Investigations of Energy Consumption and Greenhouse Gas Emissions of Fattening Farms Using Artificial Intelligence Methods

The aim o f this study was to assess artificial intelligence methods (adaptive neuro-fuzzy inference systems and artificial neural network, ANN) for modeling and predicting energy output and greenhouse gas emissions from calf fattening farms in Abyek and Alborz cities of Iran. The modeling was done based on the amount of energy input. From the total energy input of 24,003 (MJ calf(-1)), feed and fossil fuels have the most significant share. The analysis of greenhouse gas emissions showed that 1174 kg of carbon dioxide equivalent per head of calf was released for the fattening period of 6-12 months. The best model of ANN had 6-16-2 structure. The best adaptive neuro-fuzzy inference systems model was designed using four adaptive neuro-fuzzy inference systems sub-networks, which were developed at two stages. Comparison between the models showed that, due to employing fuzzy rules, the adaptive neuro-fuzzy inference systems models could model energy output and greenhouse gas emissions more accurately than the ANN model. Regression coefficient, root means square error, and mean absolute percentage error for the ANN model were 0.721, 0.055, and 0.9 for energy output, while 0.733, 0.048, and 2.49 for greenhouse gas emissions. These values for the best topology adaptive neuro-fuzzy inference systems were 0.999, 0.006, and 0.098 for energy output, while 0.996, 0.005, and 0.362 for greenhouse gas emissions. (C) 2017 American Institute of Chemical Engineers