| With the introduction of environmental protection policies in various countries,the goal of energy conservation and emission reduction planning,such as carbon peak and carbon neutrality,is gradually clear.The development of clean energy has become the consensus and inevitable trend of the current international community.Battery is the only power source for pure electric vehicles.The development of battery-related technology determines the development of electric vehicle industry,and the battery management level also determines the efficiency of electric vehicles to use batteries.In order to ensure the safe running of electric vehicles and the reliable operation of electrical equipment,Battery Management System (BMS)must be used for comprehensive Management of batteries.Among them,State of Charge (SOC)estimation,State of Power (SOP) estimation,State of Health (SOH) estimation and balance strategy module are the core functions of battery management.Accurate estimation of lithium battery SOH is of great significance to the safety,reliability and energy efficiency of electric vehicles.Accurate SOH estimation is conducive to good monitoring of the health of lithium batteries,prevention of safety accidents caused by battery aging,reasonable decommissioning or replacement of batteries in use,stepwise utilization evaluation of decommissioning power batteries,and improvement of the life service value of lithium batteries.The following studies are carried out in this paper.Basic research on lithium battery reaction principle and SOH estimation.Firstly,the types of lithium batteries,principles of electrochemical reaction and experimental steps of lithium battery aging are introduced.Then,different definitions of SOH of lithium batteries in the current industry are discussed,and the capacity is determined as the evaluation index of SOH in this paper.Then,the aging mechanism of lithium battery is explored,the inducement of aging is classified and discussed,and the aging process of lithium battery is described.Finally,the performance degradation caused by aging of lithium battery is reviewed.It lays a foundation for extracting the aging characteristic data of lithium battery and estimating SOH accurately.(1) Basic research on lithium battery reaction principle and SOH estimation.Firstly,the types of lithium batteries,principles of electrochemical reaction and experimental steps of lithium battery aging are introduced.Then,different definitions of SOH of lithium batteries in the current industry are discussed,and the capacity is determined as the evaluation index of SOH in this paper.Then,the aging mechanism of lithium battery is explored,the inducement of aging is classified and discussed,and the aging process of lithium battery is described.Finally,the performance degradation caused by aging of lithium battery is reviewed.It lays a foundation for extracting the aging characteristic data of lithium battery and estimating SOH accurately.(2) Lithium battery aging data analysis and aging feature extraction.Firstly,the aging data of lithium battery were processed and analyzed,and the relationship between external characteristics of lithium battery and aging was obtained from the relationship between voltage,current and capacity of battery under different cycles.Increment Capacity Analysis (ICA) was used to transform the voltage and Capacity relationship of the charging data in the aging experiment into Increment Capacity (IC) curve.Gaussian filter is used to reduce the noise of IC curve.At the same time,combining with the relationship between the number of cycles of lithium battery and IC curve,the evolution trend of IC curve with THE decrease of SOH and the underlying causes are analyzed from the perspective of mechanism.Finally,from the above analysis,it was determined that the data between 3.8V and 4.1V during the charging process of lithium battery were the most representative and could best characterize the SOH of lithium battery.Eleven characteristic IC points were extracted at 30mV intervals,and the extracted feature points were preprocessed according to the characteristics of the SOH estimation model built subsequently.As the model input for subsequent SOH estimation.(3) SOH estimation of lithium battery based on support vector machine.Firstly,the theory of Support Vector Regression (SVR) is deduced in detail and its theoretical basis is determined.Then,the influence of SVR model parameters on SOH estimation accuracy was analyzed,and the Particle Swarm Optimization (PSO) method was used to optimize SVR model parameters,and the SOH estimation model parameters were finally determined,so as to give better play to the ESTIMATION ability of SVR model.Finally,the model was trained and tested respectively through the data of training set and test set.The experimental results on the one hand verified that the aging characteristic points extracted from IC curve could effectively reflect the SOH of lithium battery,and on the other hand proved that SVR model could accurately estimate the SOH of lithium battery.The Root Mean Squared Error (RMSE)and Mean Absolute Error (MAE) are less than 3%,and the maximum Error (MAX) is 5%.(4) SOH estimation based on variational regression encoder.First,the structure of Auto Encoder (AE) and its dimension reduction principle are introduced,and then Variational Auto Encoder (VAE) model is proposed to improve the AE implicit variable without constraint.Then the VAE mathematical theory is determined by the variational inference theory,and the dimension reduction principle is explained from the perspective of probability.Then the VAE model is derived from the VAE model,according to the IC curve extracted SOH characteristic data characteristics,through a lot of experimental tests,determine the VAE model structural parameters.Finally,the regression VAE models were trained and tested respectively,the results show that the effectiveness of the aging characteristics extracted from IC curve to represent the SOH of lithium batteries.The estimation results of the regression VAE models are MAE and RMSE evaluation index is less than 2%,MAX evaluation index is less than 4%.Compared with SVR models,regression VAE models have higher accuracy and better performance. |
PDF Full Text Request