Hotspot Temperature Prediction of Relay Protection Equipment Based on a Physical-Model-Aided Data-Driven Method

Abstract

Relay protection equipment (RPE) is a type of automation equipment aiming to protect power systems from further damage caused by local faults. It is thus important to ensure the normal operation of RPE. As the power density of electronic components continuously increases, the overheating problem of RPE cannot be neglected. Given the difficulties in implementing direct measurement and predicting development trends of RPE temperature, a novel hotspot temperature monitoring method for RPE was proposed, which is a data-driven method. The generative adversarial network, aided by a physical model, is used to address small samples. Afterwards, a stacked ensemble model established based on random forests was used to predict the hotspot temperature of the RPE. Experiment results show that the proposed method can effectively predict hotspot temperature of RPE with the predictive error lower than 2%. And comparative results demonstrate the superiority of the proposed method compared to other methods.

Keywords

• relay protection equipment
• hotspot temperature
• generative adversarial network
• stacked ensemble