Optimization of dark fermentative hydrogen production from organic wastes using acidogenic mixed consortia

Abstract

The present study investigates the improvement of dark fermentative H2 production from organic wastes using acidogenic mixed consortia. A preliminary study was conducted to find out the suitability of the wastes as a feedstock for H2 production using acidogenic mixed consortia, which was developed by heat-shock pretreatment of anaerobically digested sludge. In the present study, starchy wastewater supplemented with groundnut de-oiled cake (GDOC) was found most suitable for H2 production. Single parameter optimization studies were performed, followed by Taguchi analysis. pH was found to be the most influential factor as compared to other physicochemical parameters. A central composite design was implemented to maximize H2 production. Additionally, the experimental results were analyzed by response surface methodology (RSM) and artificial intelligence (AI) techniques such as artificial neural network (ANN) and support vector machine (SVM). SVM was found to be better as compared to ANN and RSM in terms of prediction capability. R2 and root mean square error (RMSE) for the SVM model were estimated to be 0.988 and 0.0103, respectively. Subsequently, these AI-based models were further coupled with genetic algorithm (GA) and particle swarm optimization (PSO) for the determination of optimal process parameters. GA and PSO both offered a similar optimal value of the parameters. However, PSO was found to be faster than GA. The maximum H2 yield of 8.28 mol H2 kg−1 CODremoved was achieved under the optimized condition of pH 6.75, GDOC concentration 16.16 g L−1 and temperature 37.55 °C determined by SVM-GA and SVM-PSO models. H2 yield was improved by 2.1 folds using SVM based model as compared to un-optimized condition. This is the first attempt to implement the SVM model in biological process optimization and to compare it with ANN and RSM along with validation study.