State of Health Prediction of Lithium-Ion Batteries Based on IDO-GRU

Accurate State of Health (SOH) estimation for Lithium Batteries (LIBs) is essential for reliable energy management. However, complex nonlinear degradation and noise-sensitive data make it difficult. To address these limitations, this study proposes an advanced framework integrating the improved Dandelion Optimization (IDO) algorithm with a Gated Recurrent Unit (GRU) neural network. For a start, the proposed method extracts critical health indicators from raw battery cycling data. To mitigate the adverse effects of noise during data acquisition, a Gaussian filtering technique is applied to preprocess the extracted features. Next, the GRU network’s hyperparameters are optimized by an improved Dandelion Optimization algorithm. The IDO first adopts a Euclidean distance strategy, which is designed to enhance population diversity. Second, it integrates a golden sine search mechanism to improve the ability of local exploitation. Third, it applies adaptive inertia weights, whose main function is to balance the global exploration and local exploitation capabilities of the algorithm. Experiments on NASA and CALCE datasets show remarkable accuracy. The results show that the maximum RMSE is 0.0048 and MAPE is 0.50% on the CALCE dataset, while on the NASA dataset, RMSE is 0.0013 and MAPE is 0.11%. The proposed framework demonstrates robust adaptability across diverse battery chemistries, thereby offering a novel and scalable solution for real-world SOH monitoring in electric vehicles systems.