การพยากรณ์อุณหภูมิของสวิตช์ใบมีดในสถานีไฟฟ้าแรงสูงด้วยเทคนิคการเรียนรู้ของเครื่อง

Abstract

Large power plants are often located in remote areas due to factors such as fuel sources and environmental considerations. As a result, the power transmission system has a crucial role in connecting these power plants to urban centers through interconnected energy networks across the country. High-voltage substations serve as vital links in the transmission system and are responsible for supplying power to customers. Within these substations, there are numerous electrical connections, leading to potential maintenance challenge such as hot spots. Statistical data indicates that hot spots frequently occur with disconnecting switches. Currently, Infrared Thermography (IRT) is employed for maintenance inspections, but there is a notable shortage of both IRT equipment and trained technicians. This research aims to form a model capable of predicting the temperature at the disconnecting switch and facilitating proactive maintenance planning for high-voltage substation equipment. The study began by identifying the key factors that cause the formation of hot spots, while also collecting data for machine learning purposes. Subsequently, a regression model was constructed using five techniques. Performance evaluation revealed that the Random Forest model was the most accurate, as determined by the Root Mean Squared Error (RMSE) and Correlation Coefficient (r). Simulation tests in real operation confirmed the forecast results' accuracy and closeness to the evaluation values. This study demonstrated the potential of the model in predicting disconnecting switch temperatures and its usefulness in maintenance proactive planning.