Machine learning approaches to modeling and optimization of biodiesel production systems: State of art and future outlook

Abstract

One of the main limitations to the economic sustainability of biodiesel production remains the high feedstock cost. Modeling and optimization are crucial steps to determine if processes (esterification and transesterification) involved in biodiesel production are economically viable. Phenomenological or mechanistic models can simulate the processes. These methods have been used to simulate and manage the processes, but their broad use has been constrained by computational complexity and numerical difficulties. Therefore, it is necessary to use quick, effective, accurate, and resilient modeling methodologies to simulate and regulate such complex systems. Data-driven computational and machine-learning (ML) techniques offer a potential replacement for conventional modeling methodologies to deal with the nonlinear, unpredictable, complex, and multivariate nature of biodiesel systems. Artificial neural networks (ANN) and adaptive neuro-fuzzy inference systems (ANFIS) are the most often utilized ML tools in biodiesel research. To effectively attain maximum biodiesel yield, suitable optimization techniques based on nature-inspired optimization algorithms need to be integrated with these tools to obtain the best possible combination of various operating variables. Future research should focus on utilizing ML approaches for monitoring and managing biodiesel production systems to increase their effectiveness and promote commercial feasibility. Thus, the review discusses the various ML techniques used in modeling and optimizing biodiesel production systems.