OR-94-17-6 -- Artificial Neural Network Backpropagation Model with Three-Phase Annealing Developed for the Building Energy Predictor Shootout PDF

An artificial neural network backpropagation model with three-phase annealing was used for the first building energy prediction competition held by the American Society of Heating, Refrigerating and Air-Conditioning Engineers in 1993. Three-phase annealing is an empirical method to gradually reduce the learning rate during the training period in order to improve accuracy in a relatively short time. The preprocessing of the competition data, methods of backpropagation training (including three-phase annealing), modelling guidelines for the network, and results of the prediction are presented and discussed.

KEYWORDS: artificial intelligence, expert systems, calculating, buildings, energy consumption, competition