In this paper, a life prediction and health management model of boiler water wall based on big data analysis is proposed, which can realize real-time monitoring and early warning of water wall operation state through accurate prediction model, and provide strong support for operation and maintenance decision-making of thermal power plants. In the research, the data comes from the real-time monitoring system of thermal power plant, covering the physical parameters, basic information, historical maintenance and fault records of water walls. Through data cleaning, preprocessing and feature extraction, combined with big data analysis technologies such as machine learning, data mining and statistical analysis, life prediction model, health state evaluation model and real-time early warning threshold model are established. The improved XGBoost life prediction model is the best in 50% cross validation, with an average absolute percentage error (MAPE) of 8.2%, a root mean square error (RMSE) of 2,150 hours and a determination coefficient (R²) of 0.93. The accuracy of health status classification reaches 96.3%, which meets the design standards. The model also has the ability of real-time monitoring and early warning, and can predict the remaining life and potential failure of water wall according to real-time data, and send out early warning signals when it reaches the preset threshold. The model provides an effective technical means for the health management of boiler water wall, which is helpful to improve the safety and economy of power plant operation.
Zhang et al. (Sun,) studied this question.