Performance evaluation of artificial neural network-based variable control logic for double skin enveloped buildings during the heating season

This study describes integrated logic for an artificial neural network (ANN) to control heating devices on a continuous basis. Two ANN-based control logic systems and two conventional rule-based logic systems were developed to control a heating device and the openings of a double skin enveloped building. The ANN-based logic controls heating devices on a continuous basis according to the indoor temperature. The rule-based logic controls heating systems and openings at envelopes in two-position on/off operation. Control performance for the developed logic was numerically conducted using computer simulations for a small office space with double skin envelopes during the heating season. Analysis results indicate that the ANN-based temperature control logic resulted in a more stable temperature near the center of the comfortable range with a reduced opening period of the internal envelope. The reduced number of on/off moments of the heating device and the openings in the ANN-based logic were predicted to save energy and prevent system degradation. The use of ANN-based logic would be effective for maintaining a stable thermal environment and for system operation. Rule-based logic can be effectively used to improve building energy efficiency. In this study, two ANN-based logic types were developed for heating devices controlled on a continuous basis and their performance was compared with those of rule-based on/off logic. Thus, in order to cover the limitation of this study, further study is warranted for examining the clear difference achieved by ANN-based vs. rule-based control, when they are applied to control heating output on a continuous basis. [Display omitted] •Rule-based and artificial neural network based thermal control logic was developed.•The heating device and surface openings were controlled in an integrated manner.•Numerical computer simulation method was employed for comparative performance tests.•ANN-based logics showed the advanced features of thermal environment.