SITE QUANTITATIVE-ANALYSIS; LIBS SPECTRA; POWER-PLANTS; ASH CONTENT; WELL LOGS; VALUE GCV; CALIBRATION; PREDICTION; MODEL; PERFORMANCE

Online measurement for the gross calorific Value (GCV) of coal is important in the coal Utilization industry. This paper proposed a rapid GCV determination method that combined a laser-induced breakdown spectroscopy (LIBS) technique with artificial neural networks (ANNs) and genetic algorithm (GA). Input variables were selected according to the physical mechanism and mathematical significance to improve the prediction of the ANN. GA was applied to determine an,optirnal architecture for the network instead of a trial and error method. As a result, the mean standard deviatiori (MSD) of the GCV for four prediction set samples is,0.38 MJ/kg in 50 trials (repetitions of training the ANN with the same; input data but different random initial weights and biases), proving that the ANN model is able to provide high modeling repeatability in the GCV analysis. The mean absolute error (MAE) of the GCV for the prediction set is 039 MJ/kg. The result meets the requirements (0.8 MI/kg) for coal online analyses using the neutron activation method in the Chinese national standard (GB/T 29161-2012).