CN-122020299-A - Multi-dimensional fusion-based method, device and storage medium for analyzing electric core data

CN122020299ACN 122020299 ACN122020299 ACN 122020299ACN-122020299-A

Abstract

The embodiment of the application provides a multi-dimensional fusion-based cell data analysis method, which comprises the steps of synchronously preprocessing multi-dimensional heterogeneous original data from a cell, inputting the preprocessed multi-dimensional data into a cross-modal depth feature fusion network to generate a depth fusion feature vector, inputting the depth fusion feature vector into a pre-trained main evaluation model and a lightweight increment learning model to respectively obtain a first analysis result and a second analysis result, determining whether to update parameters of the lightweight increment learning model by using new data with a true value label according to a preset trigger condition and a confidence evaluation result of the main evaluation model, wherein parameters of the main evaluation model are kept unchanged in an updating process, and dynamically weighting and fusing the first analysis result and the second analysis result according to real-time confidence of the main evaluation model to obtain the cell analysis result. The method aims to improve the estimation accuracy of key parameters such as the health state, the residual life and the like of the battery cell.

Inventors

ZHANG HUI
Zhu Ruida
DU MINGYU
Luo Ruiqiao
ZHANG XINYONG
ZHOU YI

Assignees

深能北方能源控股有限公司
深圳市盛路物联通讯技术有限公司
深能源(深圳)创新技术有限公司

Dates

Publication Date: 20260512
Application Date: 20260128

Claims (10)

1. A method of analyzing electrical core data based on multidimensional data fusion, comprising: synchronously preprocessing multidimensional heterogeneous original data from the battery cells; Inputting the preprocessed multi-dimensional data into a cross-modal depth feature fusion network to generate a depth fusion feature vector; inputting the depth fusion feature vector into a pre-trained main evaluation model and a lightweight incremental learning model to respectively obtain a first analysis result and a second analysis result; Determining whether to update parameters of the lightweight incremental learning model by using new data with a truth value tag according to a preset triggering condition and a confidence coefficient evaluation result of the main evaluation model, wherein the parameters of the main evaluation model are kept unchanged in the updating process; and dynamically weighting and fusing the first analysis result and the second analysis result according to the real-time confidence coefficient of the main evaluation model to obtain a battery cell analysis result.
2. The method of claim 1, wherein the multi-dimensional raw data comprises at least electrical dimension data, thermal dimension data, time dimension data, and operating condition dimension data.
3. The method of claim 2, wherein the cross-modal depth feature fusion network comprises a plurality of feature extraction branches respectively adapting to different data dimensions and a cross-modal attention fusion module, wherein the cross-modal attention fusion module is used for dynamically learning and weighting and fusing the features output by the feature extraction branches to generate the depth fusion feature vector.
4. The method of claim 1, wherein the trigger condition comprises at least one of: The confidence coefficient of the main evaluation model is lower than a first threshold value, and a truth value label of the state of the battery cell is obtained; Or a significant drift in the characteristic distribution of the input data is detected.
5. The method of claim 4, wherein parameter updating the lightweight incremental learning model comprises parameter updating the lightweight incremental learning model using elastic weight cure constraints.
6. The method according to claim 5, further comprising: periodically distilling knowledge of the lightweight incremental learning model to the primary assessment model to progressively update the primary assessment model.
7. The method of claim 1, wherein the cell analysis results are used to indicate a health status, a remaining life, or fault warning information of the cell.
8. A multi-dimensional data fusion-based electrical core data analysis device, comprising: The data preprocessing module is used for synchronously preprocessing multidimensional heterogeneous original data from the battery cells; the feature fusion module is used for inputting the preprocessed multi-dimensional data into a cross-mode depth feature fusion network to generate a depth fusion feature vector; the collaborative analysis module is used for inputting the depth fusion feature vector into a pre-trained main evaluation model and a lightweight incremental learning model to respectively obtain a first analysis result and a second analysis result; The incremental learning module is used for determining whether to update parameters of the lightweight incremental learning model by using new data with a truth value tag according to a preset trigger condition and a confidence coefficient evaluation result of the main evaluation model, wherein the parameters of the main evaluation model are kept unchanged in the updating process; and the obtaining module is used for dynamically weighting and fusing the first analysis result and the second analysis result according to the real-time confidence coefficient of the main evaluation model to obtain a battery cell analysis result.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method of any one of claims 1 to 7 when the computer program is executed.
10. A computer readable storage medium, on which a computer program is stored, which computer program, when being executed by a processor, implements the method according to any of claims 1 to 7.

Description

Multi-dimensional fusion-based method, device and storage medium for analyzing electric core data Technical Field The application belongs to the technical field of battery management, and particularly relates to a method, equipment and a storage medium for analyzing electric core data based on multidimensional fusion. Background With the rapid development of electric vehicles and large-scale energy storage technologies, the battery is used as a core energy carrier, and the performance attenuation and the safety problems of the battery are of great concern. Currently, battery management systems evaluate the state of a battery cell in real time by collecting data such as voltage, current, temperature, etc., and independently analyzing the single-dimensional data or performing simple early fusion. However, since the cell data has the characteristics of multiple sources (BMS sensor, battery tester), heterogeneous (timing signal, scalar, image), and multi-scale (second-level current, hour-level capacity), if deep coupling relation and cross-scale correlation between different physical quantities (such as voltage and temperature) are not deeply mined, information loss can be caused, and estimation accuracy of the cell state is limited. Therefore, a method for analyzing the state of the battery cell, which can deeply fuse multi-source heterogeneous information and adaptively evolve along with the whole life cycle of the battery cell, is needed to improve the accuracy of the state analysis of the battery cell. Disclosure of Invention In view of the above, the embodiment of the application provides a multi-dimensional fusion-based cell data analysis method, device and storage medium, which are used for carrying out depth feature fusion on multi-dimensional data of a cell through a cross-modal attention mechanism, fully mining internal association among the multi-dimensional data, providing rich and discriminant information basis for cell state evaluation, and remarkably improving estimation precision of key parameters such as cell health state, residual life and the like. The embodiment of the application provides a method for analyzing cell data based on multi-dimensional fusion, which comprises the steps of synchronously preprocessing multi-dimensional heterogeneous original data from a cell; Inputting the preprocessed multi-dimensional data into a cross-modal depth feature fusion network to generate a depth fusion feature vector; inputting the depth fusion feature vector into a pre-trained main evaluation model and a lightweight incremental learning model to respectively obtain a first analysis result and a second analysis result; Determining whether to update parameters of the lightweight incremental learning model by using new data with a truth value tag according to a preset triggering condition and a confidence coefficient evaluation result of the main evaluation model, wherein the parameters of the main evaluation model are kept unchanged in the updating process; and dynamically weighting and fusing the first analysis result and the second analysis result according to the real-time confidence coefficient of the main evaluation model to obtain a battery cell analysis result. In an embodiment, the multi-dimensional heterogeneous raw data at least includes electrical dimension data, thermal dimension data, time dimension data, and operating condition dimension data. In an embodiment, the cross-modal depth feature fusion network comprises a plurality of feature extraction branches respectively adapting to different data dimensions and a cross-modal attention fusion module, wherein the cross-modal attention fusion module is used for dynamically learning and weighting and fusing the features output by the feature extraction branches to generate the depth fusion feature vector. In an embodiment, the trigger condition includes at least one of: The confidence coefficient of the main evaluation model is lower than a first threshold value, and a truth value label of the state of the battery cell is obtained; Or a significant drift in the characteristic distribution of the input data is detected. In one embodiment, the parameter updating of the lightweight incremental learning model includes parameter updating of the lightweight incremental learning model using elastic weight cure constraints. In an embodiment, the method further comprises: periodically distilling knowledge of the lightweight incremental learning model to the primary assessment model to progressively update the primary assessment model. In an embodiment, the battery cell analysis result is used to indicate the health status, the remaining service life, or the fault pre-warning information of the battery cell. A second aspect of an embodiment of the present application provides a device for analyzing electrical core data based on multi-dimensional fusion, including: The data preprocessing module is used for synchronously preprocessing multidimensional he