CN-121276384-B - Energy storage facility safety early warning protection method and system

CN121276384BCN 121276384 BCN121276384 BCN 121276384BCN-121276384-B

Abstract

The invention belongs to the field of energy storage monitoring, and provides a safety pre-warning protection method and system for an energy storage facility, wherein the method comprises the steps of obtaining temperature, heating rate and equivalent impedance data of each monitoring unit of a battery cluster; the method comprises the steps of constructing a physical adjacency matrix based on a physical space arrangement relation of a monitoring unit, fusing the physical adjacency matrix with an electrical adjacency matrix according to a preset weight, constructing a graph structure based on the comprehensive adjacency matrix, calculating node neighborhood residual errors based on the Laplace matrix to obtain temperature consistency residual errors, calculating node neighborhood residual errors to obtain electrical consistency residual errors based on the Laplace matrix, calculating temperature prediction residual errors under a time sequence prediction model with graph regular constraint, constructing node early abnormality scores, adaptively setting a split threshold according to healthy operation data distribution, and carrying out grading early warning judgment on the early abnormality scores. The invention can improve the accuracy of the safety pre-warning of the energy storage facility.

Inventors

HE LU
XU LULU
WANG QIAN
YANG LIU
MA YUE
TAN LIANGBIN
QU HUIMING
ZHANG YONGKUI
DENG ZHIXIAO
BI QINGQING
ZHANG YI
CAO SAIJUN

Assignees

重庆市设计院有限公司

Dates

Publication Date: 20260508
Application Date: 20251209

Claims (10)

1. The energy storage facility safety pre-warning protection method is characterized by comprising the following steps of: Acquiring temperature, heating rate and equivalent impedance data of each monitoring unit of the battery cluster, wherein the monitoring units are battery monomers or parallel branches; constructing a physical adjacency matrix based on the physical space arrangement relation of the monitoring units, and constructing an electrical adjacency matrix based on the electrical serial-parallel connection relation of the monitoring units; Fusing the physical adjacency matrix and the electrical adjacency matrix according to preset weights to obtain a comprehensive adjacency matrix, wherein rows and columns of the comprehensive adjacency matrix correspond to each monitoring unit of the battery cluster, and matrix elements represent physical or electrical adjacency relations among the monitoring units; constructing a graph structure based on the comprehensive adjacency matrix, wherein nodes of the graph structure are monitoring units of the battery cluster, and edges of the graph structure are physical adjacency relations or electrical adjacency relations among the monitoring units, and calculating a Laplace matrix of the graph; Calculating a node neighborhood residual error based on the Laplace matrix for the temperature and the temperature rising rate to obtain a temperature consistency residual error; The neighborhood residual error is the difference between the parameter value of the monitoring unit and the weighted average value of the parameter of the adjacent node, and the difference is obtained through calculation of a Laplacian matrix and is used for measuring the deviation degree of the monomer and the neighborhood; The temperature consistency residual is a neighborhood residual obtained by calculating the temperature and the temperature rising rate and is used for reflecting the deviation condition between the thermal behavior of a certain monitoring unit and the neighbors thereof, wherein the temperature vector and the temperature rising rate vector are used as input to be operated with a Laplacian matrix, and the neighborhood residual of each monitoring unit in the thermal dimension is obtained in a matrix multiplication mode; the electric consistency residual error is a neighborhood residual error obtained by calculating equivalent impedance and is used for reflecting the degree of difference between the electric state of a certain monitoring unit and the neighbors thereof; generating a consistency confidence index based on the temperature consistency residual and the electrical consistency residual, and calculating a temperature prediction residual under a time sequence prediction model with graph regular constraint; Constructing early-stage abnormal scores of nodes based on the temperature prediction residual error, the temperature consistency residual error and the electrical consistency residual error, adaptively setting a grading threshold according to healthy operation data distribution, and carrying out grading early-warning judgment on the early-stage abnormal scores; And when the early warning level reaches the intervention level or the isolation level, executing the circuit breaking operation on the target unit according to the risk propagation relationship of the physical neighborhood and the electrical neighborhood.
2. The method of claim 1, wherein the physical adjacency matrix is a matrix describing a spatial adjacency relationship between battery cells or parallel branches in a battery cluster, wherein rows and columns of the physical adjacency matrix correspond to monitoring units, when two monitoring units are adjacent in physical position in an actual structure, corresponding elements in the matrix take a value of one, and when two monitoring units are not adjacent, the corresponding elements take a value of zero; the electric adjacent matrix is used for describing a matrix of series or parallel connection relation of the battery cells or the parallel branches in a circuit connection mode, the rows and columns of the electric adjacent matrix also correspond to the monitoring units, when the two monitoring units are directly connected in series or in parallel, the corresponding element in the matrix takes a value of one, and otherwise, the value of the element in the matrix takes a value of zero.
3. The method of claim 1, wherein the integrated adjacency matrix is a, the physical adjacency matrix is Ap, the electrical adjacency matrix is Ae, and the weight coefficients are α and β, respectively, and the integrated adjacency matrix is denoted as a=αap+βae, where α is a physical adjacency weight, and β is an electrical adjacency weight.
4. The energy storage facility safety pre-warning protection method according to claim 1, wherein for each node of the comprehensive adjacency matrix construction graph structure, traversing all non-zero elements of a corresponding row of the node, accumulating the numerical values of the elements and taking the numerical values as the degree value of the node, and filling the degree value into a diagonal line of a corresponding position of the degree matrix to obtain the degree matrix reflecting the connection strength of each monitoring unit and a neighborhood thereof on topology; After the degree matrix is generated, subtracting the comprehensive adjacent matrix from the degree matrix to obtain the Laplace matrix of the graph.
5. The energy storage facility safety pre-warning protection method according to claim 1, wherein a time sequence prediction model with graph regularization constraint is realized based on a structure based on a cyclic neural network, a graph regularization term is added in a loss function of the network, calculation is carried out through the product of a Laplacian matrix and a predicted temperature vector, and the situation that excessive difference occurs between adjacent nodes in a predicted value is punished.
6. An energy storage facility safety precaution protection system, characterized by that, the system includes following module: The data acquisition module is used for acquiring temperature, heating rate and equivalent impedance data of each monitoring unit of the battery cluster, wherein the monitoring units are battery monomers or parallel branches; The matrix construction module is used for constructing a physical adjacency matrix based on the physical space arrangement relation of the monitoring units and constructing an electrical adjacency matrix based on the electrical serial-parallel connection relation of the monitoring units; The matrix fusion module is used for fusing the physical adjacency matrix and the electrical adjacency matrix according to preset weights to obtain a comprehensive adjacency matrix, rows and columns of the comprehensive adjacency matrix correspond to each monitoring unit of the battery cluster, and matrix elements represent physical or electrical adjacency relations among the monitoring units; The graph structure generation module is used for constructing a graph structure based on the comprehensive adjacency matrix, wherein nodes of the graph structure are monitoring units of the battery cluster, and edges of the graph structure are physical adjacency relations or electrical adjacency relations among the monitoring units, and a Laplace matrix of the graph is calculated according to the physical adjacency relations or the electrical adjacency relations; The residual calculation module is used for calculating a node neighborhood residual for the temperature and the heating rate based on the Laplace matrix to obtain a temperature consistency residual, and calculating a node neighborhood residual for the equivalent impedance to obtain an electrical consistency residual; The neighborhood residual error is the difference between the parameter value of the monitoring unit and the weighted average value of the parameter of the adjacent node, and the difference is obtained through calculation of a Laplacian matrix and is used for measuring the deviation degree of the monomer and the neighborhood; The temperature consistency residual is a neighborhood residual obtained by calculating the temperature and the temperature rising rate and is used for reflecting the deviation condition between the thermal behavior of a certain monitoring unit and the neighbors thereof, wherein the temperature vector and the temperature rising rate vector are used as input to be operated with a Laplacian matrix, and the neighborhood residual of each monitoring unit in the thermal dimension is obtained in a matrix multiplication mode; the electric consistency residual error is a neighborhood residual error obtained by calculating equivalent impedance and is used for reflecting the degree of difference between the electric state of a certain monitoring unit and the neighbors thereof; The prediction analysis module is used for generating a consistency confidence index based on the temperature consistency residual error and the electrical consistency residual error, and calculating a temperature prediction residual error under a time sequence prediction model with a graph regular constraint; The early warning judging module is used for constructing early abnormal scores of nodes based on the temperature prediction residual error, the temperature consistency residual error and the electrical consistency residual error, adaptively setting a grading threshold according to healthy operation data distribution, and carrying out grading early warning judgment on the early abnormal scores; and the circuit breaking control module is used for executing circuit breaking operation on the target unit according to the risk propagation relationship of the physical neighborhood and the electrical neighborhood when the early warning level reaches the intervention level or the isolation level.
7. The energy storage facility safety precaution protection system according to claim 6, wherein the matrix construction module comprises a physical adjacency matrix generation unit and an electrical adjacency matrix generation unit, the physical adjacency matrix generation unit is used for describing a matrix of a spatial adjacency relation of battery cells or parallel branches in a battery cluster, each row of the physical adjacency matrix corresponds to a monitoring unit, when two monitoring units are adjacent in physical positions in an actual structure, the corresponding element in the matrix takes a value of one, when the two monitoring units are not adjacent, the corresponding element takes a value of zero, the electrical adjacency matrix generation unit is used for describing a matrix of a series or parallel relation of battery cells or parallel branches in a circuit connection mode, each row of the electrical adjacency matrix also corresponds to a monitoring unit, when the two monitoring units are directly connected in series or parallel, the corresponding element in the matrix takes a value of one, and otherwise takes a value of zero.
8. The energy storage facility safety precaution protection system of claim 6, wherein the matrix fusion module is configured to generate a comprehensive adjacency matrix according to formula a=αap+βae, wherein a is the comprehensive adjacency matrix, ap is the physical adjacency matrix, ae is the electrical adjacency matrix, α is the physical adjacency weight, and β is the electrical adjacency weight.
9. The energy storage facility safety precaution protection system according to claim 6, wherein the graph structure generation module comprises a degree matrix calculation unit and a Laplacian matrix calculation unit, the degree matrix calculation unit is used for traversing non-zero elements of each row in the comprehensive adjacent matrix, accumulating the element values to be used as degree values of nodes, filling diagonal elements at corresponding positions of the degree matrix, and obtaining a degree matrix reflecting the topological connection strength of each monitoring unit and the neighborhood thereof; and the Laplace matrix calculation unit is used for subtracting the comprehensive adjacent matrix from the degree matrix after the degree matrix is generated, so as to obtain the Laplace matrix of the graph.
10. The energy storage facility safety pre-warning protection system according to claim 6, wherein the prediction analysis module comprises a time sequence prediction model unit with graph regular constraint, the time sequence prediction model unit is realized based on a recurrent neural network structure, a graph regularization term is added in a loss function of the network, and the time sequence prediction model unit is calculated through the product of a Laplace matrix and a predicted temperature vector and is used for punishing the situation that the predicted value has excessive difference between adjacent nodes, so that the spatial consistency constraint capacity of the predicted result is enhanced.

Description

Energy storage facility safety early warning protection method and system Technical Field The invention belongs to the field of energy storage monitoring, and particularly relates to a safety pre-warning protection method and system for energy storage facilities. Background Along with the wide application of the new energy storage system, the battery cluster plays an increasingly important role in the scenes of power grid peak shaving, standby power supply, renewable energy grid connection and the like. However, the operation environment of the energy storage facility is complex, and the single battery may have performance degradation and even potential safety hazard due to heat accumulation, electrochemical aging or abnormal connection during long-time charge and discharge. When a certain monomer is abnormally heated, internal resistance is increased or polarization is serious, if the abnormal temperature is not detected and isolated in time, the abnormal temperature can be rapidly transmitted along a thermal diffusion path or an electric coupling path to cause linkage failure, and thermal runaway accidents can be caused in serious cases. Therefore, the early abnormal identification and active safety protection of the local single bodies in the battery cluster are realized, and the key of the safe operation of the energy storage system is realized. The existing energy storage safety monitoring method is mostly dependent on macroscopic parameters such as total voltage, total current or average module temperature to carry out state judgment. And part of the system utilizes the single voltage or temperature monitoring data to realize abnormal alarm by setting a fixed threshold value. There are also studies attempting to make trend judgment on the variation of the monomer parameters by using a machine learning method through a time series prediction model or a statistical anomaly detection algorithm. However, the method has the common limitations that firstly, the interaction of each monomer in the battery cluster in space and electricity is not fully considered, so that false alarm or false alarm is easy to generate only depending on single-point data, secondly, the threshold is fixed and cannot be adaptively adjusted along with the change of the operating condition or environment, and thirdly, the prediction model lacks topological constraint, and the coupling relation of heat diffusion and current conduction is difficult to reflect. Disclosure of Invention In order to solve the problems in the prior art, the invention provides a safety pre-warning protection method for an energy storage facility, which comprises the following steps: Acquiring temperature, heating rate and equivalent impedance data of each monitoring unit of the battery cluster, wherein the monitoring units are battery monomers or parallel branches; constructing a physical adjacency matrix based on the physical space arrangement relation of the monitoring units, and constructing an electrical adjacency matrix based on the electrical serial-parallel connection relation of the monitoring units; Fusing the physical adjacency matrix and the electrical adjacency matrix according to preset weights to obtain a comprehensive adjacency matrix, wherein rows and columns of the comprehensive adjacency matrix correspond to each monitoring unit of the battery cluster, and matrix elements represent physical or electrical adjacency relations among the monitoring units; constructing a graph structure based on the comprehensive adjacency matrix, wherein nodes of the graph structure are monitoring units of the battery cluster, and edges of the graph structure are physical adjacency relations or electrical adjacency relations among the monitoring units, and calculating a Laplace matrix of the graph; Calculating a node neighborhood residual error based on the Laplace matrix for the temperature and the temperature rising rate to obtain a temperature consistency residual error; generating a consistency confidence index based on the temperature consistency residual and the electrical consistency residual, and calculating a temperature prediction residual under a time sequence prediction model with graph regular constraint; Constructing early-stage abnormal scores of nodes based on the temperature prediction residual error, the temperature consistency residual error and the electrical consistency residual error, adaptively setting a grading threshold according to healthy operation data distribution, and carrying out grading early-warning judgment on the early-stage abnormal scores; And when the early warning level reaches the intervention level or the isolation level, executing the circuit breaking operation on the target unit according to the risk propagation relationship of the physical neighborhood and the electrical neighborhood. Further, the matrix is used for describing the spatial adjacent relation of the battery cells or the parallel branches in the battery