DATA-DRIVEN TRANSIENT TEMPERATURE RISE PREDICTION MODEL FOR MEDIUM-VOLTAGE SWITCHGEAR

Abstract

Inadequate switchgear on-site installation, component deterioration and damages, although unlikely, are factors that contribute to medium-voltage switchgear malfunctions. Predicting transient temperature change under normal working conditions provides valuable information for detecting such faults, therefore improves switchgear reliability. This paper proposes predicting transient temperature rise of medium-voltage switchgears with a real-time regression model, which is adaptive to different operating conditions and setups. The proposed model is a modified transient heat balance equation with undermined thermal and energy parameters. It models the thermal transfer as a function of temperature rise and dynamic current loads. Firstly, current values, temperature at switchgear's contact points and ambient temperature are collected at each time step under normal operation. Secondly, the undermined parameters are trained with ordinary least square method (OLS) using data from each time step. Lastly, the analytic expression for current-dependent transient temperature rise at different contact points are obtained. It can iteratively predict normal transient and steady-state temperature rise under any electric load. Designed with strong thermal constraints and few parameters, the model can also prevent over-fitting caused by small and biased dataset.