A general approach for generating reduced-order models for large multi-zone buildings

Abstract

Many approaches for reducing model complexities have been suggested for characterizing the dynamic behaviour of buildings. However, previous work has focused on buildings that have a relatively small number of zones and there have been very limited comparisons with experimental data or with the predictions and computational requirements of validated simulation models that are widely used. In particular, the applicability to large multi-zone buildings has not been addressed. This is an important application, where the cost to generate a reduced-order model (ROM) can be significant, but the benefits of ROM for analysis and design for building system can be tremendous. This paper presents a general and computationally efficient approach for generating a ROM from a detailed representation of the dynamics of commercial sized multi-zone buildings and provides comparisons of predictions and computational requirements with a commercial building simulation tool.

Keywords:

• model order reduction
• multi-zone building model
• building energy simulation tool

Acknowledgments

The authors wish to acknowledge Dr George Chiu, Dr Jianghai Hu, Jie Cai, Vamsi Putta and Guangwei Zhu for their valuable suggestions.

Notes

1 In Figure 3 only internal couplings between subsystems are shown, i.e. exogenous inputs and outputs are not shown.

2 For the case of inter-zonal air mixing, in Equation (10(10) ) can be easily modified, but this is not considered.

3 The article discusses a more general approach called structure preserving MOR. Our suggested method can be thought of as an instance of the method.

4 norm on measures average discrepancy while measures maximum discrepancy over the frequency domain. The infinite integral of norm can be obtained by solving a Lyapunov equation which is a matrix equation while norm requires several iterations due to the nature of maximum or supremum. For a detailed description of the calculation of norm, see Bruinsma and Steinbuch (1990) and Zhou and Doyle (1998).

5 For other closed-loop tests, please refer to Kim et al. (2013). In this paper, the ROM was coupled with local, supervisory control and HVAC systems, and its performances were compared to the models based on TNRSYS and the Modelica Buildings library (Wetter et al. 2014).

6 Although the computational speed of EnergyPlus has improved tremendously over the past few years, it has been an ongoing issue as discussed in Hong et al. (2008) and Hong (2009).

Additional information

Funding

This work was supported by the Department of Energy through the Energy Efficient Buildings Hub (renamed the Consortium for Building Energy Innovation).