A Hybrid WOA-DeepONet Framework for Data-Driven and Physics-Guided SOH/RUL Estimation in Lithium-Ion Batteries

INTRODUCTION: For energy storage systems to be safe, effective, and reliable, it is essential to accurately forecast the State of Health (SOH) and Remaining Useful Life (RUL) of lithium-ion batteries under complicated operating situations. OBJECTIVES: This research suggests a hybrid modeling approach that combines a physics-informed Deep Operator Network (DeepONet) supplemented by encoders with a neural network optimized by the Whale Optimization Algorithm (WOA). METHODS: The framework first utilizes WOA to improve the initialization of Backpropagation (BP) neural networks, thus enhancing convergence speed and avoiding local minima in early-stage training. Then, a multi-physics informed DeepONet model is constructed to learn the spatiotemporal evolution of battery thermal and electrochemical variables from virtual heating profiles. RESULTS: By integrating boundary conditions, starting restrictions, and partial differential equation (PDE) residuals into the loss function, the model integrates physical supervision and guarantees forecast consistency with battery dynamics. CONCLUSION: Both COMSOL-generated synthetic data and publicly available battery aging datasets are used in extensive research. The suggested approach outperforms conventional MLP, LSTM, and even standard DeepONet models, according to the results, with an R2 of 0.99 and an MSE of 0.0009.