CN-121995234-A - Power battery capacity attenuation testing and predicting method and system

CN121995234ACN 121995234 ACN121995234 ACN 121995234ACN-121995234-A

Abstract

The application discloses a power battery capacity attenuation test and prediction method and system, which relate to the technical field of battery capacity, and the application constructs a multi-dimensional state characterization system by fusing electrochemical impedance related characteristics, internal physical field characteristics and macroscopic cycle data, so that more accurate fitting and prediction of a complex attenuation mode are realized, an information enrichment test which is time-consuming is triggered only when model prediction uncertainty is high or internal states are abnormal is realized by a dual-drive self-adaptive test decision mechanism, the optimal balance between test efficiency and model performance maintenance is realized, the model not only outputs capacity prediction, the internal characteristic importance analysis of the model can qualitatively or semi-quantitatively reveal dominant attenuation mechanisms in different aging stages, but also can discover the abnormal states of a battery earlier than macroscopic capacity attenuation based on real-time monitoring, and provides a prospective basis for safety early warning and maintenance decision.

Inventors

YANG ZHIYONG
SUN TAO
ZHOU ZHENG
ZHU SHANGGONG
XIU LINGLING
NI MINGYANG
YANG MINGDONG
GAO YUANWEI
LU ZHONGDE

Assignees

辽宁省交通高等专科学校

Dates

Publication Date: 20260508
Application Date: 20260312

Claims (10)

1. A power battery capacity fade testing and predicting method, comprising: s1, acquiring multi-source characteristic data of a battery, and integrating the multi-source characteristic data into a structural characteristic data set; s2, constructing and training a capacity attenuation prediction model based on the multi-source characteristic data; s3, in the battery aging process, executing a double-drive self-adaptive test decision; S4, when the dual-drive self-adaptive test decision is triggered, updating the capacity attenuation prediction model; And S5, outputting a prediction result and battery capacity attenuation mechanism analysis based on the updated capacity attenuation model.
2. The method of claim 1, wherein the multi-source signature data comprises electrochemical response signatures and internal physical field signatures of the battery.
3. The method of claim 1, wherein the acquiring multi-source feature data of the battery and integrating the multi-source feature data into a structured feature data set comprises: A1, applying current excitation containing a plurality of frequency components to a target battery based on a preset excitation current waveform, synchronously collecting terminal voltage response of the target battery, calculating an impedance value of the battery at each frequency through a frequency domain signal processing algorithm to obtain a battery impedance spectrum; A2, acquiring physical field parameters representing the internal state of the battery in the aging test process of the target battery so as to obtain an internal physical field feature vector; A3, when each charge-discharge cycle of the target battery is finished, recording the current cycle times, and measuring the actual discharge capacity of the target battery to obtain a macroscopic performance data sequence; A4, based on a preset standardization method, carrying out dimensionless treatment on the electrochemical response feature vector, the internal physical field feature vector and the macroscopic performance data sequence respectively, and combining the electrochemical response feature vector, the internal physical field feature vector and the macroscopic performance data sequence into a structural feature data set through feature fusion operation.
4. The method of claim 1, wherein said constructing and training a capacity fade prediction model based on said multi-source signature data comprises: Defining the cycle number, the historical capacity sequence, the electrochemical response characteristic of the battery and the internal physical field characteristic as an input characteristic set and the current discharge capacity as an output target based on the structured characteristic data set to form a model training sample; And training the initialized gradient lifting decision tree model based on the iterative addition modeling process of the gradient lifting decision tree algorithm, wherein the training data comprise calculation residues, fitting of a new decision tree and updating of a model to obtain a final trained gradient lifting decision tree model serving as the initial capacity attenuation prediction model.
5. The power cell capacity fade testing and predicting method of claim 1, wherein said dual drive adaptive test decision comprises a first drive and a second drive, wherein the first drive is based on a predictive uncertainty of said capacity fade prediction model and the second drive is based on a real-time varying anomaly of said internal physical field feature.
6. The power cell capacity fade test and prediction method as defined in claim 1, wherein said first driving is based on a prediction uncertainty of said capacity fade prediction model, and said second driving is based on a real-time variation anomaly of said internal physical field feature, comprising: B1, at the current aging cycle time of a battery, acquiring an input feature set of the current prediction model, wherein the input feature set comprises the current cycle times, a historical capacity sequence, an electrochemical response feature vector at the current time and an internal physical field feature vector; and B2, carrying out time sequence analysis on key parameters based on the internal physical field feature vectors at the current moment and the historical moment so as to calculate and obtain an internal physical field abnormality index.
7. The method of claim 1, wherein said performing a dual drive adaptive test decision comprises: Setting a first decision threshold and a second decision threshold, carrying out logic judgment on the prediction uncertainty measure and the internal physical field abnormality index, comparing the prediction uncertainty measure with the first decision threshold, simultaneously comparing the internal physical field abnormality index with the second decision threshold, and when the prediction uncertainty measure is larger than the first decision threshold or the internal physical field abnormality index is larger than the second decision threshold, generating an instruction for triggering the next information enrichment test, judging that new high-dimensional test data is needed to be supplemented currently to reduce model uncertainty and confirming physical field abnormality.
8. The power cell capacity fade testing and predicting method as claimed in claim 1, wherein said updating said capacity fade prediction model when said dual drive adaptive test decision triggers comprises: When an instruction for triggering the next information enrichment test is generated, performing an information enrichment test on the battery in a subsequent preset charge-discharge cycle based on the current battery cycle state to acquire new electrochemical response data and new internal physical field data so as to acquire new electrochemical response feature vectors and new internal physical field feature vectors; and adding the newly added data points into the structural feature data set to obtain an updated feature data set, and executing incremental learning on the constructed initial capacity fading prediction model to obtain an updated capacity fading prediction model.
9. The power battery capacity fade testing and predicting method as set forth in claim 1, wherein said outputting a prediction result and a battery capacity fade mechanism analysis based on the updated capacity fade model includes: Based on the updated capacity attenuation prediction model and the current cycle state, taking the updated capacity attenuation prediction model as the current model, returning to execute a dual-drive self-adaptive test decision flow, and executing the next round of monitoring and decision on the subsequent aging cycle of the battery to form a closed loop cycle path for model iteration and test execution; and when the iteration is ended, calculating and outputting a capacity attenuation curve with high confidence coefficient, a residual service life predicted value and a confidence interval thereof by using a latest capacity attenuation prediction model when the iteration is ended, and carrying out qualitative or semi-quantitative analysis on a battery attenuation dominant mechanism by combining a characteristic importance analysis result of the latest capacity attenuation prediction model so as to obtain a comprehensive evaluation report of the battery life.
10. A system for performing the power cell capacity fade testing and predicting method of any one of claims 1-9, comprising: The structural feature data set generation module is used for acquiring multi-source feature data of the battery and integrating the multi-source feature data into a structural feature data set; a capacity fading prediction model construction module for constructing and training a capacity fading prediction model based on the multi-source characteristic data; the decision module is used for executing double-drive self-adaptive test decision in the battery aging process; the model updating module is used for updating the capacity attenuation prediction model when the dual-drive self-adaptive test decision is triggered; and the result output module is used for outputting a prediction result and battery capacity attenuation mechanism analysis based on the updated capacity attenuation model.

Description

Power battery capacity attenuation testing and predicting method and system Technical Field The application relates to the technical field of battery capacity, in particular to a power battery capacity attenuation testing and predicting method and system. Background With the rapid development of electric vehicles and energy storage systems, life prediction and health management of lithium ion batteries are key challenges. The traditional capacity fading prediction method mainly relies on macro-cycle data (such as discharge capacity and cycle number) to perform empirical model fitting or simple machine learning modeling, has limited prediction precision and robustness, and is insufficient in particular when dealing with complex aging paths and sudden faults. The invention discloses a battery capacity attenuation analysis method and system as disclosed in the patent application of CN117686933A, which relate to the field of battery capacity analysis and comprise the steps of obtaining a battery unpolarized charge-discharge curve after aging and an unattenuated battery unpolarized charge-discharge curve; the method comprises the steps of obtaining an aged battery differential voltage curve and a fresh battery differential voltage curve according to an aged battery unpolarized charge-discharge curve and an unattenuated battery unpolarized charge-discharge curve, drawing the aged battery differential voltage curve and the unattenuated battery differential voltage curve under the same coordinate system and aligning, judging an attenuation mode and an attenuation amount of the aged battery according to the corresponding relation between an attenuation mode and a differential voltage, and quantitatively analyzing the attenuation mode and the attenuation amount of the aged battery according to the corresponding relation between the attenuation mode and the differential voltage of the unattenuated battery and the comparison result of the whole battery differential voltage curve under the attenuation mode. The invention can judge the specific attenuation mode and accurately quantify by applying the differential voltage curve. Aiming at the scheme, the inventor discovers that the technology at least has the following technical problems that 1, the current lack of systematically acquiring and constructing multi-source heterogeneous characteristic data can not lay a comprehensive and deep data foundation for battery health state assessment, the limitation that the traditional method depends on single macroscopic performance data (such as capacity) is not broken, the trans-scale fusion of external electrochemical response, internal physical field state and macroscopic cycle performance is not carried out, the capability of a subsequent model for learning complex aging rules from the data can not be enhanced, and key data support can not be provided for solving the difficult problem of battery attenuation multi-factor coupling and nonlinearity. 2. At present, a dual-drive self-adaptive test decision mechanism is not provided, and efficient configuration of test resources and self-adaptive management of model risks cannot be realized. The method lacks of 'data driven' decision, uncertainty of model prediction cannot be quantified, and information enrichment test with higher cost cannot be intelligently triggered when the confidence of the model is insufficient, so that 'valuable data' which is most effective in improving prediction precision cannot be obtained by the least test times, and test economy is not improved. The decision of 'mechanism driving' is absent, abnormal mutation of internal physical field characteristics is not monitored, the potential safety hazard of the battery or key nodes turned by an aging mechanism cannot be timely detected in an intervention mode, and the early warning and safety guarantee capability of the method are not enhanced. Disclosure of Invention In view of the above-mentioned technical shortcomings, the present application aims to provide a method and a system for testing and predicting capacity degradation of a power battery. In order to solve the technical problems, the application adopts the following technical scheme that the application provides a power battery capacity attenuation testing and predicting method in the first aspect, and the method comprises the following steps of S1, acquiring multi-source characteristic data of a battery and integrating the multi-source characteristic data into a structural characteristic data set. S2, constructing and training a capacity attenuation prediction model based on the multi-source characteristic data. And S3, in the battery aging process, executing a double-drive self-adaptive test decision. And S4, when the dual-drive self-adaptive test decision is triggered, updating the capacity attenuation prediction model. And S5, outputting a prediction result and battery capacity attenuation mechanism analysis based on the updated capacity attenua