CN-122026414-A - Voltage regulation and control method and device for power distribution network and electronic equipment

CN122026414ACN 122026414 ACN122026414 ACN 122026414ACN-122026414-A

Abstract

The invention discloses a voltage regulation and control method and device of a power distribution network and electronic equipment. The method comprises the steps of obtaining a power distribution network topology model, dividing a plurality of regulation areas based on electrical relevance among nodes, constructing a layered control framework comprising cloud cooperative units and edge control units of all areas, determining a cloud global decision model and edge local decision models based on the topology and the framework, generating cloud cooperative strategies according to the global model and a full network state, generating edge local regulation strategies according to the local models and a real-time state, generating and sending area activation instructions to corresponding edge units according to the cloud strategies, and generating and executing local regulation instructions according to the instructions, the local states and the strategies by the edge units. The invention solves the technical problem of complex voltage regulation and control for high-permeability photovoltaic and energy storage in the related technology.

Inventors

XU CHONGCHONG
WU YUTONG
YANG BO
ZHAO BIN
XIE LUHONG
SUN ZHENGLONG
HAO SHUYU

Assignees

国网北京市电力公司
东北电力大学

Dates

Publication Date: 20260512
Application Date: 20260123

Claims (10)

1. A voltage regulation method for a power distribution network, comprising: acquiring a power distribution network topology model corresponding to a power distribution network; dividing the power distribution network into a plurality of regulation and control areas based on electrical relevance among a plurality of nodes in the power distribution network topology model; constructing a layered control architecture corresponding to the plurality of regulation areas, wherein the layered control architecture comprises a cloud cooperative unit deployed at a cloud and an edge control unit deployed locally in the corresponding regulation area; determining a global decision model corresponding to the cloud cooperative unit and a local decision model corresponding to a plurality of edge control units respectively based on the power distribution network topology model and the layered control architecture; according to the global decision model and the whole network running state, a cloud collaborative strategy is obtained, and according to the corresponding local decision model and the local real-time running state, a local regulation strategy corresponding to the plurality of edge control units is obtained; Generating region activation instructions respectively corresponding to the plurality of edge control units according to the cloud collaborative strategy, and sending the corresponding region activation instructions to the corresponding edge control units so that the corresponding edge control units can generate and execute the corresponding local regulation and control instructions according to the received region activation instructions and the local regulation and control strategy according to the real-time operation state of the local regulation and control region.
2. The method of claim 1, wherein dividing the power distribution network into a plurality of regulatory regions based on electrical correlations among a plurality of nodes in the power distribution network topology model comprises: Determining the sensitivity of voltage between any two nodes in the power distribution network to power change according to the power distribution network topology model, and obtaining a plurality of sensitivities corresponding to the power distribution network; constructing a sensitivity matrix of voltage to power variation of all nodes according to the plurality of sensitivities; according to the sensitivity matrix, dividing the nodes into a plurality of areas by adopting a clustering algorithm to obtain the plurality of regulation areas, wherein the clustering algorithm is configured to enable the sensitivity similarity among the nodes in the same area after division to be higher than a preset coupling value, and each area at least comprises one adjustable resource node, and the adjustable resource node comprises at least one of an energy storage node and a photovoltaic node.
3. The method of claim 1, wherein determining a global decision model corresponding to the cloud collaboration unit based on the power distribution network topology model and the hierarchical control architecture comprises: constructing a global Markov decision process model as the global decision model according to the power distribution network topology model, wherein the global decision model comprises a first state space, a first action space and a first rewarding function; Determining a first state space of the global decision model, wherein the first state space comprises node voltage amplitude parameters, state of charge parameters of an energy storage system, active power parameters of a photovoltaic system and load power parameters in the power distribution network; determining a first action space of the global decision model, wherein the first action space is configured as a region activation selection vector for indicating a local regulation strategy for activating or deactivating each regulation region at each decision moment; a first reward function of the global decision model is determined, wherein the first reward function is configured to minimize a total deviation of the power distribution network voltage from a target voltage value.
4. The method of claim 1, wherein determining a local decision model for each of a plurality of edge control units based on the power distribution network topology model and the hierarchical control architecture comprises: determining region nodes and adjustable equipment information respectively corresponding to a plurality of regulation and control regions, wherein the adjustable equipment information comprises information corresponding to adjustable resource nodes; According to the regional node and the adjustable equipment information respectively corresponding to the plurality of regulation and control regions, respectively constructing a local Markov decision process model for each edge control unit as the local decision model; Determining a second state space of each local decision model, wherein the second state space comprises node voltage amplitude parameters in a corresponding regulation area, state of charge parameters of an energy storage system in the corresponding area, active power parameters of a photovoltaic system in the corresponding area and load power parameters in the corresponding area; determining a second action space of each local decision model, wherein the second action space comprises a power control set value of an adjustable resource node in a corresponding region; A second reward function for each local decision model is determined, wherein the second reward function includes a first function term that minimizes voltage deviation within the corresponding region and a second function term that minimizes a cost penalty for the energy storage action.
5. The method according to claim 1, further comprising, before determining a global decision model corresponding to the cloud coordination unit and a local decision model corresponding to each of a plurality of edge control units based on the distribution network topology model and the hierarchical control architecture: Performing joint alternate training on the initial global decision model and the initial local decision model to generate a trained global decision model and a trained local decision model, wherein the joint alternate training alternately executes edge parallel strategy updating and cloud collaborative strategy updating in each round; The edge parallel strategy updating comprises the steps that each edge control unit independently collects sample track data in a power distribution network environment of a corresponding regulation and control area based on a current strategy of a corresponding local decision model, estimates a first updating direction of strategy parameters by using a guided type alternative gradient evolution strategy, and performs near-end constraint processing on the first updating direction by using a near-end strategy optimization algorithm so as to complete gradient-free updating of the strategy parameters in the corresponding local decision model; The cloud cooperation strategy updating comprises the steps that under the condition that one round of edge parallel strategy updating is completed, the cloud cooperation unit collects global sample track data in the power distribution network based on the current strategy of the global decision model, adopts the guided alternative gradient evolution strategy to estimate a second updating direction of strategy parameters, and adopts a near-end strategy optimization algorithm to carry out near-end constraint processing on the second updating direction so as to complete gradient-free updating of the strategy parameters in the global decision model.
6. The method of claim 5, further comprising, prior to jointly and alternately training the initial global decision model and the initial local decision model: constructing a multi-scene operation data set based on the historical operation data and the typical scene data; Dividing the multi-scenario running data set into a training scenario subset and a testing scenario subset, wherein after each round of alternating iteration of the joint alternating training, the testing scenario subset is used for evaluating the performances of a current global decision model and a local decision model; if the performance index reaches the preset convergence threshold, training is completed, the trained global decision model and the trained local decision model are output, and otherwise, the next round of alternate iteration is continued.
7. The method according to any one of claims 1 to 6, wherein generating the region activation instructions corresponding to the plurality of edge control units respectively according to the cloud collaboration policy, and issuing the corresponding region activation instructions to the corresponding edge control units, so that the corresponding edge control units generate and execute the corresponding local regulation instructions according to the corresponding local regulation policies according to the received region activation instructions and the real-time running states of the local regulation regions, and the local regulation policies, including: the cloud cooperative unit collects the running state of the whole network in a first period, generates the region activation instruction based on a global decision model and sends the region activation instruction to each edge control unit; the plurality of edge control units acquire local real-time running states of the corresponding regulation and control areas respectively in a second period, wherein the second period is shorter than the first period; generating and executing equipment control instructions based on the local decision model and the local real-time running state in each second period under the condition that the area activation instruction received by the corresponding edge control unit indicates an activation state; And under the condition that the area activation instruction received by the corresponding edge control unit indicates an inactive state, generating a device control instruction by adopting a preset local bottom protection control strategy according to the local real-time running state and executing the device control instruction.
8. A voltage regulation device of a power distribution network, comprising: The acquisition module is used for acquiring a power distribution network topology model corresponding to the power distribution network; the region dividing module is used for dividing the power distribution network into a plurality of regulation and control regions based on the electrical relevance among a plurality of nodes in the power distribution network topology model; The hierarchical control architecture module is used for constructing a hierarchical control architecture corresponding to the plurality of regulation areas, wherein the hierarchical control architecture comprises a cloud cooperative unit deployed at a cloud and an edge control unit deployed locally in the corresponding regulation area; The model building module is used for determining a global decision model corresponding to the cloud cooperative unit and a local decision model corresponding to the edge control units respectively based on the power distribution network topology model and the layered control architecture; The strategy generation module is used for obtaining a cloud collaborative strategy according to the global decision model and the whole network running state, and obtaining local regulation strategies respectively corresponding to the plurality of edge control units according to the corresponding local decision model and the local real-time running state; the control module generates region activation instructions respectively corresponding to the plurality of edge control units according to the cloud collaborative strategy, and transmits the corresponding region activation instructions to the corresponding edge control units, so that the corresponding edge control units can control the real-time running state of the region according to the received region activation instructions, and the local control strategy can generate and execute the corresponding local control instructions according to the corresponding local control strategy.
9. An electronic device, comprising: A processor; a memory for storing the processor-executable instructions; wherein the processor is configured to execute the instructions to implement the voltage regulation method of the power distribution network of any one of claims 1 to 7.
10. A computer readable storage medium, characterized in that instructions in the computer readable storage medium, when executed by a processor of an electronic device, enable the electronic device to perform the voltage regulation method of the power distribution network according to any one of claims 1 to 7.

Description

Voltage regulation and control method and device for power distribution network and electronic equipment Technical Field The invention relates to the field of power distribution networks, in particular to a voltage regulation method and device of a power distribution network and electronic equipment. Background In recent years, photovoltaic (PV) and Energy Storage Systems (ESS) are accessed on a large scale in an active power distribution network, and the trend and the operation characteristics of the power distribution side are significantly changed. The output of the distributed photovoltaic has strong randomness and volatility, and the load uncertainty is superposed, so that the power distribution network needs to deal with the problems of energy fluctuation and voltage out-of-limit on a larger space-time scale. The ESS plays a key role in the aspects of slow release intermittence, peak clipping and valley filling and net load balance, but the voltage regulation problem is more complicated by the coupling control of the ESS and the multipoint photovoltaic and load. In the prior art, a model prediction control method needs a complete and accurate physical model and disturbance prediction. In a medium-large-scale power distribution network with multiple types of photovoltaic/energy storage and complex topology, the model acquisition and maintenance cost is high, the model is sensitive to parameter uncertainty, engineering application is limited, although the robust optimization method can resist uncertainty, strong conservation is generally introduced, excessive constraint is easy to generate, an operation point close to the optimal is difficult to obtain, economy and adjustability are affected, centralized Deep Reinforcement Learning (DRL) shows potential on voltage control, a centralized scheme depends on high-bandwidth communication, the state dimension is high, the network scale is large, the training time is long, convergence efficiency, robustness and on-line deployment are difficult to achieve in the large-scale power distribution network, and single-point failure and data privacy/bandwidth bottleneck problems exist at the same time. In view of the above problems, no effective solution has been proposed at present. Disclosure of Invention The embodiment of the invention provides a voltage regulation and control method and device for a power distribution network and electronic equipment, and aims to at least solve the technical problem that voltage regulation and control for high-permeability photovoltaic and energy storage are complex in the related technology. According to one aspect of the embodiment of the invention, a voltage regulation and control method of a power distribution network is provided, which comprises the steps of obtaining a power distribution network topology model corresponding to the power distribution network, dividing the power distribution network into a plurality of regulation and control areas based on electrical relevance among a plurality of nodes in the power distribution network topology model, constructing a layered control architecture corresponding to the plurality of regulation and control areas, wherein the layered control architecture comprises cloud cooperative units deployed at a cloud end and edge control units deployed locally in the corresponding regulation and control areas, determining a global decision model corresponding to the cloud cooperative units and a local decision model corresponding to the plurality of edge control units respectively based on the power distribution network topology model and the layered control architecture, obtaining a cloud end cooperative strategy according to the global decision model and a full network running state, obtaining a local regulation and control strategy corresponding to the plurality of edge control units respectively, generating an area activation instruction corresponding to the plurality of edge control units according to the cloud end cooperative strategy, and sending the corresponding area activation instruction to the corresponding edge control units, and executing the local regulation and control strategy according to the local regulation and control instruction. The power distribution network is divided into a plurality of regulation areas based on electrical relevance among a plurality of nodes in a power distribution network topology model, the power distribution network comprises the steps of determining sensitivity of voltage between any two nodes in the power distribution network to power change according to the power distribution network topology model to obtain a plurality of sensitivities corresponding to the power distribution network, constructing a sensitivity matrix of voltage of all nodes to power change according to the plurality of sensitivities, dividing the nodes into a plurality of areas according to the sensitivity matrix by adopting a clustering algorithm to obtain the plurality of r