The coal mill is an important auxiliary equipment for the operation of the boiler, and its performance safety directly affects the safety of the entire thermal power plant. Since the safety evaluation of coal mills in power plants cannot be fed back in real time, a predictive model for safety evaluation of coal mill was established by combining principal component analysis (PCA) with generalized regression neural network (GRNN).Firstly, the actual operation data of coal mill equipment were regarded as experimental samples, and principal component analysis was used to analyze the principal component for many variables affecting the safety of coal mill. Secondly, the safety evaluation and prediction model of coal mill was constructed based on generalized regression neural network (GRNN), and the crucial components were considered as the input variable, and the corresponding historical expert rating was treated as output variable, and the leave-one-out method was adopted to divide training samples and test samples to improve the training accuracy of network model. Finally, the safety evaluation prediction models were established based on GRNN neural network, PCA-BP neural network and BP neural network. The relative errors and time cost of the four prediction models were compared, respectively.The results show that the variance contribution rate of the three principal components F1, F2 and F3 extracted by PCA reaches 96.55%.Based on PCA-GRNN neural network, the average relative error of the prediction model for coal mill safety assessment is minimal and less time costly.The effectiveness of the predictive model of coal mill′s safety evaluation established by PCA-GRNN neural network is verified.