CN-122026426-A - Regional current situation distribution network development diagnosis and weak index analysis method

CN122026426ACN 122026426 ACN122026426 ACN 122026426ACN-122026426-A

Abstract

The invention relates to the technical field of intelligent analysis of power distribution networks, and discloses a regional current situation power distribution network development diagnosis and weak index analysis method. The method comprises the steps of constructing a basic cognitive model of a power distribution network based on a general large model, processing original operation data to generate a power distribution network knowledge graph comprising equipment states and network topology, fine-adjusting the knowledge graph through industry documents and standards, injecting industry rules into the model to form an industry knowledge enhancement model with professional cognition, carrying out cooperative processing on structured knowledge by utilizing a multi-agent cooperation system to generate a multi-dimensional diagnosis report, collecting evaluation deviation through a scene feedback mechanism, and driving the model and the system to carry out reinforcement learning training to form a closed loop analysis flow with iterative optimization. The method realizes the deep combination of professional knowledge and artificial intelligence, solves the problems of insufficient professional adaptability and static analysis stiffness of the traditional method, and improves the accuracy of diagnosis and the self-adaptive capacity of the system.

Inventors

ZHANG YUAN
HOU XIAOYU
LI YUEHUI
WU QIAN
YANG GUIYUN
SUN HUA
CAO YANKUN
WANG YANMIN
SUN ZHENG

Assignees

北京子衿智研科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260209

Claims (10)

1. A method for diagnosing regional current distribution network development and analyzing weak indexes, the method comprising: Constructing a basic cognitive model of a power distribution network based on a general large model, and processing original operation data of a regional power distribution network by using the basic cognitive model of the power distribution network to generate a power distribution network knowledge graph comprising equipment states and network topology; Industry documents and standards for planning, operation and maintenance and evaluation of the power distribution network are collected, industry knowledge fine adjustment is conducted on the power distribution network knowledge map, industry rules are injected into the power distribution network basic cognition model, and an industry knowledge enhancement model with power distribution network professional cognition is formed; constructing a multi-agent collaboration system comprising a data analysis agent, an index calculation agent and a diagnosis reasoning agent, and performing collaborative processing on the structured knowledge output by the industry knowledge enhancement model through the multi-agent collaboration system to generate an initial diagnosis report of the development level of the power distribution network, wherein the initial diagnosis report comprises multiple dimensions; Setting an identification scene of a power distribution network development weak link, putting the power distribution network development level initial diagnosis report into the identification scene to operate, collecting evaluation deviation of the power distribution network development level initial diagnosis report in the identification scene through a scene feedback mechanism, and generating a model optimization signal; and performing reinforcement learning training on the industry knowledge enhancement model and decision logic of the multi-agent cooperation system according to the model optimization signal, and adjusting model parameters and cooperation strategies to form an iterative optimization regional distribution network weak index analysis closed-loop flow.
2. The method for diagnosing regional current power distribution network development and analyzing weak indexes according to claim 1, wherein the method for constructing a basic cognitive model of a power distribution network based on a general large model, and using the basic cognitive model of the power distribution network to process original operation data of the regional power distribution network, generating a power distribution network knowledge graph comprising equipment states and network topology, comprises the following steps: The method comprises the steps of accessing data interfaces of a regional distribution network dispatching system and an equipment monitoring system to obtain original operation data comprising real-time load, voltage and current, switch state and protection action information; Inputting the original operation data into the power distribution network basic cognitive model, identifying data entities and attributes thereof by the power distribution network basic cognitive model, and extracting association relations among the entities; Constructing a power distribution network knowledge graph taking a transformer substation, a line, a distribution transformer and switching equipment as nodes and taking an electrical connection and operation logic relationship as a side according to the identified data entity, attribute and association relationship; the construction of the basic cognitive model of the power distribution network based on the general large model specifically comprises the following steps: acquiring pre-training weight parameters of a general large model, wherein the pre-training weight parameters comprise natural language understanding and basic logic reasoning capacity; Basic concepts and entity relation definitions of the power distribution network field are collected, and a power distribution network field dictionary and an entity relation mapping table are constructed; taking the power distribution network domain dictionary and the entity relation mapping table as domain adaptation samples, and performing domain knowledge injection training on the general large model to enable the model to master the terms and basic operation principles of power distribution network equipment; After the field adaptation training is completed, carrying out graph structure reasoning capability strengthening training on the model, so that the model can extract entities from unstructured text data and construct a relation network; and optimizing model parameters through multiple rounds of iterative training, and judging to finish the construction of a basic cognitive model of the power distribution network when the accuracy of the model on a knowledge extraction task of the power distribution network reaches a preset threshold.
3. The method for diagnosing regional current power distribution network development and analyzing weak indexes according to claim 2, wherein the collecting industry documents and standards for power distribution network planning, operation and maintenance and evaluation, performing industry knowledge fine tuning on the power distribution network knowledge map, injecting industry rules into the power distribution network basic cognition model, and forming an industry knowledge enhancement model with power distribution network professional cognition, specifically comprises the following steps: establishing an industry knowledge base, wherein the industry knowledge base stores calculation standards and threshold definitions of power distribution network reliability, line loss rate, voltage qualification rate and load transfer capacity; Extracting rules and constraint conditions from the industry knowledge base, taking the rules and constraint conditions as fine tuning samples, and performing supervised training on the power distribution network basic cognitive model to enable the model to understand industry terms and judgment logic; And re-analyzing the power distribution network knowledge graph by utilizing the fine-tuned model, mapping the industry rule to nodes and edges of the knowledge graph, and outputting an industry knowledge enhancement model containing an industry quantization label.
4. A method for diagnosing regional status quo power distribution network developments and analyzing weak indexes as claimed in claim 3, wherein said constructing a multi-agent collaboration system including data analysis agents, index calculation agents and diagnostic reasoning agents, said collaborative system is used to cooperatively process structured knowledge output by said industry knowledge enhancement model to generate an initial diagnosis report of power distribution network developments level including multiple dimensions, specifically comprising: the data analysis agent receives the structured knowledge output by the industry knowledge enhancement model, and classifies and cleans the data according to the equipment layer, the circuit layer and the network layer; The index calculation agent calls calculation standards in the industry knowledge base according to the cleaned data, and calculates a plurality of development level indexes of power supply reliability, equipment load rate, network contact rate and voltage fluctuation rate in parallel; The diagnosis reasoning agent performs comprehensive comparison and logic reasoning based on the calculated multiple development level indexes and combined with threshold definition in the industry knowledge base, generates qualitative problem description and quantitative index deviation, and collects the qualitative problem description and quantitative index deviation into an initial diagnosis report of the development level of the power distribution network.
5. The method for diagnosing regional current power distribution network development and analyzing weak indexes according to claim 4, wherein the data analysis agent receives the structured knowledge output by the industry knowledge enhancement model, and classifies and cleans data according to a device layer, a line layer and a network layer, and specifically comprises: The data analysis agent identifies entity types in the structured knowledge, and divides the entities into transformers and switches belonging to the equipment layer, feeder sections belonging to the line layer and power supply areas belonging to the network layer; checking the integrity and consistency of the data aiming at the data of each layer, interpolating the missing telemetry data, and eliminating or marking the obviously abnormal telemetry data; and packaging the cleaned layered data into a standardized data packet, attaching a data quality label, and transmitting the standardized data packet to the index computing agent.
6. The method for diagnosing regional status quo power distribution network development and analyzing weak indexes according to claim 4, wherein said index computing agent invokes computing criteria in said industry knowledge base according to the cleaned data, and calculates a plurality of development level indexes of power supply reliability, equipment load rate, network contact rate and voltage fluctuation rate in parallel, specifically comprising: The index calculation intelligent agent receives the standardized data packet and selects a corresponding calculation fault tolerance strategy according to the data quality label; synchronously reading a definition formula and required data fields of each development level index from the industry knowledge base; and starting a plurality of calculation threads in parallel, and calculating corresponding data fields in the standardized data packet according to the defined formula respectively to generate a power supply reliability index value, a device load rate index value, a network contact rate index value and a voltage fluctuation rate index value.
7. The method for diagnosing regional status quo power distribution network development and analyzing weak indexes according to claim 4, wherein said diagnostic reasoning agent performs comprehensive comparison and logical reasoning based on calculated multiple level of development indexes in combination with threshold definitions in said industry knowledge base to generate qualitative problem description and quantitative index deviation, and summarizes the qualitative problem description and quantitative index deviation as an initial diagnostic report of the level of development of the power distribution network, and specifically comprises: the diagnosis reasoning agent acquires all the index values of the development level and invokes threshold definitions corresponding to all the indexes from the industry knowledge base; Comparing each development level index value with a corresponding threshold definition, judging whether the index is out of limit, and recording the out-of-limit index and the deviation value thereof; Carrying out logic reasoning on the out-of-limit index groups with the association to determine the fundamental weak links, and generating qualitative problem description aiming at the weak links; and integrating quantitative deviation and qualitative problem description of all out-of-limit indexes, and generating a structured initial diagnosis report of the power distribution network development level according to a preset report template.
8. The method for diagnosing regional status quo power distribution network development and analyzing weak indexes according to claim 4, wherein the step of setting an identification scene of a power distribution network development weak link, putting the initial diagnosis report of the power distribution network development level into the identification scene to operate, collecting the evaluation deviation of the initial diagnosis report of the power distribution network development level in the identification scene through a scene feedback mechanism, and generating a model optimization signal, comprises the following steps: Constructing an identification scene of a power distribution network development weak link containing high load, low reliability and weak connection typical problems in a simulation platform; Inputting the problem description and index deviation in the initial diagnosis report of the power distribution network development level into the identification scene, and simulating the evolution process of the problem described by the diagnosis report in the identification scene to obtain a simulation operation result; And comparing the simulation operation result with a preset ideal operation state, calculating to obtain the evaluation deviation of the diagnosis report, and quantifying the evaluation deviation into a model optimization signal.
9. The method for diagnosing and analyzing the weak index of the regional status quo power distribution network according to claim 8, wherein the constructing a scene of identifying the weak link of the power distribution network including high load, low reliability and weak connection typical problems in the simulation platform specifically comprises: in power system simulation software, establishing a simulation model consistent with the topology of a power distribution network in a target area; Setting a high-load identification scene, a low-reliability identification scene and a weak communication identification scene in the simulation model, wherein the high-load identification scene is overloaded by gradually increasing load simulation equipment, the low-reliability identification scene calculates the power failure influence by simulating a fault event sequence, and the weak communication identification scene is insufficient in simulation transfer capability by disconnecting part communication switches; An evaluation criterion of the ideal operating state is set for each recognition scenario, the evaluation criterion including an allowable maximum load rate, a minimum power supply reliability, and a minimum contact rate.
10. The method for diagnosing regional current power distribution network development and analyzing weak indexes according to claim 9, wherein the performing reinforcement learning training on the industry knowledge enhancement model and the decision logic of the multi-agent cooperation system according to the model optimization signal adjusts model parameters and cooperation strategies to form an iterative optimized regional power distribution network weak index analysis closed loop process, and specifically comprises: Decomposing the model optimization signal into a perception optimization signal for the industry knowledge enhancement model and a decision optimization signal for the multi-agent collaboration system; Adjusting weight parameters of knowledge extraction and rule mapping in the industry knowledge enhancement model by using the perception optimization signal, and improving the recognition accuracy of the model to weak link characteristics; adjusting task triggering conditions and cooperation flows of all intelligent agents in the multi-intelligent agent cooperation system by utilizing the decision optimization signals, and optimizing the cooperation efficiency of index calculation and diagnosis reasoning; and applying the adjusted industry knowledge enhancement model and the multi-agent cooperation system to the new round of regional power distribution network data, and starting an iterative optimization regional power distribution network weak index analysis closed-loop process.

Description

Regional current situation distribution network development diagnosis and weak index analysis method Technical Field The invention relates to the technical field of intelligent analysis of power distribution networks, in particular to a regional current situation power distribution network development diagnosis and weak index analysis method. Background In the field of regional distribution network development diagnostics, the prior art relies primarily on traditional data analysis and predefined rule systems. These methods are typically based on statistical models or classical machine learning algorithms to process equipment operation data, match abnormal states using a fixed rule base, and employ a standardized index system for network evaluation. Some schemes attempt to introduce generic knowledge-graph techniques to characterize network topology. The analysis flow in the prior art is mostly of a linear structure, is executed according to the sequence of data input, feature extraction, index calculation and report output, and lacks of deep fusion of professional knowledge of a power distribution network. In the model construction process, the system is not integrated with industry specifications such as planning guidelines, operation and maintenance regulations and the like, so that deviation exists between diagnosis logic and actual demands of the power industry. The technical scheme mainly has the defects that a general data analysis model is difficult to accurately understand specific operation constraint and technical standard of a power distribution network, equipment state evaluation and network structure analysis are mutually split in a diagnosis process, a generated weak index often deviates from an industrial practical application scene, a diagnosis system adopts a static analysis framework, model parameters and evaluation thresholds are fixed for a long time once set, and a mechanism for self-adjustment according to practical operation feedback is lacked. When the structure of the power distribution network changes or a novel operation mode appears, the diagnosis accuracy of the system can be gradually reduced, and the continuous evolution analysis capability can not be formed. A novel analysis method capable of deeply fusing industry knowledge needs to be developed, so that a diagnosis model has the professional cognitive ability of the power distribution network. Meanwhile, a diagnosis system with a self-optimization function must be constructed, and continuous iteration of analysis capability is realized through a closed-loop feedback mechanism so as to adapt to the requirement of dynamic development of the power distribution network. Disclosure of Invention The invention aims to provide a regional current situation distribution network development diagnosis and weak index analysis method for solving the problems in the background technology. In order to achieve the above purpose, the present invention provides a method for diagnosing regional current situation power distribution network development and analyzing weak indexes, the method comprising: Constructing a basic cognitive model of a power distribution network based on a general large model, and processing original operation data of a regional power distribution network by using the basic cognitive model of the power distribution network to generate a power distribution network knowledge graph comprising equipment states and network topology; Industry documents and standards for planning, operation and maintenance and evaluation of the power distribution network are collected, industry knowledge fine adjustment is conducted on the power distribution network knowledge map, industry rules are injected into the power distribution network basic cognition model, and an industry knowledge enhancement model with power distribution network professional cognition is formed; constructing a multi-agent collaboration system comprising a data analysis agent, an index calculation agent and a diagnosis reasoning agent, and performing collaborative processing on the structured knowledge output by the industry knowledge enhancement model through the multi-agent collaboration system to generate an initial diagnosis report of the development level of the power distribution network, wherein the initial diagnosis report comprises multiple dimensions; Setting an identification scene of a power distribution network development weak link, putting the power distribution network development level initial diagnosis report into the identification scene to operate, collecting evaluation deviation of the power distribution network development level initial diagnosis report in the identification scene through a scene feedback mechanism, and generating a model optimization signal; and performing reinforcement learning training on the industry knowledge enhancement model and decision logic of the multi-agent cooperation system according to the model optimization sign