Research on Online Correction of SOC estimation for Power Battery Based on Neural Network

Abstract

In the power battery SOC estimation method, the Coulomb Counting (CC) method is greatly affected by the initial state and integral error of the battery, which can be improved by combining the neural network (NN) method. However, the traditional NN basically establishes an offline network through the priori data. When the battery status changes, using the original offline model will cause larger errors. In this paper, a method based on online NN for correcting the CC is proposed. The advantage of the online network model is that it can track the changes of the battery model in real time. It can not only provide a reliable initial value and correct its integration error, but also can solve the problem of available capacity change caused by battery aging effectively. Experiments show that the use of online learning to correct the CC method compares to the offline BP network, has a good effect, meet the system accuracy requirements.