The methodological approach of this review is designed to examine HVAC system performance from a system-level perspective, with particular emphasis on the interaction between design configuration, operational behaviour, and realised efficiency in real building environments. Instead of treating HVAC technologies as isolated components or control algorithms, the review framework prioritises studies that explicitly address dynamic operation, part-load behaviour, and performance divergence between design-stage expectations and in-use conditions. To achieve broad yet focused coverage, the literature search is carried out across three major academic databases: Web of Science, Scopus, and Google Scholar. These platforms are selected to capture both high-impact journal publications and interdisciplinary research spanning HVAC engineering, building physics, control systems, and building energy performance assessment. The search strategy relies on iterative keyword combinations aligned with the conceptual focus of the study. Search terms include expressions related to performance realisation and system interaction, such as “HVAC performance gap”, “part-load operation”, “system oversizing”, “sensible and latent load interaction”, “ventilation-cooling coupling”, “VRF and hybrid HVAC systems”, “dedicated outdoor air systems”, “HVAC supervisory control”, and “real-world HVAC efficiency”. Keyword groupings are progressively refined to prioritise studies addressing operational dynamics and cross-layer effects rather than nominal equipment ratings alone. The reviewed literature predominantly covers the period from approximately 2010 to 2026, reflecting the rapid expansion of research on variable-speed technologies, advanced control architectures, humidity-aware cooling strategies, and post-occupancy performance evaluation. Earlier publications are included selectively where they provide essential theoretical grounding for efficiency metrics, part-load thermodynamics, or control principles that underpin more recent developments. Articles are retained based on the following selection principles:

Non-refereed material and conference papers are generally excluded unless they offer unique system-level insights not replicated in journal literature. Following screening, the selected studies are organised into thematic groups corresponding to system architecture, operational control strategy, climatic load characteristics, and performance evaluation perspective. This classification enables cross-comparison of recurring performance degradation pathways and facilitates the synthesis of evidence supporting the proposed design-operation-performance coupling framework. Through this methodological structure, the review integrates diverse strands of HVAC research into a coherent analytical narrative, enabling a deeper understanding of why high nominal efficiency does not necessarily translate into equivalent operational performance in contemporary buildings.

Centralised HVAC systems generate heating, cooling, and ventilation at a central plant. Conditioned air or water is then distributed to different zones of the building (Korolija et al. 2011). Common examples include constant air volume (CAV) (Aktacir and Büyükalaca 2006) and variable air volume (VAV) systems supplied by air-handling units (AHUs) (Nandagopal 2024). Hydronic systems using chillers (Ding et al. 2025), boilers (Cuce et al. 2025b), and fan-coil terminals (O'Neal and Yin 2020) are also widely used.

The main advantage of centralised systems is their ability to serve large floor areas efficiently. They can also provide effective filtration, humidity control, and system monitoring (Ahmed et al. 2026). Therefore, they are commonly used in commercial offices, healthcare facilities, and institutional buildings where indoor air quality requirements are strict (Ramadan et al. 2025). However, their performance can decrease when zoning resolution is limited or when fixed air volume operation is used. In such cases, the system may respond poorly to variable occupancy, especially under part-load conditions (Li et al. 2025). Previous studies have shown that equipment oversizing, conservative design assumptions, and limited zoning control can contribute to the gap between predicted and actual HVAC performance (Walther et al. 2025).

From an operational perspective, centralised systems are strongly influenced by control strategies governing airflow rates, supply temperatures, and scheduling (Yao et al. 2025). Advanced approaches, such as variable-speed drives and demand-controlled ventilation, can improve part-load performance. However, centralised systems can still be difficult to commission, monitor, and adapt because of their high level of integration (Kim et al. 2025; Zhou et al. 2024). To clarify the operational and performance-related characteristics discussed above, Table 2 summarises the principal strengths and limitations associated with centralised HVAC systems from a system-level perspective.

Decentralised HVAC systems distribute heating and cooling capacity directly to individual zones or rooms through unitary or semi-unitary equipment, rather than relying on a central plant serving the entire building (Zhou et al. 2024; Alabdulkarem et al. 2015). Typical decentralised HVAC configurations include split and VRF systems, alongside other modular air-conditioning units. Decentralised air-conditioning systems are frequently preferred in practice owing to their modular architecture, simplified installation, and adaptability to retrofit applications (Yau et al. 2024).

A defining feature of decentralised systems is their ability to provide zone-level thermal control (Lee et al. 2026). The independent and part-space operation capability of decentralised systems allows cooling provision to be aligned with actual usage patterns, occupancy presence, and zone-level load variations (Wang et al. 2022; Kitzberger et al. 2022; Liu et al. 2022). This operational flexibility has made decentralised systems particularly attractive in office buildings, educational facilities, and small commercial premises with variable usage patterns (Faddel et al. 2021). Despite these advantages, decentralised configurations also present inherent cons (Qian et al. 2021). In many cases, ventilation and humidity control are not intrinsically integrated within the primary cooling unit (Kim et al. 2016). As a result, fresh-air supply and latent load management may depend on separate systems or simplified control strategies. This architectural separation may increase system complexity and compromise integrated control performance, particularly in buildings located in hot and humid climates (Yau and Rajput 2022).

From a performance perspective, inverter-driven compressors and variable-speed fans have improved part-load efficiency in modern decentralised systems (Hernandez III and Fumo 2020). However, the distributed nature of multiple independent units may complicate system-wide monitoring, commissioning, and long-term performance verification (Lee and Kim 2024). Without coordinated supervision, operational degradation or control mismatches can remain undetected, contributing to discrepancies between expected and actual performance (de Wilde 2014). Overall, whilst decentralised HVAC systems offer flexibility and ease of deployment, their effectiveness depends strongly on appropriate integration with ventilation strategies and careful consideration of humidity control and operational management. A comparative synthesis of centralised and decentralised HVAC systems is presented in Table 3.

Whilst centralised and decentralised HVAC configurations differ structurally, their real-world performance is increasingly shaped by the degree of system integration and control coordination rather than architecture alone (Zou et al. 2019). Contemporary building applications increasingly adopt hybrid HVAC architectures that decouple sensible and latent loads, incorporate variable-speed components, and implement coordinated supervisory control strategies to enhance energy efficiency and indoor environmental quality (Fan et al. 2023; Nawaz and Gluesenkamp 2018). In such contexts, architectural flexibility must be balanced against ventilation robustness, humidity regulation, and long-term operational stability (Rashid et al. 2025). Consequently, system effectiveness is not solely determined by nominal efficiency ratings, but by how coherently design intent, control logic, and climatic responsiveness are aligned throughout the building lifecycle (Pappalardo and Reverdy 2020). This perspective underscores the demand so as to examine HVAC performance beyond static typologies, moving toward an integrated evaluation of design-operation coupling mechanisms (Saloux et al. 2023). The interrelationship between system configuration, operational strategies, and performance outcomes is conceptually illustrated in Figure 2.

Building upon this classification and integration perspective, the following section examines the operational strategies and control approaches that shape HVAC performance in practice.

In practical applications, HVAC systems operate predominantly under part-load conditions rather than at their rated design capacity (Huang et al. 2015). This operational reality arises from the inherently dynamic nature of building thermal loads, which fluctuate in response to occupancy patterns, internal heat gains, ventilation requirements, and climatic variability (Castaldo and Pisello 2018). Consequently, system performance cannot be adequately characterised by nominal full-load efficiency metrics alone (Crespi et al. 2022). Under part-load operation, several thermodynamic mechanisms contribute to deviations between rated and realised efficiency (Anjomshoaa and Salmanzadeh 2018). First, compressor cycling introduces transient losses associated with start-up phases, refrigerant pressure stabilisation, and incomplete heat exchanger utilisation (Agharid et al. 2025). In systems lacking modulation capability, repeated on-off cycling increases specific energy consumption and diminishes effective seasonal performance (Agharid et al. 2025). Second, oversizing, often implemented as a conservative design practice, is able to exacerbate short-cycling behaviour, further widening the gap between predicted and actual efficiency (Woradechjumroen et al. 2014). Variable-speed technologies, including inverter-driven compressors and electronically commutated fan motors, have significantly improved part-load performance by enabling continuous capacity modulation and load matching (Mahmoudi et al. 2021). By maintaining stable evaporating and condensing conditions across a broader operational envelope, such systems reduce cycling losses and improve seasonal efficiency indices (Xia et al. 2021). However, these gains are contingent upon appropriate control tuning and supervisory coordination (Homod 2018). In the absence of coherent control logic, the theoretical advantages of modulation may not fully translate into operational savings (Vakiloroaya et al. 2014).

Moreover, part-load thermodynamics cannot be considered independently of ventilation strategy and latent load management (Ma et al. 2024). In humid climates, latent loads may dominate during low sensible load periods, challenging systems that prioritise sensible capacity modulation (Liang et al. 2022). This interaction underscores the importance of integrated evaluation frameworks that couple system architecture with control strategy and climatic responsiveness (Homod 2018). Hence, from a performance assessment side, effective HVAC evaluation has to extend beyond rated COP values and incorporate dynamic load behaviour, cycling characteristics, modulation stability, and system-level coordination (Alves et al. 2016). These factors collectively determine whether nominal efficiency potential is realised in operational practice (Seo et al. 2026). The degradation mechanisms discussed above do not operate in isolation but emerge across multiple hierarchical layers of HVAC system design and operation. To synthesise these interactions, Table 4 categorises the principal factors influencing the realisation of nominal efficiency under part-load conditions.

While Section 4.1 examined equipment-level thermodynamic degradation under part-load conditions, the realised operating envelope of HVAC systems is further reshaped by load composition and ventilation strategy. In practice, cooling demand is not a homogeneous scalar quantity but a composite of sensible and latent components whose relative proportions vary with climate, occupancy, and outdoor air requirements. In humid and mixed climates, latent loads may constitute a substantial fraction of total cooling demand, particularly during periods of reduced sensible load (Cao et al. 2025). Under such conditions, the sensible heat ratio (SHR) decreases, and moisture removal requirements persist even when temperature-driven cooling intensity declines (Chua et al. 2008). This decoupling between sensible demand and latent persistence alters evaporator operating temperature and compressor lift, thereby affecting system COP and operating envelope (Thu et al. 2018). This transition from sensible-dominant to latent-dominant operation fundamentally reshapes the effective thermodynamic operating envelope of HVAC systems. As conceptually illustrated in Figure 3, persistent latent demand forces operation at lower evaporator temperatures and higher compressor lift even when sensible cooling requirements are reduced. As a result, extended runtimes and moisture-driven operation led to a progressive degradation of realised efficiency that is not captured by temperature-based performance metrics alone.

Systems primarily designed around sensible capacity modulation may encounter efficiency penalties when latent control becomes the governing constraint (Winkler et al. 2019). Variable-speed compressors can reduce cooling capacity at part load; however, effective humidity control relies on extended runtimes and sustained low coil temperatures to maintain adequate latent removal (Kone and Fumo 2020). Consequently, modulation strategies optimised for temperature tracking alone may not minimise energy use under moisture-dominant scenarios. Ventilation coupling further amplifies the sensible-latent imbalance (Safdari et al. 2024). Outdoor air introduction, whether constant-volume or demand-controlled, directly increases latent cooling requirements in humid climates (Yang et al. 2020a). Elevated ventilation rates can therefore shift system operation away from optimal part-load thermodynamic conditions by increasing total enthalpy removal demands (Sekartaji et al. 2023). In decentralised or split configurations lacking dedicated outdoor air management, humidity control may rely on extended compressor operation or lower evaporating temperatures, increasing specific energy consumption (Chen et al. 2018).

Hybrid configurations incorporating DOAS attempt to mitigate this imbalance by decoupling latent and sensible loads (Mohammed et al. 2026). When properly coordinated, such architectures can stabilise part-load operation by decoupling latent and sensible control loops, thereby reducing compressor cycling frequency and improving humidity regulation stability (Cheng et al. 2023). However, the benefits of decoupling depend critically on airflow balancing, control synchronisation, and accurate humidity sensing (Li and Wang 2022). Poor integration may simply redistribute inefficiencies rather than eliminate them. These observations suggest that part-load thermodynamics cannot be assessed independently of load composition and ventilation strategy (Alves et al. 2016). Climatic humidity, occupancy-driven ventilation variation, and sensible-latent coupling collectively redefine the operating boundaries within which modulation and control strategy’s function (Yang et al. 2020b). Therefore, performance assessment methodologies must explicitly account for load composition variability rather than treating cooling demand as a purely temperature-driven parameter (Liu et al. 2021). The multi-dimensional implications of sensible-latent load variability and ventilation coupling on realised HVAC efficiency are synthesised in Table 5.

Building on the thermodynamic degradation mechanisms discussed in Section 4.1 and the sensible–latent and ventilation coupling effects examined in Section 4.2, this subsection focuses on the control layer that converts these physical constraints into operational outcomes. Control architecture determines whether design intent and part-load capability are achieved in real operation. Although thermodynamic behaviour and load composition define the physical limits of HVAC systems, the conversion of theoretical efficiency into actual performance is largely a control-related issue (Mikhail et al. 2023). Modern HVAC systems operate through multi-layered control structures. These include local equipment modulation, zone-level feedback, ventilation management, and supervisory optimisation (Raman et al. 2021). Figure 4 illustrates this hierarchy and shows how coordination problems can emerge across equipment-level, zone-level, and supervisory control layers. These layers are also affected by external drivers such as weather conditions and occupancy patterns. As a result, several points of misalignment can reduce realised performance. The degree of coordination across these layers determines whether nominal efficiency values can be achieved in practice (Kim et al. 2017). In well-integrated systems, compressor speed regulation, fan modulation, airflow balancing, and humidity control operate coherently under supervisory guidance (Chuang et al. 2019). However, in many real installations, misalignment between control layers leads to inefficiencies that are not visible in rated performance values (Du et al. 2016). Communication delays, inconsistent sampling intervals, sensor bias, and conflicting setpoint hierarchies may induce unstable modulation behaviour, unnecessary compressor ramping, or excessive ventilation delivery under part-load conditions (Su et al. 2021). These effects collectively contribute to seasonal efficiency degradation without altering equipment-level specifications (Firdaus et al. 2023).

Coordination becomes more difficult in hybrid systems or in systems where sensible and latent cooling are handled separately. In these configurations, temperature and humidity are controlled through partly coupled processes with different sensible heat ratio characteristics (Gluesenkamp and Nawaz 2021). Poor supervisory coordination can affect condenser water control, compressor loading sequences, and cycling behaviour. This may increase effective lift or keep the system operating in low-efficiency states for longer periods (Geister and Thompson 2009; Wu et al. 2021; Rafati et al. 2022). Therefore, the theoretical benefits of load decoupling may be reduced if control coordination is poor. Recent advances in distributed and predictive control strategies attempt to address these issues by embedding system-level awareness into supervisory decision-making (Jiang et al. 2016). MPC, optimisation-based scheduling, and multi-agent frameworks integrate thermal inertia, occupancy variability, and ventilation constraints into forward-looking modulation strategies (Kim et al. 2022; Jiang et al. 2016). Unlike conventional reactive loops, these approaches anticipate load evolution and reduce short-cycling tendencies (Kim et al. 2022; Taheri et al. 2024; Raman et al. 2021). Nevertheless, their effectiveness remains contingent upon sensor fidelity, computational robustness, and practical commissioning quality (Kim et al. 2022; Zhou et al. 2024; de Wilde 2014). The recurring performance gap observed between design-stage simulations and field-measured operation underscores the centrality of control architecture. Efficiency realisation is not solely a function of component technology but of how effectively modulation strategies are coordinated across the entire system hierarchy. Consequently, control architecture should be regarded not as an implementation detail but as a defining determinant of operational performance in contemporary HVAC systems.

Conventional HVAC performance assessment has historically relied on steady-state indicators such as rated COP, EER, and seasonal indices including SEER, SCOP, and IPLV (Erginer 2025). While these metrics provide standardised benchmarks for equipment comparison, they are inherently derived under controlled boundary conditions characterised by fixed inlet temperatures, prescribed airflow rates, and simplified load assumptions (Gobel et al. 2022). In contrast, real building environments are governed by continuously varying thermal loads, humidity ratios, ventilation requirements, and occupancy patterns (Yang and Becerik-Gerber 2016). Under such dynamic conditions, HVAC systems rarely operate at rated design points (Seo and Lee 2016). Instead, they function predominantly under partial-load regimes shaped by fluctuating sensible and latent demands (Kone and Fumo 2020; Seo and Lee 2016). Seasonal indices attempt to account for part-load operation through bin methods and degradation coefficients; however, these approaches remain dependent on predefined climate distributions and simplified load weighting schemes (Alves et al. 2016; Erginer 2025; Seo and Lee 2016). They do not explicitly represent cross-layer interactions between system architecture, modulation limits, supervisory coordination, and external disturbances (Shi et al. 2019; van Dronkelaar et al. 2016; Alves et al. 2016). Consequently, nominal efficiency ratings should be interpreted as equipment potential rather than realised building performance (Shi et al. 2019; van Dronkelaar et al. 2016; Alves et al. 2016; Seo et al. 2026). The divergence between rated specifications and field-measured outcomes emerges not solely from component inefficiencies, but from systemic interactions that are not embedded within conventional rating methodologies (Du et al. 2016; Su et al. 2021).

Realised HVAC efficiency is the emergent outcome of interacting determinants operating across multiple hierarchical layers. These layers, previously examined in Sections 3 and 4, interact dynamically rather than independently. At the design level, system architecture, zoning resolution, and sizing strategies establish the structural boundary within which performance can occur. Conservative oversizing and limited zoning flexibility influence runtime fractions and cycling behaviour long before operational control strategies intervene. At the equipment level, modulation capability, turndown limits, and variable-speed integration govern the stability of part-load operation. Even advanced inverter-driven systems are constrained by minimum capacity thresholds and compressor lift characteristics that shape effective operating envelopes. At the operational and control levels, supervisory coordination determines whether nominal efficiency potential is translated into practice. Misaligned setpoints, delayed feedback, inconsistent sampling intervals, or ventilation over-delivery can introduce inefficiencies that are invisible in equipment ratings yet significant in seasonal energy use. At the climatic and occupancy levels, external disturbances redefine load composition. In humid climates, latent load persistence may sustain dehumidification demand even during reduced sensible cooling periods, altering evaporator conditions and compressor lift. Ventilation-induced enthalpy shifts further reshape system operation. These determinants do not act sequentially; they act simultaneously. Realised performance, therefore, reflects the degree of alignment across design intent, modulation capability, supervisory logic, and climatic responsiveness.

In light of these interactions, HVAC performance evaluation should transition from equipment-centric benchmarking toward system-level interpretative logic. Rather than asking whether a unit achieves its rated COP, assessment should examine how effectively the overall system maintains performance under dynamic load variability and cross-layer disturbances. A system-level evaluation logic may therefore consider three integrative dimensions:

This reframing does not require abandoning nominal metrics; rather, it situates them within a broader interpretative structure that acknowledges multi-layer coupling. Nominal efficiency indicators remain essential reference points, but their translation into realised performance depends on systemic alignment rather than intrinsic equipment quality alone. By embedding performance evaluation within this design-operation-performance coupling perspective, HVAC assessment becomes a holistic exercise in system coherence. Such an approach is particularly relevant for contemporary low-carbon buildings, where marginal efficiency gains and demand flexibility depend on accurate representation of dynamic behaviour rather than static rating conditions.

The integration of renewable energy sources, like solar and wind, into HVAC systems introduces additional complexity due to supply intermittency and temporal mismatch between energy generation and building demand (Reddy et al. 2024; Wang et al. 2019). Consequently, conventional performance metrics (e.g., COP, EER, seasonal indices) remain insufficient for evaluating such systems, as they do not fully capture dynamic operational behaviour and system-level performance under real operating conditions (Alves et al. 2016; Erginer 2025).

To enable meaningful comparison between renewable-assisted and grid-interactive HVAC systems, performance evaluation should incorporate extended indicators that capture system flexibility, demand response capability, and interactions with variable energy supply conditions (Wang et al. 2019; Manz et al. 2014). These metrics reflect not only thermodynamic efficiency but also broader system-level interactions, including demand flexibility and integration with distributed energy resources (Wang et al. 2019; Schiller et al. 2020; O’Connell et al. 2020; Arteconi et al. 2019).

From a design perspective, traditional peak-load-based sizing approaches may be inadequate due to uncertainties in load prediction and the inherently dynamic nature of building operating conditions, particularly in systems influenced by variable renewable energy inputs (Huang et al. 2015; Castaldo and Pisello 2018; Alfadil et al. 2026). Instead, system sizing should incorporate load-generation matching, storage integration, and part-load adaptability, reflecting the increasing need for flexibility in building energy systems under variable and intermittent energy supply conditions (Wang et al. 2019; Elnour et al. 2024). In particular, previous studies indicate that system performance is closely influenced by the interaction between renewable energy availability and HVAC control strategies, especially in systems employing predictive or grid-responsive control approaches (Kim et al. 2022; Wang et al. 2019).

Therefore, renewable-integrated HVAC systems are increasingly evaluated within system-level frameworks that capture the interactions between energy generation, storage, control strategies, and climatic variability (Reddy et al. 2024; Elnour et al. 2024). This perspective is consistent with the need to integrate design and operational stages, as highlighted in the literature, and supports a more realistic assessment of HVAC performance in low-carbon buildings (Shi et al. 2019; van Dronkelaar et al. 2016).

Conventional sizing practices frequently rely on peak sensible load estimation and nominal equipment indicators such as COP, SEER, SCOP or IPLV. While these metrics remain useful for standardised equipment comparison, they are insufficient as standalone sizing criteria under dynamic building operation. In particular, they do not fully capture part-load instability, cycling losses, latent-load persistence, supervisory control limitations or climate-dependent shifts in the effective operating envelope. Therefore, the design-operation-performance coupling perspective proposed in this review should be interpreted as a complementary sizing and selection logic rather than a direct replacement for existing standards.

Within this framework, HVAC sizing should prioritise a wider set of quantitative parameters. At the capacity level, designers should report not only peak cooling or heating capacity, but also the distribution of expected operating hours across part-load ratios, the minimum stable capacity, the turndown ratio, runtime fraction and cycling frequency. At the efficiency level, COP or SEER should be supplemented by part-load efficiency maps, seasonal performance under representative load bins and efficiency variation under reduced sensible load conditions. For humid and mixed climates, sensible-latent performance indicators should also be reported, including SHR variation, latent removal capacity, indoor relative humidity stability, coil operating temperature and the energy penalty associated with moisture-dominated operation. In addition, control-related parameters such as set-point deviation, actuator oscillation, supervisory response time, sensor tolerance and ventilation oversupply should be considered because they determine whether nominal modulation capability can be translated into stable field performance.

The expected output of such a framework is not a single replacement efficiency value, but a multi-parameter sizing profile that supports more reliable design decisions. For designers and practitioners, this profile may include:

Such an approach would help avoid the common situation in which a system is correctly sized according to nominal peak-load criteria but performs inefficiently under the conditions that dominate most annual operating hours. It would also allow commissioning teams to compare design intent with measured operation by checking whether the system remains within its expected part-load, humidity and control-stability envelopes. In this sense, the proposed framework provides a practical bridge between conventional equipment-based sizing and real-performance-oriented HVAC design for low-carbon buildings.