Neural Network for Simulating Ventilation in a Dairy House

Artificial neural networks (ANNs) are a computational technology alternative to the classic methods based on physic modeling. They can estimate input-output functions, without a mathematical model of the reciprocal physical relationship, through a learning process based on presentation of numerical data samples. Several studies have been carried out concerning the use of neural networks for practical applications such as automation design (control systems), classification and identification problems, simulation of complex physical phenomena, etc. The present study aims to evaluate the ability of artificial neural networks in simulating some microclimatic parameters (speed and direction of air fluxes, air temperature and humidity) inside a dairy house. The data used in this paper were collected during recent experimental studies carried out in a cubicle dairy house sited in Sicily. The network architecture used is a generalized feed-forward multilayer perceptron with two hidden layers, that is one of the most used to model physical processes. The results confirm the ability of the network in simulating microclimatic parameters inside a dairy house and encourage the use of this forecasting method for the optimization of the control system of the ventilation openings.