Prediction of SOC of lead-acid battery in pure electric vehicle based on BSA-RELM

Abstract

The state of charge (SOC) of lead-acid batteries plays an important role in battery management systems. Aiming at the problem that the SOC of lead-acid batteries in pure electric vehicles cannot be measured directly, a novel algorithm based on a robust extreme learning machine (RELM) optimized by the bird swarm algorithm (BSA) is proposed in this paper. RELM overcomes the shortcoming that an extreme learning machine (ELM) cannot deal with outliers and improves the prediction accuracy of ELM. BSA is used to optimize the numbers of hidden layer nodes and the adjustment factors of RELM. The problem that the numbers of the hidden layer nodes and the adjustment factor are difficult to determine is solved, which can further improve the convergence speed of RELM and help to find the global optimal value. The main parameters which affect the SOC of batteries such as current, voltage, temperature, and internal resistance have been collected by the advanced vehicle simulation software ADVISOR to model and test. Simulation results show that compared to other algorithms such as back propagation neural network, radical basis function neural network, and fuzzy neural network, BSA-RELM has a higher prediction accuracy.