Soft-computing models for naturally ventilated buildings.

In this study, a mixed mode building, namely the Portland Building at the University of Portsmouth is considered. It combines both Natural Ventilation and conventional Heating Ventilating and Air Conditioning systems to maintain the internal comfort. The paper presents the development of Sof1 Computing models to predict the internal temperature in one of the offices using information from neighbouring rooms, corridor and the outside. To derive this model, the so called Adaptive Neuro Fuzzy lnference System method is used. This is a well established Soft Computing method using Fuzzy Logic for the modelling framework and Neural Networks to adapt the model parameters. The fuzzy model is of the Takagi-Sugeno type with linguistic if then rules in the antecedent part and linear algebraic equations in the consequent part. Regression Delay and Proportional Difference model structures are investigated which are taken from classical control theory and extended in the paper for the purpose of fuzzy modelling on the basis of sensor readings.