State of charge estimation of lithium-ion battery based on TSCSO-GRU-Attention

The precise estimation of the State of Charge (SOC) for lithium-ion batteries is paramount for guaranteeing their secure operation. To augment the precision of SOC estimation, this paper introduces a novel TSCSO-GRU-Attention model. This model amalgamates the Sand Cat Swarm Optimization (SCSO) algorithm, which is optimized by a triangular traversal strategy, and the Gated Re-current Unit (GRU) neural network, equipped with an attention mechanism. This model harnesses the robustness and accuracy of GRU and attention for SOC estimation, enhances the local search proficiency of the SCSO algorithm via the triangular traversal strategy, and employs the refined SCSO algorithm to optimize the three hyperparameters: the number of neurons, learning rate, and batch size prior to feeding into the GRU-Attention network model, thereby enhancing the precision of SOC estimation. This paper proposes a model that accurately predicts the SOC of lithium batteries. Experimental results demonstrate that when compared to the SCSO-GRU-Attention model, the proposed model achieves reductions in MAE, RMSE, and MAPE by 42.84%, 41.94%, and 2.25% points respectively, thus rendering the SOC estimation results more precise.