Prediction of gross calorific value of solid fuels from their proximate analysis using soft computing and regression analysis

The determination of gross calorific value (GCV) of solid fuel is important because GCV is frequently required in the design of most combustion and other thermal systems. However, experimental determination of GCV is time-consuming, which necessitated the development of different empirical equations to estimate GCV using the elemental composition of the solid fuels. With the growing popularity of empirical equations for estimation of GCV of solid fuels, there is a need to develop reliable and suitable models for the prediction of GCV of coal from the South African coalfields (SAC). In this study, empirical models were developed to determine the relationship between the proximate analysis of coal with its GCV, using soft computing and regression analyses. A total of 32 coal samples were used to develop three empirical models based on soft computing techniques, namely; adaptive neuro-fuzzy inference system (ANFIS), artificial neural networks (ANN), and regression analysis using multilinear regression (MLR). The performances of the proposed models were evaluated using coefficient of determination (R-2), mean absolute percentage error (MAPE), mean squared error (MSE) and variance accounted for (VAF). The R-2, MAPE, MSE and VAF for the ANFIS are 99.92%, 2.0395%, 0.0778 and 99.918% while for the ANN, they are 99.71%, 2.863%, 0.2834 and 99.703%. The R-2, MAPE, MSE and VAF for the MLR are 99.46%, 3.551%, 0.5127 and 99.460%. From the soft computing and regression analysis studies conducted, the ANFIS was found as the most suitable model for predicting the GCV for these coal samples.