CN-121114835-B - Battery health state evaluation method, system, computer equipment and medium

CN121114835BCN 121114835 BCN121114835 BCN 121114835BCN-121114835-B

Abstract

The invention provides a battery health state assessment method, a system, computer equipment and a medium, which belong to the technical field of battery health management and life prediction, and the method comprises the steps of carrying out full life cycle cyclic charge and discharge test on a reference battery, collecting data samples at fixed sampling intervals, and extracting fusion feature vectors of the reference battery samples; the method comprises the steps of training a reference prediction model by using a reference data sample, collecting fusion feature vectors of a target battery to be detected, calculating the distribution distance between the target and the characteristics of the reference battery by using an MMD algorithm, establishing a double-loss function by combining the prediction loss of the reference battery, taking the minimum total loss as a target, adjusting reference model parameters by reverse propagation iteration to obtain an optimization model of the adaptive target battery, and accurately predicting the SOH value of the target battery based on the optimization model. The method greatly reduces the data acquisition cost and the model deployment period in the migration stage, and realizes the migration of the prediction model in different battery equipment with low cost and high precision standard.

Inventors

CHEN GAIGE
DENG WEI
WANG WENWEI
LIU HUI
LIU JINSONG
HAN AIMIN
GENG JIA
YANG YAHONG
ZHANG XIAOSHEN
BAI JIAXIN
WANG QIANG
DONG ZHICHAO
YANG XIONGJI

Assignees

西安邮电大学

Dates

Publication Date: 20260508
Application Date: 20251010

Claims (10)

1. A battery state of health evaluation method, comprising: Carrying out cyclic charge and discharge test on a reference battery, synchronously collecting time-stamped charge current and charge voltage in a charge stage and time-stamped discharge current and discharge voltage in a discharge stage in each round of charge and discharge process at fixed sampling intervals; The method comprises the steps of drawing a voltage-discharge capacity curve in a discharge stage based on a discharge current and a discharge voltage, calculating a differential capacity curve by using the voltage-discharge capacity curve, determining a maximum peak value of the differential capacity curve, determining a maximum characteristic peak area according to the maximum peak value, calculating a local discharge capacity of the maximum characteristic peak area, determining a maximum voltage time point of a charging voltage, determining a time point of a current from charging to discharging according to the charging current and the discharge current, and fusing the maximum peak value, the local discharge capacity, the maximum voltage time point and the current of each cycle from charging to discharging according to weights to obtain a plurality of groups of fusion characteristic vectors; The method comprises the steps of training a reference prediction model by utilizing a plurality of groups of fusion feature vectors, outputting SOH predicted values of reference batteries, calculating the predicted losses of the SOH predicted values of the reference batteries and real labels, obtaining fusion feature vectors of current charge and discharge data of a target battery to be tested, calculating feature distribution distances of the fusion feature vectors of the target battery to be tested and the reference batteries, constructing a double-loss function based on the predicted losses and the feature distribution distances, iteratively adjusting weights and biases of the reference prediction model by using a back propagation algorithm with the minimum double-loss function as a target, and obtaining an optimized prediction model when the double-loss function value is smaller than a set threshold; And based on the optimized prediction model, evaluating the SOH value of the health state of the target battery to be tested.
2. The method for evaluating the state of health of a battery according to claim 1, wherein the maximum peak value, the local discharge capacity, the maximum voltage time point and the current are weighted and fused by using a CNN network to obtain a fused feature vector.
3. The method of claim 1, wherein the reference prediction model is a gated loop unit GRU network, and further comprising evaluating a prediction result of the gated loop unit GRU network using a mean square error, a root mean square error, a mean absolute error, and a decision coefficient.
4. The method for evaluating the state of health of a battery according to claim 1, wherein the feature distribution distance of the fusion feature vector of the target battery to be tested and the reference battery is calculated by using an MMD algorithm.
5. The method of claim 1, wherein the step of plotting the voltage-discharge capacity curve of the discharge phase based on the discharge current and the discharge voltage comprises integrating the discharge current of the discharge phase over time to obtain the discharge capacity, and plotting the voltage-discharge capacity curve of the discharge phase using the voltage and the discharge capacity of the discharge phase.
6. The method for evaluating the health state of a battery according to claim 1, wherein the method for determining the maximum characteristic peak area according to the maximum peak value is characterized by calculating the local discharge capacity of the maximum characteristic peak area, and specifically comprises the steps of determining minimum value points adjacent to the maximum peak value, calculating the area of a closed area formed by connecting the maximum peak value with the minimum value points adjacent to the left and the right, wherein the area of the closed area is the local discharge capacity of the maximum characteristic peak area, and the local discharge capacity represents the total capacity change in a voltage interval corresponding to the maximum characteristic peak.
7. The method for evaluating the state of health of a battery according to claim 1, wherein before the fusion feature vector of the current charge and discharge data of the battery to be tested is obtained and the feature distribution distance of the fusion feature vector of the battery to be tested and the reference battery is calculated, the method further comprises the step of constructing historical cyclic charge and discharge data of the battery to be tested, wherein the historical cyclic charge and discharge data comprises charging current and charging voltage with time stamps in a charging stage and discharging current and discharging voltage with time stamps in a discharging stage of N-round charge and discharge of the battery to be tested, and the reference prediction model is optimized by utilizing the feature set of the battery to be tested.
8. A battery state of health evaluation system, comprising: The data acquisition module is used for carrying out cyclic charge and discharge test on the reference battery, synchronously acquiring the charging current and the charging voltage with the time stamp in the charging stage and the discharging current and the discharging voltage with the time stamp in the discharging stage in the charge and discharge process of each round at fixed sampling intervals; The characteristic fusion module is used for drawing a voltage-discharge capacity curve in a discharge stage based on a discharge current and a discharge voltage, calculating a differential capacity curve by utilizing the voltage-discharge capacity curve, determining a maximum peak value of the differential capacity curve, determining a maximum characteristic peak area according to the maximum peak value, calculating a local discharge capacity of the maximum characteristic peak area, determining a maximum voltage time point of a charging voltage, determining a time point of a current from charging to discharging according to the charging current and the discharge current, and fusing the maximum peak value, the local discharge capacity, the maximum voltage time point and the current of each cycle from charging to discharging according to weights to obtain a plurality of groups of fusion characteristic vectors; The model training module is used for training a reference prediction model by utilizing the multiple groups of fusion feature vectors and outputting SOH predicted values of the reference battery, calculating the predicted loss of the SOH predicted values of the reference battery and a real label, obtaining the fusion feature vector of the current charge and discharge data of the target battery to be tested, calculating the feature distribution distance of the fusion feature vector of the target battery to be tested and the reference battery, constructing a double-loss function based on the predicted loss and the feature distribution distance, taking the minimum double-loss function as a target, iteratively adjusting the weight and the bias of the reference prediction model through a back propagation algorithm, obtaining an optimized prediction model when the double-loss function value is smaller than a set threshold, and evaluating the SOH value of the health state of the target battery to be tested based on the optimized prediction model.
9. A computer device comprising a memory, a processor and a computer program stored on the memory, characterized in that the processor executes the computer program to carry out the steps of the method according to any one of claims 1 to 7.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when loaded by a processor, is able to carry out the steps of the method according to any one of claims 1 to 7.

Description

Battery health state evaluation method, system, computer equipment and medium Technical Field The invention belongs to the technical field of battery health management and life prediction, and particularly relates to a battery health state assessment method, a system, computer equipment and a medium. Background With the rapid development of new energy automobiles, renewable energy storage systems and intelligent electronic devices, lithium ion batteries are increasingly used as key energy storage elements. The accurate assessment and residual life prediction of the State of Health (SOH) of the battery are key functions of the battery management system, and have important significance for improving the service efficiency of the battery, reducing the maintenance cost and prolonging the service life of the system. Currently, lithium ion battery SOH prediction methods are mainly divided into three types, namely a direct measurement method, a model driving method and a data driving method. The direct measurement method can realize high-precision estimation under certain ideal working conditions by monitoring physical quantities such as internal resistance, electric conduction and the like of the battery, but is easily influenced by sensor precision and environmental factors in complex application scenes, thereby limiting the practicability. The model driving method generally depends on an equivalent circuit model or an electrochemical mechanism model, and has a certain theoretical basis, but the modeling process has higher requirements on professional knowledge and parameter identification precision, and is difficult to adapt to different battery types. In contrast, the data driving method becomes the main stream direction of current research and engineering practice by virtue of the advantages of independence of physical modeling, strong generalization capability and the like, and is widely used for battery state estimation and degradation trend modeling especially under the support of deep learning and big data technology. The prior art has a certain limitation in predicting the health state of the lithium ion battery, and comprises the steps that a training model often depends on data under specific types or working conditions, so that the training model is difficult to migrate to other batteries or scenes, part of the method is complex in calculation, difficult to operate efficiently on an embedded platform with limited resources, and unfavorable for actual engineering deployment. Disclosure of Invention The invention provides a battery state of health assessment method, a system, computer equipment and a medium, which are used for solving the problems of difficult migration, low precision and poor suitability of the existing battery state of health assessment model when adapting to different battery types. In order to achieve the above object, the present invention provides a battery state of health evaluation method, comprising: Performing cyclic charge and discharge test on a reference battery, and synchronously collecting time-stamped charge current and charge voltage in a charge stage and time-stamped discharge current and discharge voltage in a discharge stage in each charge and discharge process at fixed sampling intervals; and synchronously collecting the SOH real label of the battery health state corresponding to each cycle. Drawing a voltage-discharge capacity curve in a discharge stage based on a discharge current and a discharge voltage, calculating a differential capacity curve by using the voltage-discharge capacity curve, determining a maximum peak value of the differential capacity curve, determining a maximum characteristic peak area according to the maximum peak value, calculating a local discharge capacity of the maximum characteristic peak area, determining a maximum voltage time point of a charging voltage, determining a time point of a current from charging to discharging according to the charging current and the discharge current, and fusing the maximum peak value, the local discharge capacity, the maximum voltage time point and the current of each cycle from charging to discharging according to weights to obtain a plurality of groups of fusion characteristic vectors. The method comprises the steps of training a reference prediction model by utilizing a plurality of groups of fusion feature vectors, outputting SOH predicted values of reference batteries, calculating the predicted losses of the SOH predicted values of the reference batteries and real labels, obtaining fusion feature vectors of current charge and discharge data of a target battery to be detected, calculating feature distribution distances of the fusion feature vectors of the target battery to be detected and the reference batteries, and adjusting parameters of the reference prediction model by utilizing the predicted losses and the feature distribution distances to obtain an optimized prediction model. And based on the o