Optimization of biofuel production from biomass using montmorillonite catalyst by development of predictive models

Abstract

Fossil fuel depletion and their environmental problems lead to new research for production of alternative fuels. Biofuels are promising nontoxic and biodegradable energies obtained from renewable sources. Experimental studies of biofuel production and optimizing the properties are very time-consuming, cost-intensive, and difficult. Thus, recently there have been major interests in application of Machine Learning (ML) techniques to develop models for predicting biofuel production process and optimization. This study proposed novel ML approach based on Support Vector Machines (SVM), Multi-Layer Perceptron (MLP), and Adaboost on SVM base (Adaboost + SVM) algorithms optimized the production of biofuel from waste cooking oil through transesterification process and under montmorillonite (MMT) catalyst. The input variables for the models were chosen to be the temperature of the reaction (ranging from 100 to 200 °C), treatment time (ranging from 3 to 9 h), catalyst loading (ranging from 2 to 6 wt%), and methanol to oil molar ratio (ranging from 8 to 16). Meanwhile, the output of the models was selected to be the yield of biofuel production. The obtained results of three models were evaluated and compared and it was confirmed that the Adaboost + SVM model with highest regression coefficient (R2) of 0.971, was more accurate and could predict the biofuel production more properly. Additionally, the SVM model exhibited the highest maximum error with an MAE of 6.504 and an RMSE of 7.462. On the other hand, the Adaboost + SVM model showed the lowest error, with an MAE of 3.151 and an RMSE of 4.484. The MLP model had an MAE of 4.423 and an RMSE of 6.008, which falls between the other two models in terms of error. The optimization results obtained by the Adaboost + SVM model demonstrated that the optimal conditions giving the highest biofuel production of 96.79 % were found to be: reaction time of 8.62 h, catalyst loading of 4.88 wt%, methanol/oil molar ratio of 11.2, and reaction temperature of 152 °C.