CN-121978534-A - Lithium battery micro-internal short circuit fault diagnosis method based on dynamic modal decomposition and radial basis function neural network

CN121978534ACN 121978534 ACN121978534 ACN 121978534ACN-121978534-A

Abstract

The application discloses a lithium battery micro-internal short circuit fault diagnosis method based on dynamic modal decomposition and a radial basis function neural network. The method comprises the following steps of S1, preprocessing original voltage time sequence data acquired by a lithium battery in an operation process to construct an input matrix suitable for dynamic modal decomposition, S2, decomposing the input matrix based on a dynamic modal decomposition method, extracting a dominant dynamic mode, screening and recombining key modes based on Pearson correlation analysis to generate a low-dimensional and high-sensitivity fault feature vector, S3, taking the feature vector as input of a radial basis neural network, taking an internal short circuit equivalent resistance value as a label to construct a nonlinear mapping relation between a feature space and an internal short circuit state, and S4, comparing an internal resistance predicted value output by the radial basis neural network with a set threshold or a reference value, calculating a predicted error and evaluating diagnosis accuracy. The application avoids misdiagnosis caused by noise and aging effect.

Inventors

LIANG LONG
Xiong Liangqian
HAN HUA
LI ZIHE
LIU DONG
HE WEI
LI LONG
Long Tengwu
HE YAOTING
ZHANG JIE
ZENG HUAILING
TANG CHENFENG

Assignees

长沙矿山研究院有限责任公司
中南大学

Dates

Publication Date: 20260505
Application Date: 20260129

Claims (9)

1. A lithium battery micro-internal short circuit fault diagnosis method based on dynamic modal decomposition and radial basis function neural network is characterized by comprising the following steps: S1, preprocessing original voltage time sequence data acquired in the operation process of a lithium battery, and constructing an input matrix suitable for dynamic modal decomposition; S2, decomposing the input matrix based on a dynamic mode decomposition method, extracting a dominant dynamic mode, screening and recombining key modes based on Pearson correlation analysis, and generating a fault feature vector with low dimension and high sensitivity; S3, taking the characteristic vector as the input of the radial basis function neural network, taking the internal short circuit equivalent resistance value as a label, and constructing a nonlinear mapping relation between the characteristic space and the internal short circuit state; And S4, comparing the predicted value of the internal resistance output by the radial basis function neural network with a set threshold or reference value, calculating a predicted error, and evaluating the diagnosis precision, so that quantitative identification and early warning of the micro-internal short circuit fault of the lithium ion battery are realized.
2. The method for diagnosing a lithium battery micro-internal short circuit fault based on dynamic modal decomposition and radial basis function neural network according to claim 1, wherein the preprocessing in the step S1 comprises normalization processing and resampling processing of voltage time sequence data, and the normalization method adopts a min-max normalization method to map the voltage data to a [0,1] interval.
3. The method for diagnosing a micro-internal short circuit fault of a lithium battery based on dynamic modal decomposition and radial basis function neural network according to claim 1, wherein the decomposing of the input matrix based on the dynamic modal decomposition method in the step S2 is to construct the input matrix of dynamic modal decomposition by a sliding window mode.
4. The method for diagnosing a micro-internal short circuit fault of a lithium battery based on dynamic modal decomposition and radial basis function neural network as set forth in claim 1, wherein the voltage time sequence data preprocessed in step S1 is used for constructing an input matrix of dynamic modal decomposition through a sliding window, and setting a window size and an overlapping rate of the sliding window to generate a forward matrix And backward matrix 。
5. The method for diagnosing a micro-internal short circuit fault of a lithium battery based on dynamic modal decomposition and radial basis function neural network according to claim 1, wherein the step S2 of decomposing the input matrix is to decompose the singular value of the input matrix and truncate the decomposition result according to the singular value energy ratio.
6. The method for diagnosing a lithium battery micro-internal short circuit fault based on dynamic modal decomposition and radial basis function neural network according to claim 1, wherein the fault feature vector extracted in the step S2 includes dominant modal energy duty ratio, modal frequency characteristics, modal energy concentration degree, stable modal proportion and signal reconstruction error.
7. The method for diagnosing a micro-internal short circuit fault of a lithium battery based on dynamic modal decomposition and radial basis function neural network according to claim 1, wherein the step S2 further comprises performing correlation analysis on fault feature vectors, and rejecting feature parameters with correlation with an internal short circuit equivalent resistance lower than a preset threshold.
8. The method for diagnosing a lithium battery micro-internal short circuit fault based on dynamic modal decomposition and radial basis function neural network according to claim 1, wherein the radial basis function neural network in the step S3 comprises an input layer, an hidden layer and an output layer, and the hidden layer adopts a radial basis gaussian function as an activation function.
9. The lithium battery micro-internal short circuit fault diagnosis method based on dynamic modal decomposition and radial basis function neural network according to claim 1, wherein the radial basis function neural network in the step S3 is trained by a supervised learning mode, and accuracy and stability of internal short circuit equivalent resistance prediction are improved by adjusting network parameters.

Description

Lithium battery micro-internal short circuit fault diagnosis method based on dynamic modal decomposition and radial basis function neural network Technical Field The invention belongs to the field of lithium ion battery micro-internal short circuit fault diagnosis, and particularly relates to a lithium battery micro-internal short circuit fault diagnosis method based on dynamic modal decomposition and a radial basis function neural network. Background The lithium battery is a battery using lithium metal or lithium alloy as a negative electrode material and a nonaqueous electrolyte solution, and comprises a mining power lithium battery and the like. Along with the continuous promotion of the level of dynamism and intelligence of mine equipment, the mining power lithium battery has been widely applied to key mobile equipment such as underground electric locomotives, trackless rubber-tyred transport vehicles, heading machine auxiliary power supplies and the like because of the advantages of high energy density, high power output capability, zero emission and the like, and gradually replaces the traditional lead-acid battery or an internal combustion power system. However, the underground working environment of the coal mine has the characteristics of closed space, limited ventilation condition, existence of flammable and explosive gases such as gas, coal dust and the like, and extremely strict requirements are put on the safety of the power lithium battery system. In the actual operation process, the mining power lithium battery often faces complex working conditions such as frequent start-stop, high-rate charge and discharge, mechanical vibration, temperature fluctuation and the like, and micro-scale short circuit faults in the battery are easy to induce. Such failures are typically caused by manufacturing defects, dendrite penetration through the separator, localized lithium precipitation, or long-term cyclical aging, and are highly latent and hidden. In the early stage, the internal short circuit only causes weak current abnormality, local temperature rise or voltage deviation, the signal characteristics are easily covered by normal working condition noise, and the signal characteristics are difficult to effectively identify through a conventional monitoring means. Once the internal short circuit is not found in time and continuously evolved, local overheating is caused, so that thermal runaway chain reaction is triggered, and fire and even explosion accidents are caused. In view of the characteristics of closed underground space, difficult evacuation, limited rescue and the like, the safety accident has serious consequences, threatens the life safety of operators and can cause serious production interruption and property loss. At present, although the diagnosis technology for the internal short-circuit fault of the mining power lithium battery is improved, obvious defects still exist. The method mainly focuses on two aspects, namely, on one hand, the existing method depends on steady-state electrical parameter threshold judgment and is difficult to adapt to frequently-changed load conditions and dynamic operation characteristics of a battery under mining working conditions, on the other hand, the initial fault characteristics of internal short circuits are weak, nonlinearity is strong, signal to noise ratio is low, voltage abnormality caused by the initial fault characteristics is extremely easy to be mutually coupled with slow-change characteristics and complex working condition noise generated in the normal aging process of the battery, and an effective characteristic decoupling and extracting mechanism is lacked, so that aging effects, noise interference and real internal short circuit fault characteristics are difficult to accurately distinguish, and therefore diagnosis sensitivity is insufficient, false alarm or missing report rate is high. The bulletin number is CN119986409B, which provides a battery micro-short circuit fault diagnosis method and system, wherein the method and system are characterized in that a characteristic point matrix is constructed by carrying out variable decomposition denoising, dynamic reference voltage sequence calculation and characteristic value and correlation coefficient extraction, and then an abnormality score is obtained based on an improved French distance, so that fault judgment is realized by comparing the abnormality score with a threshold value; the patent with publication number of CN117849622A discloses a method for detecting short circuit faults in a battery based on variation modal decomposition and a support vector machine, which is used for acquiring voltage, current and surface temperature data by establishing a charge-discharge model, calculating internal temperature auxiliary diagnosis, carrying out variation modal decomposition on the voltage data to obtain modal components, calculating sample entropy and then inputting the sample entropy into the suppo