Prediction Modeling of Biogas Production with Classification and Regression Tree at Wastewater Treatment Plants

Abstract

Predicting biogas production is important for energymanagement in wastewater treatment plants (WWTPs). Biogas production quantitydepends on its production system variables, such as, influent flow rate,process temperature, alkalinity, volatile fatty acid, sludge retention time,total suspended solid, etc. WWTPs keep the records of wastewater treatment processvalues with supervisory control and data acquisition (SCADA) system on aregular basis. The relationship between the biogas production and itsproduction system variables, which are measured continuously with SCADA system,can be identified with classification and regression tree (CART) algorithm byusing the existing data. In this paper, CART approach is presented for theprediction of biogas production at WWTPs. Standard CART algorithm is used toselect split predictor. Curvature and interaction tests are also applied in themodel to search for reducing split predictor selection bias and improving thedetection of important interactions among each predictor and response and amongeach pair of predictors and response in turn.