An Early Prediction Method for Battery Pack SOH Based on Spatio-Temporal Multi GCN

With the widespread application of battery energy storage systems, predicting the State of Health (SOH) of battery packs is crucial for system safety and lifespan management. Existing methods primarily focus on individual cells rather than the entire battery pack, neglecting the multidimensional correlations and synergistic effects between cells. In this paper, we propose a spatio-temporal multi-graph convolutional network (GCN) driven early prediction framework for battery pack SOH. By constructing multidimensional graph structures, such as current, voltage, and temperature, to characterize the heterogeneous spatial correlations between cells, and integrating GCN, LSTM, and Transformer models, we jointly capture the spatio-temporal dependencies. This framework enables collaborative optimization prediction for the entire battery pack, synchronously capturing both the degradation features of individual cells and the dynamic interactions within the pack. Experimental results demonstrate that the proposed method significantly improves prediction accuracy compared to traditional models, offering a new approach for managing the health of complex battery systems while considering multidimensional spatio-temporal characteristics.