Battery State of Health Estimation Based on Multiple Linear Regression by Particle Swarm Optimization

Abstract

State of health of battery is crucial for electric vehicles, which has been widely studied through many different methods where incremental capacity analysis (ICA) is a common method for battery state of health estimation. However, previous methods are difficult to find suitable feature information from multiple features to make the estimation result better, especially considering that the ICA curves of different batteries have different peak numbers and positions. In this paper, in order to acquire accurate peak coordinates, the Gaussian filtering is used to make the incremental capacity (IC) curve smoother, and then the method that using multiple linear regression by particle swarm optimization (PSO) is proposed to solve the problems mentioned above. Compared with the traditional multiple linear regression, multiple linear regression by particle swarm optimization selects the most appropriate variables to improve the accuracy of estimation. In the experiment, a group of peak data of series module of battery is selected for linear regression to get the optimal variables and corresponding parameters, and then the corresponding verification is carried out by using the data of six different aging states of another series module of battery. The experimental results show that except the sixth group of the state of health estimation error is 2.09%, the estimation error of the remaining five groups is less than 2%, so this method has high accuracy and adaptability for series module of battery.