A heating value estimation of refuse derived fuel using the genetic programming model

•Proximate analysis for refuse derived fuel samples were performed.•HHV values of refuse derived fuel samples were measured.•Two estimation models were trained using pre-determined experimental data.•Models were tested using other estimation models.•High R2 values (0.9951–0.9988) were obtained indicating the efficiency of the models. Refuse Derived Fuel (RDF) makes an increasingly important contribution to sustainable waste management as an energy source in cement kilns. The most important parameter of RDF in an evaluation of its performance as a fuel is Higher Heating Value (HHV). The two methods of HHV determination are the direct method and the indirect method. The direct method requires the use of a calorimetric bomb and the indirect method requires ultimate or proximate analysis. As in the direct method, the ultimate analysis based indirect method requires the use of specific equipment and a skilled analyst. Most cement plants do not have special equipment. From this point of view, this study aims to predict the HHVs of RDF samples using the results of proximate analysis. Two Genetic Programming (GP) Models, namely GP Model #1 and GP Model #2 are used for the prediction. GP Model #1 denotes a modest nonlinear mapping function used for the prediction of HHVs, whereas GP Model #2 is a more inclusive nonlinear correlation analysis model as an improved version of GP Model #1. To assess the developed models, the test data is simulated and statistical results to the estimation of HHVs are reported as R2 equal to 0.9951 and 0.9988, Root Mean Square Error (RMSE) equal to 1.4126 and 0.6971 and Average Absolute Error (AAE) equal to 0.0543 and 0.0251, for GP Model #1 and GP Model #2, respectively. It can be seen that GP Model #2 may be confidently used for HHV estimation.