Trend reasoning approach with multi-perception mechanism for battery SOH estimation

As a critical component of renewable energy systems, battery energy storage systems (BESS) play a pivotal role in ensuring the overall operational efficiency through their safety and reliability. Accurate estimation of the battery's State of Health (SOH) is essential for enabling intelligent health management and optimizing maintenance strategies. To this end, this paper proposes an SOH estimation method based on a multi-perception mechanism and trend reasoning. The proposed approach comprises three functional modules: a multi-frequency degradation feature construction module that captures degradation features across different frequency domains; a degradation dependency perception module that models the dynamic interaction structure among key state variables; and a long-term degradation trend modeling module that extracts essential trend information underlying the temporal evolution of SOH. Experimental evaluations on real-world battery datasets demonstrate that the proposed method outperforms benchmark models in key metrics such as RMSE and MAE, consistently delivering accurate and stable predictions. Furthermore, consistent results across different data samples validate the robustness of the method.