Algorithm of BPNN‐UKF based on a fusion model for SOC estimation in lithium‐ion batteries

Lithium-ion batteries (LIBs) are widely used as energy storage devices. Accurately estimating the state of charge (SOC) is critical for safe operation LIBs, but complex internal characteristics LIBs make this a difficult task. In study, an algorithm called BPNN-UKF, which combines backpropagation neural network (BPNN) and unscented Kalman filtering (UKF), proposed. First, fusion model construction method incorporating AdaBoost recursive least squares (RLS) developed to enhance accuracy generalization battery model. Next, applied create reduces dependence UKF training requirements BPNN. method, estimates SOC, then BPNN learns nonlinear relationship between SOC estimation error process variables in preliminary estimation. The output update vector real time during new process. Experimental results reveal that proposed can drastically improve compared with simple combined it exhibits considerable common disturbances various working conditions.