Optimisation of biochar dose in anaerobic co-digestion of green algae and cattle manure using artificial neural networks and response surface methodology

This study aimed to optimise and model biogas yields using artificial neural networks (ANN) and response surface method (RSM) of biogas yields after adding different doses of ozonation and ultrasonic pretreated biochar (BC) doses to the co-digestion of cattle manure and green algae substrates, with and without S.Parkle culture in the environment. The results of the RSM-D optimal design indicated that, in the absence of S.Parkle culture in the environment, the optimum ultrasonic pretreated biochar (UPBC) doses were 29.23 mg (for 100 mL digestion volume) and the corresponding optimum biogas yield was 340.50 mL/g volatile solids (VS). In the case of S. Parkle culture in the environment, the optimum ozonation pretreated biochar (OPBC) dose was 55.96 mg and the corresponding optimum biogas yield was 660.40 mL/g VS. ANN and RSM model performances were analyzed using different statistical indicators and ANN result obtain showed absolute percentage error (MAPE) value of 2.1, mean average error (MAE) value of 0.347, standard error of prediction (SEP) value of 0.590, and root-meansquare error (RMSE) value of 0.55, mean squared error (MSE) value of 0.31, the sum of the squared estimate of errors (SSE) value of 4.9 and coefficient of determination (R2) value of 0.9999, and RSM result reveals MAPE of 0.0030, MAE value of 30.693, SEP value of 45.57, RMSE of 42.63, MSE value of 1817.6, SSE value of 29,081 and R2 of 0.9612. The model performance indicators and estimation results indicate that the ANN performs relatively better than the RSM in modelling the process. It is recommended that the optimum OPBC and UPBC doses be tested and verified on different substrates.