Coal Mill Fault Warning Method Based on Optimized XGBoost

Coal mills in thermal power units have problems such as harsh operating conditions and frequent failures. In order to prevent the abnormal operating condition of coal mill in time and effectively, reduce the economic loss of thermal power plants and improve the maintenance system of power plants, a coal mill fault warning method based on Bayesian optimization limit gradient boosting algorithm is proposed; firstly, the principal component analysis method and wavelet transform method are used to optimize, reduce and denoise the historical data of coal mill to select the main state parameters of coal mill and improve the data quality Then the XGBoost technique is used to model the historical data of the coal mill under normal operating conditions, and the Bayesian technique is used to optimize the hyperparameters of XGBoost to determine the operating state of the coal mill. Finally, a medium-speed coal mill in a power plant in Zhejiang Province is used as an example to build a fault warning model, and after comparison and verification, the results show that the method can provide effective warning for coal mills.