Modelling Energy Efficiency in Broiler Chicken Production Units Using Artificial Neural Network (ANN)
he purpose of the current study is to determine the energy consumption and to investigate the relationship between input and output
energy of producing units for broilers. Accordingly, data was collected from 50 broiler chicken production units using personal questionnaires
in the winter 2013. The total energy and output were estimated at ~220.02 and 30.25 GJ, respectively, per 1000 birds. The most important
energy inputs were gasoline, food, gas and electricity explaining 43.03%, 25.56%, 20.81% and 10.07%, respectively, of the total energy used.
Minimal amounts of energy was inputs including day old birds, equipment and labor explaining 0.27%, 0.16% and 0.10%, respectively, of the
total energy used. Energy index, energy ratio, energy efficiency, specific energy and net energy added were calculatedat0.15, 0.01 Kg/MJ,
76.59 MJ/Kg and 189.77 MJ per 1000 birds. Determination of different forms of energy also revealed that the contribution of direct energies
(26%) was higher than indirect energies (74%); in addition, almost all the energy sources used in the production of broilers in Alborz Province
were nonrenewable (99.90% renewable and 0.10% nonrenewable). Various neural networks (approximately 600 networks) were assessed to
estimate the relative amount of energy used in production units. The results showed that feedforward neural network with two hidden layers (2
and 16 neurons for energy model) provided the best results; thus, it can be used for more accurate estimation of energy. The optimal model
performance was evaluated using measures such as the coefficient of determination (R2), MSE, MAPE and MAE. The correlation coefficient
for both energy models was reported at 99.