Neural networks applied to temperature estimation in school buildings

This paper present an Artificial Neural Network (NN) applied to the modelling of inside air temperature in a building school. This modelling is a function of outside air temperature and solar radiation, inside air humidity and state of windows and doors. This NN is a one step-ahead predictive model, and is intended to be the basis model for longer prediction horizons. The NN model employed was the Radial Basis Functions Neural Network (RBFNN, trained using the Levenberg-Maquardt algorithm. The structure selection of the best fitted model RBFNN was accomplished by multiobjective genetic algorithms (MOGA)