A combined improved dung beetle optimization and extreme learning machine framework for precise SOC estimation

Accurate estimation of the state of charge (SOC) of lithium-ion batteries (LiBs) proportionally impacts the efficiency of battery management systems (BMS) considering the dynamic and non-linear behavior of LiBs. Changes in the activities of the cathode and anode materials and internal resistance tend to impact the battery capacity. When the battery is operated at high or low temperatures or under the HWFET condition, battery capacity tends to deteriorate drastically. Therefore, high-precision SOC estimation is required to ensure safe and stable operation. In this work, we propose a combined Improved Dung Beetle Optimization (IDBO) and Extreme Learning Machine (ELM) framework for SOC estimation and evaluate the efficiency of the BMS. The novelty of the model stems from the application of the IDBO algorithm, which incorporating Circle chaotic mapping, the Golden sine strategy, and the Levy flight strategy, for hyper-parameter optimization. This effectively resolves the problems of inconsistent performance and instability arising from randomly initialized hidden layer weights and biases in ELM, resulting in enhanced prediction accuracy. The proposed IDBO-ELM method is validated in the context of five parameters, namely, different ambient temperatures, operating conditions, battery materials, initial SOC values, and running time. The experimental results show that the error ranges of both MAE and RMSE of the proposed model for SOC estimation under different conditions are around 1.4%, demonstrating high precision and robustness. The MAE and RMSE of SOC estimation decreased by more than 30%, respectively, compared to those by DBO-ELM. The model provides strong support for the safe and efficient application of LiBs under various practical conditions.