System Identification for Indoor Climate Control

TL;DR

This paper explores the use of system identification techniques to efficiently model building dynamics for indoor climate control, reducing simulation time and capturing fast dynamic behaviors from discrete data.

Contribution

It demonstrates the applicability of system identification for modeling building heat, air, and moisture dynamics using hourly data, offering a faster alternative to traditional simulation software.

Findings

System identification accurately models building dynamics at lower frequencies.

Models significantly reduce computation time compared to traditional simulation software.

Applicable for climate control design using discrete hourly data.

Abstract

The study focuses on the applicability of system identification to identify building and system dynamics for climate control design. The main problem regarding the simulation of the dynamic response of a building using building simulation software is that (1) the simulation of a large complex building is time consuming, and (2) simulation results often lack information regarding fast dynamic behaviour (in the order of seconds), since most software uses a discrete time step, usually fixed to one hour. The first objective is to study the applicability of system identification to reduce computing time for the simulation of large complex buildings. The second objective is to research the applicability of system identification to identify building dynamics based on discrete time data (one hour) for climate control design. The study illustrates that system identification is applicable for the identification of building dynamics with a frequency that is smaller as the maximum sample frequency as used for identification. The research shows that system identification offers good perspectives for the modelling of heat, air and moisture processes in a building. The main advantages of system identification models compared to the modelling of building dynamics using building simulation software are, that (1) the computing time is reduced significantly, and (2) system identification models run in a MATLAB environment, in which many building simulation tools have been developed