State-of-Charge Estimation of Automotive Lithium-Ion Batteries during Lack of Input Data Update

Accurate states monitoring and timely diagnosis of energy storage device for electric vehicles are important, and they depend on the data collected by a large number of sensors. High-frequency vibration and loose connectors make the local sensors fail in the long term, which causes the abnormal collection of data and states monitoring. In this article, a state estimation method based on the bidirectional long-short term memory (Bi-LSTM) and least square support vector regression (LSSVR) is proposed to correct the abnormal data when measured data are missing or stagnant. The simultaneous input and state estimation (SISE) algorithm is used in state monitoring. This method combines SISE algorithm and the dual data-driven correction mechanism based on Bi-LSTM and LSSVR, and SISE algorithm is used to solve the problem of input data anomaly in state-of-charge estimation, which enables simultaneous estimation and correction of both system input and state. In experiments, the error of the proposed method is kept at about 5% at six cases of mixed abnormal test conditions, thus it is proved to be effective.