints are determined. Among them, based on the task constraints, tasks are divided into tasks that can be deloaded, tasks that must be guaranteed, and tasks that can be moved. Based on the computing power utilization rate, video memory occupancy rate, instantaneous power consumption of devices, and total cluster load in the computing power load and energy consumption data, the real-time power demand is calculated.

3. The method according to claim 1, characterized in that, The method further includes: Obtain electricity price data for the same period in previous years; Based on the power output data and the electricity price data for the same period in previous years, a routine charging and discharging plan for the energy storage system is determined. The routine charging and discharging plan includes at least the following: charging the energy storage system through the power grid during a first time period when the power output data is greater than or equal to a first set value and the electricity price is lower than a second set value; and discharging the energy storage system to supply power to the computing center during a second time period when the power output data is less than a third set value and the electricity price is higher than a fourth set value. The first set value is greater than the third set value, and the second set value is less than the fourth set value.

4. The method according to claim 3, characterized in that, The method further includes: When the time reaches a preset time point before the first and second time periods of the routine charging and discharging plan, the real-time electricity price data corresponding to the preset time point is obtained; The real-time electricity price data is compared with the average electricity price in the first time period of the same period in previous years and the average electricity price in the second time period of the same period in previous years, respectively. If the electricity price difference is less than the preset electricity price difference, the routine charging and discharging plan will be executed after the time reaches the time point corresponding to the routine charging and discharging plan. If the electricity price difference is greater than or equal to the preset electricity price difference, the routine charging and discharging plan is adjusted based on the real-time electricity price data, and the energy storage system is charged and discharged based on the adjusted routine charging and discharging plan.

5. The method according to claim 1, characterized in that, The energy storage system includes a voltage stabilizing unit, and the method further includes: Real-time monitoring of voltage fluctuations in power grid output data; If the voltage fluctuation reaches the preset fluctuation threshold, and the computing center is currently powered by the power grid, then the remaining power of the energy storage system is detected. If the remaining power of the energy storage system is greater than or equal to the first power threshold, the connection between the power grid and the computing center is cut off, and the energy storage system is switched to supply power to the computing center. If the remaining power of the energy storage system is less than the first power threshold but greater than the second power threshold, the grid will be connected to the energy storage system so that the power output from the grid can be stabilized by the voltage stabilization unit in the energy storage system to supply power to the computing center. The second power threshold is a low power warning value or a discharge protection threshold, and the second power threshold is less than the first power threshold.

6. The method according to claim 5, characterized in that, The computing center is also equipped with voltage stabilizing equipment, and the method further includes: If the remaining power of the energy storage system is less than the second power threshold, the power supply connection between the energy storage system and the computing center is disconnected, the voltage stabilizing device is activated, and the voltage stabilizing device is connected between the power grid and the computing center so that the power output from the power grid is stabilized by the voltage stabilizing device before being supplied to the computing center; and The energy storage system is connected to the power grid so that the power grid can charge the energy storage system. During the charging process, the voltage of the power output from the power grid is regulated by a voltage stabilizing unit in the energy storage system.

7. The method according to claim 1, characterized in that, The method further includes: Real-time acquisition of abnormal data, including: abnormal power load data, historical power outage records, power grid fault data or power equipment maintenance notices, temporary power outage notices, and major event notices where the power load reaches a set value. Abnormal data is uploaded to a large model, which then makes predictions and outputs the estimated time when the power grid will experience a power outage. During normal grid power supply periods, the energy storage system is charged, and a dedicated reserve identifier is set for it. This identifier indicates that the energy storage system will only be activated when preset emergency conditions are met; and / or, If the energy storage system is detected to have a dedicated storage identifier and is in the process of discharging, the discharge operation will be stopped. If the time reaches the routine discharge time of the energy storage system, the discharge operation will not be performed until the preset emergency conditions are met, at which point the energy storage system will be activated to power the computing center. If the power grid is in a power outage state after the energy storage system is activated, the load operation strategy is calculated through a large model based on the power data of the energy storage system, the power demand of each load in all the mixed loads of the computing center, and the priority of each load. The mixed loads include rigid loads and adjustable flexible loads, and the priority of rigid loads is higher than that of adjustable flexible loads. The priority of different loads in the adjustable flexible loads is positively correlated with the real-time requirements of the corresponding loads and negatively correlated with the energy consumption of the corresponding loads. The operating status of each load is controlled based on the load operation strategy. The load operation strategy can ensure the normal operation of rigid loads. Among the flexible adjustable loads, each load is determined to operate normally or stop operating according to priority. The lower the priority of the flexible adjustable load, the more likely it is to stop operating. This is so that the power grid ends the power outage and restores power supply before the energy storage system's power data is exhausted.

8. A computing power and power coordinated dispatch system, characterized in that, The computing power and power coordinated dispatch system is connected to the computing power center and the power grid, respectively. A large-scale model is deployed within the computing power and power coordinated dispatch system. The computing power center is equipped with an energy storage system and is powered by the power grid or the energy storage system. The computing power and power coordinated dispatch system includes: The first acquisition module is used to acquire the real-time power supply margin of the power grid and the supply and demand fluctuation trend data within a future preset time period; The second acquisition module is used to acquire the computing power resource reserves, mixed load characteristics, task constraints and real-time power demand of the computing power center. The data acquisition module is used to collect power data from the energy storage system of the computing center in real time. The input module is used to input the real-time power supply margin of the power grid and the supply and demand fluctuation trend data within a preset time period, the computing power resource margin of the computing center, the characteristics of mixed loads, task constraints and real-time power demand, and the power data of the energy storage system as multi-dimensional data into the large model. The large model is used to perform data fusion processing on the received multidimensional data, construct an integrated dataset of power supply, computing power demand, task constraints and energy consumption loss characteristics, and perform deep correlation feature mining based on the integrated dataset. The deep correlation features are used to characterize the time-series fluctuation pattern of power output, the fluctuation trend of power output in the future preset time period, the time period of power supply surplus and the corresponding power surplus data, and the time period of power supply shortage and the corresponding power shortage data. The large model also intelligently generates a computing and power collaborative scheduling strategy based on the received multidimensional data and the mined deep correlation features. The scheduling module performs coordinated scheduling of computing power and electricity based on a computing-electricity coordinated scheduling strategy.

9. An electronic device, characterized in that, include: Memory and processor, and the connection between memory and processor; Memory, used to store computer programs; A processor for invoking a computer program stored in memory to perform the method as claimed in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It stores a computer program, which, when executed by a computer, performs the method as described in any one of claims 1-7.

Images