A Unique Variable Selection Approach in Fuzzy Modeling to Predict Biogas Production in Upflow Anaerobic Sludge Blanket Reactor (UASBR) Treating Distillery Wastewater

The upflow anaerobic sludge blanket reactor is known to carry out a complex high-rate anaerobic process used to treat distillery wastewater and is met with many conflicts because of continuous fluctuations in quantity and quality of wastewater, and therefore, it incorporates a lot of uncertainties in operating, controlling and measuring different parameters. In this paper, a multiple-input and single-output fuzzy knowledge-based model was developed to predict biogas production in real-scale upflow anaerobic sludge blanket reactor treating distillery wastewater incorporating seven input variables such as pH (effluent), COD load, COD reduction, temperature, alkalinity-to-acidity ratio, pH (influent) and spent flow rate. Trapezoidal and triangular membership functions were classified to represent the fuzzy sets, and a Mamdani type of fuzzy inference system was used in Matlab fuzzy toolbox. A total of 270 IF-THEN rules have been generated in the fuzzy rule editor using a knowledge-based system. Furthermore, an innovative sequential variable selection approach has been proposed to recognize the most significant parameters in the fuzzy model to predict biogas production which makes the model more practical, manageable and efficient. As a result of the sequential variable selection approach, a combination of five variables such as temperature, COD reduction, COD load, pH(I) and alkalinity-to-acidity ratio has been chosen as the optimal set of variables. The results of the root mean square error and coefficient of determination clearly indicated the better predictive ability of the fuzzy model with the five most important input variables obtained from the sequential variable selection approach than the one with all seven variables.