A methodology for generating reduced-order models for large-scale buildings using the Krylov subspace method

Abstract

Developing a computationally efficient but accurate building energy simulation (BES) model is important in order to accelerate building design optimizations, retrofit analysis, and development and evaluation of advanced control algorithms where a number of iterations over a long simulation period are required. For this purpose, identification approaches that develop simplified models from building simulation datasets could replace detailed energy simulation software. However, those approaches require extensive computational time at the front to generate necessary data sets and train models for sufficient reliability. An alternative approach that utilizes model order reduction (MOR) methods to directly extract a lower dimensional model from a detailed physics-based model consisting of a number of differential equations is attractive to avoid the pre-simulation requirement. Among many MOR approaches, the balanced truncation method has been most reliably and popularly applied to the building science field. However, it can not be practically applied to a large-scale building with many zones due to computational and data storage requirements. To overcome the problem, this paper introduces the Krylov subspace method to the building science field. Technical issues of applying the Krylov subspace method to building applications are addressed and a suitable algorithm that overcomes those challenges is presented. To demonstrate reliability of the algorithm, comparisons between the resulting reduced-order model (ROM) and a high fidelity model developed from a commercially available BES software for a 60-zone case study building are provided. The ROM was a factor of 100 faster than the high fidelity model but with high accuracy.

A physics-based domain adaptation framework for modeling and forecasting building energy systems

“…While the demonstrative example in this paper assumes a simple building component, the physicsbased SDA framework is scalable to model more intricate building systems and or thermal zones. There exists dedicated work to automate the generation of RC networks from 3D models of thermal zones (Kim et al, 2018). Furthermore, the energy balance equations assumed to inform the SSM are well known by the energy community and straightforward to implement.…”

Knowledge Acquisition and Representation for High-Performance Building Design: A Review for Defining Requirements for Developing a Design Expert System

“…Analysis model Inverse gray box reduced order model [86], Reduced order model [87], VRF and air handler [88] KR-III-2.…”

Development of control-oriented models for a building under regular heating, ventilation and air-conditioning operation - a comparative simulation study and an experimental validation

“…Grey-box models are developed using physical principles and are calibrated using measurement data. The resistance-capacitance (RC) grey-box models (Li et al, 2010;Kim et al, 2020) have been the focus of most works on MPC as applied to buildings. A brief summary of recent works on the development of linear time invariant (LTI) building thermal models is provided next.…”