A machine learning-genetic algorithm based models for optimization of intensification process through microwave reactor: A new approach for rapid and sustainable synthesis of biodiesel from novel Hiptage benghalensis seed oil

The depletion of fossil fuels necessitates the need for sustainable and energy-efficient biodiesel production techniques. Microwave techniques have recently emerged as a promising alternative to conventional biodiesel production due to their enhanced conversion rates and reduced processing times. However, the complex chemical processes and reaction kinetics involved in biodiesel production present modeling challenges. Machine learning (ML) is well-suited for capturing intricate and nonlinear behaviors that are challenging to model using traditional mechanistic approaches. This study uses eight different machine learning (ML) models to analyze experimental datasets of biodiesel production from Hiptage benghalensis seed oil through a microwave reactor at different input parameters and operating conditions. Furthermore, tree-based pipeline optimization tools (TPOT) and genetic algorithms (GA) were used to optimize the input parameters. The results, including R2: 0.9756, MAE: 1.124, RMSE: 1.429, and MAPE: 0.014, show the gradient-boosting model's superior predictive accuracy. It was found that a 320-W microwave power, a 5-minute reaction time, a 6:1 M ratio, and a 0.75- wt% catalyst concentration were the optimum conditions for a maximum biodiesel yield of 98.62%. Comparing conventional and microwave techniques reveals that the MW method achieves significantly higher energy efficiency, with an improvement of 8.63 times. Additionally, the reaction time is expedited by 12-15 times, and the biodiesel yield is enhanced by 6.02% compared to the conventional method.