Research On Health State Estimation And Remaining Useful Life Prediction Of Lithium-ion Batteries

In recent years,driven by national policies and the market,the industrialization of new energy vehicles has developed rapidly.In the process of promotion and application of new energy vehicles,the health status and performance indicators of power battery have an important impact on the vehicle’s power performance and endurance,but also seriously affect the safety of the vehicle.Therefore,to carry out research and development of new energy automobile power battery management system,achieve more accurate health state estimation and residual life prediction of lithium-ion batteries for vehicles,and then formulate a scientific and reasonable battery charging and discharging strategy,is great significance to improve the comprehensive performance of vehicles,ensure the safe operation of battery packs and delay battery aging,and is also one of the key issues to speed up the further popularization of new energy vehicles.Based on a school-enterprise cooperation project,this paper takes lithium-ion battery for vehicle as the research object.Aiming at the problem that it is difficult to obtain the internal state in the process of li-ion battery life decline,the paper focuses on the health feature construction and state estimation algorithm development of li-ion battery,the method of battery health state estimation and remaining useful life prediction based on data drive was deeply studied.The health characteristics of automotive lithium-ion batteries were constructed by using incremental capacity analysis,and the health state estimation method based on long short-term memory networks and the remaining life prediction method based on Gaussian process regression were developed.The results of the algorithm are verified by the measured battery cycle charge-discharge data.The research and experimental results show that the proposed algorithm can effectively improve the accuracy of battery health state estimation and remaining useful life prediction.The specific research of this paper are as follows:From the theoretical level,the working principle and decay mechanism of lithium ion battery were analyzed.Based on the battery experimental data set,the influence of discharging ratio,average SOC and depth of discharge on battery cycle life was studied.The influence of SOC and temperature on battery calendar life was also studied.It provides reference and guidance for the extension of battery life and the development of BMS algorithm,and also provides theoretical and data support for the following chapters.Based on the incremental capacity analysis and the internal decay mechanism,an improved calculation method of incremental capacity analysis is used to calculate the battery charge data,which solves the problems of noise and feature disappearing easily in the application of traditional methods.According to the obtained results,the decay mode in the battery was analyzed.At the same time,the battery health features were extracted from the calculated results,and the correlation between the health features and the battery health state was verified by the correlation analysis method,which provided a foundation for the subsequent battery health state estimation.Using the health characteristics and NASA battery data as the research object,in view of the battery health decline of the phenomenon of long-term dependence on timing,long short-term memory networks model is introduced,the characteristics of health as the input,battery state of health as the output,the health state of the battery is used to estimate the real-time.Feedforward neural network and support vector regression were introduced to compare the estimation accuracy,and different batteries were used to verify the mobility of the model.The results show that the error of the proposed algorithm is less than 1% when applied to its own SOH estimation,and less than 2% when transferred to other batteries.The proposed algorithm has better estimation accuracy than the other two algorithms,and also has strong portability.Taking CALCE battery data as the research object,a prediction process was designed based on Gaussian process regression algorithm and combining with specific problems of battery life prediction.With the number of cycles as the input and the remaining useful life as the output,the model kernel was configured and the model parameters were optimized by maximum likelihood estimation and conjugate gradient descent method.Finally,the single point prediction results and long-term prediction results of batteries remaining useful life are obtained.In order to verify the prediction effect of the model,the single exponential model in the empirical model is introduced as a comparison.The results show that the average error of residual life prediction obtained by the model based on Gaussian process regression algorithm is less than 8%,which is more than 30%less than that of the single exponential model.Meanwhile,the model based on Gaussian process regression algorithm has the advantages of strong interpretability,adaptability and uncertainty expression ability.This paper provides an accurate,efficient and adaptable method for health state estimation and residual life prediction of lithium-ion batteries,and its research results can provide a reference for the development of battery management system for electric vehicles.