Process modelling of biodiesel production process using genetic programming

Abstract

Biodiesel production is a complex process that involves a number of process parameters. These process parameters must be carefully optimized to increase its yield. In this paper, a process model for biodiesel production is developed by using genetic programming (GP). A case study on biodiesel production from palm oil transesterification is selected from the literature to analyze the proposed methodology. The dataset from literature is based on a central composite design (CCD) with four process parameters (namely methanol-to-oil ratio (x1), catalyst loading (x2), reaction time (x3) and reaction temperature (x4)). The FAME yield % is the target response. The available CCD data is divided into dedicated training (90 %) and testing (10 %) sets to analyze and evaluate the predictive power of the GP metamodel. The GP metamodel is trained using the tuned parameters and validated on the testing data using various statistical metrics. A high accuracy with an R2 of 0.9676 for training data and 0.8834 for testing data is obtained. This study shows that the GP metamodel is a robust and accurate approach to model biodiesel production processes.