A SOC estimation method based on LSTM with polarization characteristics

Inhalt:
Utilizing the emerging technology neural network and other related methods to predict the state of charge (SOC) of battery has begun to become a popular technique. In this paper, an improved long short-term memory (LSTM) network is designed to predict the SOC of battery considering the potential polarization characteristics. The existing LSTM has the defect of insufficient capture of battery characteristics. This paper proposes a new memory structure to store the extracted potential polarization characteristics of battery SOC characteristics. Specifically, for engineering implementation, the method proposed in this paper is implemented by using the popular deep learning framework, Pytoch, and simulation experiments are carried out on the NASA battery data set. The experiments demonstrate that the method proposed in this paper is more accurate than the existing methods in SOC estimation.