Simultaneously maximizing microalgal biomass and lipid productivities by machine learning driven modeling, global sensitivity analysis and multi-objective optimization for sustainable biodiesel production

Abstract

Simultaneous enhancement in the productivities of microalgal biomass and lipids that are inversely correlated to each other has been a long-standing challenge before the scientists engaged in algal biodiesel research and innovation. This study, develops an Artificial Neural Network (ANN) model and uses multi-objective optimization approach to determine the process parameters, namely, light intensity, CO2%, air flow rate, and C/N ratio to maximize biomass and lipid productivities of microalga, Chlorella sorokiniana, simultaneously. Experimental data based on central composite design (CCD) were used to train a feed-forward multilayer ANN with four critical parameters. The global sensitivity analysis was performed on the trained ANN model using Sobol's method to assess the relative significance of the four process variables. Since the objectives of maximizing biomass and lipid productivities conflict with each other, multi-objective optimization approach was used for deriving the optimal process parameters, experimental validation of which resulted in approximately 3 times increase in biomass productivity and 7 times increase in lipid productivity. On transesterification of the lipid, the biodiesel product with saturated to unsaturated fatty acids ratio of 48:52 conformed very well to the international standards, ASTM D6751 and EN14214, thereby making it an environment friendly green biofuel. Thus, the present study showcases the successful application of machine learning tool in analysing the global sensitivity analysis of process variables, and employ it for multi-objective optimization towards achieving maximum microalgal biomass and lipid productivities concomitantly for potentially sustainable production of biodiesel.