Modeling methods for chiller and heat pump systems in buildings: A systematic review in the context of digital twin applications

Building cooling and heating account for around 25% of global greenhouse gas emissions, with a rising trend. While the growing cooling demand is largely met by chillers, decarbonization and electrification efforts are boosting the adoption of heat pumps. Both technologies rely on electricity-intensive vapor compression cycles, which contribute to environmental impacts and place stress on power grids. To mitigate these effects, previous research has suggested improving the performance of vapor compression systems (VCSs) through optimization and fault detection. In this regard, digital twins (DTs) offer a promising approach by enabling real-time simulation with minimal manual effort. However, their adoption in the building sector remains limited. This paper presents a systematic literature review of modeling methods for VCSs, discussing their strengths and limitations in the context of DT applications. Additionally, it introduces a set of evaluation criteria to analyze the performance of DTs. The findings highlight that (i) model selection is highly dependent on contextual factors such as load and weather variations, (ii) commonly used modeling methods often face challenges when applied to DT scenarios, and (iii) focusing only on error-based metrics may overlook other aspects such as generalizability, resiliency, scalability, and interpretability which might be crucial in final applications. These insights facilitate the development and validation of VCS DTs and supports their adaptation in the building sector. • A systematic review of existing modeling methods for vapor compression systems (VCSs) is conducted. • Strengths and limitations of VCS modeling methods are discussed in the context of digital twinning. • A structured set of evaluation criteria for DT-oriented modeling is introduced. • The importance and role of each evaluation criterion are discussed in relation to practical DT applications. • A case study comparing seven commonly used modeling methods in terms of accuracy, reliability, robustness, scalability, and interpretability is provided. • Results suggest that model selection should consider multiple aspects beyond accuracy alone.