State of Charge Estimation for Zinc-Bromine Flow Batteries by Improved Long Short-Term Memory Network and Kalman Filter

In order to improve the accuracy of estimating the state of charge (SOC) of zinc-bromine flow batteries (ZBFB) in the discharge stage and overcome the problems caused by sudden voltage changes, this paper proposes an algorithm for estimating and correcting the SOC. The core contribution is the proposal of an improved Long Short-Term Memory (LSTM) neural network algorithm, which considers the effect of the circulating pump speed on the battery voltage during the discharge stage, and establishes a neural network model for the battery. The estimation result of the improved LSTM network has a mean square error of 3.322, representing a 12% improvement in estimation accuracy compared to the traditional LSTM network. Additionally, Kalman filtering is employed for error mitigation to further enhance estimation accuracy. After applying the proposed method, the maximum error between the estimated SOC and the actual SOC is 1.91%, and the average error is 0.30%. This indicates that the proposed method improves the accuracy of SOC estimation through enhanced real-time LSTM network estimation and Kalman filtering error cancellation, particularly during voltage sudden change in the discharge of ZBFB.