Prediction of effective substrate concentration and its impact on biogas production using Artificial Neural Networks in Hybrid Upflow anaerobic Sludge Blanket reactor for treating landfill leachate

Abstract

In this study, a lab scale Hybrid Upflow Anaerobic Sludge Blanket reactor (HUASB) is used to treat landfill leachate at varying the Organic Loading Rates (OLRs) from 0.41 to 19.52 kg COD/ m3 d, over a period of 365 days. A three layer back propagation Artificial Neural Network (ANN) model was developed to assess the effective substrate concentration and maximum biogas yield. The proposed ANN model prediction accuracy was assessed based on the following factors, such as, Mean Square Error (MSE), Coefficient of Regression (R), Coefficient of determination (R2), Frictional Variances (FV) and Index of Agreement (IA). The results showed that OLR of 16.27 kg COD/ m3 d, is considered as an optimum. The highest biogas production of 30.07 L/d with methane content of 64% and COD removal of 89.6% is observed at this condition. In addition to this, the profile of alkalinity and Volatile Fatty Acid (VFA) supports the optimum OLR condition. The concentration of VFA shows fourfold increase from 710 to 2800 mg/L beyond optimum OLR. Further increasing the OLR beyond optimum conditions shows ineffective biogas yield and reactor operation. Similarly, the ratio between VFA/alkalinity reached critical limit, which indicates deterioration of reactor performance.