Cross‐domain state‐of‐health estimation of Li‐ion batteries based on transfer neural network with soft‐dynamic time warping

The success of deep learning in the field state-of-health (SOH) estimation relies on a large amount battery data and fact that all possess same probability distribution. While real situations, model based one working condition set may not be valid for another due to distribution differences. Therefore, this article proposes transfer method using soft-dynamic time warping (soft-DTW) as statistical feature method, called soft-DTW domain adaptation network (SDDAN). By combining prediction error with time-series gap training process, transformation can make obtained results more similar source results, which help us obtain better target domain. Experimental show SDDAN effectively predict SOH Li-ion batteries significantly improve performance knowledge transfer.