Optimal BP neural network algorithm for state of charge estimation of lithium-ion battery using PSO with Levy flight

The battery state of charge (SoC) of lithium batteries for electric vehicles is highly non-linear and time-varying. The convergence speed is slow and the accuracy is low when using ordinary neural network models for SoC estimation. The particle swarm optimization (PSO) algorithm based on Levy's flight strategy (LPSO) is proposed to optimize the weights and thresholds of BP neural network, which would improve the prediction accuracy of SoC. According to the mechanism of lithium battery charging and discharging, voltage, current and temperature are selected as input vectors and SoC is selected as output vector. The comparison of the model before and after optimization is carried out by using NASA lithium battery charging and discharging data, which shows this method has better generalization ability and high prediction accuracy. It has practical application significance for SoC estimation.