Kalman Filter Improved Neural Network State of Charge Estimation Strategy of Zinc Bromine Flow Battery Considering Sampling Inaccuracy
Aiming at the difficulty in state of charge estimation of Zinc Bromine flow battery caused by its own characteristics,a Kalman Filter improved neural network state of charge estimation strategy is proposed.The core of this work is to solve the nonlinear characteristic caused by the variable pump speed through the neural network model.Meanwhile,the self-discharge current is incorporated into the state equation of Kalman filter to solve the self-discharge and sampling noise problems of Zinc Bromine flow battery.Validated by site operation data,the mean absolute error of proposed strategy is 0.98%and the max absolute error is 3.73%,which is superior to the individual use of Ah integral,the unscented Kalman filter and the neural network methods.Furthermore,the proposed strategy can maintain good result when single sensor sampling is inaccurate,which can meet the needs of long-term energy storage applications.
Zinc Bromine flow batterystate of charge estimationKalman filterneural networkinaccurate sampling
万方数据
维普
