توپولوژی تحلیل حساسیت یک شبکه عصبی مصنوعی ساده ورودی انرژی در شبکه های عصبی مصنوعی Topology of a simple artificial neural network Sensitivity analysis of energy inputs in crop production using artificial neural networks

1. Introduction Agriculture and energy are closely related since efficient use of energy is a key factor in sustainable agricultural production. Increasing requirement of higher food production has led to intensive use of agricultural and natural resources (Khoshroo, 2014). However, bio-energy has placed agriculture in the position of energy consumer and energy supplier (Esengun et al., 2007). Efficient energy use in agriculture is a pathway toward decreasing environmental hazards and improving agricultural sustainability (Izadikhah and Khoshroo, 2018). Energy demand in agriculture can be classified into direct and indirect energies or renewable and non-renewable energies (Ozkan et al., 2004a). Direct energy consists of human labor, diesel fuel, electricity and water for irrigation, while farmyard manure (FYM), chemicals and machinery are considered indirect energy. Renewable energy includes human labor, FYM and water for irrigation whereas machinery, diesel fuel and chemicals are considered non-renewable forms of energy (Demircan et al., 2006; Ozkan et al., 2004a). The established method to determine energy efficiency of production systems is the input-output analysis. Using this type of analysis, researchers have studied energy consumption in the production of fruits such as citrus (Ozkan et al., 2004), grape (Ozkan et al., 2007), apple (Gokdogan and Baran, 2017; Taghavifar and Mardani, 2015), prune (Tabatabaie et al., 2013), walnut (Khoshroo and Mulwa, 2014) and pomegranate (Houshyar et al., 2017). Modeling crop yield based on energy consumption is an interesting issue for researchers. Prediction of agricultural production is useful for farmers, governments, and agribusiness industries. It helps farmers to make marketing decision. Government requires forecasts of the crop yield to implement policies that provide technical and market support for the agricultural sector. Processors of food, and others in the marketing chain, need forecasts for their purchasing and storing decisions. Various approaches and methods have been used to model energy consumption (Arabi et al., 2017; Jebaraj and Iniyan, 2006; Laha and Chakraborty, 2017; Say and Yücel, 2006; Tso and Yau, 2007). Traditionally, econometric models, based on Cobb-Douglass production function were the most popular modeling technique for investigating functional relations between input energy and various crop yield (Hamedani et al., 2011; Hatirli et al., 2006; Houshyar et al., 2015).