Fuzzy clustering based multi-model support vector regression state of charge estimator for lithium-ion battery of electric vehicle

Based on fuzzy clustering and multi-model support vector regression, a novel lithium-ion battery state of charge (SOC) estimating model for electric vehicle is proposed. Fuzzy C-means and Subtractive clustering combined algorithm is employed to implement the fuzzy partition for the input space with the input vectors sampled in UDDS drive cycle, temperature, current, load voltage of the lithium-ion battery pack. For each cluster of training samples, support vector regression is applied to achieve the estimating sub-model dependent on the corresponding cluster centre. Then SOC estimating model is determined by the synthesis of all the submodels with the introduction of fuzzy membership values. Simulation results indicate that this model is able to effectively reduce the negative influence from outliers and the mean relative training error and the validating error fall by respectively 22% and 27.3%, compared to counterparts of the standard support vector regression model, which proves the achieved SOC estimating model has a high accuracy.