Abstract: In order to achieve real-time monitoring of the operating status of the unit and improve its reliability and stability, this paper will take the thermal power generation unit as an example and design an online monitoring method based on machine learning algorithms during its operation. Establish a model of the operating system of thermal power generation units using steady-state equations and dynamic equations, and use recurrent neural networks in machine learning to predict the operating conditions of thermal power generation units; build a human-computer interaction interface for real-time online monitoring of key parameters during unit operation. The comparative experimental results show that after applying the designed method, online monitoring of the operating efficiency of thermal power generation units can be achieved. The consistency between the monitoring results and the manual monitoring results(expected output results) is high, indicating its strong reliability.