Research On SOH Estimation Algorithm Of Power Battery Based On LSTM-RNN

The problem of environmental pollution and energy crisis is becoming more and more serious,which has attracted the attention of people from all walks of life at home and abroad.New energy electric vehicles have gradually become the long-term strategic direction of China’s automobile industry.The power battery system is the biggest bottleneck restricting the development of new energy electric vehicles in China,and the power battery management technology is the key technology to guarantee the high efficiency,safety and long-term stable operation of electric vehicles,as well as the core technology of electric vehicles competing with each other in various countries.Due to the limited parameters that can be accurately measured,the coupling between the characteristics and the strong nonlinearity of the power battery,the accurate estimation of SOH has always been the focus of academic research and the technical breakthrough difficulty.Aiming at the problem that it is difficult to accurately estimate the power battery SOH,based on the characteristic parameters extracted from the constant voltage charging(CV)curve,a SOH estimation algorithm based on Long Short-Term Memory Recurrent Neural Network(LSTM-RNN)was designed in this paper to improve the accuracy of SOH estimation.The main work of this paper is as follows:(1)The domestic and foreign SOH estimation methods are divided into two categories,and the accuracy and complexity of various SOH estimation methods are compared.Several common battery models are introduced.Based on the basic performance test of the battery,the main reasons for the degradation of lithium ion battery life are analyzed comprehensively from two aspects,including internal failure mechanism and external influencing factors.Furthermore,the parameters that can characterize the decline of battery capacity are summarized.(2)Extract data from a set of battery aging experiment data published by NASA,draw CV curves under different cycle times,and extractr four feature parameter,such as CV phase average current,CV phase duration,CV phase and other current drop time difference,CV phase charging capacity;then use the gray correlation analysis method to calculate the correlation degree and compare the correlation degree of each feature with SOH;finally,determine the best feature parameter according to the correlation degree with SOH.(3)Design an LSTM-RNN neural network model to estimate SOH,with CV curve features as input and SOH as output,the structure of the model is determined by the number of input features,and the structure of the SOH neural network estimation model is optimized using particle swarm optimization,And optimize each key parameter at the same time.(4)A set of battery aging data published by NASA was used to verify the designed LSTM-RNN neural network SOH estimation algorithm.Taking the life cycle data of B0005 battery in the cyclic aging data,the first 100 samples are used as the training set,and the rest are used as the test set to train the SOH estimation model.Enter the characteristic parameters of the B0007 battery into the trained SOH estimation model to verify the algorithm.The experimental results show that under the same conditions,the multi-featured LSTM-RNN model has higher SOH estimation accuracy,and the estimation accuracy reaches 2.14%.The features of this article are as follows:(1)In this paper,an LSTM-RNN neural network estimation model is designed based on the constant voltage charging data of the battery.This method can obtain the characteristic parameters from the battery voltage and current data collected by the battery management system,and establish the SOH estimation model through the offline data,so as to realize the online real-time SOH estimation.(2)Particle swarm optimization is used to optimize the key parameters of the LSTM-RNN estimation model to ensure that the key parameters are globally optimal.