A thermal twin of medium-voltage switchgears for predictive maintenance

The thermal supervision of switchgears is a central element of any condition-based maintenance concept. High current connections, especially those mounted on-site, such as the cable connection, are one of the most endangered locations during operation. Due to the non-linearity of a switchgear’s thermal behaviour, the probability of severe, irreparable thermal issues, resulting from extreme weather conditions, load peaks or general aging of the equipment, may increase in the future. To solve these challenges and avoid unexpected downtimes, predictive maintenance is necessary. To anticipate and mitigate thermally induced faults, temperatures at critical locations in the switchgear, as well as other operational data such as the primary current, can be used as a model input, enabling the required predictive capabilities. A challenge is the creation and parametrization of such models. This paper presents a new grey box approach, utilizing OEM data from the design phase with the help of large language models and enhancing the approach during the operational phase. The entire workflow is demonstrated on an OEM test setup, demonstrating both the applicability of the approach and its advantages.