Modeling of biogas production from cattle manure with co-digestion of different organic wastes using an artificial neural network

Abstract

The present study utilizes an artificial neural network (ANN) as an estimation model of biogas production from laboratory-scale up-flow anaerobic sludge blanket (UASB) reactors treating cattle manure with co-digestion of different organic wastes. It can be estimated depending on working days, influent chemical oxygen demand, influent pH, influent alkalinity, influent ammonia, influent total phosphorus, hydraulic retention time, waste adding ratio, pretreatment and additive waste sorts. The suitable architecture of an ANN for use in biogas prediction consists of 10 input factors, tangent sigmoid transfer function (tansig) at the four hidden layer neurons and a linear transfer function (purelin) at the output layer neuron. The R 2 was found to equal 0.89, 0.79 and 0.75 in the training, validation and testing steps, respectively. ANN estimation modeling can effectively predict the biogas production performance of laboratory-scale UASB reactors. These results indicate that biogas production was optimized to occur in the 20–30% addition range with different organic wastes.