CN-120490875-B - Electric vehicle battery pack SOH estimation method for extracting polarization characteristics from dynamic working conditions of real vehicle

CN120490875BCN 120490875 BCN120490875 BCN 120490875BCN-120490875-B

Abstract

The invention relates to an SOH estimation method for an electric vehicle battery pack, which extracts polarization characteristics from a dynamic working condition of a real vehicle, and belongs to the technical field of batteries. Aiming at the problems that the existing method depends on laboratory data and complete charging segments and cannot effectively utilize polarization characteristics caused by current switching in dynamic working conditions, the invention analyzes SOC distribution at the current switching moment based on a multi-stage constant current charging curve by sorting battery operation data and preprocessing and extracting the charging segments, screens a characteristic extraction interval and sets a dynamic threshold value, extracts characteristics such as polarization voltage difference, current variation and the like from a representative monomer, and constructs a two-stage machine learning model to realize SOH estimation. The method breaks through the limitation of the traditional static characteristics, and the instantaneous polarization effect of current switching under the dynamic working condition is utilized to remarkably improve the utilization rate of real vehicle data and the characteristic characterization capability, effectively improve the evaluation precision of the health state of the battery pack and realize the large-scale rapid detection requirement of the electric vehicle.

Inventors

HU XIAOSONG
Wu Ranglei
LIU JIA
Liu Hongao
HUANG RUI
YI JINGSONG
LI JIACHENG
ZHANG KAI
LI CHANG
XIE YANG

Assignees

重庆大学

Dates

Publication Date: 20260512
Application Date: 20250514

Claims (7)

1. The method for estimating the SOH of the battery pack of the electric vehicle by extracting the polarization characteristic from the dynamic working condition of the real vehicle is characterized by comprising the following steps of: s1, arranging operation data from a battery pack of a hybrid electric vehicle and establishing a battery data set; S2, analyzing a battery data set, preprocessing, extracting a more complete charging segment, and calculating the capacity of the label by using an ampere-hour integration method; s3, analyzing current and SOC changes in the multistage constant-current charging curve based on dynamic working conditions, setting feature screening conditions, and selecting current switching time for extracting features, wherein the method specifically comprises the following steps: S31, analyzing corresponding SOC distribution conditions when the current of a charging segment of the battery pack is switched based on a rule that the current changes along with the SOC in the charging process, and dividing a feature extraction interval according to the SOC; S32, based on the selected characteristic extraction interval, counting current change conditions during current switching in the interval, and setting characteristic screening conditions according to the current change and the SOC; s33, counting the number of charging fragments at each moment of current switching, and selecting the current switching moment of the extracted features; S4, analyzing the conditions of each monomer of the battery pack, selecting a representative monomer, and extracting polarization characteristics of the representative monomer based on the selected current switching moment, wherein the method specifically comprises the following steps: S41, analyzing the conditions of each single body in the battery pack according to the selected current switching moment, and finding out a plurality of representative single bodies which can represent capacity fading of the battery pack at each moment; S42, extracting polarization characteristics of each representative monomer from the selected current switching time to form a polarization characteristic set; selection criteria for the representative monomers include: The first 3 monomers with the largest voltage change amplitude at the moment of current switching; the first 3 monomers with the smallest voltage change amplitude at the current switching moment; a single body with a voltage difference exceeding a set threshold before and after current switching; An abnormal monomer with a historical capacity attenuation rate exceeding 20% of the group average attenuation rate; And S5, establishing a machine learning estimation model, and taking the extracted polarization feature set as input to obtain an SOH estimation result of the battery pack.
2. The method for estimating SOH of battery pack of electric vehicle according to claim 1, wherein S2 is specifically: s21, analyzing charging data in a battery data set, preprocessing the charging data, and screening more complete charging fragments which accord with the capacity of a label calculated later; s22, calculating the current maximum available capacity of the battery pack by utilizing an ampere-hour integration method according to the extracted charging fragment data to obtain a capacity label trained by a subsequent model, wherein the mathematical expression for calculating the label capacity based on the ampere-hour integration method is as follows: Wherein, the And The battery packs are respectively arranged at time steps And Is used for the control of the SOC of (c), Is that the battery pack is in time step Is set to be a current of (a); is the maximum available capacity of the battery pack.
3. The method for estimating SOH of an electric vehicle battery pack for extracting polarization features from dynamic conditions of a real vehicle according to claim 1, wherein the polarization features comprise a maximum voltage difference among all the monomers during current switching, a voltage of a monomer corresponding to a maximum voltage difference during current switching at a previous time, a minimum voltage difference among all the monomers during current switching, a voltage of a monomer corresponding to a minimum voltage difference during current switching at a previous time, a maximum monomer voltage value during current switching, a voltage difference corresponding to a maximum monomer voltage value during current switching, a minimum monomer voltage value during current switching, a voltage difference corresponding to a minimum monomer voltage value during current switching, a voltage average value of all the monomers during current switching, a voltage difference corresponding to a voltage average value of all the monomers during current switching, a current value corresponding to a previous time, and a current difference corresponding to a current switching.
4. The method for estimating SOH of battery pack of electric vehicle according to claim 3, wherein S5 is specifically: S51, establishing a machine learning estimation model, inputting training data of the extracted polarization feature set into the model for training, and estimating SOH of the battery pack; S52, combining the selected current switching moments, verifying the effectiveness and universality of the extracted polarization characteristics, and realizing the rapid detection of the SOH of the vehicle.
5. The method for estimating SOH of battery pack for electric vehicle with polarization feature extracted from dynamic condition of real vehicle according to claim 1, wherein the feature screening condition is set such that current variation at current switching time is greater than 2A, and SOC value is at least two consecutive intervals selected from the group consisting of (61.5%, 66%), (71%, 76%), (80%, 85%), (95%, 97.1%) intervals.
6. The method for estimating SOH of battery pack of electric vehicle by extracting polarization features from dynamic working condition of real vehicle as claimed in claim 1, wherein SOC values corresponding to the selected current switching time are 71%, 85% and 95% of three feature points, and the number of charging segments corresponding to each feature point is not less than 100 historical charging cycles.
7. The method for estimating SOH of battery pack of electric vehicle by extracting polarization features from dynamic working condition of real vehicle as claimed in claim 3, wherein said machine learning estimation model adopts a two-stage architecture based on feature importance, the first stage filters polarization feature subset by random forest, the second stage adopts long-term memory network (LSTM) with attention mechanism to conduct SOH sequence prediction, wherein the attention weight distribution function is: Wherein, the A hidden state vector representing LSTM at time step t, with dimensions of Representing a memory coding result of the model on the historical polarization characteristic sequence; Representing a trainable parameter matrix for learning the importance mapping relationship of different time step features; A matching degree scalar value representing the calculation time step t feature and the global attention pattern; the expression is normalized for the matching degree of T for all time steps s=1, 2.. And is also provided with ; The attention weight of time step t is represented, reflecting the intensity of contribution of the polarization feature at that moment to the current SOH prediction.

Description

Electric vehicle battery pack SOH estimation method for extracting polarization characteristics from dynamic working conditions of real vehicle Technical Field The invention belongs to the technical field of batteries, and relates to an SOH estimation method for an electric vehicle battery pack, which extracts polarization characteristics from a real vehicle dynamic working condition. Background The current estimation method of the SOH of the lithium ion battery mainly comprises a model-based method and a data driving method. Based on the method of the model, a mathematical model of the battery degradation phenomenon is established on the basis of deep research of electrochemical mechanism, model parameter identification is carried out by utilizing a least square method and other optimization methods, and then SOH estimation of the battery is carried out on the basis of a Kalman filtering method and other methods, wherein the models comprise an equivalent circuit model and an electrochemical model. The data driving method regards the battery as a black box, and establishes a mapping relation between the characteristics and SOH (establishes an estimation model by using machine learning) by searching a large number of characteristics reflecting aging information of the battery (namely extracting health characteristics of the battery as input). Although a large number of SOH estimation methods exist at present, most of them are based on laboratory data and are difficult to apply to real vehicles. Furthermore, in terms of feature engineering, current methods rely on complete charge segments and do not take into account battery dynamic response characteristics due to current switching during real world vehicle operation. When the automobile is running, the concentration of an internal example of the battery is unbalanced in a short time due to the reciprocating movement of lithium ions between the positive pole and the negative pole in the current switching process, a polarization phenomenon is formed outwards, the phenomenon is more obvious along with the aging of the battery, and the battery can be used for evaluating the SOH of the battery. Most of the current researches focus on depolarization after full charge, and the method needs to be kept still for a long time after full charge, and in the real world, the vehicle is usually not recorded any more after full charge, so that relaxation voltage data are difficult to acquire, or the vehicle directly starts to run, so that large-scale application is difficult. Disclosure of Invention Therefore, the invention aims to provide the method for estimating the SOH of the battery pack of the electric vehicle, which is used for extracting the polarization characteristics from the dynamic working condition of the actual vehicle, can extract the polarization mechanism characteristics representing the battery aging process from the multistage constant current charging curve, realizes the accurate estimation of the SOH of the battery pack, and is applied to the large-scale SOH rapid detection of the electric vehicle. In order to achieve the above purpose, the present invention provides the following technical solutions: an electric vehicle battery pack SOH estimation method for extracting polarization characteristics from real vehicle dynamic working conditions comprises the following steps: s1, arranging operation data from a battery pack of a hybrid electric vehicle and establishing a battery data set; S2, analyzing a battery data set, preprocessing, extracting a more complete charging segment, and calculating the capacity of the label by using an ampere-hour integration method; S3, analyzing current and SOC changes in the multistage constant-current charging curve based on dynamic working conditions, setting feature screening conditions, and selecting current switching time for extracting features; S4, analyzing the conditions of each monomer of the battery pack, selecting a representative monomer, and extracting polarization characteristics of the representative monomer based on the selected current switching moment; And S5, establishing a machine learning estimation model, and taking the extracted polarization feature set as input to obtain an SOH estimation result of the battery pack. Further, the S2 specifically is: s21, analyzing charging data in a battery data set, preprocessing the charging data, and screening more complete charging fragments which accord with the capacity of a label calculated later; s22, calculating the current maximum available capacity of the battery pack by utilizing an ampere-hour integration method according to the extracted charging fragment data to obtain a capacity label trained by a subsequent model, wherein the mathematical expression for calculating the label capacity based on the ampere-hour integration method is as follows: Wherein, SOC (t) and SOC (t 0) are the SOC of the battery pack at time steps t and t 0, respectively, I (t)