Preliminary performance tests on artificial neural network models for opening strategies of double skin envelopes in winter

Abstract

This study aims to develop Artificial Neural Network (ANN) models to examine the thermal performance of double skin-enveloped buildings under different opening conditions. Performance tests of the ANN models, which were developed for integrated temperature control logics, were conducted for a space with a double skin envelope in a one-storey building during the winter.

ANN models were embedded in the logic for predictive and adaptive controls in order to ensure comfortable, energy-efficient indoor temperature conditions. Four ANN models were developed to predict future indoor temperatures under different opening conditions of the internal and external envelopes. Their performances were preliminarily tested by comparing them with conventional non-ANN-based methods in terms of thermal control and energy efficiency.

The comparative analysis revealed that the ANN models were properly organized to predict future indoor temperature conditions. Based on the prediction accuracy, the optimal opening conditions and heating system operations could be determined to guarantee advanced methods for effective thermal control and energy efficiency. Thus, ANN models are expected to be applied to the temperature control logic for double skin-enveloped buildings in order to improve their thermal control performance and energy efficiency.