A new energy and computing power collaborative optimization scheduling method, device and storage medium

Abstract

The present application belongs to the technical field of power dispatching, and particularly relates to a new energy and computing power collaborative optimization dispatching method, device and storage medium. The method divides the dispatching into medium and long term and short term based on the computing power task duration of the data center and the power dispatching cycle, and the step of the medium and long term dispatching cycle is the short term dispatching cycle; the total computing task quantity is allocated according to the maximum computing task quantity supported by the new energy power generation power prediction value in each medium and long term dispatching step, and the computing task quantity allocated in each medium and long term dispatching step is obtained; a short term computing power collaborative optimization dispatching model is established based on the computing task quantity allocated in each medium and long term dispatching step, the new energy power generation power prediction value corresponding to the short term dispatching step and the total load; after solving, the decision variable of each short term dispatching step is obtained, and the complexity of modeling and solving in the medium and long term new energy and computing power collaborative dispatching is reduced.

Description

Technical Field

[0001] This invention belongs to the field of power dispatching technology, specifically relating to a collaborative optimization dispatching method, device, and storage medium for new energy sources and computing power. Background Technology

[0002] With the rapid development of big data technology, computing centers (data centers) are growing in scale, leading to a surge in electricity consumption in the computing sector. Data center loads differ from flexible loads such as air conditioning and electric heating; they are not only time-adjustable but can also transfer workloads through communication networks, enabling spatial transfer of power loads. In other words, data center loads are spatiotemporally adjustable and can participate in grid demand response. Coordinated scheduling of new energy sources and computing power is considered a key path to improve green electricity consumption, reduce computing power costs, and ensure energy security. Industry forecasts indicate that future computing power energy consumption will be enormous, making the cost reduction and carbon reduction benefits of coordinated scheduling of new energy sources and computing power substantial.

[0003] Currently, the system scheduling practices for new energy and computing power mainly rely on cross-regional scheduling based on real-time or short-term signals (such as electricity prices and instantaneous power generation from new energy sources). When electricity prices are low or the instantaneous power generation from new energy sources is high, the load on the data center can be increased; when electricity prices are high or the instantaneous power generation from new energy sources is low, the computing load on the data center can be reduced. However, there are no effective and feasible practical methods at the local level. There are three main reasons for this: First, the IT equipment in data centers (such as servers) has high requirements for power supply reliability, and frequent adjustments to the computing load affect the performance of IT equipment; second, data centers have service quality constraints, and adjusting the load carries the risk of tasks not being completed on schedule; third, the existing operating model of most data centers is based on leasing services, and adjusting the computing speed affects the user experience.

[0004] Theoretical research focuses on current-day collaborative scheduling methods, which optimize modeling by coupling the mapping relationship between computing power and load and adding constraints on computing task completion. For example, Chinese patent application CN120896156A discloses a new energy computing power collaborative optimization scheduling method and system based on global economic optimality. This method constructs a mapping relationship between computing power tasks and power consumption, and based on this mapping relationship, builds a demand response model for computing units; based on the demand response model, it constructs a new energy model; considering the time-of-use differences in wind power, photovoltaic power, grid-connected electricity prices, and energy-consuming electricity prices, it solves the new energy model with global economic optimality as the optimization objective to determine the optimal operating mode for energy storage and adjustable computing power load, and performs new energy scheduling based on this optimal operating mode, ensuring optimal economy after all computing power tasks are completed throughout the day. However, in the overall operation of a data center, computing tasks are distributed across various computing devices, and the power consumption of multiple devices collectively constitutes the external characteristics of the data center's total power consumption. Adjusting a single computing task is typically achieved through small-batch batch processing on a few servers. However, when single-task adjustment cannot meet power regulation requirements, it involves coupling multiple tasks with multiple servers, forming a more complex adjustment relationship. These methods are only suitable for intraday or shorter-cycle collaborative scheduling. With the development of data processing technology, the duration of a single computing task in a data center can reach one week or even one month. However, facing these medium- to long-term scheduling cycles, the current scheduling methods have many co-variable factors and excessively high model complexity, resulting in very high modeling and solution complexity. This leads to high regulation complexity and stringent accuracy requirements, making practical application extremely difficult and hindering the full realization of the synergistic potential of new energy sources and computing power. Summary of the Invention

[0005] The purpose of this invention is to provide a method, device, and storage medium for the collaborative optimization scheduling of new energy sources and computing power, in order to solve the problems of excessive model complexity and difficulty in solving existing collaborative scheduling of new energy sources and computing power.

[0006] To address the aforementioned technical problems, the first aspect of this invention provides a method for the coordinated optimization scheduling of new energy sources and computing power, comprising:

[0007] The scheduling is divided into medium- and long-term and short-term. The medium- and long-term scheduling cycle is determined based on the duration of the computing power tasks of the data center, while the short-term scheduling cycle is determined based on the power scheduling cycle. The step size of the medium- and long-term scheduling cycle is the same as that of the short-term scheduling cycle.

[0008] Determine the maximum amount of computational tasks that the predicted power generation of new energy sources can support within each medium- and long-term scheduling step, and allocate the total amount of computational tasks according to the maximum amount of computational tasks to obtain the amount of computational tasks allocated to each medium- and long-term scheduling step.

[0009] A short-term computational-coordinated optimization scheduling model is established based on the computational workload allocated for each medium- and long-term scheduling step, the predicted value of new energy power generation corresponding to the short-term scheduling step, and the total load; the total load includes the load power of the data center and the charging power of energy storage.

[0010] The short-term computing-power collaborative optimization scheduling model is solved to obtain the decision variables for each short-term scheduling step. The decision variables include the number of computing devices and the charging and discharging power of energy storage in each operating state.

[0011] In one possible implementation, the total computational workload is allocated based on the principle of proximity in time sequence. The allocation method is as follows: if the maximum computational workload that the predicted renewable energy power generation value can support within the entire scheduling period is greater than or equal to the total computational workload, then the computational workload is arranged forward; if the maximum computational workload that the predicted renewable energy power generation value can support within the entire scheduling period is less than the total computational workload, then the computational workload is arranged backward under the constraint of computational workload completion.

[0012] In one possible implementation, if the maximum computational workload supported by the predicted renewable energy power generation over the entire scheduling cycle is greater than or equal to the total computational workload, then the computational workload allocated for each medium- and long-term scheduling step is:

[0013]

[0014]

[0015] Among them, v l,i The computational workload allocated for the i-th medium- to long-term scheduling step, V total This represents the total computational workload. Let j be the maximum computational workload that can be supported by the predicted power generation of new energy sources within the i-th medium- and long-term scheduling step, and j be the critical point.

[0016] In one possible implementation, if the maximum computational workload supported by the predicted renewable energy power generation over the entire scheduling cycle is less than the total computational workload, then the computational workload allocated for each medium- and long-term scheduling step is:

[0017]

[0018]

[0019] Among them, v l,i The computational workload allocated for the i-th medium- to long-term scheduling step, V total This represents the total computational workload. Let j be the maximum computational workload that the predicted power generation of new energy sources can support within the i-th medium-to-long-term scheduling step, and j be the critical point. This represents the maximum amount of computing tasks that the data center can complete within the medium- to long-term scheduling step.

[0020] In one possible implementation, the short-term computational-coordinated optimization scheduling model includes an objective function and constraints; the objective function is determined based on the predicted power generation of new energy sources and the total load corresponding to the short-term scheduling step size.

[0021] In one possible implementation, the constraints include data center constraints, grid interaction constraints, and energy storage constraints; data center constraints include task completion constraints, equipment balancing constraints, and equipment quantity constraints; grid interaction constraints require that the power drawn from the grid be less than or equal to the maximum allowable downlink power of the grid; and energy storage constraints include energy storage power constraints and energy storage SOC constraints.

[0022] In one possible implementation, the continuous operating state of data center computing devices is discretized into three operating states: standby, default utilization, and maximum utilization. The power and computing power corresponding to each operating state of a single computing device are determined. Based on the power and computing power corresponding to each operating state of a single computing device, the load power of the data center is determined. The objective function is then solved to obtain the number of computing devices operating at default utilization, the number of computing devices operating at maximum utilization, and the charging and discharging power of energy storage for each short-term scheduling step.

[0023] In one possible implementation, the method further includes: calculating the deviation between the amount of computational tasks allocated before the current short-term scheduling period and the amount of computational tasks actually executed, and updating the amount of computational tasks in the current short-term scheduling period based on the deviation of the amount of computational tasks, so as to correct the execution deviation of the amount of computational tasks.

[0024] To address the aforementioned technical problems, a second aspect of the present invention provides a collaborative optimization scheduling device for new energy and computing power, comprising a processor that executes a computer program to implement the steps of the above-described method.

[0025] To address the aforementioned technical problems, a third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0026] The beneficial effects of this invention are as follows: By introducing a multi-timescale collaborative optimization mechanism, this invention divides the collaboration between new energy and computing power into two levels: medium-to-long-term and short-term. It clarifies the dominant collaborative elements for different periods: in the medium-to-long term, computing power task allocation is the primary factor, while in the short term, energy storage regulation is the core. Based on the medium-to-long-term task allocation results, a short-term computing-power collaborative optimization scheduling model is established to solve for the number of computing devices and energy storage charging and discharging plans for each operating state. Through the coupled collaboration of medium-to-long-term and short-term, the dominant variables at each time scale are modeled with reduced dimensionality. This not only enhances the new energy absorption potential of data centers but also reduces the complexity of modeling and solving, overcoming the problem of excessive model complexity and difficulty in solving the collaborative scheduling of new energy and computing power in the medium-to-long term, and significantly reducing computational complexity.

[0027] The time-series near-end priority allocation strategy proposed in this invention improves the local consumption rate and utilization efficiency of new energy by prioritizing the consumption of near-term green electricity.

[0028] This invention also addresses the deviation in task completion caused by volatility and uncertainty during task execution, effectively ensuring the reliable completion of the total computing power tasks. Attached Figure Description

[0029] Figure 1 This is a flowchart of the collaborative optimization scheduling process of new energy and computing power in this invention;

[0030] Figure 2 This is a block diagram of the collaborative optimization scheduling device for new energy and computing power of the present invention. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of the present invention clearer, the specific embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0032] This invention introduces a multi-timescale collaborative optimization mechanism, dividing the scheduling cycle into medium-to-long-term and short-term layers. At the medium-to-long-term scale, renewable energy output is predicted with short-term cycles as the step size. The maximum computational workload that the predicted power can support is calculated through the state combination of computing devices. Based on the relationship between the maximum computational workload and the total computational workload, tasks are allocated according to the principle of proximity in time sequence. At the short-term scale, based on the computational workload allocation results of the medium-to-long-term period, a short-term computing-power collaborative optimization scheduling model is established, considering the regulatory role of energy storage. This model solves for the number of computing devices operating at default and maximum utilization rates. Through the coupling and collaboration of the medium-to-long-term and short-term scales, the renewable energy absorption potential of data centers is enhanced. Simultaneously, the complexity of modeling and solving is reduced to account for changes in collaborative elements across different cycles.

[0033] Implementation of the Cooperative Optimization Scheduling Method of New Energy and Computing Power

[0034] A collaborative optimization scheduling method for new energy sources and computing power is proposed for reliable new energy power supply scenarios in data centers. For example... Figure 1 As shown, the method includes the following steps:

[0035] 1. Discretize the continuous operating state of data center computing equipment into three typical states: standby, default utilization, and maximum utilization, and generalize the external characteristics of power.

[0036] For a computing center with N central devices (computing devices), determine the total amount of computing tasks V to be allocated within the total scheduling cycle. total Based on historical operating data and power tests, the functional relationship between computing power and power of a single computing device was determined:

[0037] p it =f(c)

[0038] Where, p it denoted as , where is the power of a single computing device; c is the computing power of a single computing device, i.e., the computing task capacity of a single computing device running in the data center. The computing task capacity of a single computing device is the product of the computing power of the single computing device and the time.

[0039] The power of a single computing device in this invention includes the power of the computing device performing the computing task and the cooling energy consumption and other energy consumption distributed across the device, i.e., p it This refers to the total energy consumption, including refrigeration, etc.

[0040] The functional relationship between computing power and power of a single operating computing device in a computing center is determined based on specific circumstances. For example, for a given computing center, the functional relationship between computing power and power of a single computing device is as follows:

[0041]

[0042] Where k1, k2, and k3 are fitting parameters for the power of the computing device and the computing power during operation. In the example, a quadratic function is used to fit the power of the device and the computing power during operation. The fitting function can be adjusted by combining the actual operating data of different devices.

[0043] Determine the power consumption of a single computing device under three operating states: standby, default utilization, and maximum utilization.

[0044] Standby mode refers to a device that is powered on but not performing computing tasks. In standby mode, the power consumption of a single computing device is [standby power consumption]. The computing power of a single computing device is 0.

[0045] Under default utilization, the power consumption of a single computing device is the default power. The computing power of a single computing device is c. def.

[0046] Under maximum utilization operation, the power consumption of a single computing device is its maximum power. The computing power of a single computing device is c. max .

[0047] 2. Determine the scheduling cycle and step size for hierarchical scheduling.

[0048] Under normal operating conditions, the power supply for a data center includes: renewable energy generation, mains power, and energy storage devices.

[0049] Scheduling is divided into two layers: medium-to-long-term and short-term. The medium-to-long-term scheduling cycle is determined based on the duration of the computing task, and the medium-to-long-term scheduling cycle is longer than 24 hours. It can be selected from quarterly, monthly, or weekly cycles based on the duration of the computing task.

[0050] Data centers have numerous tasks, and the medium- to long-term scheduling cycle can be determined based on the duration of a certain number of individual computing tasks. For example, if more than 30% of the computing tasks in a data center last for 4 days to 1 week, then the medium- to long-term scheduling cycle can be set to 1 week. If more than 30% of the computing tasks in a data center last for 2 weeks to 1 month, then the medium- to long-term scheduling cycle can be set to 1 month.

[0051] The short-term dispatch cycle is consistent with the power dispatch cycle, generally one day. If the short-term dispatch step is one hour, the short-term dispatch cycle is divided into 24 points; if the short-term dispatch step is 15 minutes, the short-term dispatch cycle is divided into 96 points. The short-term dispatch cycle can also be customized according to requirements.

[0052] The medium- and long-term scheduling period is taken as the total scheduling period, denoted as T. l The medium- to long-term scheduling step size is Δt. l The number of medium- and long-term scheduling nodes is n l :

[0053]

[0054] The short-term scheduling period is T s The medium- and long-term scheduling step size is the same as the short-term scheduling period, i.e., T. s =Δt l The short-term scheduling step size is Δt. s The number of short-run scheduling nodes is n s :

[0055]

[0056] 3. Predict the new energy power generation corresponding to each medium- and long-term scheduling step size to obtain the medium- and long-term new energy power generation prediction sequence.

[0057] On a medium- to long-term timescale, new energy power output is predicted using a short-term cycle as the step size, resulting in a medium- to long-term new energy power generation prediction sequence:

[0058]

[0059] This represents the predicted renewable energy power generation at the i-th medium-to-long-term scheduling node, i.e., the predicted renewable energy power generation at the i-th medium-to-long-term scheduling step. The predicted renewable energy power generation is obtained based on historical data.

[0060] 4. Determine the maximum computational workload that the predicted power generation of new energy sources can support within each medium- and long-term scheduling step.

[0061] The maximum computational workload that can be carried by calculating the predicted power generation of new energy sources within each medium- and long-term scheduling step is determined by combining the state of the equipment.

[0062] 1) When This indicates that the output of new energy sources can support all equipment operating at maximum power. At this point, the maximum computational workload supported by the predicted power generation of new energy sources within the medium- to long-term scheduling step is:

[0063]

[0064] in, This represents the maximum computational workload that can be supported by the predicted power generation of new energy sources within the i-th medium-to-long-term scheduling step.

[0065] 2) When This indicates that the output of new energy sources can support all equipment to operate at a combination of maximum utilization and default utilization. At this time, the maximum power output N of equipment that new energy sources can support within the medium-to-long-term scheduling step is... max Calculate using the following formula:

[0066]

[0067] N max It is a virtual quantity and can be a decimal.

[0068] The number of operating equipment N that can be supported by renewable energy output within the medium- and long-term dispatching step size (default power). def Calculate using the following formula:

[0069]

[0070] The number of operating equipment N that can be supported by renewable energy output within the medium- and long-term dispatching step size (default power). def It can also be calculated using the following formula:

[0071] N def =N-N max

[0072] At this point, the maximum computational workload that the predicted power generation of new energy sources within the medium- to long-term scheduling step size can support is:

[0073]

[0074] 3) When This indicates that the output of new energy sources can support all equipment to operate at a combination of standby and default power. At this time, the number N of equipment operating at default power that can be supported by the output of new energy sources within the medium- to long-term scheduling step is... def for:

[0075]

[0076] At this point, the maximum computational workload that the predicted power generation of new energy sources within the medium- to long-term scheduling step size can support is:

[0077]

[0078] 5. Allocate the total computational workload based on the maximum computational workload that can be supported by the predicted power generation of new energy sources within each medium- and long-term scheduling step, and obtain the computational workload allocated to each medium- and long-term scheduling step.

[0079] Since most data centers use electrochemical energy storage, the regulatory role of electrochemical energy storage in the medium and long term is not considered. Instead, the computational workload is allocated based on the relationship between the maximum computational workload that can be supported by the predicted power generation of new energy sources within each medium and long term scheduling step and the total computational workload. This allocation is based on the principle of proximity in time.

[0080] 1) When When this occurs, it indicates that the maximum amount of computational tasks that new energy sources can support during the entire scheduling cycle is greater than or equal to the total amount of computational tasks, and the difference in computational tasks is:

[0081]

[0082] At this time, some of the electricity generated by new energy sources faces the risk of being curtailed. Therefore, the calculation of the workload should be arranged as early as possible to ensure the more certain absorption of new energy in the early stages and leave sufficient room for adjustment in the later stages.

[0083] Specifically, the critical point j is determined to satisfy:

[0084]

[0085] The computational workload for allocating each medium- and long-term scheduling step size is as follows:

[0086]

[0087] Among them, v l,iThe amount of computational tasks allocated for the i-th medium-to-long-term scheduling step.

[0088] at this time, , Satisfy the maximum computational task constraint.

[0089] 2) When This indicates that the maximum computational workload supported by new energy sources during the entire scheduling cycle is less than the total computational workload, and the excess tasks... Mains power supply is required. Under the constraint of completing the computing tasks, excess computing tasks should be scheduled to be postponed as much as possible.

[0090] Specifically, the critical point j is determined to satisfy:

[0091]

[0092] That is, the first j medium- and long-term scheduling nodes are arranged according to... During runtime, the following n l -j nodes The execution will just complete the total task. Therefore:

[0093]

[0094] at this time, , Satisfy the maximum computational task constraint, This represents the maximum amount of computing tasks that the data center can complete within the medium- to long-term scheduling step. By scheduling the computing tasks after the critical point, when the maximum computing task supported by new energy sources during the current forecast scheduling cycle is less than the total computing task, while ensuring the total computing task, more new energy power generation can be appropriately used to support the computing task when the predicted power of new energy generation increases in the later stages, thereby consuming more new energy power.

[0095] 6. In the short term, based on the computational workload allocated for each medium- and long-term scheduling step, and combined with the regulating role of energy storage, a short-term computing-power collaborative optimization scheduling model is established to solve the decision variables for each short-term scheduling step. The decision variables include the number of computing devices and the charging and discharging power of energy storage in each operating state.

[0096] For example, to determine the number of computing devices running at default utilization. Number of computing devices operating at maximum utilization and the charging and discharging power of energy storage .

[0097] Specifically, based on the previous steps, the total number of computational tasks to be completed in a certain short-term scheduling cycle is known to be... For example, the total number of computational tasks in the first short-term scheduling cycle. .

[0098] The short-term computing-power collaborative optimization scheduling model includes an objective function and constraints.

[0099] 1) On a short-term scale, predict the renewable energy power generation corresponding to the short-term scheduling step size to obtain the short-term renewable energy power generation prediction sequence:

[0100]

[0101] This represents the predicted value of renewable energy power generation at the i-th short-term scheduling node, i.e., the predicted value of renewable energy power generation at the i-th short-term scheduling step.

[0102] 2) Determine the objective function of the short-term computing and power-coordinated optimization scheduling model.

[0103] The objective function is determined by minimizing the sum of squares of the differences between the predicted power generation of new energy sources corresponding to the short-term scheduling step size and the total load (data center load + energy storage charging).

[0104]

[0105] The charging power for the energy storage of the data center of the i-th short-term scheduling node (charging is positive and discharging is negative). Let be the load power of the data center of the i-th short-term scheduling node.

[0106] The load power of a data center can be calculated using the following formula:

[0107]

[0108] The load power of a data center can also be calculated using the following formula:

[0109]

[0110] + + =N

[0111] Let be the number of computing devices running at the default utilization rate for the i-th short-lived scheduling node. Let be the number of computing devices that the i-th short-term scheduling node is running at maximum utilization. Let N be the number of computing devices in standby mode of the i-th short-term scheduling node. The total number of computing devices in the three operating states is N.

[0112] 3) Determine the constraints.

[0113] The constraints include data center constraints, grid interaction constraints, and energy storage constraints.

[0114] ① Data center constraints.

[0115] Data center constraints include task completion constraints, device balancing constraints, and device quantity constraints.

[0116] The task completion constraint means that the sum of the computational tasks allocated for each short-term scheduling step is greater than or equal to the total computational tasks.

[0117] Equipment balance constraints include computational equipment state space constraints and time balance constraints.

[0118] The state space constraint of computing devices is that at any given moment, there cannot be both a device with zero computing power and a device with maximum computing power in the space; that is, there cannot be a computing device in standby mode and a computing device operating at maximum utilization at the same time.

[0119] The time balance constraint is that the difference in computing power between computing devices at adjacent time points is less than or equal to the difference in maximum computing power.

[0120] The device quantity constraint refers to the fact that the number of computing devices operating in each state is a positive integer, and the sum of the quantities is less than or equal to the total number of computing devices in the data center.

[0121] ② Power grid interaction constraints.

[0122] The power drawn from the mains is less than or equal to the maximum downstream power allowed by the power grid.

[0123] ③ Energy storage constraints.

[0124] Energy storage constraints include energy storage power constraints and energy storage SOC constraints.

[0125] Energy storage power constraints: maximum and minimum operating power constraints and maximum ramp power constraints.

[0126] The maximum and minimum operating power constraints are: the operating power of the energy storage device shall not exceed the maximum allowable operating power of the energy storage device and shall not be less than the minimum allowable operating power of the energy storage device. That is:

[0127]

[0128] The maximum ramp power constraint is: the absolute value of the difference between the energy storage operating power at two consecutive moments must not exceed the maximum allowable operating power difference (maximum power change) of the energy storage device. That is:

[0129]

[0130] In the formula, This refers to the minimum permissible operating power of energy storage devices. This refers to the maximum permissible operating power of the energy storage device. This represents the maximum permissible power change of an energy storage device per unit time.

[0131] Energy storage SOC constraints: energy storage SOC continuity constraints, energy storage SOC state constraints, and energy storage SOC initial and final state constraints.

[0132] The continuity constraint of energy storage SOC is: the energy storage SOC state at time t is the sum of the SOC value at time t-1 and the difference between the amount of electricity charged (or the amount of electricity released) during the period from t to t-1, that is:

[0133] when At that time, energy storage charging, ;

[0134] when At that time, the stored energy is discharged. .

[0135] The SOC (State of Charge) constraint for energy storage is: at any given time, the SOC of the energy storage device shall not exceed the maximum allowable SOC of the energy storage device, nor be lower than the minimum allowable SOC of the energy storage device, that is:

[0136]

[0137] The initial and final state constraints of energy storage SOC are: the SOC value at the beginning of the first time step within a short-term scheduling cycle is equal to the SOC value at the end of the last time step, i.e.:

[0138]

[0139]

[0140] in, For energy storage charging and discharging efficiency, , The minimum and maximum SOC values ​​are those allowed by the equipment or set by operators based on experience to protect the equipment. and These are the fixed energy storage SOC values ​​at the beginning and end of the short-term scheduling cycle, respectively.

[0141] 4) Solve for decision variables using intelligent optimization algorithms such as particle swarm optimization.

[0142] As an alternative implementation, other optimization algorithms can be used to solve for the decision variables of each short-term scheduling step. The decision variables include the number of devices operating at default utilization, the number of devices operating at maximum utilization, and the charging and discharging power of the energy storage.

[0143] 7. A single-step look-ahead amortization correction strategy is adopted to perform rolling correction of the execution deviation of the calculation task.

[0144] Hierarchical scheduling includes one medium- to long-term scheduling cycle and nl A short scheduling cycle. The deviation between the allocated (and planned) computational tasks and the actual executed computational tasks within a short scheduling cycle will affect the completion of the entire cycle. Therefore, it is necessary to correct the execution deviation of the computational tasks. A single-step look-ahead amortization correction strategy can be used to correct the execution deviation of the computational tasks. This includes:

[0145] 1) At the m-th medium-to-long-term scheduling node, i.e., after m-1 short-term schedulings have been completed, determine the execution deviation of the computational task volume.

[0146]

[0147] 2) Calculate the average computational workload of each of the remaining medium- and long-term scheduling nodes.

[0148]

[0149] 3) For the short-term scheduling of the m-th medium-to-long-term scheduling node, update the target of the computational task volume for that short-term scheduling period based on the amortized amount.

[0150]

[0151] in, This refers to the cumulative execution deviation of the computational workload when the task reaches the m-th medium-to-long-term scheduling node. Let represent the actual computational workload executed by the i-th medium- to long-term scheduling node. Let be the average computational workload of the m-th medium-to-long-term scheduling node. This represents the computational workload after the m-th medium-to-long-term scheduling node is updated.

[0152] As another implementation method, the single-step look-ahead amortization correction strategy may not be adopted. That is, instead of distributing the execution deviation of the computation task equally to all remaining medium- and long-term scheduling nodes, some of the remaining medium- and long-term scheduling nodes are selected to distribute the execution deviation of the computation task.

[0153] In one embodiment, the collaborative optimization scheduling method for new energy sources and computing power proposed in this invention includes:

[0154] The scheduling is divided into medium- and long-term and short-term. The medium- and long-term scheduling cycle is determined based on the duration of the data center's computing power tasks, while the short-term scheduling cycle is determined based on the power dispatch cycle. The step size of the medium- and long-term scheduling cycle is the same as the short-term scheduling cycle. The maximum amount of computing tasks that the predicted power generation of new energy sources can support within each medium- and long-term scheduling step size is determined. The total amount of computing tasks is allocated according to the maximum amount of computing tasks, resulting in the amount of computing tasks allocated to each medium- and long-term scheduling step size. A short-term computing-power collaborative optimization scheduling model is established based on the amount of computing tasks allocated to each medium- and long-term scheduling step size, the predicted power generation of new energy sources corresponding to the short-term scheduling step size, and the total load. The total load includes the load power of the data center and the charging power of energy storage. The short-term computing-power collaborative optimization scheduling model is solved to obtain the decision variables for each short-term scheduling step size. The decision variables include the number of computing devices in each operating state and the charging and discharging power of energy storage.

[0155] The present invention also proposes a strategy for allocating the total amount of computing tasks based on the principle of proximity in time sequence: if the maximum amount of computing tasks that the predicted value of new energy power generation can support within the entire scheduling period is greater than or equal to the total amount of computing tasks, then the computing tasks are arranged forward; if the maximum amount of computing tasks that the predicted value of new energy power generation can support within the entire scheduling period is less than the total amount of computing tasks, then the computing tasks are arranged backward under the constraint of completing the computing tasks.

[0156] The method further includes: calculating the deviation between the amount of computational tasks allocated before the current short-term scheduling period and the amount of computational tasks actually executed, and updating the amount of computational tasks in the current short-term scheduling period based on the deviation of the amount of computational tasks, so as to correct the execution deviation of the amount of computational tasks.

[0157] Implementation Method of Collaborative Optimization Scheduling Device for New Energy and Computing Power

[0158] A collaborative optimization scheduling device for new energy and computing power, such as Figure 2 As shown, the system includes a memory, a processor, a system bus, and a computer program stored in the memory. The processor and memory communicate and interact with each other via the system bus. The processor executes the computer program to implement the steps of the new energy and computing power collaborative optimization scheduling method of the present invention. The processor can be a microprocessor (MCU) or other processing device; the memory can be various types of memory that store information using electrical energy, such as non-volatile storage media (including computer programs, databases), or other types of memory.

[0159] The collaborative optimization scheduling methods for new energy sources and computing power include:

[0160] The scheduling is divided into medium- and long-term and short-term. The medium- and long-term scheduling cycle is determined based on the duration of the data center's computing power tasks, while the short-term scheduling cycle is determined based on the power dispatch cycle. The step size of the medium- and long-term scheduling cycle is the same as the short-term scheduling cycle. The maximum amount of computing tasks that the predicted power generation of new energy sources can support within each medium- and long-term scheduling step size is determined. The total amount of computing tasks is allocated according to the maximum amount of computing tasks, resulting in the amount of computing tasks allocated to each medium- and long-term scheduling step size. A short-term computing-power collaborative optimization scheduling model is established based on the amount of computing tasks allocated to each medium- and long-term scheduling step size, the predicted power generation of new energy sources corresponding to the short-term scheduling step size, and the total load. The total load includes the load power of the data center and the charging power of energy storage. The short-term computing-power collaborative optimization scheduling model is solved to obtain the decision variables for each short-term scheduling step size. The decision variables include the number of computing devices in each operating state and the charging and discharging power of energy storage.

[0161] If the maximum computational workload supported by the predicted renewable energy power generation within the entire scheduling period is greater than or equal to the total computational workload, then the computational workload will be scheduled forward; if the maximum computational workload supported by the predicted renewable energy power generation within the entire scheduling period is less than the total computational workload, then the computational workload will be scheduled backward under the constraint of completing the computational workload.

[0162] Specifically, the method includes the following steps:

[0163] 1. Discretize the continuous operating state of data center computing equipment into three typical states: standby, default utilization, and maximum utilization, and generalize the external characteristics of power.

[0164] For a computing center with N central devices (computing devices), determine the total amount of computing tasks V to be allocated within the total scheduling cycle. total Based on historical operating data and power tests, the functional relationship between computing power and power of a single computing device was determined:

[0165] p it =f(c)

[0166] Where, p it denoted as 'c', where 'c' represents the power of a single computing device; 'c' represents the computing power of a single computing device, i.e., the computing task capacity of a single computing device in a data center.

[0167] The power of a single computing device in this invention includes the power of the computing device performing the computing task and the cooling energy consumption and other energy consumption distributed across the device, i.e., p it This refers to the total energy consumption, including refrigeration, etc.

[0168] Determine the power consumption of a single computing device under three operating states: standby, default utilization, and maximum utilization.

[0169] Standby mode refers to a device that is powered on but not performing computing tasks. In standby mode, the power consumption of a single computing device is [standby power consumption]. The computing power of a single computing device is 0.

[0170] Under default utilization, the power consumption of a single computing device is the default power. The computing power of a single computing device is c. def .

[0171] Under maximum utilization operation, the power consumption of a single computing device is its maximum power. The computing power of a single computing device is c. max .

[0172] 2. Determine the scheduling cycle and step size for hierarchical scheduling.

[0173] Under normal operating conditions, the power supply for a data center includes: renewable energy generation, mains power, and energy storage devices.

[0174] The scheduling is divided into two layers: medium-to-long-term and short-term.

[0175] The medium- and long-term scheduling cycle is determined based on the duration of the computing task, and the medium- and long-term scheduling cycle is greater than 24 hours.

[0176] The short-term dispatch cycle is determined based on the power dispatch cycle. The medium- and long-term dispatch cycles are taken as the total dispatch cycle, denoted as T. l The medium- to long-term scheduling step size is Δt. l The number of medium- and long-term scheduling nodes is n l :

[0177]

[0178] The short-term scheduling period is T s The medium- and long-term scheduling step size is the same as the short-term scheduling period, i.e., T. s =Δt l The short-term scheduling step size is Δt. s The number of short-run scheduling nodes is n s :

[0179]

[0180] 3. Predict the new energy power generation corresponding to each medium- and long-term scheduling step size to obtain the medium- and long-term new energy power generation prediction sequence.

[0181] On a medium- to long-term timescale, new energy power output is predicted using a short-term cycle as the step size, resulting in a medium- to long-term new energy power generation prediction sequence:

[0182]

[0183] This represents the predicted value of renewable energy power generation at the i-th medium- and long-term scheduling node, i.e., the predicted value of renewable energy power generation at the i-th medium- and long-term scheduling step.

[0184] 4. Determine the maximum computational workload that the predicted power generation of new energy sources can support within each medium- and long-term scheduling step.

[0185] The maximum computational workload that can be carried by calculating the predicted power generation of new energy sources within each medium- and long-term scheduling step is determined by combining the state of the equipment.

[0186] 1) When This indicates that the output of new energy sources can support all equipment operating at maximum power. At this point, the maximum computational workload supported by the predicted power generation of new energy sources within the medium- to long-term scheduling step is:

[0187]

[0188] in, This represents the maximum computational workload that can be supported by the predicted power generation of new energy sources within the i-th medium-to-long-term scheduling step.

[0189] 2) When This indicates that the output of new energy sources can support all equipment to operate at a combination of maximum utilization and default utilization. At this time, the maximum power output N of equipment that new energy sources can support within the medium-to-long-term scheduling step is... max Calculate using the following formula:

[0190]

[0191] N max It is a virtual quantity and can be a decimal.

[0192] The number of operating equipment N that can be supported by renewable energy output within the medium- and long-term dispatching step size (default power). def Calculate using the following formula:

[0193]

[0194] The number of operating equipment N that can be supported by renewable energy output within the medium- and long-term dispatching step size (default power). def It can also be calculated using the following formula:

[0195] N def =N-N max

[0196] At this point, the maximum computational workload that the predicted power generation of new energy sources within the medium- to long-term scheduling step size can support is:

[0197]

[0198] 3) When This indicates that the output of new energy sources can support all equipment to operate at a combination of standby and default power. At this time, the number N of equipment operating at default power that can be supported by the output of new energy sources within the medium- to long-term scheduling step is... def for:

[0199]

[0200] At this point, the maximum computational workload that the predicted power generation of new energy sources within the medium- to long-term scheduling step size can support is:

[0201]

[0202] 5. Allocate the total computational workload based on the maximum computational workload that can be supported by the predicted power generation of new energy sources within each medium- and long-term scheduling step, and obtain the computational workload allocated to each medium- and long-term scheduling step.

[0203] Since most data centers use electrochemical energy storage, the regulation effect of electrochemical energy storage is not considered for data centers using electrochemical energy storage. The calculation workload is allocated based on the relationship between the maximum calculation workload that can be supported by the predicted power generation of new energy sources within each medium and long-term scheduling step and the total calculation workload, and the principle of proximity in time.

[0204] 1) When When this occurs, it indicates that the maximum amount of computational tasks that new energy sources can support during the entire scheduling cycle is greater than or equal to the total amount of computational tasks, and the difference in computational tasks is:

[0205]

[0206] At this time, some of the electricity generated by new energy sources faces the risk of being curtailed. Therefore, the calculation of the workload should be arranged as early as possible to ensure the more certain absorption of new energy in the early stages and leave sufficient room for adjustment in the later stages.

[0207] Specifically, the critical point j is determined to satisfy:

[0208]

[0209] The computational workload for allocating each medium- and long-term scheduling step size is as follows:

[0210]

[0211] Among them, v l,i The amount of computational tasks allocated for the i-th medium-to-long-term scheduling step.

[0212] at this time, , Satisfy the maximum computational task constraint.

[0213] 2) When This indicates that the maximum computational workload supported by new energy sources during the entire scheduling cycle is less than the total computational workload, and the excess tasks... Mains power supply is required. Under the constraint of completing the computing tasks, excess computing tasks should be scheduled to be postponed as much as possible.

[0214] Specifically, the critical point j is determined to satisfy:

[0215]

[0216] That is, the first j medium- and long-term scheduling nodes are arranged according to... During runtime, the following n l -j nodes The execution will just complete the total number of tasks. Therefore:

[0217]

[0218] at this time, , Satisfy the maximum computational task constraint, This represents the maximum amount of computing tasks that the data center can complete within a medium- to long-term scheduling step.

[0219] 6. In the short term, based on the computational workload allocated for each medium- and long-term scheduling step, and combined with the regulating role of energy storage, a short-term computation-power collaborative optimization scheduling model is established to solve the decision variables for each short-term scheduling step. The decision variables include the number of computing devices and the charging and discharging power of energy storage in each operating state.

[0220] For example, to determine the number of computing devices running at default utilization. Number of computing devices operating at maximum utilization and the charging and discharging power of energy storage .

[0221] Specifically, based on the previous steps, the total number of computational tasks to be completed in a certain short-term scheduling cycle is known to be... For example, the total number of computational tasks in the first short-term scheduling cycle. .

[0222] The short-term computing-power collaborative optimization scheduling model includes an objective function and constraints.

[0223] 1) On a short-term scale, predict the renewable energy power generation corresponding to the short-term scheduling step size to obtain the short-term renewable energy power generation prediction sequence:

[0224]

[0225] This represents the predicted value of renewable energy power generation at the i-th short-term scheduling node, i.e., the predicted value of renewable energy power generation at the i-th short-term scheduling step.

[0226] 2) Determine the objective function of the short-term computing and power-coordinated optimization scheduling model.

[0227] The objective function is determined by minimizing the sum of squares of the differences between the predicted power generation of new energy sources corresponding to the short-term scheduling step size and the total load (data center load + energy storage charging).

[0228]

[0229] The charging power for the energy storage of the data center of the i-th short-term scheduling node (charging is positive and discharging is negative). Let be the load power of the data center of the i-th short-term scheduling node.

[0230] The load power of a data center can be calculated using the following formula:

[0231]

[0232] The load power of a data center can also be calculated using the following formula:

[0233]

[0234] + + =N

[0235] Let be the number of computing devices running at the default utilization rate for the i-th short-lived scheduling node. Let be the number of computing devices that the i-th short-term scheduling node is running at maximum utilization. Let N be the number of computing devices in standby mode of the i-th short-term scheduling node. The total number of computing devices in the three operating states is N.

[0236] 3) Determine the constraints.

[0237] The constraints include data center constraints, grid interaction constraints, and energy storage constraints.

[0238] ① Data center constraints.

[0239] Data center constraints include task completion constraints, device balancing constraints, and device quantity constraints.

[0240] The task completion constraint means that the sum of the computational tasks allocated for each short-term scheduling step is greater than or equal to the total computational tasks.

[0241] Equipment balance constraints include computational equipment state space constraints and time balance constraints.

[0242] The state space constraint of computing devices is that at any given moment, there cannot be both a device with zero computing power and a device with maximum computing power in the space; that is, there cannot be a computing device in standby mode and a computing device operating at maximum utilization at the same time.

[0243] The time balance constraint is that the difference in computing power between computing devices at adjacent time points is less than or equal to the difference in maximum computing power.

[0244] The device quantity constraint refers to the fact that the number of computing devices operating in each state is a positive integer, and the sum of the quantities is less than or equal to the total number of computing devices in the data center.

[0245] ② Power grid interaction constraints.

[0246] The power drawn from the mains is less than or equal to the maximum downstream power allowed by the power grid.

[0247] ③ Energy storage constraints.

[0248] Energy storage constraints include energy storage power constraints and energy storage SOC constraints.

[0249] Energy storage power constraints: maximum and minimum operating power constraints and maximum ramp power constraints.

[0250] The maximum and minimum operating power constraints are: the operating power of the energy storage device shall not exceed the maximum allowable operating power of the energy storage device and shall not be less than the minimum allowable operating power of the energy storage device. That is:

[0251]

[0252] The maximum ramp power constraint is: the absolute value of the difference between the energy storage operating power at two consecutive moments must not exceed the maximum allowable operating power difference (maximum power change) of the energy storage device. That is:

[0253]

[0254] In the formula, This represents the minimum permissible operating power for energy storage devices. This refers to the maximum permissible operating power of the energy storage device. This represents the maximum permissible power change of an energy storage device per unit time.

[0255] Energy storage SOC constraints: energy storage SOC continuity constraints, energy storage SOC state constraints, and energy storage SOC initial and final state constraints.

[0256] The continuity constraint of energy storage SOC is: the energy storage SOC state at time t is the sum of the SOC value at time t-1 and the difference between the amount of electricity charged (or the amount of electricity released) during the period from t to t-1, that is:

[0257] when At that time, energy storage charging, ;

[0258] when At that time, the stored energy is discharged. .

[0259] The SOC (State of Charge) constraint for energy storage is: at any given time, the SOC of the energy storage device shall not exceed the maximum allowable SOC of the energy storage device, nor be lower than the minimum allowable SOC of the energy storage device, that is:

[0260]

[0261] The initial and final state constraints of energy storage SOC are: the SOC value at the beginning of the first time step within a short-term scheduling cycle is equal to the SOC value at the end of the last time step, i.e.:

[0262]

[0263]

[0264] in, For energy storage charging and discharging efficiency, , The minimum and maximum SOC values ​​are those allowed by the equipment or set by operators based on experience to protect the equipment. and These are the fixed energy storage SOC values ​​at the beginning and end of the short-term scheduling cycle, respectively.

[0265] 4) Solve for decision variables using intelligent optimization algorithms such as particle swarm optimization.

[0266] As an alternative implementation, other optimization algorithms can be used to solve for the decision variables for each short-term scheduling step. The decision variables include the number of devices operating at default utilization, the number of devices operating at maximum utilization, and the charging and discharging power of the energy storage.

[0267] 7. A single-step look-ahead amortization correction strategy is adopted to perform rolling correction of the execution deviation of the calculation task.

[0268] Hierarchical scheduling includes one medium- to long-term scheduling cycle and n l A short scheduling cycle. The deviation between the allocated (and planned) computational tasks and the actual executed computational tasks within a short scheduling cycle will affect the completion of the entire cycle. Therefore, it is necessary to correct the execution deviation of the computational tasks. A single-step look-ahead amortization correction strategy can be used to correct the execution deviation of the computational tasks. This includes:

[0269] 1) At the m-th medium-to-long-term scheduling node, i.e., after m-1 short-term schedulings have been completed, determine the execution deviation of the computational task volume.

[0270]

[0271] 2) Calculate the average computational workload of each of the remaining medium- and long-term scheduling nodes.

[0272]

[0273] 3) For the short-term scheduling of the m-th medium-to-long-term scheduling node, update the target of the computational task volume for that short-term scheduling period based on the amortized amount.

[0274]

[0275] in, This refers to the cumulative execution deviation of the computational workload when the task reaches the m-th medium-to-long-term scheduling node. Let represent the actual computational workload executed by the i-th medium- to long-term scheduling node. Let be the average computational workload of the m-th medium-to-long-term scheduling node. This represents the computational workload after the m-th medium-to-long-term scheduling node is updated.

[0276] As another implementation method, the single-step look-ahead amortization correction strategy may not be adopted. That is, instead of distributing the execution deviation of the computation task equally to all remaining medium- and long-term scheduling nodes, some of the remaining medium- and long-term scheduling nodes are selected to distribute the execution deviation of the computation task.

[0277] Computer-readable storage medium implementation

[0278] The present invention provides a computer-readable storage medium storing a computer program thereon. When the computer program is executed by a processor, it implements a method for the coordinated optimization scheduling of new energy sources and computing power. The specific implementation process of this method has been described in detail in the implementation of the method for the coordinated optimization scheduling of new energy sources and computing power, and will not be repeated here.

[0279] This invention introduces a multi-timescale collaborative optimization mechanism, dividing the collaboration between new energy and computing power into two levels: medium-to-long-term and short-term. It identifies differentiated dominant collaborative elements for different periods: the medium-to-long-term is dominated by computing power task arrangement, while the short-term (such as the day-ahead) is centered on energy storage regulation, supplemented by planned scheduling. This enables effective guidance for the collaborative optimization of new energy and computing power in data centers under the conditions of infrequent load adjustment, task completion as a constraint, and support for a leasing service model.

[0280] This method has four main beneficial effects. First, it employs a multi-scale coupled nested optimization method, which layers and couples medium- and long-term renewable energy output forecasting, computing power task planning, and short-term real-time scheduling with energy storage charging and discharging strategies. This fully stimulates the synergistic potential between renewable energy and computing power, deeply explores the long-term adjustment potential of the computing power system, and achieves deep synergy from the long term to the short term in the renewable energy-computing power system. Second, by clarifying the dominant variables at each time scale for dimensionality reduction modeling, it overcomes the problems of excessive system model complexity and difficulty in solving, significantly reducing computational complexity. Third, it proposes a time-series near-end priority allocation strategy, which improves the local consumption rate and utilization efficiency of renewable energy by prioritizing the consumption of recent green electricity. Fourth, it introduces a single-step look-ahead amortization correction strategy to handle the task completion deviation caused by volatility and uncertainty during task execution, effectively ensuring the reliable completion of the total computing power task.

Claims

1. A collaborative optimization scheduling method for new energy sources and computing power, characterized in that, include: The scheduling is divided into medium- and long-term and short-term. The medium- and long-term scheduling cycle is determined based on the duration of the computing power tasks of the data center, while the short-term scheduling cycle is determined based on the power scheduling cycle. The step size of the medium- and long-term scheduling cycle is the same as that of the short-term scheduling cycle. Determine the maximum amount of computational tasks that the predicted power generation of new energy sources can support within each medium- and long-term scheduling step, and allocate the total amount of computational tasks according to the maximum amount of computational tasks to obtain the amount of computational tasks allocated to each medium- and long-term scheduling step. A short-term computational-coordinated optimization scheduling model is established based on the computational workload allocated for each medium- and long-term scheduling step, the predicted value of new energy power generation corresponding to the short-term scheduling step, and the total load; the total load includes the load power of the data center and the charging power of energy storage. The short-term computing-power collaborative optimization scheduling model is solved to obtain the decision variables for each short-term scheduling step. The decision variables include the number of computing devices and the charging and discharging power of energy storage in each operating state.

2. The collaborative optimization scheduling method for new energy and computing power according to claim 1, characterized in that, The total computational workload is allocated based on the principle of proximity in time sequence. The allocation method is as follows: if the maximum computational workload that the predicted power generation of new energy sources can support within the entire scheduling cycle is greater than or equal to the total computational workload, then the computational workload is arranged forward; if the maximum computational workload that the predicted power generation of new energy sources can support within the entire scheduling cycle is less than the total computational workload, then the computational workload is arranged backward under the constraint of completing the computational workload.

3. The collaborative optimization scheduling method for new energy and computing power according to claim 2, characterized in that, If the maximum computational workload supported by the predicted renewable energy power generation during the entire scheduling cycle is greater than or equal to the total computational workload, then the computational workload allocated for each medium- and long-term scheduling step is: Among them, v l,i The computational workload allocated for the i-th medium- to long-term scheduling step, V total This represents the total computational workload. Let j be the maximum computational workload that can be supported by the predicted power generation of new energy sources within the i-th medium- and long-term scheduling step, and j be the critical point.

4. The collaborative optimization scheduling method for new energy and computing power according to claim 2, characterized in that, If the maximum computational workload supported by the predicted renewable energy power generation during the entire scheduling cycle is less than the total computational workload, then the computational workload allocated for each medium- and long-term scheduling step is: Among them, v l,i The computational workload allocated for the i-th medium- to long-term scheduling step, V total This represents the total computational workload. Let j be the maximum computational workload that the predicted power generation of new energy sources can support within the i-th medium-to-long-term scheduling step, and j be the critical point. This represents the maximum amount of computing tasks that the data center can complete within a medium- to long-term scheduling step.

5. The collaborative optimization scheduling method for new energy and computing power according to claim 1, characterized in that, The short-term computing-powered collaborative optimization scheduling model includes an objective function and constraints; The objective function is determined based on the predicted value of new energy power generation and the total load corresponding to the short-term scheduling step size.

6. The collaborative optimization scheduling method for new energy and computing power according to claim 5, characterized in that, The constraints include data center constraints, grid interaction constraints, and energy storage constraints. Data center constraints include task completion constraints, equipment balancing constraints, and equipment quantity constraints. Grid interaction constraints require that the power drawn from the grid be less than or equal to the maximum allowable downlink power of the grid. Energy storage constraints include energy storage power constraints and energy storage SOC constraints.

7. The collaborative optimization scheduling method for new energy and computing power according to claim 5 or 6, characterized in that, The continuous operating state of data center computing equipment is discretized into three operating states: standby, default utilization, and maximum utilization. The power and computing power corresponding to each operating state of a single computing device are determined. Based on the power and computing power corresponding to each operating state of a single computing device, the load power of the data center is determined. The objective function is solved to obtain the number of computing devices operating at default utilization, the number of computing devices operating at maximum utilization, and the charging and discharging power of energy storage for each short-term scheduling step.

8. The collaborative optimization scheduling method for new energy and computing power according to claim 1, characterized in that, The method further includes: calculating the deviation between the amount of computing tasks allocated before the current short-term scheduling period and the amount of computing tasks actually executed, and updating the amount of computing tasks in the current short-term scheduling period based on the deviation of the amount of computing tasks, so as to correct the execution deviation of the amount of computing tasks.

9. A collaborative optimization scheduling device for new energy and computing power, comprising a processor, characterized in that, The processor executes a computer program to implement the steps of the method according to any one of claims 1-8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the method according to any one of claims 1-8.

Images