Artificial Neural Network for Controlling the Openings of Double Skin Envelopes and Cooling Systems

Abstract

This study aimed at developing an artificial neural network (ANN)-based temperature control method for buildings with a double skin envelope. For this objective, logic for controlling the opening conditions of inlets and outlets of the double facade as well as the cooling system’s operation was developed employing the ANN model for predictive and adaptive controls. For the optimal ANN model’s structure and learning methods, a parametrical optimization process was conducted in terms of the number of hidden layers, the number of neurons in the hidden layers, learning rate, and moment followed by the performance tests of this optimized model. Analysis of the performance tests proved predictability and adaptability of the developed ANN model for diverse background conditions in terms of a stable Root Mean Square and Mean Square Error values. Results of the study indicated that the developed ANN model could potentially be applied to control temperature of double skin envelope buildings.