TCN method development for SOC prediction of Li-ion batteries

Abstract

Lithium battery state-of-charge (SOC) estimation is a core function of battery management system, which directly affects the safety and range performance of electric vehicles. Aiming at the nonlinear modelling limitations of traditional methods under dynamic operating conditions and battery aging scenarios, temporal convolutional networks (TCNs) have become a research hotspot in the field of SOC estimation. This paper reviews the recent progress of TCN methods: through the introduction of attention mechanism, migration learning and hybrid architecture, it effectively solves the challenges of data missing sensitivity and poor cross-cell generalisation; combined with genetic algorithm, grey wolf optimisation and other strategies, it further optimises the network structure and hyperparameters.