CN-121997752-A - Design optimization method and system for power supply circuit of permanent magnet recloser based on deep learning

CN121997752ACN 121997752 ACN121997752 ACN 121997752ACN-121997752-A

Abstract

The application relates to the technical field of power supply circuits, and discloses a design optimization method and system for a power supply circuit of a permanent magnet recloser based on deep learning. The method comprises the steps of collecting electrical parameters of all nodes of a permanent magnet recloser by an intelligent sensor of a power distribution network to form a power supply state data set, extracting power supply path characteristics by a special topology sensing network of the permanent magnet recloser to generate a topology characteristic matrix, carrying out multi-objective constraint reconstruction based on the characteristic matrix to obtain a circuit topology reconstruction scheme, optimizing the power supply parameters by a deep reinforcement learning algorithm to obtain power supply circuit optimization configuration, and generating intelligent power supply control signals by closed loop adjustment processing. The application solves the technical problem that the design of the power supply circuit of the existing permanent magnet recloser lacks intelligent topology sensing, multi-objective coordination optimization and self-adaptive parameter adjustment capability.

Inventors

ZHANG XINHONG
SUN ZHIYIN
ZHANG LONGBIAO
LI HAO
MA YINHUA

Assignees

中宝电气有限公司

Dates

Publication Date: 20260508
Application Date: 20260128

Claims (10)

1. The design optimization method for the power supply circuit of the permanent magnet recloser based on deep learning is characterized by comprising the following steps of: The method comprises the steps that multidimensional electrical parameter acquisition processing is carried out on each power supply node of a permanent magnet recloser through an intelligent sensor of a power distribution network, and a power supply state data set containing node voltage amplitude, branch current and power factor is obtained; Inputting the power supply state data set into a special topology sensing network of a permanent magnet recloser to perform power supply path feature extraction processing to obtain a power supply path feature matrix containing local power supply relevance and global network topology; Performing multi-objective constraint reconstruction processing on the power supply circuit topology of the permanent magnet recloser according to the power supply path characteristic matrix to obtain a circuit topology reconstruction scheme comprising an optimal power supply path combination and a load distribution strategy; Inputting the circuit topology reconstruction scheme into a deep reinforcement learning algorithm to perform power supply parameter optimizing treatment to obtain power supply circuit optimizing configuration containing an optimal power supply path switching strategy and voltage regulation parameters; And performing closed-loop adjustment processing on the real-time power supply parameters of the permanent magnet recloser according to the power supply circuit optimal configuration to obtain an intelligent power supply control signal containing a fault prediction result and a power supply switching instruction.
2. The method for optimizing the design of the power supply circuit of the permanent magnet recloser based on deep learning according to claim 1, wherein the step of performing multidimensional electric parameter acquisition processing on each power supply node of the permanent magnet recloser by using the intelligent power distribution network sensor to obtain a power supply state data set containing node voltage amplitude, branch current and power factor comprises the following steps: Performing self-adaptive time sequence segmentation processing on the original electrical signals of all power supply nodes of the power distribution network based on a permanent magnet recloser action response window to obtain segmented electrical data containing pre-reclosing precursor features and post-reclosing steady-state features; Inputting the segmented electrical data into a power supply node relevance analysis algorithm to perform topology relevance screening processing to obtain a core power supply node electrical parameter set with redundant node information removed; Carrying out importance labeling treatment on the core power supply node electrical parameter set according to a permanent magnet recloser power supply path weight distribution strategy to obtain layered electrical characteristic data comprising a main power supply path and a standby power supply path; And carrying out working condition grouping processing on the layered electrical characteristic data based on a power supply state similarity clustering method to obtain a power supply state data set for training a deep learning model.
3. The deep learning-based permanent magnet recloser power supply circuit design optimization method according to claim 1, wherein the step of inputting the power supply state data set into a permanent magnet recloser dedicated topology aware network to perform power supply path feature extraction processing to obtain a power supply path feature matrix including local power supply relevance and global network topology comprises the steps of: inputting the power supply state data set into a power supply adjacent matrix construction layer of a special topology perception network of the permanent magnet recloser to perform node association degree calculation processing to obtain a dynamic weighting adjacent matrix based on the electric distance and the fault propagation probability; Carrying out small-range power supply topology convolution processing on a local feature extraction layer of a special topology perception network of the permanent magnet recloser based on the dynamic weighting adjacency matrix to obtain a local topology feature map containing power supply association features of adjacent nodes; inputting the partial topological feature map into a medium receptive field expansion layer of a special topological sensing network of a permanent magnet recloser to perform medium distance power supply network convolution processing to obtain a medium topological feature map containing power supply coordination relations in an area; Carrying out large-range power supply network convolution processing on a global receptive field expansion layer of a special topology perception network of the permanent magnet recloser according to the medium topology feature map to obtain a global topology feature map containing cross-region power supply influence relation; And carrying out multi-scale feature fusion processing on the local topological feature map, the medium topological feature map and the global topological feature map based on a power supply path attention weight mechanism to obtain a power supply path feature matrix containing local power supply relevance and global network topology.
4. The method for optimizing power supply circuit design of permanent magnet recloser based on deep learning according to claim 3, wherein the step of inputting the power supply state data set into a power supply adjacency matrix construction layer of a topology aware network special for the permanent magnet recloser to perform node association calculation processing to obtain a dynamic weighted adjacency matrix based on electric distance and fault propagation probability comprises the following steps: Performing electrical distance measurement calculation processing on each power supply node in the power supply state data set based on the switching state transfer characteristic of the permanent magnet recloser to obtain an inter-node electrical distance matrix reflecting the influence range of the permanent magnet recloser; performing fault propagation path analysis processing on the electric distance matrix according to the fault response time sequence of the permanent magnet recloser to obtain a fault propagation probability matrix containing the probability of diffusing faults from a source node to a target node; Inputting the electric distance matrix and the fault propagation probability matrix into a weighted fusion algorithm to perform weight coefficient distribution processing to obtain node importance weight vectors showing the power supply priority of the permanent magnet recloser; Dynamically adjusting the node importance weight vector based on the permanent magnet recloser power supply switching logic to obtain a self-adaptive weight coefficient changing along with the running state of the permanent magnet recloser; and carrying out weighted assignment processing on the power supply node adjacency relation according to the self-adaptive weight coefficient to obtain a dynamic weighted adjacency matrix based on the electric distance and the fault propagation probability.
5. The deep learning-based permanent magnet recloser power supply circuit design optimization method according to claim 1, wherein the performing multi-objective constraint reconstruction processing on the permanent magnet recloser power supply circuit topology according to the power supply path feature matrix to obtain a circuit topology reconstruction scheme including an optimal power supply path combination and a load distribution strategy comprises: Performing power supply path feasibility screening processing on the power supply path feature matrix based on the power supply reliability constraint condition of the permanent magnet recloser to obtain a candidate power supply path set meeting the minimum action frequency requirement of the permanent magnet recloser; Inputting the candidate power supply path set into a multi-objective optimization algorithm to perform power supply quality constraint analysis processing to obtain voltage deviation control constraint and harmonic content control constraint; carrying out load distribution strategy calculation processing on the voltage deviation control constraint and the harmonic content control constraint according to the load bearing capacity limit of the permanent magnet recloser to obtain the optimal power distribution proportion of each power supply branch; performing power supply time sequence coordination optimization processing on the optimal power distribution proportion based on the power supply path switching time constraint of the permanent magnet recloser to obtain a main and standby power supply path switching sequence and a switching time interval; And carrying out topology reconstruction integration processing on the switching sequence of the main power supply paths and the standby power supply paths and the optimal power distribution proportion to obtain a circuit topology reconstruction scheme.
6. The method for optimizing power supply circuit design of a permanent magnet recloser based on deep learning according to claim 1, wherein the step of inputting the circuit topology reconstruction scheme into a deep reinforcement learning algorithm to perform power supply parameter optimizing processing to obtain power supply circuit optimizing configuration including an optimal power supply path switching strategy and voltage regulation parameters comprises the steps of: Performing state vector coding processing on the circuit topology reconstruction scheme based on the permanent magnet recloser power supply state space definition to obtain a state vector reflecting the current power supply network topology and load distribution; Performing executable action screening processing on the state vector according to the power supply action space constraint of the permanent magnet recloser to obtain a discrete action set of power supply path switching action and voltage regulating action; Inputting the state vector and the discrete action set into a deep Q network to perform state-action value evaluation processing to obtain a long-term gain predicted value corresponding to each power supply action; performing strategy gradient optimization processing on the long-term gain predicted value based on a permanent magnet recloser power supply rewarding mechanism to obtain an optimal power supply path switching strategy and a voltage regulation parameter; And carrying out power supply configuration parameter integration processing according to the optimal power supply path switching strategy and the voltage regulation parameter to obtain power supply circuit optimal configuration.
7. The method for optimizing the design of the power supply circuit of the permanent magnet recloser based on deep learning according to claim 1, wherein the performing closed-loop adjustment processing on the real-time power supply parameters of the permanent magnet recloser according to the power supply circuit optimization configuration to obtain an intelligent power supply control signal including a fault prediction result and a power supply switching instruction comprises: Performing real-time parameter deviation detection processing on the power supply circuit optimal configuration based on the permanent magnet recloser running state monitoring to obtain deviation amount of an actual value and a target value of a power supply parameter; Inputting the deviation value into a fault symptom recognition algorithm to perform abnormal mode analysis processing to obtain potential fault types and fault occurrence probability of the permanent magnet recloser; Performing preventive regulation strategy calculation processing on the permanent magnet recloser power supply switching decision according to the potential fault type to obtain a pre-fault preventive power supply path switching scheme; Performing real-time execution instruction generation processing on the preventive power supply path switching scheme based on a permanent magnet recloser closed-loop control feedback mechanism to obtain a fault prediction result and a power supply switching instruction; And carrying out control signal packaging processing on the fault prediction result and the power supply switching instruction to obtain an intelligent power supply control signal.
8. A deep learning-based permanent magnet recloser power supply circuit design optimization system for implementing the deep learning-based permanent magnet recloser power supply circuit design optimization method according to any one of claims 1-7, the deep learning-based permanent magnet recloser power supply circuit design optimization system comprising: The acquisition module is used for carrying out multidimensional electrical parameter acquisition processing on each power supply node of the permanent magnet recloser through the intelligent sensor of the power distribution network to obtain a power supply state data set containing node voltage amplitude, branch current and power factor; the extraction module is used for inputting the power supply state data set into a special topology sensing network of the permanent magnet recloser to perform power supply path feature extraction processing to obtain a power supply path feature matrix containing local power supply relevance and global network topology; The reconstruction module is used for carrying out multi-objective constraint reconstruction processing on the power supply circuit topology of the permanent magnet recloser according to the power supply path characteristic matrix to obtain a circuit topology reconstruction scheme comprising an optimal power supply path combination and a load distribution strategy; the optimizing module is used for inputting the circuit topology reconstruction scheme into a deep reinforcement learning algorithm to perform power supply parameter optimizing processing to obtain power supply circuit optimizing configuration containing an optimal power supply path switching strategy and voltage regulation parameters; And the adjusting module is used for performing closed-loop adjustment processing on the real-time power supply parameters of the permanent magnet recloser according to the power supply circuit optimal configuration to obtain an intelligent power supply control signal containing a fault prediction result and a power supply switching instruction.
9. A deep learning-based permanent magnet recloser power supply circuit design optimization device comprising a memory and a processor, the memory storing a computer program executable on the processor, the processor implementing the deep learning-based permanent magnet recloser power supply circuit design optimization method of any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, causes the processor to perform the deep learning based permanent magnet recloser supply circuit design optimization method according to any one of claims 1 to 7.

Description

Design optimization method and system for power supply circuit of permanent magnet recloser based on deep learning Technical Field The application relates to the technical field of power supply circuits, in particular to a design optimization method and system for a power supply circuit of a permanent magnet recloser based on deep learning. Background The design of the power supply circuit of the permanent magnet recloser mainly depends on a traditional power system analysis method and a rule-based control strategy, and automatic power restoration after power distribution network faults is realized through a preset protection setting value and fixed reclosing logic. The conventional method adopts static circuit parameter configuration and single fault discrimination criteria, combines a time-current characteristic curve to coordinate the action of the permanent magnet recloser, and simultaneously utilizes an SCADA system to carry out basic operation state monitoring and manual scheduling decision. However, the existing data acquisition system often has the problems of data loss, noise interference, incomplete coverage of working conditions and the like, so that the operation state of the permanent magnet recloser is not accurately known, secondly, the model interpretation and reliability problems are solved, the traditional rule-based control method lacks deep understanding of the topological relation of a complex power supply network, the nonlinear coupling relation among multiple nodes is difficult to process, thirdly, the real-time performance and adaptability problems are solved, the fixed parameter configuration cannot adapt to the factors such as dynamic load change and equipment aging of the power distribution network, and the enough flexibility is lacking when the new operation environment or abnormal working conditions are faced. The problem of the prior art is that a comprehensive technical scheme capable of intelligently sensing the topology change of the power distribution network, adaptively learning the optimal control strategy of the permanent magnet recloser and realizing multi-objective coordination optimization is lacking. Specifically, how to establish a topology sensing mechanism special for a permanent magnet recloser to accurately identify the relevance between power supply paths, how to construct a circuit topology reconstruction method under multi-objective constraint to balance power supply reliability and power quality, and how to realize parameter self-adaptive optimization based on deep learning to cope with complex and variable operation conditions are needed to be solved. Disclosure of Invention The application provides a design optimization method and system for a permanent magnet recloser power supply circuit based on deep learning, which are used for solving the technical problem that the existing permanent magnet recloser power supply circuit design lacks intelligent topology sensing, multi-objective coordination optimization and self-adaptive parameter adjustment capability. The application provides a deep learning-based permanent magnet recloser power supply circuit design optimization method, which comprises the steps of carrying out multidimensional electric parameter acquisition processing on power supply nodes of a permanent magnet recloser through a power distribution network intelligent sensor to obtain a power supply state data set containing node voltage amplitude, branch current and power factor, inputting the power supply state data set into a special topology sensing network of the permanent magnet recloser to carry out power supply path feature extraction processing to obtain a power supply path feature matrix containing local power supply relevance and global network topology, carrying out multi-objective constraint reconstruction processing on the permanent magnet recloser power supply circuit topology according to the power supply path feature matrix to obtain a circuit topology reconstruction scheme containing optimal power supply path combination and load distribution strategy, inputting the circuit topology reconstruction scheme into a deep reinforcement learning algorithm to carry out power supply parameter optimizing processing to obtain power supply circuit optimizing configuration containing optimal power supply path switching strategy and voltage adjusting parameters, and carrying out closed-loop adjustment processing on the real-time power supply parameters of the permanent magnet recloser according to obtain a power supply circuit optimizing configuration to obtain an intelligent power supply control signal containing a fault prediction result and a switching instruction. Optionally, the multi-dimensional electrical parameter collection processing is performed on each power supply node of the permanent magnet recloser by the power distribution network intelligent sensor to obtain a power supply state data set including node voltage amplitude, bran