Protection and Safety in Battery Energy Storage Systems: Modeling Fault Detectability, Isolation Latency, and Thermal Runaway Escalation Under Sensor and Control Uncertainty

Abstract

This article presents a reliability-centered framework for BESS protection and safety that treats detection, decision, and isolation as an end-to-end process, quantifying how uncertainty propagates through sensing and control logic to determine the probability of undetected faults, time-to-isolation, likelihood of thermal runaway propagation beyond a module, nuisance trip rate, and expected downtime cost under operational constraints. A scenario-based quantitative study is developed using Monte Carlo simulation of representative fault classes, including internal cell short development, connection resistance growth, and coolant loss, with explicit modeling of sensor noise and drift, estimation uncertainty, and actuation delays. Four architectures are compared, spanning baseline BMS thresholding, redundant sensing with model-based diagnostics, fast hardware interlocks with conservative trip logic, and a governance-optimized two-tier decision architecture that couples early warning with verification and staged isolation. Results show that (i) escalation risk is governed more by detection latency distributions than by mean detectability, (ii) moderate sensor bias can materially increase false stability and delay isolation during incipient faults even when average alarms look stable, and (iii) hybrid governance that controls nuisance alarms while enabling early staged intervention provides the best cost–risk balance for grid-deployed BESS, reducing propagation probability without driving excessive operational trips. The study provides copy-ready tables and figure prompts suitable for Techne submission and for adaptation to site-specific datasets.