A Comparative Study Between Artificial Neural Netwo rks and Adaptive Neuro-Fuzzy Inference Systems for Modeling Energy Consumption in Greenhouse Tomato Production: A Case Study in Isfahan Province

Abstract

In this study greenhouse tomato production was inve stigated from energy consumption and greenhouse gas (GHG) emission point of views. M oreover, artificial neural networks (ANNs) and adaptive neuro-fuzzy inference systems ( ANFIS) were employed to model energy consumption for greenhouse tomato production . Total energy input and output were calculated as 1,316.14 and 281.1 GJ ha -1 . Among the all energy inputs, natural gas and electricity had the most significant contributi on to the total energy input. Evaluations of GHG emission illustrated that the total GHG emis sion was estimated at 34,758.11 kg CO 2 eq ha -1 and, among all the inputs, electricity played the most important role, followed by natural gas. Comparison between ANN and ANFIS mo dels showed that, due to employing fuzzy rules, the ANFIS-based models could model output energy more accurately than ANN models. Accordingly, correlatio n coefficient (R), Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MA PE) for the best ANFIS architecture were calculated as 0.983, 0.025, and 0 .149, respectively, while these performance parameters for the best ANN model were computed as 0.933, 0.05414, and 0.279, respectively. Keywords : Neuro-fuzzy, Greenhouse crop production , GHG emission.