The power grid in Guangdong Province is densely distributed and covers a wide area. The surface environment is complex, so it is vulnerable to natural disasters. As one of the disasters, wildfires have a great impact on the electric transmission lines, and it is of great significance to realize the monitoring and early warning of wildfires in a timely and effective manner. In this paper, based on the principles of MODIS mountain fire identification, the satellite remote sensing monitoring of mountain fires in Guangdong is carried out. In order to improve the monitoring accuracy, the brightness and background brightness temperature of typical fire points in 2017 in Guangdong Province are calculated based on factors such as time, space and cloud coverage. According to the statistical results, the threshold value applicable to the local fire point identification is obtained. It is verified that the accuracy of this method can reach 905% in the automatic identification of fire spots in Guangdong Province and surrounding areas on 3 April 2017, and this method has played an early warning role in the monitoring of mountain fires in Guangdong Province.