Battery Health Prediction Based on Multi-Parameter Charging Feature Identification

The estimation of battery health plays a crucial role in battery management and system safety. With the widespread use of batteries in various applications, such as electric vehicles, portable devices, and renewable energy systems, accurately extracting battery charging characteristics has become essential. This paper proposes a multi-channel charging feature extraction method. This method involves extracting degradation features from the voltage and current charging curves as health indicators. Heatmaps are used to demonstrate the strong correlations among these features. The method was validated using the NASA battery dataset and evaluated through various representative machine learning techniques to achieve accurate battery health assessment. Multiple performance metrics were employed for validation. The experimental results indicate that, compared to models trained with single features, the models trained with the proposed features exhibit higher accuracy, with prediction metrics falling below 4%.