Application of OOA-BP Neural Network in Battery Management Systems for SOH Prediction

Lithium-ion Battery life prediction is critical in utilizations like electric transportation, energy storage, and mobile equipment. Accurate prediction is essential for improving battery safety, reducing operational and maintenance costs, enhancing performance and reliability, and supporting sustainable development. Optimization algorithms have been extensively applied in State of Health (SOH) prediction to enhance model accuracy and reliability. However, these algorithms often face significant challenges when applied to complex battery systems. Traditional optimization algorithms often face challenges, including slow convergence, susceptibility to local optima, and high computational demands. These issues become even more problematic when dealing with complex, high-dimensional problems, where the solution space may not be fully explored. To tackle these issues, this paper introduces an innovative a new State of Health (SOH) prediction model using the Osprey Optimization Algorithm-Backpropagation Neural Network (OOA-BPNN) and tests its performance using the NASA battery dataset. The results show that the RMSE of the SOH prediction results for the B0005 and B0006 batteries is less than $\mathbf{0. 9 \%}$, and the MAPE of the SOH prediction results is under 0.8%. Despite the significant capacity recovery observed in the B0007 battery, leading to a slight increase in error, the RMSE remains below 1.6%, and the MAPE stays under 1.35%, both within acceptable limits. These findings reveal that the suggested model is effective in predicting battery SOH.