Application of artificial neural network for prediction of energy flow in wheat production based on mechanization development approach

Due to the inadequate use of limited energy resources, it is necessary to study the energy consumption patterns in agriculture. The purpose of this study was to investigate the process of energy consumption of wheat in two mechanized and semi-mechanized production systems. The statistical population included all wheat farmers in Mazandaran province in 2018-2019. Multilayer perceptron artificial neural network (MLP) was used to find the best model to predict the wheat yield. The results indicated that the average energy consumption of wheat production in Mazandaran province was 20581.46 MJ ha-1, which was higher in mechanized systems. Chemical fertilizer input by 51.64% had the highest share of energy consumption, which was higher in mechanized systems than semi-mechanized systems. The energy efficiency and energy productivity values for the average production were determined to be 2.02 and 0.14 kg MJ-1, respectively, in which both indices were higher in the mechanized systems. Assessing the energy indices highlighted that energy management of wheat production in mechanized systems was better than semi-mechanized ones. In the study of energy flow modeling in mechanized systems, the model performed best with the tangent sigmoid as the activation function and nine neurons by R2 value of 0.994. In the semi-mechanized systems, the model had the best performance with logarithmic sigmoid function as the activation function and eight neurons by R2 value of 0.997.