Abstract:

The prediction of the state of charge (SOC) of lithium-ion batteries by the least squares support vector machine (LSSVM) shows a faster convergence speed and gives an extraordinary method for the global optimal solution. The prediction ability is enhanced more than ever before. However, the parameter selection of the LSSVM will greatly affect the prediction result. A prediction method for the SOC of lithium-ion batteries by estimating the distribution algorithm (EDA) with an LSSVM is proposed herein. The operating voltage, current, and temperature of the lithium-ion batteries are used as the input quantities. The SOC of the batteries is used as the output quantity. Moreover, a non-linear system model is built using the LSSVM. The EDA is designed to optimize the regularization parameter and the radial basis kernel width of the model. We then obtain the optimal model. The simulation results show that compared with the conventional prediction model for the SOC of lithium-ion batteries, the proposed EDA-LSSVM method has a higher prediction accuracy for the SOC.

Key words: lithium-ion battery, prediction for state of charge, estimation of distribution algorithm, least squares support vector machine