Estimation of biomass higher heating value (HHV) based on the proximate analysis: Smart modeling and correlation

Abstract

In order to evaluate the potential and make a technical assessment of biomass energy, it is crucial to determine the higher heating value (HHV) of biomass fuels. Thus, multilayer perceptron artificial neural network (MLP-ANN) genetic algorithm-adaptive neuro fuzzy inference system (GA-ANFIS) differential evolution-ANFIS (DE-ANFIS), GA-radial basis function (GA-RBF), least square support vector machine (LSSVM) methods and an empirical correlation (multivariate polynomial regression (MPR)) were employed for the estimation of the HHV of biomass fuels. The comparisons of results show that GA-RBF and MPR models have higher accuracy as coefficients of regression (R2) values equal to 0.9591 and 0.9597, respectively. The average Absolute Relative Errors (% AARD) were obtained as 3.9547 for GA-RBF and 3.9791 for MPR models. The results show that proposed techniques are working efficiently in the estimation of HHV of different sources of biomass.