Prediction models of calorific value of coal based on wavelet neural networks

New prediction models based on wavelet neural networks (WNNs) have been proposed to estimate the gross calorific value (GCV) of coals. The input sets for the prediction models are involved of the proximate and ultimate analysis components of coal and the oxide analyses of ash. The coal samples, which have been employed to develop and verify the prediction models, are from United States Geological Survey (USGS) and China Huaneng Group. Some published methods have also been employed and redeveloped to make a comparison with the models proposed in this paper. The comparison reveals that the WNN models proposed here based on the proximate (ultimate) analysis components of coal, are consistently better than the published ones. The WNN models based on the oxide analyses of ash have higher accuracy in estimating the GCV of Chinese coals than US coals. Here we also analyze the possible reasons that could lead to the low estimated accuracy.