Chemical Reaction Optimization (CRO) of Deep Neural Network (DNN) Model for Characterization of Algae Drying Kinetics

Abstract

Drying is an essential step needed to improve the extraction of lipids and other valuable compounds in the algae for biodiesel production. However, there is a limited amount of information available regarding its drying kinetics. Previous studies have used computational intelligence e.g., artificial neural networks (ANN) and deep neural networks (DNN) to characterize the drying kinetics of algae. Chemical Reaction optimization (CRO), a recently introduced metaheuristic optimization approach, is employed in this study to identify the ideal number of neurons to use in a Deep Neural Network (DNN) model that will produce the lowest root mean squared error (RMSE). CRO can reduce the computational time since the population does not need to be coordinated in each computing units. The molecular structure in the CRO contains the set of neurons, while the potential energy (II) corresponds to the RMSE of the DNN model. At a minimum RMSE value, the accuracy of the moisture removal rate prediction increases given maximum temperature, sample temperature, time of drying, heat rate, and percent weight of the remaining algae. The DNN model created obtained an RMSE value of 4.9430 x$10^{-4}$ which corresponds to R -value of 0.9996 and 0.99958 in the training and validation phases.