Aiming at the different sampling rates of multiple sensors for the real-time detection system of heat faults in high-voltage equipment in substation, the step by step prediction fusion algorithm based on synchronous sampling data was used to rapidly calibrate the location of thermal faults. And a new method that using infrared thermal imager together with CCD (Charge Coupled Device) camera and wireless sensor networks was proposed to implement heat faults detection. The energy harvesting technique was employed to obtain solar energy, which achieved the independent power supply of sensor nodes. The experimental results in a substation showed that the system proposed in this paper could effectively detect overheating faults of the transformers and could improve the operation reliability of the electrical equipment..