Application of real valued genetic algorithm on prediction of higher heating values of various lignocellulosic materials using lignin and extractive contents

Abstract

The higher heating values (HHVs) of 11 non-wood lignocellulosic materials from Turkey were measured experimentally and calculated incorporating various theoretical models with the values of both lignin and extractive contents. Multiple linear regression (MLR) and real valued genetic algorithm (RVGA) were used to derive the theoretical models. A non-linear RVGA6 model was determined as the best non-linear model considering the experimental results with a regression coefficient of 92% coefficient of determination (R2), 0.301 sum of squared errors (SSE), 0.301 mean squared errors (MSE), 0.548 root mean squared errors (RMSE) and 0.0187 mean absolute percentage error (MAPE) and is proposed as a better alternative for theoretical HHV calculations to the multiple linear modellings such as MLR and RVGA1.