Artificial neuro-fuzzy intelligent prediction of techno-economic parameters of computer-aided scale-up for palm kernel oil based biodiesel production

Abstract

Palm kernel oil (PKO) is one of the promising starting materials for biodiesel production. Economic viability of large-scale biodiesel production from PKO happens to be the major challenge, as investors would like to know the overall cost-benefit value before making decisions. Therefore, this study develops artificial intelligence (AI) techno-economic models for predicting overall cost-benefit value which will provide fundamental investment decisions for potential investors. The two AI techniques used in this study were artificial neural networks (ANN) and adaptive neuro-fuzzy inference system (ANFIS). The input-output data for modelling was gotten from a previous work which based solely on experimental design for PKO for biodiesel production. The input variables are Methanol:oil ratio, temperature, catalyst quantity, residence time and catalyst calcination temperature, while return on investment (ROI), payback time (PBT), net present value (NPV) and production capacity (PC) are the responses. ANN and Fuzzy Logic Toolboxes in MATLAB R2013a were used for model implementation. The developed models were appraised using statistical indices such as coefficient of determination (R2) and root mean square error (RMSE). The results showed that, trimf based ANFIS models (ROI- R2: 0.9999; RMSE: 7.39 × 10−7; PBT- R2: 0.9999; RMSE: 5.32 × 10−7; NPV- R2: 0.9999; RMSE: 5.89 × 10−7; PC- R2: 0.9999; RMSE: 5.89 × 10−7) performed marginally better than ANN models (ROI- R2: 0.9496; RMSE: 0.0599; PBT- R2: 0.9945; RMSE: 0.0373; NPV- R2: 0.9957; RMSE: 0.0384; PC- R2: 0.9959; RMSE: 0.0376). Also, the relative significance of input parameters based on sensitivity analysis showed catalyst calcination temperature (CT) as the most significant input parameter. These findings show that both the ANFIS and ANN models are effective in predicting techno-economic parameters.