An Innovative Method for the State of Charge Estimation Utilizing an Enhanced Firework Algorithm and Levenberg-Marquardt Optimization Neural Network

The state of charge (SOC) denotes a fundamental aspect of battery management systems (BMS). Reliable SOC calculation is vital to guarantee the optimal performance and risk-free use of electrical instruments. This research presents an advanced method for SOC estimation, utilizing a Backpropagation Neural Network refinement using the Levenberg-Marquardt algorithm (LMBPNN) for preliminary estimation. Building upon the LMBPNN, this study employs an enhanced Fireworks Algorithm (IFWA) for further optimization, conducting a comparative analysis with Particle Swarm Optimization (PSO) and the conventional Fireworks Algorithm (FWA). Comparative experiments were conducted using three typical driving conditions for cross-validation. The proposed intelligent SOC estimation method demonstrated high accuracy under the New European Driving Cycle (NEDC), as evidenced by the lessening of the average absolute error (AAE) to 1.4% and the root mean square error (RMSE) to 1.7%.