Cross-transformer-area cooperative control method and device for power distribution network and electronic equipment

Abstract

The invention provides a cross-transformer-area cooperative control method and device for a power distribution network and electronic equipment, and belongs to the technical field of intelligent power distribution networks. The method comprises the steps of performing spatial-temporal feature extraction based on power distribution network data of all transformer areas to obtain spatial-temporal features of a plurality of transformer areas of a power distribution network; constructing a cross-transformer-area coupling coefficient matrix based on the power distribution network data of all transformer areas; respectively inputting the spatial-temporal characteristics, the cross-court coupling coefficient matrix, the power distribution network data and the environment prediction data into a deep reinforcement learning model to obtain a power distribution network initial reconstruction scheme output by the deep reinforcement learning model; and based on the local decision schemes of the plurality of transformer areas and the initial reconstruction scheme of the power distribution network, determining a target reconstruction scheme of the power distribution network, thereby enabling the plurality of transformer areas of the power distribution network to perform cooperative control based on the target reconstruction scheme of the power distribution network. The method is used for solving the problem that a traditional power distribution network is poor in operation stability and reliability under complex working conditions under the conditions of distributed energy intermittency, volatility, cross-court load imbalance and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of smart distribution networks, and in particular to a distribution network cross-zone collaborative control method, a distribution network cross-zone collaborative control device, an electronic device, a machine-readable storage medium, and a computer program product. Background Art

[0002] With the increasing penetration of distributed energy (such as solar energy and wind energy) in distribution networks and the growth of diversified electricity demand on the user side, the operating characteristics of distribution networks have become increasingly complex.

[0003] Traditional distribution networks are primarily powered by centralized power sources, with relatively simple network structures and control strategies focused on ensuring power supply reliability. However, faced with the intermittent and fluctuating nature of distributed energy resources and uneven load distribution across substations, traditional distribution networks face operational stability and reliability issues under complex operating conditions. Summary of the Invention

[0004] The purpose of the embodiments of the present invention is to provide a method, device and electronic equipment for coordinated control of distribution networks across substations, so as to solve the problems of poor operating stability and reliability of traditional distribution networks under complex working conditions when faced with intermittent and fluctuating distributed energy resources and unbalanced loads across substations.

[0005] To achieve the above objectives, an embodiment of the present invention provides a method for coordinated control of distribution networks across substations, including: Acquire sensor data of multiple substations in the distribution network; the sensor data of the multiple substations at least includes distribution network data and environmental prediction data of all substations; Extracting spatiotemporal features based on the distribution network data of all the substations to obtain spatiotemporal features of multiple substations in the distribution network; Constructing an inter-area coupling coefficient matrix based on the distribution network data of all the areas; the inter-area coupling coefficient matrix includes the inter-area coupling coefficients of all the areas, and the inter-area coupling coefficients represent the sensitivity of the local area voltage fluctuation caused by the load change of other areas; Inputting the spatiotemporal characteristics, the cross-station coupling coefficient matrix, the distribution network data, and the environmental prediction data into a deep reinforcement learning model respectively, and obtaining an initial distribution network reconstruction plan output by the deep reinforcement learning model; Based on the local decision plans of multiple substations and the distribution network initial reconstruction plan, a distribution network target reconstruction plan is determined, so that multiple substations of the distribution network are coordinated and controlled based on the distribution network target reconstruction plan.

[0006] Optionally, the cross-substation coupling coefficient is calculated based on a voltage change when a voltage fluctuation occurs in the first substation, a load change in the remaining substations, and a per-unit value of branch impedances between the first substation and the remaining substations; The first substation is any one of all substations in the distribution network; and the remaining substations are all substations except the first substation.

[0007] Optionally, the cross-stage coupling coefficient is calculated using the following formula: ; in, C i represents the cross-station coupling coefficient of the ith station; represents the voltage change of the ith station; represents the load change of the j-th substation; represents the per-unit value of the branch impedance between the i-th substation and the j-th substation; n represents the number of the remaining substations.

[0008] Optionally, the distribution network data of all substations include topological structure data and distribution network operation data of all substations; and the spatiotemporal feature extraction based on the distribution network data of all substations to obtain the spatiotemporal features of multiple substations of the distribution network includes: Graph convolutional neural network is used to extract spatial features from the topological structure data of all substations to obtain spatial features; Use deep learning models to extract time features from the distribution network operation data of all substations to obtain time features; The spatial features and the temporal features are fused using an attention fusion method to obtain the spatiotemporal features of the multiple substations in the distribution network.

[0009] Optionally, the local decision-making plan for each substation is obtained by the following steps: Controlling each edge agent in the station area to generate a random waiting time through a trusted execution environment; The edge agent with the shortest random waiting time is determined as the target edge agent, and the decision-making plan of the target edge agent is determined as the local decision-making plan of the station area.

[0010] Optionally, the local decision-making plan of each edge intelligent agent is based on the distribution network operation data of the substation and the initial reconstruction plan of the distribution network, and is obtained by using a fuzzy logic controller to make a decision.

[0011] Optionally, the sensor data of the multiple substations include distribution network data, environmental forecast data, distributed energy output data, and user electricity usage behavior data of all substations; The step of inputting the spatiotemporal characteristics, the cross-station coupling coefficient matrix, the distribution network data, and the environmental prediction data into a deep reinforcement learning model to obtain an initial distribution network reconstruction plan output by the deep reinforcement learning model includes: The spatiotemporal characteristics, the cross-station coupling coefficient matrix, the distribution network data, the environmental prediction data, the distributed energy output data, and the user electricity consumption behavior data are respectively input into a deep reinforcement learning model to obtain an initial distribution network reconstruction plan output by the deep reinforcement learning model.

[0012] Optionally, determining a target distribution network reconstruction plan based on the local decision plans of the multiple substations and the initial distribution network reconstruction plan includes: When the distribution network initial reconstruction plan conflicts with the local decision-making plan of the first substation, determining the distribution network initial reconstruction plan as the distribution network target reconstruction plan of the first substation; Among them, in the event of an emergency occurring in the first substation, the first substation will give priority to handling the emergency and then execute the distribution network target reconstruction plan; the emergency event includes a voltage over-limit event and / or an equipment overload event, and the first substation is any one of the multiple substations.

[0013] On the other hand, an embodiment of the present invention further provides a distribution network cross-station coordinated control device, comprising: A data acquisition module is used to obtain sensor data of multiple substations in the distribution network; the sensor data of the multiple substations at least includes distribution network data and environmental prediction data of all substations; A model training and optimization module is used to extract spatiotemporal features based on the distribution network data of all the substations to obtain spatiotemporal features of multiple substations in the distribution network; A model evaluation module is configured to construct an inter-area coupling coefficient matrix based on the distribution network data of all the areas; the inter-area coupling coefficient matrix includes inter-area coupling coefficients of all areas, and the inter-area coupling coefficients represent the sensitivity of the local area voltage fluctuation caused by load changes in other areas; An upper-layer global optimization module is used to input the spatiotemporal characteristics, the cross-station coupling coefficient matrix, the distribution network data, and the environmental prediction data into a deep reinforcement learning model, respectively, to obtain an initial distribution network reconstruction plan output by the deep reinforcement learning model; The conflict resolution strategy module is used to determine the distribution network target reconstruction plan based on the local decision plans of multiple substations and the distribution network initial reconstruction plan, so that multiple substations of the distribution network can be coordinated and controlled based on the distribution network target reconstruction plan.

[0014] Optionally, the cross-substation coupling coefficient is calculated based on a voltage change when a voltage fluctuation occurs in the first substation, a load change in the remaining substations, and a per-unit value of branch impedances between the first substation and the remaining substations; The first substation is any one of all substations in the distribution network; and the remaining substations are all substations except the first substation.

[0015] Optionally, the cross-stage coupling coefficient is calculated using the following formula: ; in, C i represents the cross-station coupling coefficient of the ith station; represents the voltage change of the ith station; represents the load change of the j-th substation; represents the per-unit value of the branch impedance between the i-th substation and the j-th substation; n represents the number of the remaining substations.

[0016] Optionally, the distribution network data of all substations include topological structure data and distribution network operation data of all substations; and the spatiotemporal feature extraction based on the distribution network data of all substations to obtain the spatiotemporal features of multiple substations of the distribution network includes: Graph convolutional neural network is used to extract spatial features from the topological structure data of all substations to obtain spatial features; Use deep learning models to extract time features from the distribution network operation data of all substations to obtain time features; The spatial features and the temporal features are fused using an attention fusion method to obtain the spatiotemporal features of the multiple substations in the distribution network.

[0017] Optionally, the local decision-making plan for each substation is obtained by the following steps: Controlling each edge agent in the station area to generate a random waiting time through a trusted execution environment; The edge agent with the shortest random waiting time is determined as the target edge agent, and the decision-making plan of the target edge agent is determined as the local decision-making plan of the station area.

[0018] Optionally, the local decision-making plan of each edge intelligent agent is based on the distribution network operation data of the substation and the initial reconstruction plan of the distribution network, and is obtained by using a fuzzy logic controller to make a decision.

[0019] Optionally, the sensor data of the multiple substations include distribution network data, environmental forecast data, distributed energy output data, and user electricity usage behavior data of all substations; The step of inputting the spatiotemporal characteristics, the cross-station coupling coefficient matrix, the distribution network data, and the environmental prediction data into a deep reinforcement learning model to obtain an initial distribution network reconstruction plan output by the deep reinforcement learning model includes: The spatiotemporal characteristics, the cross-station coupling coefficient matrix, the distribution network data, the environmental prediction data, the distributed energy output data, and the user electricity consumption behavior data are respectively input into a deep reinforcement learning model to obtain an initial distribution network reconstruction plan output by the deep reinforcement learning model.

[0020] Optionally, determining a target distribution network reconstruction plan based on the local decision plans of the multiple substations and the initial distribution network reconstruction plan includes: When the distribution network initial reconstruction plan conflicts with the local decision-making plan of the first substation, determining the distribution network initial reconstruction plan as the distribution network target reconstruction plan of the first substation; Among them, in the event of an emergency occurring in the first substation, the first substation will give priority to handling the emergency and then execute the distribution network target reconstruction plan; the emergency event includes a voltage over-limit event and / or an equipment overload event, and the first substation is any one of the multiple substations.

[0021] On the other hand, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned cross-station coordinated control method of the distribution network when executing the program.

[0022] On the other hand, the present invention also provides a machine-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned distribution network cross-station coordinated control method.

[0023] On the other hand, the present invention also provides a computer program product, including a computer program, which implements the above-mentioned distribution network cross-station coordinated control method when executed by a processor.

[0024] Through the above technical solution, the embodiment of the present invention inputs the spatiotemporal characteristics, the cross-station coupling coefficient matrix, the distribution network data and the environmental prediction data into the deep reinforcement learning model respectively, and obtains the initial distribution network reconstruction plan output by the deep reinforcement learning model; then, based on the local decision-making plans of multiple stations and the initial distribution network reconstruction plan, the distribution network target reconstruction plan is determined, so that multiple stations in the distribution network can be collaboratively controlled based on the distribution network target reconstruction plan. The embodiment of the present invention realizes efficient cross-station energy scheduling and precise coordinated voltage control through the technical means of multimodal data fusion, deep reinforcement learning and distributed collaborative decision-making, thereby improving the operational stability and reliability of the distribution network under complex working conditions when facing situations such as intermittent and fluctuating distributed energy and unbalanced loads across stations.

[0025] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following detailed description, they are used to explain the embodiments of the present invention, but do not constitute a limitation of the embodiments of the present invention. In the accompanying drawings: Figure 1 This is one of the flow charts of the cross-station coordinated control method of the distribution network provided by the present invention; Figure 2 This is the second flow chart of the cross-station coordinated control method of the distribution network provided by the present invention; Figure 3 This is one of the structural diagrams of the cross-station coordinated control device for the distribution network provided by the present invention; Figure 4 This is the second structural diagram of the cross-station coordinated control device for the distribution network provided by the present invention; Figure 5 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0027] The following describes the specific implementation of the embodiment of the present invention in detail with reference to the accompanying drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the embodiment of the present invention and is not used to limit the embodiment of the present invention.

[0028] Method Example Please refer to Figure 1 The embodiment of the present invention provides a method for coordinated control of a distribution network across substations, including: Step 100: Obtain sensor data from multiple substations in the distribution network.

[0029] Electronic devices can obtain sensor data from multiple distribution network substations via a communication network. A distribution network substation can be understood based on the following scenario: a feeder line in the distribution network connects multiple substation transformers, with each substation independently supplying power. A residential community or village is typically powered by one or more substations, each independently responsible for voltage conversion, power distribution, and operational monitoring for users within its area. The sensor data from these multiple substations includes at least distribution network data and environmental prediction data for all substations. In one embodiment, to maximize multimodal data, the sensor data from these multiple substations includes distribution network data, environmental prediction data, distributed energy output data, and user electricity usage behavior data for all substations. These sensors are distributed across key locations in the distribution network, such as busbars, feeders, and distributed energy access points. They can collect real-time, heterogeneous information from multiple sources, including topological data (reflecting the electrical connections between substations), distribution network operational data (e.g., voltage, current, and power), environmental prediction data (e.g., future light intensity, temperature, and wind speed), user behavior data (e.g., electricity usage patterns, load curves, and electric vehicle charging patterns), and distributed energy output data.

[0030] In addition, please refer to Figure 2 After data collection, the electronic device can also perform data preprocessing on the collected data. For example, the electronic device can clean the collected data to remove noise and outliers, and perform data standardization, such as averaging, to make different types of data have a unified dimension and range.

[0031] Step 200: Extract spatiotemporal features based on the distribution network data of all the substations to obtain spatiotemporal features of multiple substations in the distribution network.

[0032] The distribution network data for all substations includes topological structure data and distribution network operation data for all substations. The electronic device first extracts spatial features from the topological structure data of all substations to obtain spatial features; then extracts temporal features from the distribution network operation data of all substations to obtain temporal features; and finally, fuses the spatial features and temporal features to obtain spatiotemporal features for multiple substations in the distribution network.

[0033] In one embodiment, step 200, performing spatiotemporal feature extraction based on the distribution network data of all the substations to obtain spatiotemporal features of multiple substations in the distribution network, includes: using a graph convolutional neural network to extract spatial features from the topological structure data of all the substations to obtain spatial features; using a deep learning model to extract temporal features from the distribution network operation data of all the substations to obtain temporal features; and using an attention fusion method to fuse the spatial features and the temporal features to obtain the spatiotemporal features of multiple substations in the distribution network.

[0034] Among them, the topological structure data of all distribution network substations is constructed with nodes (transformers) and edges (line connections) for each substation. G =( V , E ), node characteristics include voltage and load, and edge weights are electrical distance or impedance. The matrix is ​​represented as follows: adjacency matrix A Represents the connection relationship between nodes and neighboring nodes, distance matrix D Calculates the Euclidean distance based on the node's geographic coordinates (latitude and longitude).

[0035] When processing distribution network topology data, electronic devices use a graph convolutional neural network (GCN) to extract spatial features. Through multi-layer graph convolution operations, the GCN aggregates information about a node and its neighboring nodes, thereby extracting spatial features that reflect the electrical connectivity between substations. The electronic devices then use various deep learning models (for example, the Transformer model) to extract temporal features, arranging distribution network operation data (such as voltage, current, and power) in chronological order to form time series data. In other embodiments, the electronic devices may also arrange distribution network operation data (such as voltage, current, and power), environmental prediction data, and user electricity usage behavior data in chronological order to form time series data. Using a multi-head attention mechanism, the Transformer model can capture the dependencies between different time steps in the time series data, thereby extracting temporal features. Finally, the electronic devices use attention fusion to fuse the spatial features extracted by the GCN with the temporal features extracted by the Transformer model. Specifically, the spatial and temporal features are first linearly transformed and mapped to the same dimension. Attention weights are then calculated between the spatial and temporal features, and the spatial and temporal features are weightedly fused based on the attention weights. Therefore, the embodiment of the present invention generates a comprehensive spatiotemporal feature matrix for the distribution network data of all substations in the distribution network in this way, providing rich input information for the subsequent deep reinforcement learning model.

[0036] Step 300: Construct an inter-substation coupling coefficient matrix based on the distribution network data of all substations.

[0037] The electronic device constructs an inter-area coupling coefficient matrix based on the distribution network data of all the areas. The inter-area coupling coefficient matrix includes inter-area coupling coefficients of all the areas. The inter-area coupling coefficients represent the sensitivity of the local area voltage fluctuation caused by load changes in other areas.

[0038] In one embodiment, the cross-area coupling coefficient is calculated based on the voltage change when the voltage fluctuation occurs in the first area, the load change in the remaining areas, and the branch impedance standard value between the first area and the remaining areas. The first area is any one of all the areas in the distribution network; the remaining areas are all areas other than the first area. Among them, the voltage change when the voltage fluctuation occurs in the first area can be the voltage change detected in a detection cycle. The load change in the remaining areas can be the load change detected in the above detection cycle. Specifically, for example, the cross-area coupling coefficient is calculated by the following formula: ; in, C i represents the cross-station coupling coefficient of the ith station; represents the voltage change of the ith station; Represents the load change of the j-th substation; the load change can be represented by the power change. The per-unit value of the branch impedance between the i-th and j-th substations is the ratio of the actual impedance value (nominal value) between the i-th and j-th substations to the selected reference impedance value. n represents the number of substations in the remaining substations, that is, the number of substations in the remaining substations is the difference between the total number of substations and 1.

[0039] This embodiment of the present invention quantifies inter-substation voltage sensitivity by calculating the cross-substation coupling coefficient, reflecting the degree to which load changes in one substation affect voltages in other substations. By inputting this coefficient into a deep reinforcement learning model as a reference factor, the model can more comprehensively understand the electrical connection characteristics of the distribution network and the interactions between substations.

[0040] Step 400: Input the spatiotemporal characteristics, the cross-station coupling coefficient matrix, the distribution network data, and the environmental prediction data into a deep reinforcement learning model respectively to obtain an initial distribution network reconstruction plan output by the deep reinforcement learning model.

[0041] The electronic device can maximize overall system performance (e.g., minimizing voltage deviation, maximizing distributed energy consumption, optimizing energy storage charging and discharging strategies, etc.). The spatiotemporal features obtained in step 200, the cross-substation coupling coefficient matrix obtained in step 300, and the distribution network data and environmental prediction data obtained in step 100 are respectively input into a deep reinforcement learning model to obtain an initial distribution network reconstruction plan output by the deep reinforcement learning model. The distribution network data may include the global state of the distribution network (e.g., the power distribution of each feeder, the state of charge of the energy storage, etc.). The environmental prediction data may include meteorological forecast data for the distribution network location for a period of time (e.g., light intensity, temperature, wind speed, etc.). The distribution network reconstruction plan includes at least one of a sequence of switch states for all substations in the distribution network (switches need to be opened and closed to adjust the grid topology), charge and discharge power instructions for the energy storage system, and an output regulation coefficient for the distributed energy resources. In one embodiment, the distribution network reconstruction plan includes a sequence of switch states for all substations in the distribution network (switches need to be opened and closed to adjust the grid topology), charge and discharge power instructions for the energy storage system, and an output regulation coefficient for the distributed energy resources.

[0042] In other embodiments, the sensor data from the multiple substations includes distribution network data, environmental prediction data, distributed energy output data, and user electricity usage behavior data for all substations. The spatiotemporal characteristics, the cross-substation coupling coefficient matrix, the distribution network data, and the environmental prediction data are each input into a deep reinforcement learning model to obtain an initial distribution network reconstruction plan output by the deep reinforcement learning model. This includes: inputting the spatiotemporal characteristics, the cross-substation coupling coefficient matrix, the distribution network data, the environmental prediction data, the distributed energy output data, and the user electricity usage behavior data into the deep reinforcement learning model to obtain an initial distribution network reconstruction plan output by the deep reinforcement learning model. In one embodiment, the distribution network reconstruction plan includes a sequence of switch states for all substations in the distribution network (switches need to be opened and closed to adjust the grid topology), charge and discharge power instructions for the energy storage system, and output adjustment coefficients for the distributed energy resources. The multimodal data input into the deep reinforcement learning model is considered, including the spatiotemporal characteristics, the cross-substation coupling coefficient matrix, the distribution network data, the environmental prediction data, the distributed energy output data, and the user electricity usage behavior data. The deep reinforcement learning model can provide a more comprehensive understanding of the electrical connection characteristics of the distribution network and the interactions between substations, which will help further improve the operational stability and reliability of the distribution network under complex working conditions.

[0043] In one embodiment, the present invention defines a state space (referred to as "state") as the historical spatiotemporal characteristics of the distribution network, the historical cross-substation coupling coefficient matrix, historical distribution network data, and historical environmental forecast data. The possible action space (referred to as "action") is defined as the historical distributed energy output data of the distribution network and historical user electricity usage behavior data. The reward function (reward) is defined to maximize overall system performance, which can balance multiple objectives. For example, minimizing voltage deviation, maximizing distributed energy consumption, and optimizing energy storage charging and discharging strategies can be objectives. The deep reinforcement learning model outputs a sequence of substation switch states (switches need to be opened and closed to adjust the grid topology), charge and discharge power commands for the energy storage system, and output regulation coefficients for the distributed energy resources.

[0044] Deep reinforcement learning combines deep neural networks with Q learning, using deep neural networks to approximate the Q-value function. In one embodiment, the deep reinforcement learning model of the embodiment of the present invention is obtained by the following steps: Initialize the parameters of the experience replay pool, deep Q network, and deep reinforcement learning model; Repeat the following steps until the set conditions are achieved: Obtain the status (i.e., sample fault data) and select an action based on the ε-greedy strategy (historical distributed energy output data and historical user electricity usage behavior data, etc.): Execute actions and observe the rewards and new states calculated based on the reward function; Store the experience (state, action, reward, and new state) into the replay pool; Randomly sample a set number of experiences from the replay pool; Calculate the target Q value and predicted Q value based on each sample; Based on the target Q value and predicted Q value of each sample, the loss function is calculated using mean square error. Based on the loss function, the model parameters of the deep Q network are updated through back propagation and an optimizer (such as Adam). The model parameters of the deep Q network are copied to the deep reinforcement learning model every fixed number of steps (for example, every 100 training steps).

[0045] Therefore, the embodiment of the present invention realizes outputting a distribution network reconstruction plan for all substations in the distribution network based on a deep reinforcement learning model.

[0046] Step 500: Determine a distribution network target reconstruction plan based on the local decision plans of multiple substations and the distribution network initial reconstruction plan, so that multiple substations of the distribution network are collaboratively controlled based on the distribution network target reconstruction plan.

[0047] The electronic device can collaboratively consider the local decision plans of multiple substations and the initial distribution network reconstruction plan to determine a target distribution network reconstruction plan. The local decision plan for each substation can be determined by using an edge agent (such as an intelligent control unit integrated in a smart meter, distributed energy inverter, or other device) in the substation. Based on locally collected distribution network operating data (such as local voltage deviation, load fluctuation, etc.) and the distribution network reconstruction plan obtained in the above steps, a fuzzy logic controller is used to quickly make a local response decision to obtain a local decision plan. In the embodiment of the present invention, when there is a conflict between the local decision plans of multiple substations and the initial distribution network reconstruction plan, a conflict resolution strategy is used to determine the target distribution network reconstruction plan. The electronic device issues instructions based on the distribution network target reconstruction plan, and issues control instructions (such as switch opening and closing instructions, reactive compensation device switching instructions, distributed energy output adjustment instructions, etc.) generated by the distribution network target reconstruction plan to the corresponding execution device.

[0048] In one embodiment, a target reconstruction plan of the distribution network is determined based on the local decision plans of multiple substations and the initial reconstruction plan of the distribution network, including: when the initial reconstruction plan of the distribution network conflicts with the local decision plan of the first substation, the initial reconstruction plan of the distribution network is determined as the target reconstruction plan of the distribution network of the first substation; wherein, in the event of an emergency in the first substation, the first substation gives priority to handling the emergency and then executes the target reconstruction plan of the distribution network; the emergency includes a voltage over-limit event and / or an equipment overload event, and the first substation is any one of the multiple substations.

[0049] This embodiment of the present invention employs a conflict resolution strategy when a conflict arises between the initial distribution network reconfiguration plan and the local decision-making plan of the first substation. The global directive of the initial distribution network reconfiguration plan has the highest priority and is executed first. For example, in one implementation scenario, the initial distribution network reconfiguration plan aims to increase the photovoltaic power consumption rate (target: consumption rate > 95%) and instructs substation C to operate the photovoltaic inverter at full power (1MW). However, the local decision-making plan for substation C states that the terminal voltage of substation C is approaching 80% of the upper limit, requiring substation C to derate the photovoltaic inverter. In this case, the initial distribution network reconfiguration plan and the local decision-making plan of substation C conflict. The global directive of the initial distribution network reconfiguration plan has the highest priority, determining the initial distribution network reconfiguration plan as the target distribution network reconfiguration plan, and thus prioritizing the strategy of operating the photovoltaic inverter at full power (1MW) in substation C. Therefore, when a conflict arises between the initial distribution network reconfiguration plan and the local decision-making plans of each substation, this embodiment of the present invention handles the conflict according to the rule that the global directive of the initial distribution network reconfiguration plan takes precedence over the local decision-making plan.

[0050] It should be noted that before substation C prioritizes executing the initial distribution network reconstruction plan, if an emergency occurs. For example, an emergency event occurs in which the voltage exceeds the limit (±5% threshold) and / or the equipment is overloaded (110% rated current). Substation C prioritizes handling the emergency event and then executes the strategy of full power output (1MW) of the photovoltaic inverter in substation C. In this embodiment of the present invention, the local substation prioritizes handling the emergency event before processing the global initial distribution network reconstruction plan, thereby ensuring the safe and stable operation of the power grid in various complex situations.

[0051] In an embodiment of the present invention, the spatiotemporal characteristics, the cross-substation coupling coefficient matrix, the distribution network data, and the environmental prediction data are respectively input into a deep reinforcement learning model to obtain an initial distribution network reconstruction plan output by the deep reinforcement learning model; then, based on the local decision plans of multiple substations and the initial distribution network reconstruction plan, a distribution network target reconstruction plan is determined, so that multiple substations in the distribution network can be collaboratively controlled based on the distribution network target reconstruction plan. In an embodiment of the present invention, efficient cross-substation energy scheduling and precise coordinated voltage control are achieved through the technical means of multimodal data fusion, deep reinforcement learning, and distributed collaborative decision-making, thereby improving the operational stability and reliability of the distribution network under complex working conditions when faced with intermittent and volatile distributed energy and unbalanced loads across substations.

[0052] In other aspects of the embodiments of the present invention, the local decision-making plan of each area is obtained through the following steps: controlling each edge agent in the area to generate a random waiting time through a trusted execution environment; determining the edge agent with the shortest random waiting time as the target edge agent, and determining the decision-making plan of the target edge agent as the local decision-making plan of the area.

[0053] In this embodiment of the present invention, a consensus algorithm is used to achieve efficient coordination among multiple edge agents in each substation. Edge agents generate random waiting times through a trusted execution environment, and the edge agent with the shortest waiting time is prioritized to submit a local decision plan. Each edge agent's local decision plan is based on the substation's distribution network operating data and the initial distribution network reconstruction plan, and is determined using a fuzzy logic controller.

[0054] Specifically, each edge agent in the same substation (such as the intelligent control units integrated in smart meters, distributed energy inverters and other devices) generates a random waiting time through a trusted execution environment. The edge agent with the shortest waiting time submits the local decision-making plan first and becomes the temporary "leader". During the waiting process, other edge agents verify the validity and rationality of the submitted local decision-making plans. If no better proposal is received within the time limit, the local decision-making plan will be accepted to avoid conflicts caused by simultaneous proposals from multiple edge agents. In this way, the embodiment of the present invention ensures that the local decision-making plans of multiple edge agents can be efficiently and consistently executed in a coordinated manner, and can still ensure the normal operation of the system in the face of failures or malicious behaviors of some edge agents, with a fault tolerance rate of 33%.

[0055] In other aspects of embodiments of the present invention, the method for coordinated control of distribution networks across substations further includes: continuously monitoring the operating status of the distribution network, and feeding back the actual execution results and the latest operating data to the electronic device. Based on the feedback data, the control effect is evaluated. If the expected target is not achieved, for example, if the voltage deviation of the distribution network or the distributed energy consumption rate is lower than expected, the data processing and feature extraction steps of steps 200 and 300 are returned to optimize and adjust the model and decision making. Thus, the embodiment of the present invention forms a closed-loop control, thereby continuously improving the effect of cross-substation energy scheduling and voltage coordinated control.

[0056] Device embodiment Please refer to Figure 3 In another aspect, an embodiment of the present invention further provides a cross-zone coordinated control device for a distribution network. The device includes a data acquisition module 301, a model training and optimization module 302, a model evaluation module 303, an upper-level global optimization module (or upper-level global optimizer) 304, and a conflict resolution strategy module 305.

[0057] The data acquisition module is used to acquire sensor data from multiple distribution network substations; the sensor data from these multiple substations includes at least distribution network data and environmental forecast data from all substations. The module has data acquisition capabilities and acquires data from various sensors via a communication network. These sensors are located at key locations in the distribution network, such as busbars, feeders, and distributed energy access points. They can collect real-time, multi-source heterogeneous information, including distribution network operating data (such as voltage, current, and power), environmental forecast data (such as light intensity, temperature, and wind speed), user behavior data (such as power consumption patterns, load curves, and electric vehicle charging patterns), and distributed energy output data.

[0058] The model training and optimization module is used to extract spatiotemporal features based on the distribution network data of all the substations, thereby obtaining spatiotemporal features of multiple substations in the distribution network. Specifically, the model training and optimization module integrates multi-source data from multiple substations to generate a comprehensive and accurate spatiotemporal feature matrix for data feature extraction.

[0059] A model evaluation module is used to construct an inter-area coupling coefficient matrix based on the distribution network data of all the areas; the inter-area coupling coefficient matrix includes the inter-area coupling coefficients of all areas, and the inter-area coupling coefficients represent the sensitivity of the local area voltage fluctuation caused by load changes in other areas.

[0060] The upper-level global optimization module is configured to input the spatiotemporal characteristics, the cross-station coupling coefficient matrix, the distribution network data, and the environmental prediction data into a deep reinforcement learning model, thereby obtaining an initial distribution network reconstruction solution output by the deep reinforcement learning model. Specifically, the upper-level global optimization module outputs a target distribution network reconstruction solution based on the deep reinforcement learning model.

[0061] A conflict resolution strategy module is configured to determine a target distribution network reconstruction plan based on the local decision plans of multiple substations and the initial distribution network reconstruction plan, so that multiple substations in the distribution network can be coordinated and controlled based on the target distribution network reconstruction plan. Specifically, when a decision conflict occurs, it is handled according to the rule that global instructions take precedence over local decision plans, with local decisions taking precedence over emergency situations such as voltage exceeding the limit and / or equipment overload.

[0062] In some embodiments, the cross-area coupling coefficient is calculated based on a voltage change when a voltage fluctuation occurs in the first area, a load change in the remaining areas, and a per-unit value of branch impedance between the first area and the remaining areas; The first substation is any one of all substations in the distribution network; and the remaining substations are all substations except the first substation.

[0063] Optionally, the cross-stage coupling coefficient is calculated using the following formula: ; in, C i represents the cross-station coupling coefficient of the ith station; represents the voltage change of the ith station; represents the load change of the j-th substation; represents the per-unit value of the branch impedance between the i-th substation and the j-th substation; n represents the number of the remaining substations.

[0064] In some embodiments, the distribution network data of all substations include topological structure data and distribution network operation data of all substations; and the spatiotemporal feature extraction based on the distribution network data of all substations to obtain the spatiotemporal features of multiple substations of the distribution network includes: Graph convolutional neural network is used to extract spatial features from the topological structure data of all substations to obtain spatial features; Use deep learning models to extract time features from the distribution network operation data of all substations to obtain time features; The spatial features and the temporal features are fused using an attention fusion method to obtain the spatiotemporal features of the multiple substations in the distribution network.

[0065] In some embodiments, the local decision solution for each station is obtained by the following steps: Controlling each edge agent in the station area to generate a random waiting time through a trusted execution environment; The edge agent with the shortest random waiting time is determined as the target edge agent, and the decision-making plan of the target edge agent is determined as the local decision-making plan of the station area.

[0066] In some embodiments, the local decision-making plan of each edge intelligent agent is based on the distribution network operation data of the substation and the initial reconstruction plan of the distribution network, and is obtained by using a fuzzy logic controller to make a decision.

[0067] In some embodiments, the sensor data of the plurality of substations include distribution network data, environmental prediction data, distributed energy output data, and user electricity usage behavior data of all substations; The step of inputting the spatiotemporal characteristics, the cross-station coupling coefficient matrix, the distribution network data, and the environmental prediction data into a deep reinforcement learning model to obtain an initial distribution network reconstruction plan output by the deep reinforcement learning model includes: The spatiotemporal characteristics, the cross-station coupling coefficient matrix, the distribution network data, the environmental prediction data, the distributed energy output data, and the user electricity consumption behavior data are respectively input into a deep reinforcement learning model to obtain an initial distribution network reconstruction plan output by the deep reinforcement learning model.

[0068] In some embodiments, determining the target distribution network reconstruction plan based on the local decision plans of the multiple substations and the initial distribution network reconstruction plan includes: When the distribution network initial reconstruction plan conflicts with the local decision-making plan of the first substation, determining the distribution network initial reconstruction plan as the distribution network target reconstruction plan of the first substation; Among them, in the event of an emergency occurring in the first substation, the first substation will give priority to handling the emergency and then execute the distribution network target reconstruction plan; the emergency event includes a voltage over-limit event and / or an equipment overload event, and the first substation is any one of the multiple substations.

[0069] The distribution network cross-station collaborative control device includes a processor and a memory. The above-mentioned data acquisition module 301, model training and optimization module 302, model evaluation module 303, upper-level global optimization module 304, and conflict resolution strategy module 305 are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize corresponding functions.

[0070] The processor includes a kernel, which retrieves the corresponding program unit from the memory. There can be one or more kernels.

[0071] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0072] The cross-station coordinated control device of the distribution network is designed to solve the dynamic optimization and coordinated control problems of the distribution network under the high proportion of distributed energy access. Through multimodal data fusion, deep reinforcement learning and distributed collaborative decision-making technology, it can achieve efficient cross-station energy scheduling and precise coordinated voltage control. Figure 4 In one embodiment, the cross-station collaborative control device of the distribution network adopts a four-layer architecture design, including a perception layer, a platform layer, an application layer and an execution layer. The perception layer includes the above-mentioned data acquisition module, a data processing module and a communication interface. The platform layer includes the above-mentioned model training and optimization module, the above-mentioned model evaluation module, the security and privacy protection module and the data twin module. The application layer includes the above-mentioned upper-layer global optimization module, the lower-layer edge intelligent agent module, the consensus algorithm module and the above-mentioned conflict resolution strategy module. Among them, the upper-layer global optimization module and the lower-layer edge intelligent agent module constitute a two-layer optimization decision module. Execution layer: refers to execution devices such as feeder switches, reactive compensation devices, and load controllers. Receive the distribution network target reconstruction method from the application layer and perform specific control operations.

[0073] The data processing module performs preliminary processing on the data collected by the data acquisition module, including data cleaning and format conversion. The communication interface transmits the collected data to the platform layer via high-speed communication networks (such as 5G or fiber optics). The security and privacy protection module ensures the security of multi-source data transmission through encrypted communication protocols, preventing data theft or tampering. The data twin module utilizes advanced simulation technology to construct a virtual model that is highly consistent with the physical distribution network, mapping the physical grid state in real time. It can also simulate the distribution network's operation under various extreme scenarios (such as typhoons, heavy rains, and load surges), providing a virtual environment for developing and validating optimization strategies. The lower-layer edge agent module is responsible for local decision-making, enabling rapid local responses based on real-time data on local voltage deviations and load fluctuations to generate local decision solutions. The consensus algorithm module enables efficient coordination among multiple edge agents. Edge agents generate random waiting times through a trusted execution environment, with nodes with the shortest waiting times prioritized for decision submission.

[0074] Therefore, the distribution network cross-station collaborative control device of the embodiment of the present invention solves the shortcomings of existing distribution network technology in dynamic adaptability, multi-source data utilization and cross-station collaborative control, realizes real-time perception and precise regulation of the operating status of the distribution network, improves the distributed energy absorption rate, ensures voltage quality, enhances the flexibility and reliability of the distribution network under complex working conditions, and improves the overall operating efficiency and economy.

[0075] Figure 5 An example of a physical structure diagram of an electronic device is shown below. Figure 5As shown, the electronic device may include: a processor (processor) 510, a communication interface (Communications Interface) 520, a memory (memory) 530 and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other through the communication bus 540. The processor 510 can call the logic instructions in the memory 530 to execute the distribution network cross-zone collaborative control method, which includes: obtaining sensor data of multiple zones in the distribution network; the sensor data of the multiple zones include at least distribution network data and environmental prediction data of all zones; performing spatiotemporal feature extraction based on the distribution network data of all zones to obtain spatiotemporal features of the multiple zones in the distribution network; constructing a cross-zone coupling coefficient matrix based on the distribution network data of all zones; the cross-zone coupling coefficient matrix includes the cross-zone coupling coefficients of all zones, and the cross-zone coupling coefficient characterizes the sensitivity of the local zone voltage fluctuation caused by load changes in other zones; inputting the spatiotemporal features, the cross-zone coupling coefficient matrix, the distribution network data and the environmental prediction data into a deep reinforcement learning model respectively to obtain an initial distribution network reconstruction plan output by the deep reinforcement learning model; determining a distribution network target reconstruction plan based on the local decision plans of the multiple zones and the initial distribution network reconstruction plan, so that the multiple zones in the distribution network are collaboratively controlled based on the distribution network target reconstruction plan.

[0076] Furthermore, the logic instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0077] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a machine-readable storage medium. When the computer program is executed by a processor, the computer can execute a distribution network cross-zone collaborative control method, the method including: obtaining sensor data of multiple zones in the distribution network; the sensor data of the multiple zones include at least distribution network data and environmental prediction data of all zones; performing spatiotemporal feature extraction based on the distribution network data of all zones to obtain spatiotemporal features of the multiple zones in the distribution network; constructing a cross-zone coupling coefficient matrix based on the distribution network data of all zones; the cross-zone coupling coefficient matrix includes cross-zone coupling coefficients of all zones, and the cross-zone coupling coefficient characterizes the sensitivity of local zone voltage fluctuations caused by load changes in other zones; inputting the spatiotemporal features, the cross-zone coupling coefficient matrix, the distribution network data and the environmental prediction data into a deep reinforcement learning model respectively to obtain an initial distribution network reconstruction plan output by the deep reinforcement learning model; determining a distribution network target reconstruction plan based on the local decision plans of the multiple zones and the initial distribution network reconstruction plan, so that the multiple zones in the distribution network are collaboratively controlled based on the distribution network target reconstruction plan.

[0078] On the other hand, the present invention also provides a machine-readable storage medium having a computer program stored thereon, which is implemented by a processor to perform a method for coordinated control of a distribution network across substations, the method comprising: obtaining sensor data of multiple substations in the distribution network; the sensor data of the multiple substations include at least distribution network data and environmental prediction data of all substations; extracting spatiotemporal features based on the distribution network data of all substations to obtain spatiotemporal features of the multiple substations in the distribution network; constructing a cross-substation coupling coefficient matrix based on the distribution network data of all substations; the cross-substation coupling coefficient matrix includes cross-substation coupling coefficients of all substations, and the cross-substation coupling coefficient characterizes the sensitivity of the local substation voltage fluctuation caused by load changes in other substations; inputting the spatiotemporal features, the cross-substation coupling coefficient matrix, the distribution network data and the environmental prediction data into a deep reinforcement learning model respectively to obtain an initial distribution network reconstruction plan output by the deep reinforcement learning model; determining a target distribution network reconstruction plan based on the local decision plans of the multiple substations and the initial distribution network reconstruction plan, so that the multiple substations in the distribution network are coordinated and controlled based on the target distribution network reconstruction plan.

[0079] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0080] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for coordinated control of distribution networks across substations, characterized in that: include: Obtain sensor data from multiple substations in the distribution network; The sensor data of the plurality of substations include at least the distribution network data and environmental prediction data of all substations; Extracting spatiotemporal features based on the distribution network data of all the substations to obtain spatiotemporal features of multiple substations in the distribution network; Constructing an inter-substation coupling coefficient matrix based on the distribution network data of all substations; The cross-area coupling coefficient matrix includes cross-area coupling coefficients of all areas, and the cross-area coupling coefficients represent the sensitivity of the local area voltage fluctuation caused by the load change of other areas; Inputting the spatiotemporal characteristics, the cross-station coupling coefficient matrix, the distribution network data, and the environmental prediction data into a deep reinforcement learning model respectively, and obtaining an initial distribution network reconstruction plan output by the deep reinforcement learning model; Based on the local decision plans of multiple substations and the distribution network initial reconstruction plan, a distribution network target reconstruction plan is determined, so that multiple substations of the distribution network are coordinated and controlled based on the distribution network target reconstruction plan.

2. The method for coordinated control of distribution network across substations according to claim 1, characterized in that: The cross-substation coupling coefficient is calculated based on the voltage change when the voltage fluctuation occurs in the first substation, the load change of the other substations, and the per-unit value of the branch impedance between the first substation and the other substations; The first substation is any one of all substations in the distribution network; and the remaining substations are all substations except the first substation.

3. The method for coordinated control of distribution network across substations according to claim 2, characterized in that: The cross-stage coupling coefficient is calculated by the following formula: ； in, C i represents the cross-station coupling coefficient of the ith station; represents the voltage change of the ith station; represents the load change of the j-th substation; represents the per-unit value of the branch impedance between the i-th substation and the j-th substation; n represents the number of the remaining substations.

4. The method for coordinated control of distribution network across substations according to claim 1, characterized in that: The distribution network data of all substations includes topological structure data and distribution network operation data of all substations; the spatiotemporal feature extraction based on the distribution network data of all substations to obtain spatiotemporal features of multiple substations of the distribution network includes: Graph convolutional neural network is used to extract spatial features from the topological structure data of all substations to obtain spatial features; Use deep learning models to extract time features from distribution network operation data of all substations to obtain time features; The spatial features and the temporal features are fused using an attention fusion method to obtain the spatiotemporal features of the multiple substations in the distribution network.

5. The cross-station coordinated control method of the distribution network according to claim 4, characterized in that: The local decision-making plan for each substation is obtained through the following steps: Controlling each edge agent in the station area to generate a random waiting time through a trusted execution environment; The edge agent with the shortest random waiting time is determined as the target edge agent, and the decision-making plan of the target edge agent is determined as the local decision-making plan of the station area.

6. The cross-station coordinated control method of the distribution network according to claim 5, characterized in that: The local decision-making plan of each edge intelligent agent is based on the distribution network operation data of the substation and the initial reconstruction plan of the distribution network, and is obtained by using a fuzzy logic controller to make a decision.

7. The method for coordinated control of distribution network across substations according to claim 1, characterized in that: The sensor data of the multiple substations include distribution network data, environmental forecast data, distributed energy output data, and user electricity consumption behavior data of all substations; The step of inputting the spatiotemporal characteristics, the cross-station coupling coefficient matrix, the distribution network data, and the environmental prediction data into a deep reinforcement learning model to obtain an initial distribution network reconstruction plan output by the deep reinforcement learning model includes: The spatiotemporal characteristics, the cross-station coupling coefficient matrix, the distribution network data, the environmental prediction data, the distributed energy output data, and the user electricity consumption behavior data are respectively input into a deep reinforcement learning model to obtain an initial distribution network reconstruction plan output by the deep reinforcement learning model.

8. The method for coordinated control of distribution network across substations according to claim 1, characterized in that: The determining of the target distribution network reconstruction plan based on the local decision-making plans of the multiple substations and the initial distribution network reconstruction plan includes: When the distribution network initial reconstruction plan conflicts with the local decision-making plan of the first substation, determining the distribution network initial reconstruction plan as the distribution network target reconstruction plan of the first substation; Among them, in the event of an emergency occurring in the first substation, the first substation will give priority to handling the emergency and then execute the distribution network target reconstruction plan; the emergency event includes a voltage over-limit event and / or an equipment overload event, and the first substation is any one of the multiple substations.

9. A cross-station coordinated control device for a distribution network, characterized in that: include: Data acquisition module, used to obtain sensor data from multiple substations in the distribution network; The sensor data of the plurality of substations include at least the distribution network data and environmental prediction data of all substations; A model training and optimization module is used to extract spatiotemporal features based on the distribution network data of all the substations to obtain spatiotemporal features of multiple substations in the distribution network; A model evaluation module, configured to construct an inter-substation coupling coefficient matrix based on the distribution network data of all the substations; The cross-area coupling coefficient matrix includes cross-area coupling coefficients of all areas, and the cross-area coupling coefficients represent the sensitivity of the local area voltage fluctuation caused by the load change of other areas; An upper-layer global optimization module is used to input the spatiotemporal characteristics, the cross-station coupling coefficient matrix, the distribution network data, and the environmental prediction data into a deep reinforcement learning model, respectively, to obtain an initial distribution network reconstruction plan output by the deep reinforcement learning model; The conflict resolution strategy module is used to determine the distribution network target reconstruction plan based on the local decision plans of multiple substations and the distribution network initial reconstruction plan, so that multiple substations of the distribution network can be coordinated and controlled based on the distribution network target reconstruction plan.

10. The cross-station coordinated control device for a distribution network according to claim 9, characterized in that: The cross-substation coupling coefficient is calculated based on the voltage change when the voltage fluctuation occurs in the first substation, the load change of the other substations, and the per-unit value of the branch impedance between the first substation and the other substations; The first substation is any one of all substations in the distribution network; and the remaining substations are all substations except the first substation.

11. The cross-station coordinated control device for a distribution network according to claim 10, characterized in that: The cross-stage coupling coefficient is calculated by the following formula: ； in, C i represents the cross-station coupling coefficient of the ith station; represents the voltage change of the ith station; represents the load change of the j-th substation; represents the per-unit value of the branch impedance between the i-th substation and the j-th substation; n represents the number of the remaining substations.

12. The cross-station coordinated control device for a distribution network according to claim 9, characterized in that: The distribution network data of all substations includes topological structure data and distribution network operation data of all substations; the spatiotemporal feature extraction based on the distribution network data of all substations to obtain spatiotemporal features of multiple substations of the distribution network includes: Graph convolutional neural network is used to extract spatial features from the topological structure data of all substations to obtain spatial features; Use deep learning models to extract time features from distribution network operation data of all substations to obtain time features; The spatial features and the temporal features are fused using an attention fusion method to obtain the spatiotemporal features of the multiple substations in the distribution network.

13. The cross-station coordinated control device for a distribution network according to claim 12, characterized in that: The local decision-making plan for each substation is obtained through the following steps: Controlling each edge agent in the station area to generate a random waiting time through a trusted execution environment; The edge agent with the shortest random waiting time is determined as the target edge agent, and the decision-making plan of the target edge agent is determined as the local decision-making plan of the station area.

14. The cross-station coordinated control device for a distribution network according to claim 13, characterized in that: The local decision-making plan of each edge intelligent agent is based on the distribution network operation data of the substation and the initial reconstruction plan of the distribution network, and is obtained by using a fuzzy logic controller to make a decision.

15. The cross-station coordinated control device for a distribution network according to claim 9, characterized in that: The sensor data of the multiple substations include distribution network data, environmental forecast data, distributed energy output data, and user electricity consumption behavior data of all substations; The step of inputting the spatiotemporal characteristics, the cross-station coupling coefficient matrix, the distribution network data, and the environmental prediction data into a deep reinforcement learning model to obtain an initial distribution network reconstruction plan output by the deep reinforcement learning model includes: The spatiotemporal characteristics, the cross-station coupling coefficient matrix, the distribution network data, the environmental prediction data, the distributed energy output data, and the user electricity consumption behavior data are respectively input into a deep reinforcement learning model to obtain an initial distribution network reconstruction plan output by the deep reinforcement learning model.

16. The cross-station coordinated control device for a distribution network according to claim 9, characterized in that: The determining of the target distribution network reconstruction plan based on the local decision-making plans of the multiple substations and the initial distribution network reconstruction plan includes: When the distribution network initial reconstruction plan conflicts with the local decision-making plan of the first substation, determining the distribution network initial reconstruction plan as the distribution network target reconstruction plan of the first substation; Among them, in the event of an emergency occurring in the first substation, the first substation will give priority to handling the emergency and then execute the distribution network target reconstruction plan; the emergency event includes a voltage over-limit event and / or an equipment overload event, and the first substation is any one of the multiple substations.

17. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the distribution network cross-station coordinated control method according to any one of claims 1 to 8 is implemented.

18. A machine-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the distribution network cross-station coordinated control method according to any one of claims 1 to 8 is implemented.

19. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the distribution network cross-station coordinated control method according to any one of claims 1 to 8 is implemented.

Images