Adaptive Safety Modeling for Improved Battery Protection via Integrated State-of-Safety Estimation

Lithium-ion batteries play an essential role in modern energy storage systems (ESS), making proactive and robust safety management increasingly critical. Traditional state of safety (SOS) estimation methods commonly use uniform thresholds, failing to reflect the individual risks associated with each safety indicator, thus limiting sensitivity and accuracy. To address this, we propose an advanced battery protection system (BPS) utilizing adaptive safety modeling based on multifaceted safety indicators (MSI). Our approach dynamically adjusts thresholds per indicator, significantly enhancing anomaly detection accuracy. We introduce an adaptive sigmoid-based SOS function with adjustable sensitivity parameters to precisely identify hazardous battery conditions. Experimental results from an ESS testbed demonstrate that our proposed system reduces misclassification errors associated with conventional SOS methods, effectively detecting battery anomalies in real-time and thereby improving ESS reliability and safety without modifying existing battery management systems (BMS).