Soft computing modeling and multiresponse optimization for production of microalgal biomass and lipid as bioenergy feedstock

Abstract

Microalga biomass is a reliable bioenergy feedstock to produce green fuel owing to its high lipid and organic content. On the other hand, the microalgal biomass productivity as well as lipid accumulation widely depends on various cultivation factors - including nitrogen/phosphorus ratio and light-dark cycles (LD). This study investigated the effects of LD and NaNO3 (nitrogen) dose on the specific growth rate (SGR), biomass productivity (P), and intracellular lipid productivity (LP) of Chlorella kessleri. Response surface methodology (RSM) and support vector regression (SVR) based nonlinear empirical models were developed to forecast SGR, P, and LP. The laboratory data acquired based on central composite design (CCD) matrix, was utilized to establish the adequacy of the models. Bayesian optimization algorithm (BOA) was coupled with SVR to tune the hyperparameters automatically. The performance of the hybrid intelligence model (BOA-SVR) was better than RSM model for anticipating all the responses. Lastly, the crow search algorithm was combined with BOA-SVR to achieve the global optimal solution for maximizing SGR, P and LP, simultaneously. The maximum SGR, P, and LP were found to be 0.302 d−1, 45.31 mgL−1d−1, and 16.3 mgL−1d−1, respectively at the operating environments of LD of 12/12 (h/h) and NaNO3 dose of 10.92 gL-1.