Since Energy Storage Systems (ESS) store massive amount of electricity, repeated recharging on cells with vulnerability, defect, or fatigue can pump up excessive heat, which possibly leads to large-scale fire accidents. It is necessary to establish a system capable of detecting anomalies at an early stage in order to safely manage them and prevent accidents.
Approach
The characteristics of thermal runaway process are analyzed using normal ESS operation data and test data on abnormal-conditioned batteries collected from test beds by battery research centers. To predict customer-defined failure situations, we devlope predictive models using quasi-supervised anomaly detection techniques.
Results
Safety indicator provided by developed predictive model can signal occurrence of full-scale thermal runaway about 30 minutes ahead of time. It is possible to detect serious anomalies 12 hours before fire accidents and monitor abnormal conditions using the rack drop-out precursor detection model.
Related Use Cases
Chemical
Anomaly detection
Polymer Reactor shutdown analysis
Data-based anomaly detection models can help prevent emergency shutdown in petrochemical plants.
AI algorithms learn the pattern of sensor data from semiconductor manufacturing equipment and identify possible anomalies during the manufacturing process.