Virtual power plant cloud side end operation control method and system

Abstract

The invention discloses a virtual power plant cloud side end operation control method and system, and belongs to the field of cloud side end operation control methods, and the method comprises the following steps: 1, data collection and preprocessing; step 2, edge calculation and quick response: aggregating operation data of the charging piles in each area at edge nodes, and executing local optimization and quick response; step 3, cloud global optimization and scheduling: a cloud platform predicts the charging demand in the next 24 hours based on a long and short term memory model, and solves a multi-objective optimization model to generate a scheduling plan; 4, issuing and executing an instruction; and 5, performing dynamic feedback and closed-loop optimization. According to the invention, through the cloud side-end collaborative architecture and the multi-level optimization strategy, the problems of response delay, local optimization deficiency, poor multi-target collaboration and the like of an existing virtual power plant technology in a large-scale charging pile scene are solved, and an innovative solution is provided for efficient operation of a smart power grid and bidirectional interaction of user requirements.

Description

Technical Field

[0001] The present invention relates to the field of cloud-edge-end operation control methods, and specifically to a cloud-edge-end operation control method and system for a virtual power plant. Background Art

[0002] By integrating distributed energy resources (such as renewable energy, energy storage systems, electric vehicle charging piles, etc.) and leveraging information and communication technologies to achieve efficient coordination of resources, virtual power plants have become an important means to enhance the flexibility and stability of the power grid. However, the following problems still exist in the existing technologies for real-time scheduling and optimization of large-scale electric vehicle charging piles: 1. Insufficient response speed: Traditional virtual power plants rely on centralized cloud scheduling, with a long data processing link, making it difficult to meet the demand for rapid power regulation of charging piles (such as second-level response), resulting in a lag in regulation when the dynamic load of the power grid fluctuates.

[0003] 2. Lack of local optimization: Existing solutions mostly adopt global optimization strategies, but lack the ability to perform real-time data analysis and make rapid decisions on the edge side, and cannot effectively address problems such as overload of distribution network transformers in local areas or sudden increases in charging demand.

[0004] 3. Poor multi-objective coordination: Existing optimization models have insufficient coordination of multiple objectives such as power grid power purchase cost, peak power suppression, and satisfaction of user charging demands, and the comprehensiveness of constraint conditions (such as charging completion time deviation, energy storage SOC range, voltage stability) needs to be improved.

[0005] Therefore, those skilled in the art have provided a cloud-edge-end operation control method and system for a virtual power plant to solve the problems raised in the above background art. Summary of the Invention

[0006] The purpose of the present invention is to provide a cloud-edge-end operation control method and system for a virtual power plant. Through a cloud-edge-end collaborative architecture and a multi-level optimization strategy, the present invention solves problems such as response delay, lack of local optimization, and poor multi-objective coordination in the existing virtual power plant technologies in large-scale charging pile scenarios, and provides an innovative solution for the efficient operation of the smart grid and the two-way interaction of user needs, so as to solve the problems raised in the above background art.

[0007] To achieve the above purpose, the present invention provides the following technical solutions: A cloud-edge-end operation control method for a virtual power plant, comprising the following steps: Step 1: Data collection and preprocessing; Real-time collection of charging power, battery state of charge (SOC), user reservation information, and distribution network node voltage data through sensors deployed at the charging pile terminals, and filtering of outliers and protocol conversion of the original data through an edge gateway; Step 2: Edge computing and rapid response; Aggregate the operation data of charging piles in each region at the edge node, and perform local optimization and rapid response; Step 3: Cloud global optimization and scheduling; The cloud platform predicts the charging demand for the next 24 hours based on the Long Short-Term Memory (LSTM) model, and solves the multi-objective optimization model to generate a scheduling plan; Step 4: Instruction issuance and execution; Decompose the scheduling plan generated by the cloud into power instructions at 15-minute intervals, and send them to the charging pile terminal through the 5G ultra-reliable low-latency communication (URLLC) channel, and adjust the charging power in real time through the Pulse Width Modulation (PWM) signal, with a response time ≤ 1 second; Step 5: Dynamic feedback and closed-loop optimization; Real-time monitor the deviation between the actual power of the charging pile and the instruction. If the Mean Absolute Percentage Error (MAPE) > 8%, trigger the adjustment mechanism.

[0008] As a further solution of the present invention: The local optimization in the said step 2 includes the following sub-steps: S1: Collect the current total load of the distribution network transformer in the target area, and compare it with the load value in the previous 15 minutes; S2: According to the comparison result, divert the vehicles to be charged and not confirmed for reservation in the target area to the charging station with the highest priority; S3: Extract the charging station that the target vehicle intends to go to from the user reservation information, and obtain the load of the distribution network transformer of this charging station; S4: Allocate charging piles for user vehicles according to the preset allocation rules, and send the allocation result to the user vehicles; S5: If the total load in the current target area exceeds the preset threshold, reduce the power of the charging pile according to the preset priority.

[0009] As a further solution of the present invention: The specific process of charging station diversion in the said step S2 is as follows: Extract the loads of the distribution network transformers of each charging station in the target area, and sort them from small to large to generate the first list; Extract the distances between each charging station and the target vehicle, and sort them from small to large to generate the second list; According to the current load change amplitude, assign dynamic weight factors a and b to the first list and the second list respectively; Calculate the priority parameter S of each charging station = list ranking weight * a + distance ranking weight * b, and select the charging station with the smallest S as the charging station with the highest priority; Send the location of the charging station with the highest priority to the target vehicle to guide it to go.

[0010] As a further solution of the present invention: The assignment rules for the dynamic weight factors a and b are as follows: If the current total load increases by ≥5% compared to the previous 15 minutes, then a = 3 and b = 1; If the increase is <5%, then a = 3 and b = 2; If the decrease is <5%, then a = 2 and b = 3; If the decrease is ≥5%, then a = 1 and b = 3.

[0011] As a further solution of the present invention: If the target vehicle accepts the guidance and arrives at the highest-priority charging station, contribution points are accumulated for it according to the driving mileage.

[0012] As a further solution of the present invention: The preset allocation rule in step S4 is as follows: According to the load status of the target charging station, calculate the dispersion weight value V of each idle charging pile; If the load is within the preset interval, V is the number of idle charging piles within a radius of 5 meters; If the load exceeds the limit, V is the number of idle charging piles within a radius of 3 meters; If the load is insufficient, V is the number of idle charging piles within a radius of 10 meters; Select the charging pile corresponding to the maximum V value and allocate it to the user vehicle.

[0013] As a further solution of the present invention: In step S5, the specific process of reducing the power of the charging pile according to the preset priority is as follows: Reduce the power of non-bus charging piles in the current target area by 50%; Collect the load of the distribution network transformer after the power reduction in the target area and mark it as C; On the premise of ensuring that the bus charging piles operate at full power, calculate the load idle value Q = (c1 - C) / c2, where c1 and c2 are both preset values; Extract the integer of the load idle value Q and mark it as q; Extract the contribution points of the remaining charging vehicles in the target area and sort them from largest to smallest to obtain the third list; Send an offer to the first q charging vehicles in the third list, and the content of the offer is "Whether to consume ten contribution points to increase the current charging power"; If the charging vehicle agrees to the offer, deduct the corresponding contribution points of the charging vehicle and adjust the charging pile corresponding to the charging vehicle to full power operation; If the charging vehicle does not agree to the offer, send the offer to the charging vehicle that is next in the third list.

[0014] As a further solution of the present invention: In step 3, the model expression of the multi-objective optimization model is: ; Among them, refers to the grid power purchase cost during time period t, refers to the total charging power of all charging piles within time period t, refers to the weight parameter for the peak power, refers to the maximum value of the charging power within the optimization period, and T refers to the total time period of the optimization period; The constraint conditions of the multi-objective optimization model include: the deviation of the user's charging completion time ≤ 10%, the energy storage SOC is maintained in the range of 20% - 90%, and the voltage deviation of the distribution network node does not exceed ±7%.

[0015] As a further solution of the present invention: the adjustment mechanism includes: (1), Enable the standby energy storage system to compensate for the power gap; (2), Update the weights of the LSTM model through the federated learning framework and roll-optimize the subsequent scheduling plan every 15 minutes.

[0016] This application also discloses a virtual power plant cloud-edge-end operation control system, which adopts the virtual power plant cloud-edge-end operation control method, including: The data acquisition module is used to perform data acquisition and preprocessing; The edge computing module is used to perform edge computing and rapid response; The cloud decision-making module is used to perform cloud global optimization and scheduling; The instruction execution module is used to issue and execute power instructions; The dynamic feedback module is used to perform dynamic feedback and closed-loop optimization.

[0017] Compared with the prior art, the beneficial effects of the present invention are: 1. This application realizes real-time monitoring and rapid diversion of local loads through edge computing, alleviating the risk of overload of the distribution network transformer.

[0018] 2. This application uses cloud intelligent prediction (such as the LSTM model) to generate a scheduling plan for multi-objective optimization, balancing economy and grid stability.

[0019] 3. This application designs a dynamic feedback mechanism, combines the standby energy storage system with the rolling update of model parameters, and improves the robustness of the system.

[0020] 4. This application introduces incentive mechanisms such as user contribution points to optimize the charging pile allocation strategy and improve user participation and resource utilization. Description of the Drawings

[0021] Figure 1 It is a flowchart of the virtual power plant cloud-edge-end operation control method; Figure 2 It is a structural block diagram of the operation control system for the cloud-edge-terminal of the virtual power plant. Specific implementation manners

[0022] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the scope of protection of the present invention.

[0023] As mentioned in the background art of this application, through research, it is found that there are still problems in the prior art such as insufficient response speed, lack of local optimization, and poor multi-objective coordination in the real-time scheduling and optimization of large-scale electric vehicle charging piles, and there are certain defects.

[0024] To solve the above defects, this application discloses a virtual power plant cloud-edge-terminal operation control method and system. Through the cloud-edge-terminal collaborative architecture and multi-level optimization strategy, it solves the problems of response delay, lack of local optimization, and poor multi-objective coordination in the prior virtual power plant technology in large-scale charging pile scenarios, and provides an innovative solution for the efficient operation of the smart grid and the two-way interaction of user needs.

[0025] The following will introduce in detail how the solution of this application solves the above technical problems in conjunction with the accompanying drawings.

[0026] Please refer to Figure 1, in the embodiment of the present invention, the virtual power plant cloud-edge-end operation control method includes the following steps: Step 1: Data collection and preprocessing; real-time collect the charging power, State of Charge (SOC) of the battery, user reservation information, and distribution network node voltage data through sensors deployed at the charging pile terminals, and perform outlier filtering and protocol conversion on the original data through the edge gateway; Step 2: Edge computing and fast response; aggregate the operation data of charging piles in each area at the edge node, and perform local optimization and fast response; Step 3: Cloud global optimization and scheduling; the cloud platform predicts the charging demand for the next 24 hours based on the Long Short-Term Memory (LSTM) model, and solves the multi-objective optimization model to generate a scheduling plan; Step 4: Instruction issuance and execution; decompose the scheduling plan generated by the cloud into power instructions at 15-minute intervals, and send them to the charging pile terminals through the 5G Ultra-Reliable Low-Latency Communication (URLLC) channel, and adjust the charging power in real time through the Pulse Width Modulation (PWM) signal, with a response time ≤ 1 second; Step 5: Dynamic feedback and closed-loop optimization; monitor the deviation between the actual power of the charging pile and the instruction in real time. If the Mean Absolute Percentage Error (MAPE) > 8%, trigger the adjustment mechanism. Real-time scheduling and closed-loop control are achieved through a hierarchical architecture (fast response on the edge side and global optimization on the cloud side), solving the problem of response delay in traditional solutions and improving the dynamic regulation ability of the power grid.

[0027] In this embodiment, the local optimization in Step 2 includes the following sub-steps: S1: Collect the current total load of the distribution network transformer in the target area and compare it with the load value in the previous 15 minutes; S2: According to the comparison result, divert the vehicles to be charged and not confirmed for reservation in the target area to the charging station with the highest priority; S3: Extract the charging station that the target vehicle intends to go to from the user reservation information, and obtain the current load of the distribution network transformer of this charging station; S4: Allocate charging piles for user vehicles according to the preset allocation rules, and send the allocation result to the user vehicles; S5: If the total load in the current target area exceeds the preset threshold, reduce the power of the charging piles according to the preset priority. Rapidly adjust the charging strategy based on the real-time data of the edge node, preferentially relieve the overload risk in local areas, and ensure the stable operation of the distribution network.

[0028] In this embodiment, the specific process of charging station drainage in step S2 is as follows: extract the distribution network transformer loads of each charging station in the target area, sort them from small to large to generate a first list; extract the distances between each charging station and the target vehicle, sort them from small to large to generate a second list; according to the current load change range, assign dynamic weight factors a and b to the first list and the second list respectively; calculate the priority parameter S of each charging station = list ranking weight * a + distance ranking weight * b, and select the charging station with the smallest S as the highest priority charging station; send the location of the highest priority charging station to the target vehicle to guide it to go there. Dynamically balance the charging station load and user convenience, optimize resource allocation, reduce the local power grid pressure, and improve the user charging efficiency.

[0029] In this embodiment, the assignment rules for the dynamic weight factors a and b are as follows: if the current total load increases by ≥5% compared with the previous 15 minutes, then a = 3 and b = 1; if the increase is <5%, then a = 3 and b = 2; if the decrease is <5%, then a = 2 and b = 3; if the decrease is ≥5%, then a = 1 and b = 3. This setting can flexibly respond to the power grid load fluctuations, give priority to ensuring the power grid stability in high-load scenarios, and at the same time take into account the charging convenience of users.

[0030] In this embodiment, if the target vehicle accepts the guidance and arrives at the highest priority charging station, contribution points will be accumulated for it according to the driving mileage. This setting encourages users to cooperate with the scheduling strategy, promotes the spatio-temporal balanced distribution of charging demands, and improves the resource utilization rate and user participation.

[0031] In this embodiment, the preset allocation rule in step S4 is as follows: according to the load status of the target charging station, calculate the dispersion weight value V of each idle charging pile; if the load is within the preset interval, V is the number of idle charging piles within a radius of 5 meters; if the load exceeds the limit, V is the number of idle charging piles within a radius of 3 meters; if the load is insufficient, V is the number of idle charging piles within a radius of 10 meters; select the charging pile corresponding to the maximum V value and allocate it to the user vehicle. This setting can optimize the spatial distribution of charging piles, avoid local congestion, improve the operation efficiency of the charging station, and extend the service life of the equipment.

[0032] In this embodiment, in step S5, the specific process of reducing the power of the charging pile according to the preset priority is as follows: reduce the power of non-bus charging piles in the current target area by 50%; collect the load of the distribution network transformer after the power reduction in the target area and mark it as C; on the premise of ensuring the full-power operation of the bus charging pile, calculate the load idle value Q = (c1 - C) / c2, where both c1 and c2 are preset values; extract the integer of the load idle value Q and mark it as q; extract the contribution scores of the remaining charging vehicles in the target area and sort them from largest to smallest to obtain the third list; send an offer to the first q charging vehicles in the third list, and the content of the offer is "whether to consume ten contribution scores to increase the current charging power"; if the charging vehicle agrees to the offer, deduct the corresponding contribution scores of the charging vehicle and adjust the charging pile corresponding to the charging vehicle to full-power operation; if the charging vehicle does not agree to the offer, send the offer to the charging vehicle that is next in line in the third list. This setting quickly reduces the load when the emergency load exceeds the limit, and at the same time, through the integral mechanism, users can make autonomous choices to balance the grid stability and user experience.

[0033] In this embodiment, in step 3, the model expression of the multi-objective optimization model is: ; Among them, refers to the grid power purchase cost in time period t, refers to the total charging power of all charging piles in time period t, refers to the weight parameter for the peak power, refers to the maximum value of the charging power within the optimization period, and T refers to the total time period of the optimization period; the constraint conditions of the multi-objective optimization model include: the deviation of the user's charging completion time ≤ 10%, the energy storage SOC is maintained in the range of 20% - 90%, and the voltage deviation of the distribution network node does not exceed ±7%. Through mathematical modeling, multi-objective collaborative optimization of economy and grid security is achieved to ensure that the scheduling plan takes into account both user needs and grid constraints.

[0034] In this embodiment, the adjustment mechanism includes: (1) enabling the standby energy storage system to compensate for the power gap; (2) updating the weights of the LSTM model through the federated learning framework and rolling and optimizing the subsequent scheduling plan every 15 minutes. This setting dynamically compensates for the operation deviation, combines data-driven optimization to continuously improve the prediction accuracy, and enhances the system robustness and adaptive ability.

[0035] Please refer to Figure 2, this application also discloses a virtual power plant cloud-edge-end operation control system, which adopts the virtual power plant cloud-edge-end operation control method, including: a data acquisition module for performing data acquisition and preprocessing; an edge computing module for performing edge computing and rapid response; a cloud decision-making module for performing cloud global optimization and scheduling; an instruction execution module for issuing and executing power instructions; and a dynamic feedback module for performing dynamic feedback and closed-loop optimization.

[0036] To further illustrate the present invention, the following describes in detail the virtual power plant cloud-edge-end operation control method and system provided by the present invention in conjunction with embodiments.

[0037] Embodiment 1: Charging station drainage and power regulation The edge node collects the data of charging piles in the target area and calculates the load change rate of the distribution network transformer; Dynamically adjust the charging station priority weights (a, b) according to the load change rate to generate a list of the highest priority charging stations; Send guiding information to the target vehicle and allocate charging piles according to the user response; If the total regional load exceeds the limit, reduce the power of non-bus charging piles according to the preset rules and dynamically adjust the user charging priority through the contribution points mechanism.

[0038] Embodiment 2: Cloud multi-objective optimization The cloud predicts the charging demand in the next 24 hours based on the LSTM model; Solve the weighted sum of minimizing the grid power purchase cost and peak power of the objective function, while satisfying the user charging time, energy storage SOC, and voltage stability constraints; Update the model parameters every 15 minutes through federated learning to roll-optimize the subsequent scheduling plan.

[0039] The present invention realizes real-time monitoring of local load and rapid drainage through edge computing, alleviating the risk of overload of the distribution network transformer. And uses cloud intelligent prediction (such as the LSTM model) to generate a multi-objective optimized scheduling plan to balance economy and grid stability. By designing a dynamic feedback mechanism, combined with a backup energy storage system and rolling update of model parameters, the system robustness is improved. At the same time, incentive mechanisms such as user contribution points are introduced to optimize the charging pile allocation strategy, improving user participation and resource utilization.

[0040] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these changes and modifications.

[0041] The above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, making equivalent substitutions or changes, shall be covered by the protection scope of the present invention.

Claims

1. A virtual power plant cloud-edge operation control method, characterized in that: The following steps are involved: Step 1: Data collection and preprocessing; The sensors deployed at the charging pile terminals collect charging power, battery state of charge, user reservation information, and distribution network node voltage data in real time, and perform outlier filtering and protocol conversion on the raw data through the edge gateway; Step 2: Edge computing and rapid response; Aggregate charging pile operation data in various regions at the edge nodes, and perform local optimization and rapid response; Step 3: Global optimization and scheduling in the cloud; The cloud platform predicts the charging demand for the next 24 hours based on the long short-term memory model and solves the multi-objective optimization model to generate a scheduling plan; Step 4: Instruction issuance and execution; The scheduling plan generated in the cloud is decomposed into power instructions at 15-minute intervals, which are sent to the charging pile terminal through the 5G communication channel, and the charging power is adjusted in real time through pulse width modulation signals, with a response time of ≤1 second; Step 5: Dynamic feedback and closed-loop optimization; Monitor the deviation between the actual power of the charging pile and the command in real time. If the average absolute percentage error is greater than 8%, the adjustment mechanism will be triggered.

2. The virtual power plant cloud edge operation control method according to claim 1 is characterized in that: The local optimization in step 2 includes the following sub-steps: S1: Collect the current total load of the distribution network transformers in the target area and compare it with the load value of the previous 15 minutes; S2: Based on the comparison results, vehicles waiting to be charged in the target area and without confirmed reservations are directed to the highest priority charging station; S3: extract the charging station that the target vehicle intends to go to from the user reservation information, and obtain the current distribution network transformer load of the charging station; S4: Allocate a charging pile to the user's vehicle according to a preset allocation rule, and send the allocation result to the user's vehicle; S5: If the total load in the current target area exceeds the preset threshold, the power of the charging pile is reduced according to the preset priority.

3. The virtual power plant cloud edge operation control method according to claim 2 is characterized in that: The specific process of charging station drainage in step S2 is as follows: Extract the distribution network transformer loads of each charging station in the target area, and generate a first list by sorting the loads from small to large; Extract the distance between each charging station and the target vehicle, and generate a second list by sorting the distance from small to large; According to the current load change range, dynamic weight factors a and b are assigned to the first list and the second list respectively; Calculate the priority parameter S of each charging station = list ranking weight * a + distance ranking weight * b, and select the charging station with the smallest S as the highest priority charging station; The location of the highest priority charging station is sent to the target vehicle to guide it there.

4. The virtual power plant cloud edge operation control method according to claim 3 is characterized in that: The assignment rules of the dynamic weight factors a and b are as follows: If the current total load increases by ≥5% compared to the previous 15 minutes, then a=3, b=1; If the increase is less than 5%, then a=3, b=2; If the decrease is less than 5%, then a=2, b=3; If the decrease is ≥5%, then a=1, b=3.

5. The virtual power plant cloud edge operation control method according to claim 4 is characterized in that: If the target vehicle accepts the guidance and arrives at the highest priority charging station, contribution points will be accumulated for it based on the mileage.

6. The virtual power plant cloud edge operation control method according to claim 5 is characterized in that: The preset allocation rule in step S4 is: According to the load status of the target charging station, the decentralized weight value V of each idle charging pile is calculated; If the load is within the preset range, V is the number of idle charging piles within a radius of 5 meters; If the load exceeds the limit, V is the number of idle charging piles within a radius of 3 meters; If the load is insufficient, V is the number of idle charging piles within a radius of 10 meters; Select the charging pile corresponding to the maximum V value and assign it to the user's vehicle.

7. The virtual power plant cloud edge operation control method according to claim 6 is characterized in that: In step S5, the specific process of reducing the power of the charging pile according to the preset priority is: Reduce the power of non-public transportation charging stations in the current target area by 50%; Collect the transformer load of the distribution network after the power reduction in the target area and mark it as C; Under the premise of ensuring that the bus charging pile operates at full power, calculate the load idle value Q = (c1-C) / c2, where c1 and c2 are both preset values; Extract the integer of the load idle value Q, marked as q; Extract the contribution points of the remaining charging vehicles in the target area, and sort them from large to small according to the contribution points to obtain a third list; Send an offer to the first q charging vehicles in the third list, where the offer content is "whether to consume ten contribution points to increase the current charging power"; If the charging vehicle agrees to the offer, the contribution points corresponding to the charging vehicle will be deducted, and the charging pile corresponding to the charging vehicle will be adjusted to full power operation; If the charging vehicle disagrees with the offer, the offer is sent to the charging vehicles in the next descending order in the third list.

8. The virtual power plant cloud edge operation control method according to claim 7 is characterized in that: In step 3, the model expression of the multi-objective optimization model is: ； in, Refers to the cost of purchasing electricity from the power grid during time period t, Refers to the total charging power of all charging piles within time period t. Refers to the weight parameter for peak power, Refers to the maximum value of charging power during the optimization cycle, and T refers to the total time period of the optimization cycle; The constraints of the multi-objective optimization model include: the user charging completion time deviation is ≤10%, the energy storage SOC is maintained in the range of 20%-90%, and the distribution network node voltage deviation does not exceed ±7%.

9. The virtual power plant cloud edge operation control method according to claim 8 is characterized in that: The adjustment mechanism includes: (1) Enable the backup energy storage system to compensate for the power gap; (2) Update the LSTM model weights through the federated learning framework and optimize the subsequent scheduling plan every 15 minutes.

10. The virtual power plant cloud-edge operation control system is characterized by: The virtual power plant cloud-edge operation control method according to any one of claims 1 to 9 comprises: A data acquisition module, used to perform data acquisition and preprocessing; Edge computing module, used to perform edge computing and rapid response; Cloud decision module, used to perform cloud global optimization and scheduling; An instruction execution module, used for issuing and executing power instructions; Dynamic feedback module, used to perform dynamic feedback and closed-loop optimization.

Images