Abstract: The thermal control system not only involves complex control logic, but also contains a large number of measuring instruments, which inevitably encounter various faults during long-term operation. Therefore, in order to accurately diagnose and solve faults in thermal control systems, a research on fault diagnosis of power plant thermal control systems based on temporal convolutional residual networks is proposed. Firstly, the fault signals of the thermal control system are extracted to identify potential faults. Secondly, a fault diagnosis model based on temporal convolutional residual networks is constructed, and the data of the power plant thermal control system to be diagnosed is input into the model. Finally, information such as fault type and severity is output to achieve intelligent diagnosis of faults in the power plant thermal control system. The experimental results show that the fault diagnosis method for power plant thermal control systems based on temporal convolutional residual networks can accurately diagnose abnormal faults that occur in actual power plant thermal control systems, thus proving that this method has high fault recognition accuracy and classification ability.