Intelligent management and control method for flexible power distribution network based on cloud side-end cooperation

Abstract

The invention provides a flexible power distribution network intelligent management and control method based on cloud side-end cooperation, and belongs to the technical field of intelligent power grids. An intelligent decision-making model is constructed and issued to an edge server through the cloud, the power grid state is monitored, and a management and control strategy is formulated. The edge servers exchange strategy information and work cooperatively. The sending end edge server computes policy utility based on multi-edge collaboration. And if the strategy is rejected, the sending end edge server updates the intelligent regulation and control judgment and decision model parameters based on the abnormal operation index probability. And if the strategy is accepted and the strategy utility exceeds the utility threshold, the edge server updates the model according to the operation condition of each regional power grid. If the strategy is rejected for multiple times or does not meet the multi-threshold stability requirement, the cloud side uses a ternary loss function containing multiple deviation factors to update the model, and the self-adaptive learning rate and the ternary loss function are introduced by judging the multi-domain stability requirement. According to the invention, the real-time performance and accuracy of decision making are improved.

Description

Technical Field

[0001] The present invention provides an intelligent control method for a flexible distribution network based on cloud-edge-end collaboration, which belongs to the technical field of smart grids. Background Art

[0002] With the continuous development of smart grid technology, intelligent management and control of distribution networks has become the key to improving grid operation efficiency, ensuring power supply reliability and optimizing energy utilization. Traditional distribution network management and control methods often rely on fixed, centralized control centers, making it difficult to achieve flexible and efficient resource scheduling and fault response. In particular, in the face of large-scale access to distributed renewable energy, diversification and rapid changes in power loads, traditional methods are unable to quickly adapt to these dynamic changes, resulting in reduced grid operation efficiency and threats to power supply reliability. Cloud-edge-end collaborative technology provides a new solution for intelligent management and control of distribution networks by integrating the powerful data processing capabilities of cloud computing and the real-time response capabilities of edge computing, as well as the flexible access of terminal devices.

[0003] The intelligent control method of cloud-edge flexible distribution network can intelligently adjust the operation strategy of the power grid according to the changes in power load and the access of distributed renewable energy, so as to achieve optimal utilization of energy and energy conservation and emission reduction. However, the current intelligent control of cloud-edge flexible distribution network still has the following problems: On the one hand, the traditional method fails to make full use of the cooperation between edge servers, resulting in information isolation and inability to integrate data and strategies. The lack of close cooperation between the cloud and the edge leads to slow response of power grid status monitoring and strategy formulation. The failure to calculate the policy utility in a multi-edge collaborative way leads to inaccurate and timely decision-making, which makes it difficult to cope with the complexity of the power grid environment. On the other hand, the traditional method fails to fully combine the advantages of cloud and edge computing to coordinate and optimize the model, and does not consider multi-domain stability, which affects the long-term safe operation of the power grid. At the same time, the traditional method does not update the model based on the improvement of loss function and learning rate based on multiple factors, resulting in inaccurate model updates. Therefore, it is urgent to propose an intelligent control method of flexible distribution network based on cloud-edge collaboration to improve the operation efficiency of the power grid, ensure power supply reliability and optimize energy utilization.

[0004] CN118659894A provides a distribution network network security management and control method and system based on device fingerprints. By real-time scanning of distribution network equipment, abnormal equipment and unknown equipment are monitored and warned, thereby preventing distribution network intrusion. However, this technology fails to fully utilize the cooperation between edge servers, resulting in information isolation and the inability to integrate data and strategies. The lack of close cooperation between the cloud and the edge, and the failure to use a multi-edge collaborative approach to calculate the utility of the strategy, results in inaccurate and in-time decision-making, making it difficult to cope with the complexity of the power grid environment.

[0005] Application No. 202410335908 provides a distribution network control method, system, terminal and storage medium based on the Internet of Things. By establishing a communication connection with the Internet of Things terminal, the power grid monitoring data is obtained from the Internet of Things terminal, and the abnormality of the power grid monitoring data is checked. However, this technology fails to fully combine the advantages of cloud and edge computing to coordinate and optimize the model, and does not consider multi-domain stability, which affects the long-term safe operation of the power grid. And there is no ternary loss function and adaptive learning rate based on multiple factors to update the model, resulting in inaccurate model updates. Summary of the invention

[0006] The present invention provides an intelligent control method for a flexible distribution network based on cloud-edge-end collaboration, which solves the following two technical problems:

[0007] 1) Traditional methods fail to fully utilize the cooperation between edge servers and do not use multi-edge collaboration to calculate the utility of strategies, resulting in inaccurate and in-time decisions and difficulty in coping with the complexity of the power grid environment.

[0008] 2) Traditional methods fail to fully combine the advantages of cloud and edge computing with multi-domain stability, and do not construct a ternary loss function based on multiple factors to update the model, resulting in inaccurate model updates.

[0009] A flexible distribution network intelligent management and control method based on cloud-edge-end collaboration, including utility calculation of distribution network intelligent management and control strategy based on multi-edge collaboration and update of cloud-edge collaborative management and control model based on multi-domain stability;

[0010] (1) Utility calculation of intelligent management and control strategy of distribution network based on multi-edge collaboration;

[0011] First, the cloud server trains and sends intelligent control judgment and decision-making models to the edge server to support intelligent decision-making and distribution network operation status monitoring. Secondly, the sending edge server calculates the grid operation status and control strategy based on the model, and interacts with the receiving edge server for strategy. Finally, the sending edge server calculates the strategy utility based on multi-edge collaboration.

[0012] (2) Update of cloud-edge collaborative management and control model based on multi-domain stability;

[0013] First, when the strategy is rejected by the receiving server, the sending edge server updates the parameters of the intelligent control judgment and decision model based on the probability of abnormal operation indicators of each receiving edge server. Secondly, if the strategy is received by the receiving server, and the strategy utility of the sending edge server based on multi-edge collaborative computing is greater than the utility threshold, the intelligent control judgment and decision model is updated based on the operation status of the power grid in the control area of ​​each receiving server. Finally, if the strategy is rejected multiple times, or the strategy utility does not meet the multi-domain stability requirements, a ternary loss function and adaptive learning rate are constructed on the cloud side to update the intelligent control judgment and decision model.

[0014] Specific:

[0015] (1) The utility calculation of the intelligent management and control strategy of the distribution network based on multi-edge collaboration includes the following steps:

[0016] Step 1.1: The cloud server trains and issues intelligent control judgment and decision-making models;

[0017] The cloud server obtains an intelligent control judgment and decision-making model adapted to the intelligent management and control of flexible distribution networks based on data training in the power grid database. The model analyzes the operating status of the power grid according to the real-time data uploaded by the terminal and the edge collaborative parameters, and gives corresponding control strategies.

[0018] The cloud server will train the intelligent control judgment and decision model M cloud It is sent to each edge server. The edge server set is represented as {d1,...,d j ,...,d J}, where dj represents the jth edge server and J is the total number of edge servers. j First, update your intelligent control judgment and decision model according to the model parameters issued by the cloud server, that is, w edge,j =w cloud Then, the distribution network operation status is evaluated based on the regional distribution network operation data and models uploaded by the terminals in the control area and the corresponding control strategies are generated. j Intelligent control judgment and decision model M edge,j It is expressed as:

[0019] output=M edge,j (w edge,j ,input)

[0020] input={D j ,ζ}

[0021]

[0022] Among them, the input parameters include two parts, one is dj The regional distribution network operation data uploaded by the terminals in the control area is represented by D j ,The other part is the edge collaboration input parameter, which is used when edge collaboration is performed between edge servers.,The control strategy is input here to analyze the effectiveness of the strategy.,When edge collaboration is not performed, this parameter is set to empty, expressed as ζ.,S1, S2, and S3 respectively indicate that the power grid operation status is normal, good, and abnormal. is d j The corresponding probabilities of the three power grid operation states in the control area. S1: It is a key-value pair, indicating the current d j The probability that the power grid in the control area is in state S1 is Grid control strategy.

[0023] Edge Server j According to the grid operation status data calculated by the model, it is judged whether the current regional grid needs to be controlled. According to historical experience, the grid operation safety threshold is set to P th ,when When edge server d j If it is determined that the operating status of the power grid in the current control area is abnormal, the server will send the control strategy to the adjacent server as the sending server, and evaluate and optimize the strategy through edge collaboration.

[0024] Step 1.2: Control policy interaction between edge servers

[0025] The edge server d determines the abnormal operation status of the power grid in the current control area j As a sending edge server to the adjacent edge server d j′ , Send control policy in, is d j A collection of adjacent edge servers. j′ After receiving the control policy, the receiving server will j The control strategy sent is input into the model M as the edge collaboration input parameter edge,j′ In , the model input and output are expressed as:

[0026]

[0027] In the formula, the output of the model is is the updated power grid operation state obtained based on the edge collaborative control strategy. Let Θ1, Θ2, Θ3 be the transition thresholds of the three states. or This shows that the operation status of the power grid has deteriorated after the control strategy is adopted, and it cannot meet the requirements of power grid security and stability.j′ The control strategy will be rejected and various operating indicators and corresponding probabilities will be Updated power grid operation status deviation obtained based on edge collaborative control strategy Send back to the sending edge server d j . Complete the interaction process. Otherwise, directly combine the various operating indicators and the corresponding probabilities Updated power grid operation status deviation obtained based on edge collaborative control strategy Send back to the sending edge server d j .

[0028] Step 1.3: Utility calculation of intelligent management and control strategy of distribution network based on multi-edge collaboration

[0029] Sending edge server d j Based on multi-edge collaborative computing strategy utility ψ j On the one hand, the policy utility takes into account the control strategy adopted by the receiving server After that, the power grid is in three operating states. If the probability of abnormal state increases after adopting the strategy, it means that for d j′ The utility of this strategy in the region is poor. Feedback to the utility of the multi-edge collaborative computing strategy is reflected as follows: when the probability of abnormal state increases or the probability of normal and good is low, the utility will also decrease. On the other hand, considering various operating indicators, if the probability of voltage over-limit, line overcurrent and reverse overload exceeds the threshold, it means that the strategy has poor coordination effect among edge servers and the strategy utility should be reduced, which is specifically expressed as:

[0030]

[0031] In the formula, is the number of edge servers at the receiving end, Ω1, Ω2, and Ω3 are the exponential factors of the three power grid operation states, which are used to distinguish the influence differences of probability deviations of different states. v , T r , T b are the probability thresholds of voltage over-limit, line over-current and reverse overload, ω v ,ω r ,ω b They are the weights of voltage over-limit, line over-current and reverse overload respectively. is the indicator function, when It is 1 when the probability of voltage over-limit, line over-current and reverse overload exceeds the threshold, and a penalty term is imposed on the strategy utility, and the greater the threshold is exceeded, the greater the penalty term is.

[0032] (2) Updating the cloud-edge collaborative management and control model based on multi-domain stability includes the following steps:

[0033] Step 2.1: Update the intelligent control judgment and decision model on the edge server side;

[0034] If the strategy The receiving server rejects the update, and the sending edge server obtains the updated power grid operation status according to the edge collaborative management and control strategy. and the corresponding probability Update the intelligent control judgment and decision model parameters, expressed as:

[0035]

[0036] In the formula, is the gradient function and η is the learning rate.

[0037] If the strategy is received by the receiving server, and d j Strategic utility ψ based on multi-edge collaborative computing j Greater than the utility threshold ψ th , then execute the policy And record the multi-edge collaboration situation.

[0038] Step 2.2: Multi-domain stability judgment and cloud-side intelligent control judgment and decision model update;

[0039] If the strategy is rejected multiple times, or the strategy utility ψ j If the multi-domain stability requirement is not met, the edge server d j Grid operation and decision making Upload to the cloud. Multi-domain stability includes high-frequency fluctuation domain, large-range fluctuation domain and trust domain: the high-frequency fluctuation domain can reflect whether the policy utility fluctuates repeatedly with high frequency near the threshold; the large-range fluctuation domain can reflect whether the policy utility fluctuates frequently and in a large range; the trust domain reflects the credibility of the effectiveness of the control strategy, which is quantified by the state change of the power grid after the edge server area of ​​each receiving end executes the strategy. Multi-domain stability cannot be achieved if the conditions of any domain are not met.

[0040] Consider a total of T interaction iterations, and obtain T interaction strategy utilities, the set of which is represented by ψ j ={ψ j,1 ,...,ψ j,t ,...,ψ j,T}, the case where the multi-domain stability requirement is not met is expressed as:

[0041]

[0042] In the formula, Indicates the mean of the strategy utility The deviation from the utility threshold does not exceed a constant δ, Indicates that the frequency of strategy utility sets that are less than the utility threshold exceeds Num th,1 , both of which are satisfied at the same time, which can reflect that the strategy utility fluctuates repeatedly with high frequency near the threshold and does not meet the multi-domain stability requirements. Indicates that the deviation between the strategy utility in the strategy utility set and the threshold exceeds the maximum range The frequency is greater than Num th,2 , indicating that the strategy utility frequently fluctuates over a large range and does not meet the multi-domain stability requirements. It represents the relative value of the state change of the power grid after the edge server area of ​​each receiving end executes the strategy. If the probability of the power grid being in a good and normal state increases slightly and the probability of being in an abnormal state decreases slightly, it means that the effectiveness is poor. Therefore, when This indicates that the strategy has poor effectiveness, low confidence, and does not meet the multi-domain stability requirements.

[0043] The cloud side constructs a ternary loss function based on the policy utility ψ j With the utility threshold ψ th The deviation of each receiving edge server is based on various operating indicators and the corresponding probability And the receiving server adopts the control strategy The changes of the power grid in three operating states The intelligent control judgment and decision-making model is updated to improve the model accuracy, which is specifically expressed as follows:

[0044]

[0045] In the formula, d(ψ j ,ψ th ) represents the strategic utility ψ j With the utility threshold ψ th The deviation, η cloud The learning rate is adaptive on the cloud side. The worse the effect of the edge interaction strategy is, the higher the learning rate is, thereby increasing the speed of model convergence.

[0046] The present invention proposes an intelligent control method for a flexible distribution network based on cloud-edge-end collaboration, which has the following technical effects:

[0047] (1) Utility calculation of intelligent management and control strategies for distribution networks based on multi-edge collaboration: Build and send intelligent decision-making models to edge servers through the cloud to monitor the status of the power grid and formulate management and control strategies. Edge servers exchange strategy information and work together to achieve intelligent management and control, improving decision-making speed and accuracy. Finally, the sending edge server calculates the utility of the strategy based on multi-edge collaboration, further improving the real-time and accuracy of decision-making.

[0048] (2) Update of cloud-edge collaborative control model based on multi-domain stability; if the strategy is rejected, the sending edge server updates the parameters of the intelligent control judgment and decision-making model based on the probability of abnormal operation indicators of each receiving edge server. If the strategy is accepted and the strategy utility exceeds the utility threshold, the edge server will further update the model according to the operation status of the power grid in each region. If the strategy is rejected multiple times or does not meet the multi-threshold stability requirements, the cloud will use a ternary loss function containing multiple deviation factors to update the model. By judging the multi-domain stability requirements and introducing adaptive learning rate and ternary loss function, it provides a comprehensive and accurate optimization direction for the model to improve the efficiency and accuracy of the update. This method takes advantage of the advantages of cloud and edge computing and optimizes the model update process. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0050] The present invention proposes an intelligent control method for a flexible distribution network based on cloud-edge-end collaboration. Figure 1 As shown in the figure, it includes the utility calculation of the intelligent management and control strategy of the distribution network based on multi-edge collaboration and the update of the cloud-edge collaborative management and control model based on multi-domain stability, which are introduced as follows:

[0051] (1) Utility calculation of intelligent management and control strategy of distribution network based on multi-edge collaboration

[0052] First, the cloud server trains and sends intelligent control judgment and decision-making models to the edge server to support intelligent decision-making and distribution network operation status monitoring. Secondly, the sending edge server calculates the grid operation status and control strategy based on the model, and interacts with the receiving edge server for strategy. Finally, the sending edge server calculates the strategy utility based on multi-edge collaboration.

[0053] Step 1.1: The cloud server trains and sends the intelligent control judgment and decision model

[0054] A cloud server is a virtualized server based on cloud computing technology, which provides sufficient computing resources, storage space and application services through the Internet. The cloud server is trained based on data in the power grid database to obtain an intelligent control judgment and decision-making model that is adapted to the intelligent control of flexible distribution networks. The model can analyze the operation status of the power grid based on the real-time data uploaded by the terminal and the edge coordination parameters, and give corresponding control strategies.

[0055] The cloud server will train the intelligent control judgment and decision model M cloud It is sent to each edge server. The edge server set is represented as {d1,...,d j ,...,d J}, where dj represents the jth edge server and J is the total number of edge servers. j First, update your intelligent control judgment and decision model according to the model parameters issued by the cloud server, that is, w edge,j =w cloud Then, the distribution network operation status is evaluated based on the regional distribution network operation data and models uploaded by the terminals in the control area and the corresponding control strategies are generated. j Intelligent control judgment and decision model M edge,j It is expressed as:

[0056] output=M edge,j (w edge,j ,input)

[0057] input={D j ,ζ}

[0058]

[0059] Among them, the input parameters include two parts, one is d j The regional distribution network operation data uploaded by the terminals in the control area, including voltage, current, active power, reactive power, power quality, etc., is expressed as D j ,The other part is the edge collaboration input parameter, which is used when edge collaboration is performed between edge servers. The management and control strategy can be input here to analyze the effectiveness of the strategy. When there is no edge collaboration, this parameter is set to empty, denoted as ζ. S1, S2, and S3 respectively indicate that the operation status of the power grid is normal, good, and abnormal. is d j The corresponding probabilities of the three power grid operation states in the control area. S1: It is a key-value pair, indicating the current d j The probability that the power grid in the control area is in state S1 is Grid control strategy.

[0060] Edge Server j According to the grid operation status data calculated by the model, it is judged whether the current regional grid needs to be controlled. According to historical experience, the grid operation safety threshold is set to P th ,when When edge server d j If it is determined that the operating status of the power grid in the current control area is abnormal, the server will send the control strategy to the adjacent server as the sending server, and evaluate and optimize the strategy through edge collaboration.

[0061] Step 1.2: Control policy interaction between edge servers

[0062] The edge server d determines the abnormal operation status of the power grid in the current control area j As a sending edge server to the adjacent edge server d j′ , Send control policy in, is d j A collection of adjacent edge servers. j′ After receiving the control policy, the receiving server will j The control strategy sent is input into the model M as the edge collaboration input parameter edge,j′ In the above example, the model input and output are expressed as

[0063]

[0064] In the formula, the output of the model is is the updated power grid operation state obtained based on the edge collaborative control strategy. Let Θ1, Θ2, Θ3 be the transition thresholds of the three states. or This shows that the operation status of the power grid has deteriorated after the control strategy is adopted, and it cannot meet the requirements of power grid security and stability. j′ The control strategy will be rejected and various operating indicators such as voltage limit, reverse overload index and corresponding probabilities will be Updated power grid operation status deviation obtained based on edge collaborative control strategy Send back to the sending edge server d j Otherwise, directly calculate the voltage over-limit, reverse overload index and other operating indicators and their corresponding probabilities. Updated power grid operation status deviation obtained based on edge collaborative control strategy Send back to the sending edge server d j .

[0065] Step 1.3: Utility calculation of intelligent management and control strategy of distribution network based on multi-edge collaboration

[0066] Sending edge server d j Based on multi-edge collaborative computing strategy utility ψ j On the one hand, the policy utility takes into account the control strategy adopted by the receiving server After that, the power grid is in three operating states. If the probability of abnormal state increases after adopting the strategy, it means that for d j′The utility of this strategy in the region is poor. Feedback to the utility of the multi-edge collaborative computing strategy is reflected as follows: when the probability of abnormal state increases or the probability of normal and good is low, the utility will also decrease. On the other hand, considering various operating indicators such as voltage over-limit and reverse overload indicators, if the probability of voltage over-limit, line overcurrent and reverse overload exceeds the threshold, it means that the strategy has poor coordination effect among edge servers, and the strategy utility should be appropriately reduced, which is specifically expressed as:

[0067]

[0068] In the formula, is the number of edge servers at the receiving end, Ω1, Ω2, and Ω3 are the exponential factors of the three power grid operation states, which are used to distinguish the influence differences of probability deviations of different states. v , T r , T b are the probability thresholds of voltage over-limit, line over-current and reverse overload, ω v ,ω r ,ω b They are the weights of voltage over-limit, line over-current and reverse overload respectively. is the indicator function, when It is 1 when the probability of voltage over-limit, line over-current and reverse overload exceeds the threshold, and a penalty term is imposed on the strategy utility, and the greater the threshold is exceeded, the greater the penalty term is.

[0069] (2) Update of cloud-edge collaborative management and control model based on multi-domain stability

[0070] First, when the strategy is rejected by the receiving server, the sending edge server updates the parameters of the intelligent control judgment and decision model based on the probability of abnormal operation indicators of each receiving edge server. Secondly, if the strategy is received by the receiving server, and the strategy utility of the sending edge server based on multi-edge collaborative computing is greater than the utility threshold, the intelligent control judgment and decision model is updated based on the operation status of the power grid in the control area of ​​each receiving server. Finally, if the strategy is rejected multiple times, or the strategy utility does not meet the multi-domain stability requirements, a ternary loss function and adaptive learning rate are constructed on the cloud side to update the intelligent control judgment and decision model.

[0071] Step 2.1: Update the intelligent control judgment and decision model on the edge server side

[0072] If the strategy The receiving server rejects the update, and the sending edge server obtains the updated power grid operation status according to the edge collaborative management and control strategy. and the corresponding probability Update the intelligent control judgment and decision model parameters, expressed as:

[0073]

[0074] In the formula, is the gradient function and η is the learning rate.

[0075] If the strategy is received by the receiving server, and d j Strategic utility ψ based on multi-edge collaborative computing j Greater than the utility threshold ψ th , you can execute the policy And record the multi-edge collaboration situation.

[0076] Step 2.2: Multi-domain stability judgment and cloud-side intelligent control judgment and decision model update

[0077] If the strategy is rejected multiple times, or the strategy utility ψ j If the multi-domain stability requirement is not met, the edge server d j Grid operation and decision making Upload to the cloud. Multi-domain stability includes high-frequency fluctuation domain, large-range fluctuation domain and trust domain: the high-frequency fluctuation domain can reflect whether the policy utility fluctuates repeatedly with high frequency near the threshold; the large-range fluctuation domain can reflect whether the policy utility fluctuates frequently and in a large range; the trust domain reflects the credibility of the effectiveness of the control strategy, which can be quantified by the state change of the power grid after the edge server area of ​​each receiving end executes the strategy. Multi-domain stability cannot be achieved if the conditions of any domain are not met.

[0078] Consider a total of T interaction iterations, and obtain T interaction strategy utilities, the set of which is represented by ψ j ={ψ j,1 ,...,ψ j,t ,...,ψ j,T}, the case where the multi-domain stability requirement is not met is expressed as:

[0079]

[0080] In the formula, The mean value of the strategy utility The deviation from the utility threshold does not exceed a constant δ, Indicates that the frequency of strategy utility sets that are less than the utility threshold exceeds Num th,1 , both of which are satisfied at the same time, which can reflect that the strategy utility fluctuates repeatedly with high frequency near the threshold and does not meet the multi-domain stability requirements. Indicates that the deviation between the strategy utility in the strategy utility set and the threshold exceeds the maximum range The frequency is greater than Num th,2 , indicating that the strategy utility frequently fluctuates over a large range and does not meet the multi-domain stability requirements. It represents the relative value of the state change of the power grid after the edge server area of ​​each receiving end executes the strategy. If the probability of the power grid being in a good and normal state increases slightly and the probability of being in an abnormal state decreases slightly, it means that the effectiveness is poor. Therefore, when This indicates that the strategy has poor effectiveness, low confidence, and does not meet the multi-domain stability requirements.

[0081] The cloud side constructs a ternary loss function based on the policy utility ψ j With the utility threshold ψ th The deviation of each receiving edge server is based on various operating indicators such as voltage limit, reverse overload indicators, and the corresponding probability And the receiving server adopts the control strategy The changes of the power grid in three operating states The intelligent control judgment and decision-making model is updated to improve the model accuracy, which is specifically expressed as follows:

[0082]

[0083] In the formula, d(ψ j ,ψ th ) represents the strategy utility ψ j With the utility threshold ψ th The deviation, η cloud The adaptive learning rate on the cloud side, the worse the effect of the edge interaction strategy, the higher the learning rate, thereby increasing the speed of model convergence. The adaptive learning rate and the ternary loss function can more accurately and comprehensively quantify the deviation between the current model and the optimal model, providing a more accurate direction for model learning calibration.

Claims

1. A method for intelligent management and control of a flexible distribution network based on cloud-edge-end collaboration, characterized in that: Including utility calculation of intelligent management and control strategies of distribution networks based on multi-edge collaboration and update of cloud-edge collaborative management and control models based on multi-domain stability; (1) Utility calculation of intelligent management and control strategy of distribution network based on multi-edge collaboration; First, the cloud server trains and sends intelligent control judgment and decision models to the edge server to support intelligent decision-making and distribution network operation status monitoring; Secondly, the edge server at the sending end calculates the grid operation status and control strategy based on the model, and interacts with the edge server at the receiving end for strategy; finally, the edge server at the sending end calculates the strategy utility based on multi-edge collaboration; (2) Update of cloud-edge collaborative management and control model based on multi-domain stability; First, when the strategy is rejected by the receiving server, the sending edge server updates the parameters of the intelligent control judgment and decision-making model based on the probability of abnormal operation indicators of each receiving edge server; secondly, if the strategy is received by the receiving server, and the strategy utility based on multi-edge collaborative computing of the sending edge server is greater than the utility threshold, the intelligent control judgment and decision-making model is updated based on the operation status of the power grid in the control area of ​​each receiving server; finally, if the strategy is rejected multiple times, or the strategy utility does not meet the multi-domain stability requirements, a ternary loss function and adaptive learning rate are constructed on the cloud side to update the intelligent control judgment and decision-making model.

2. According to claim 1, a method for intelligent management and control of a flexible distribution network based on cloud-edge-end collaboration is characterized in that: (1) The utility calculation of the intelligent management and control strategy of the distribution network based on multi-edge collaboration includes the following steps: Step 1.1: The cloud server trains and issues intelligent control judgment and decision-making models; The cloud server obtains an intelligent control judgment and decision-making model adapted to the intelligent control of flexible distribution networks based on data training in the power grid database. The model analyzes the operation status of the power grid based on the real-time data uploaded by the terminal and the edge coordination parameters, and gives corresponding control strategies. The cloud server will train the intelligent control judgment and decision model M cloud It is sent to each edge server. The edge server set is represented as {d1,...,d j ,...,d J }, where dj represents the jth edge server, J is the total number of edge servers; edge server d j First, update your intelligent control judgment and decision model according to the model parameters issued by the cloud server, that is, w edge,j =w cloud ; Then, based on the regional distribution network operation data and models uploaded by the terminals in the control area, the distribution network operation status is evaluated and the corresponding control strategy is generated; the edge server d j Intelligent control judgment and decision model M edge,j It is expressed as: output=M edge,j (w edge,j ,input) input={D j ,ζ} Among them, the input parameters include two parts, one is d j The regional distribution network operation data uploaded by the terminals in the control area is represented by D j , the other part is the edge collaboration input parameter, which is used when edge collaboration is performed between edge servers. The control strategy is input here to analyze the effectiveness of the strategy. When edge collaboration is not performed, the parameter is set to empty, expressed as ζ; S1, S2, and S3 respectively indicate that the power grid operation status is normal, good, and abnormal; is d j The probabilities corresponding to the three power grid operation states in the control area; S1: It is a key-value pair, indicating the current d j The probability that the power grid in the control area is in state S1 is Grid control strategy; Edge Server j According to the grid operation status data calculated by the model, it is judged whether the current regional grid needs to be controlled; according to historical experience, the grid operation safety threshold is set as P th ,when When edge server d j If the power grid in the current control area is judged to be abnormal, the server will send the control strategy to the adjacent servers as the sending server, and evaluate and optimize the strategy through edge collaboration; Step 1.2: Control policy interaction between edge servers The edge server d determines the abnormal operation status of the power grid in the current control area j As a sending edge server to the adjacent edge server Send control policy in, is d j A collection of adjacent edge servers; d j′ After receiving the control policy, the receiving server will j The control strategy sent is input into the model M as the edge collaboration input parameter edge,j′ In , the model input and output are expressed as: In the formula, the output of the model is is the updated power grid operation state obtained based on the edge collaborative control strategy; let Θ1, Θ2, Θ3 be the transition thresholds of the three states, if or This shows that the operation status of the power grid has deteriorated after the control strategy is adopted, and it cannot meet the requirements of power grid security and stability. j′ The control strategy will be rejected and various operating indicators and corresponding probabilities will be Updated power grid operation status deviation obtained based on edge collaborative control strategy Send back to the sending edge server d j ; Complete the interaction process; otherwise, directly combine the various operating indicators and the corresponding probabilities Updated power grid operation status deviation obtained based on edge collaborative control strategy Send back to the sending edge server d j ; Step 1.3: Utility calculation of intelligent management and control strategy of distribution network based on multi-edge collaboration Sending edge server d j Based on multi-edge collaborative computing strategy utility ψ j On the one hand, the policy utility takes into account the control strategy adopted by the receiving server After that, the power grid is in three operating states. If the probability of abnormal state increases after adopting the strategy, it means that for d j′ The utility of this strategy in the region is poor; feedback to the utility of the multi-edge collaborative computing strategy is reflected as follows: when the probability of abnormal state increases or the probability of normal and good is low, the utility will also decrease; on the other hand, considering various operating indicators, if the probability of voltage over-limit, line overcurrent and reverse overload exceeds the threshold, it means that the strategy has poor coordination effect among edge servers and the strategy utility should be reduced, which is specifically expressed as: In the formula, is the number of edge servers at the receiving end, Ω1, Ω2, and Ω3 are the exponential factors of the three power grid operation states, which are used to distinguish the influence differences of probability deviations of different states; T v , T r , T b are the probability thresholds of voltage over-limit, line over-current and reverse overload, ω v ,ω r ,ω b are the weights of voltage over-limit, line over-current and reverse overload respectively; is the indicator function, when It is 1 when the voltage exceeds the limit, the line overcurrent and the reverse overload probability exceed the threshold, and a penalty term is imposed on the strategy utility, and the greater the threshold is exceeded, the greater the penalty term is.

3. According to claim 1, a flexible distribution network intelligent management and control method based on cloud-edge-end collaboration is characterized in that: (2) Updating the cloud-edge collaborative management and control model based on multi-domain stability includes the following steps: Step 2.1: Update the intelligent control judgment and decision model on the edge server side; If the strategy The receiving server rejects the update, and the sending edge server obtains the updated power grid operation status according to the edge collaborative management and control strategy. and the corresponding probability Update the intelligent control judgment and decision model parameters, expressed as: In the formula, is the gradient function, η is the learning rate; If the strategy is received by the receiving server, and d j Strategic utility ψ based on multi-edge collaborative computing j Greater than the utility threshold ψ th , then execute the policy And record the multi-edge collaboration situation; Step 2.2: Multi-domain stability judgment and cloud-side intelligent control judgment and decision model update; If the strategy is rejected multiple times, or the strategy utility ψ j If the multi-domain stability requirement is not met, the edge server d j Grid operation and decision making Upload to the cloud; Multi-domain stability includes high-frequency fluctuation domain, large-range fluctuation domain and trust domain: the high-frequency fluctuation domain can reflect whether the strategy utility fluctuates repeatedly with high frequency near the threshold; the large-range fluctuation domain can reflect whether the strategy utility fluctuates frequently and in a large range; the trust domain reflects the credibility of the effectiveness of the control strategy, which is quantified by the state changes of the power grid after the execution of the strategy in each receiving edge server area; Multi-domain stability cannot be achieved if the conditions of any domain are not met; Consider a total of T interaction iterations, and obtain T interaction strategy utilities, the set of which is represented by ψ j ={ψ j,1 ,...,ψ j,t ,...,ψ j,T }, the case where the multi-domain stability requirement is not met is expressed as: In the formula, The mean value of the strategy utility The deviation from the utility threshold does not exceed a constant δ, Indicates that the frequency of strategy utility sets that are less than the utility threshold exceeds Num th,1 ,The simultaneous satisfaction of both can reflect that the strategy utility fluctuates repeatedly at a high frequency near the threshold, which does not meet the multi-domain stability requirements; Indicates that the deviation between the strategy utility in the strategy utility set and the threshold exceeds the maximum range The frequency is greater than Num th,2 , indicating that the strategy utility frequently fluctuates over a large range, and does not meet the multi-domain stability requirements; It represents the relative value of the state change of the power grid after the edge server area of ​​each receiving end executes the strategy. If the probability of the power grid being in a good and normal state increases slightly and the probability of being in an abnormal state decreases slightly, it means that the effectiveness is poor. Therefore, when This indicates that the strategy has poor effectiveness, low confidence, and does not meet the multi-domain stability requirements; The cloud side constructs a ternary loss function based on the policy utility ψ j With the utility threshold ψ th The deviation of each receiving edge server is based on various operating indicators and the corresponding probability And the receiving server adopts the control strategy The changes of the power grid in three operating states The intelligent control judgment and decision-making model is updated to improve the model accuracy, which is specifically expressed as follows: In the formula, d(ψ j ,ψ th ) represents the strategic utility ψ j With the utility threshold ψ th The deviation, η cloud The learning rate is adaptive on the cloud side. The worse the effect of the edge interaction strategy is, the higher the learning rate is, thereby increasing the speed of model convergence.

Images