Abstract: In response to the problem of long time and high cost in coal element analysis experiments, a GA-BP neural network is used to predict the results of coal element analysis. The GA-BP neural network model is constructed using 289 sets of typical thermal coal industrial analysis results as known quantity prediction element analysis. The prediction results are compared with the prediction results obtained from traditional multiple regression correlations, and the prediction results are evaluated through determination coefficients and root mean square errors.The results show that the determination coefficients of the GA-BP neural network prediction model for carbon, hydrogen, oxygen, nitrogen, and sulfur are 0.935, 0.811, 0.779, 0.430 and 0.051, respectively, with root mean square errors of 2.503, 0.458, 2.551, 0.245 and 0.824, which are better than the prediction results of traditional multiple regression correlations for each element. Finally, the element analysis prediction model is applied to calculate the boiler thermal efficiency and carbon emissions, and the error between it and the actual data obtained from element analysis is calculated. The average relative error of the boiler thermal efficiency is 0.19%, and the average relative error of carbon emissions is 3.56%.

Key words: coal element analysis, GA-BP neural network, boiler thermal efficiency, combustion carbon emissions