HVAC characterisation of existing Canadian buildings for decarbonisation retrofit identification

2.1 HVAC characterisation approaches or archetypes

HVAC system characterisation across buildings frequently uses a three-stage process (Johari et al. 2020; Sokol et al. 2017): building data collection, building classification, and system characterisation. Data such as building geometry, constructions, systems and usage schedules are typically collected through surveys and used to segment buildings into subgroups using deterministic, probabilistic or clustering approaches, depending on data availability. Deterministic approaches classify buildings using simulated energy consumption using reference models or archetypes, where real data is unavailable. Probabilistic classification further subdivides buildings based on historical energy use, improving building categorisation (Cerezo Davila et al. 2016). Clustering is an unsupervised method that groups buildings by energy consumption patterns. To characterise HVAC systems, field surveys and energy code requirements are often used (Johari et al. 2020), using deterministic and/or probabilistic classification. Deterministic methods are appropriate where building characteristics, for example geometry and typology, are known, while probabilistic classification is more appropriate when these are uncertain (Sokol, Cerezo and Reinhart 2017). Table 1 shows HVAC characterisation from BTAP archetypes based on ASHRAE Standard 90.1-2004 (ASHRAE 2004) using deterministic studies informed by field data (Deru et al. 2011).

2.2 Evolution of North American HVAC systems

Figure 1 shows the timeline of the HVAC system development in North America, based on a seminal US study (US DOE 2015). This is also relevant in the Canadian context owing to the shared market.

Figure 1

Figure 1 shows that, while heating systems such as wood stoves in smaller buildings and furnaces in larger ones have been around since the early 1800s, cooling systems were not mainstream until 1935, when they started to be sold commercially in North America (Carrier 2012). Electric resistance heating in the form of baseboards became popular in the 1960s and is still used in some locations where electricity is inexpensive. The development of electric boilers has supported their retrofit to replace gas boilers in some buildings to support decarbonisation, while heat pumps – whether air-to-air (AAHP), air-to-water (AWHP), ground source (GSHP) or water-source (WWHP) – are common retrofits for residential (Gaur et al. 2021) and some commercial buildings (Toronto Hydro 2023), with AAHPs being most popular for smaller structures (Staffell et al. 2012). For these stated reasons, heat pumps are the technological retrofits proposed in this study.

3.1 Characterisation of HVAC systems

Based on previous studies (NRCan 1997b; US DOE 2015), a more granular set of periods was developed to guide this analysis: pre-1935, 1935 to 1945, 1945 to 1970, 1970 to 1985, and 1985 to 2004. These periods were informed by the changes in HVAC systems over the last two centuries. For each period, deterministic classification had to be used owing to the lack of publicly disclosed energy consumption. Typical systems and their efficiencies were initially estimated based on reference models (US DOE 2012) and building code requirements (NRCan 1997b; NRCan 2011a) and guidance documents (NRCan 2011b).

From this analysis, a preliminary spreadsheet was developed for expert review in using the Delphi technique. The Delphi technique is a procedural research method that involves repeated rounds of input by experts on a specific research problem until a consensus is reached (Barrett & Heale 2020; Linstone 1985). It is based on the principle that forecasts or decisions are more accurate from a structured group of individuals, such as a body of experts, than those from unstructured sources such as surveys. The Delphi technique is widely used to develop best practice guidance when data is limited, ethically or logistically difficult to obtain, or conflicting (Barkhordari Ahmadi et al. 2023). While heavily used in social science, characterisation and energy modelling studies are increasingly adopting this approach (see Jayawardena et al. 2022; Phichetkunbodee et al. 2023). In this study, two consulting firms supporting this constituted the body of experts, reviewing the spreadsheet data against their internal energy model databases (200+ per firm), providing clarifications and corrections in rounds until consensus was reached.

3.2 Retrofit characterisation

Retrofit identification and characterisation followed a similar methodology to the HVAC system characterisation, with additional constraints related to building geometry. ASHRAE 90-1-2022 efficiency requirements (COP, SEER etc.) were used as these constitute the code minimum and are thus conservative in nature. Because recent economic conditions make it impossible to estimate representative costs across a diversity of buildings, this paper excludes explicit return-on-investment calculations, though relative cost is considered based on the difficulty or extent of the retrofit. Instead, CO₂ emission reductions are used, following ASHRAE (2022), and AEE (2023), which is consistent with government metrics (Government of Canada 2022). Other drivers for owners to invest in retrofits include reputational benefit, regulatory pressures, investor value, equipment end of life, and avoiding obsolescence (Brooks & McArthur 2019). Recognising these, three levels of retrofit have been developed as-follows. R1 represents the ‘easiest’ or most cost-effective retrofit according to market research, typically a like-for-like equipment replacement with fuel-switching. R1 exists for building owners who are replacing equipment at the end of its life or who are seeking to minimise retrofit costs. R2 represents the HVAC system retrofit offering the most significant energy savings; for example, replacing a gas-fired air handler with VRF systems will both avoid direct CO₂ emissions and increase efficiency from 96% to a coefficient of performance (COP) of 4 or higher. R2 reflects the ambitious or sustainability-minded building owners more concerned with long-term energy savings and reputational benefits than initial capital costs. R3 represents the potential available to building owners in presence of a medium-temperature (40–70°C) district heating system, supporting the installation of high-temperature water-to-water heat pumps to replace boilers. R3 can support broader district energy or thermal network feasibility studies, which are increasingly gaining traction as a means of large-scale urban decarbonisation (Fry et al. 2024). Because such systems can be implemented using central equipment upgrades rather than full system replacement, this represents a moderate level of investment.

3.3 Sensitivity analysis

EnergyPlus was used to explore the impact of grid carbon intensity on retrofit effectiveness. For each building type and period, energy models were created for the baseline condition and three levels of retrofit using OpenStudio. All models used the geometry, envelope assemblies and occupancy schedules from the DOE reference building models (US DOE 2024a; US DOE 2024b), changing only the HVAC system types and characteristics.

Once simulated, the expected energy consumption and associated CO₂ emissions were calculated (Equation 1), considering both direct emissions from on-site fossil fuel combustion and indirect emissions from grid-supplied electricity. A range of real urban power grid carbon intensities were used to allow the impact of grid decarbonisation – or recarbonisation – on retrofit CO₂ impacts to be evaluated.

where ɛ_total is the total CO₂ emissions from an archetype characterisation or retrofit choice, e_NG is the emission factor of the natural gas grid, C_NG is the annual energy consumption of HVAC systems that require natural gas, e_elec is the emission factor for the electric grid, and C_Elec is the annual energy consumption of HVAC systems that require natural gas. All retrofits developed fully electrify the building, including domestic hot water systems, so C_NG = 0 for all retrofits.

To explore the impact of grid sensitivity, each of the baseline and retrofits was modelled in the following regions, representing the full range of grid carbon intensities across climate zone 5 A in North America: very low, based on Quebec (0.472 kgCO₂/GJ (ECC Canada 2022)); low, based on Toronto (6.67 kgCO₂/GJ (ECC Canada 2022)); medium, based on New York (112 kgCO₂/GJ (US EIA 2023)); high, based on Ohio (112.6 kgCO₂/GJ (US EIA 2023)); and very high, based on Michigan (208 kgCO₂/GJ (US EIA 2023)). We define a low-carbon grid as one where e_elec < e_NG, which is approximated as 51 kgCO₂/GJ (ECC Canada 2022; US EIA 2023) in all jurisdictions. A hypothetical decarbonised electric grid with an emission factor of zero kgCO₂/GJ was also considered. The energy model outputs used for the sensitivity analysis are provided as supplementary data.

4.1 HVAC characterisation

Six major clusters of systems were identified based on the typology and vintage: rooftop package and DX cooling (PRT-DX), gas-fired recirculating multi-zone (RMZ) air handling unit (AHU), non-condensing boiler with natural air convection, non-condensing boiler with a RMZ AHU and local fan coil units (FCUs), and non-condensing boiler with air conditioners (ACs).

According to the historical survey (US DOE 2015) and the Delphi analysis, HVAC technology advanced every 10 to 15 years and HVAC equipment has a lifespan of 15 to 25 years on average (Lee & Ahn 2018). Canadian building codes did not specify required efficiencies for HVAC systems until 1997 (NRCan 1997b) and these were increased in 2011 (NRCan 2011a), though the common market with the US would have driven equipment manufacturers to design equipment compliant with ASHRAE 90.1-2004. Accounting for actual equipment service lives, in the absence of more precise information, central equipment efficiencies are assumed to at least meet NECB 1997, while terminal units – with shorter lives – are assumed to meet or exceed those prescribed by ASHRAE 90.1-2004 (ASHRAE 2004), which was mandated in 2011 (NRCan 2011a).

The consensus from the Delphi method was that 70% of AHU systems use VAV, while 30% remain CAV owing to regulation shifts in the 1980s; this is assumed to be the default for any AHU systems encountered after 1980, while pre-1980 buildings may have CAV or VAV systems with a higher distribution of VAV. All MURBs are expected to utilise some form of non-condensing boiler for heating owing to high envelope loads and space constraints requiring higher operating temperatures. Small MURBs typically either do not have cooling or have window ACs. Mid- and high-rise MURBs will be expected to use water chillers and small MURBs either have non-existent cooling systems if built before 1970 owing to the cost of installation and reduction of space, or occupant-installed window-mounted ACs.

From the 1980s, mid- and high-efficiency (83–87%) systems became more popular, and some existing buildings were converted from low-efficiency cast-iron sectional oil boilers to natural gas boilers. The change in focus on HVAC efficiency was driven by the energy crisis of the early 1970s, which placed a greater emphasis on energy efficiency (McPhie & Caouette 2007). Code minimum HVAC efficiencies have been assumed for all periods, shown in Figure 2 (heating) and Figure 3 (cooling). Nine subclusters were finalised from this analysis, summarised in Table 2. Condensing boilers became popular in Canada only after 2005 and then generally as pre-heat, since minimising the size and cost of heating terminal devices such as baseboard convectors was paramount in older buildings and thus systems were designed to supply water at 82°C and return at 71°C.

Figure 2

Figure 3

4.2 Retrofit characterisation

Four main factors affect retrofit selection: HVAC system characterisation, space availability, electrification capacity and installation cost. However, for this study, cost was not considered owing to the factors detailed in Section 3.2 of this paper. Both the electrical capacity in a building and the available grid capacity impact the viability of electrification retrofits and must be considered on a location-by-location basis.

HVAC systems take up varying amounts of space, both within and around the building, so retrofit feasibility had to consider space availability to avoid requiring costly renovations. Further, any retrofits installed in mechanical rooms will require examination of existing pipework. AAHPs that require wall mounting (i.e. ductless AAHPs) or rooftop installation (singular vertical HPs) should be evaluated for spatial compatibility with a building that contains an air source system.

Ideally, new systems would be compatible with existing distribution systems to avoid extra expenses for system redesign. In addition, the operational temperature ranges and distribution medium have an impact on overall system energy efficiency and should be considered in retrofit selection. An AHU can only be retrofitted with AAHPs, FCUs require water systems that operate on medium (70°C) to high (90°C) temperatures, and any system with radiator units requires high-temperature systems such as high-temperature WWHPs. The ASHRAE 90-1-2022 minimum efficiency requirements were considered for all retrofit options as a baseline, with AAHPs operating on a COP_H of 2.5 and WWHPs on a COP_H of 4 (ASHRAE 2022; NRCan 2022). Some retrofit options that are easy to install but low in efficiency will need to use auxiliary heating in subzero (below –5°C) outdoor temperatures.

For the ‘easy’ (R1) retrofits, the gas-fired AHUs typically found in clusters H1 and H3a can be replaced with those using electric resistance coils. The boiler systems that use FCUs can be replaced with air-to-water HPs if they have fan coil units operating on medium temperatures (H2a and H6), while boilers found in H2b, H4 and H5 can be retrofitted with high-efficiency (99%) electric boilers to avoid local unit replacement and minimise cost. Mid- to large-sized MURBs and commercial buildings (H3b, H5–6) have the potential to use GSHPs as a retrofit alternative, as a number of them have sufficient land for installation (Toronto Hydro 2023).

For the ‘most efficient’ (R2) retrofits, AAHP AHUs replace the PRT-DX AHUs of clusters H1 and H3a. Buildings with non-condensing boilers with hydronic heating (H2a–b, H4–H6) can be retrofitted with high-temperature WWHPs. Where space allows, ductless AAHPs could also be installed for cooling for H2b, H4 and H5. ‘District energy enabled’ (R3) retrofits focus on central heating and cooling plant equipment compatible with such systems. Table 3 summarises these retrofit options as well as their minimum code efficiencies for cooling and heating (ASHRAE 2022).

4.3 Grid carbon intensity sensitivity analysis results

In small commercial buildings built after 1960 (H1), all retrofits reduce emissions for all grids. Figure 4 represents this graphically.

Figure 4

Even with considerable emissions reduction on both low-intensity (QC and TO) and higher-intensity grids, the reduction is less substantial on high-intensity grids. This trend is most prominent in the highest energy impact approach (R2), as it has the highest emissions values of any retrofit of all high-intensity grids, and this is closely followed by the easiest retrofit approach (R1). For H1, the district heating retrofit (R3) is most suitable for all grids, as it produces the lowest emissions for all cases.

Pre-1960 mid-size commercial buildings with FCUs (H2a) have reduced emissions with all retrofits on the low-carbon grids. However, the high-carbon grids see rise in emissions with all retrofits when compared with baseline emissions. Figure 5 shows this trend in detail.

Figure 5

With H2a buildings on low-carbon grids, emissions fall below baseline values for all retrofits, with R2 producing the lowest emissions. High-carbon grids show a different trend; all retrofits exceed the baseline on all grids, with R1 producing more than 100% of the baseline emissions. R2 would be the best retrofit for low-carbon grids, but no retrofit approach would be suitable for H2a on high-carbon grids.

In pre-1960 mid-size commercial buildings with hydronic heating (H2b), the highest emission reductions can be seen with all retrofits on all low-carbon grids. The high-carbon grids, however, show a different trend. There is a marked increase in emissions on all high-carbon grids with R1, while R2 and R3 only decrease emissions on the medium and high grids, as shown in Figure 6.

Figure 6

All retrofit emissions fall below baseline values on both low-intensity grids, with R2 and R3 producing the lowest emissions. Similar to the pattern observed in H2a, carbon emissions are much greater than baseline emissions for R1 on high-intensity grids. R2 and R3 emissions fall below baseline values on all high-intensity grids except those of very high intensity, where emissions exceed the baseline by 38%. Both R2 and R3 prove to be effective retrofit approaches for low-carbon grids. Both approaches also apply to high-intensity grids with high intensity or lower, though these approaches provide minimal emissions savings.

In mid-size commercial building stock established after 1960 (H3a), the largest emission reductions are seen on the low-carbon grids. The only reductions on high-carbon grids, while minimal, are from R1 and R3 on the medium grid, as observed in Figure 7.

Figure 7

Retrofits for low-carbon grids have substantially lower emissions in comparison to their baseline. R1 has the lowest of all retrofits on these grids, followed closely by R1. R2 exceeds baseline emissions on all high-carbon grids, R1 exceeds baseline emissions on high- and very-high-intensity grids and R3 exceeds emissions on medium-intensity grids. This trend makes sense, as R2 produces the highest emissions on both low- and high-carbon grids. R1 would be the most effective retrofit approach for low-intensity grids, while R3 might only be marginally effective for high-intensity grids with medium intensity or lower.

For large commercial buildings built after 1960 (H3b), the trend of considerably high emissions reductions continues on low-carbon grids. There are minimal emission savings on all high-carbon grids for all retrofits except R2, where there is a substantial increase in emissions. Figure 8 shows this trend in detail.

Figure 8

On all low-carbon grids, all retrofit methods considerably fall below baseline values, with R1 and R3 producing the lowest. On all high-carbon grids, only R1and R3 fall below baseline emissions. However, the higher the intensity, the less the reductions. R2 produces higher emissions than the baseline for all high-carbon grids. R1 is determined to be the most effective retrofit approach for H3b on both low-carbon and high-carbon grids, though its effectiveness is remarkably lower on high-carbon grids. R3 is also an effective retrofit choice for low-carbon grids, as it has the same emissions values as R1.

In mid-rise MURBs built after 1970 that utilise hydronic heating (H5), emissions are significantly reduced with all retrofits on all low-intensity grids, as seen in Figure 9. Note that because of limitations on modelling window AC units, the baseline electrical consumption has known errors and should be considered only as a relative value for comparison with retrofits rather than as an absolute quantity.

Figure 9

On all low-carbon grids, all H5 and H6 retrofits reduce emissions considerably, with R3 maximising these reductions. For H5, R1 emissions exceeds baseline values for all high-carbon grids, and the higher the grid intensity the higher the emissions increase. R3 is the only retrofit approach with any emission decrease in comparison to the baseline; the medium- and high-intensity grids both have lower emissions on R3, while the very-high-intensity grid slightly exceeds baseline emissions on R3. R3 would be the most effective retrofit method for low-carbon grids, and could be somewhat applicable on high-carbon grids with high intensity or lower. In post-1980 mid- and high-rise MURBs that utilise FCUs as terminal units (H6), trends similar to those observed in H5 can be seen.

Figure 10 shows this graphically. For H6, R2 exceeds baseline emissions on all high-carbon grids, while R3 would be the most effective retrofit method for low-carbon grids and could be somewhat applicable on high-carbon grids with high intensity or lower.

Figure 10

Overall, all retrofits are suitable choices for all alternatives in low-carbon grids. R3 is the lowest of all original retrofit selections for all archetypes on low-carbon grids, meaning that district heating could be a viable step towards net-zero emissions. On some grids, the energy-efficient option R2 has much higher emissions than the cost-effective choice, notably where R2 introduces VRF systems. Longer cooling durations in VRF systems can result in decreased equipment efficiency and thus an increase in electrical consumption. This will need to be considered when suggesting retrofit strategies and occupancy setpoints will need to be readjusted. H1 retrofits saw reductions on all grids, though the amount of reduction increased with grid decarbonisation. However, there were minimal emissions for medium-carbon grids and emissions actually increased with fuel-switching/electrification in the most carbon-intensive grids.

These results show that most electric HVAC retrofits would not be appropriate for high-carbon-intensive grids; instead, grid decarbonisation should be addressed first, particularly in areas with a higher concentration of mid-size to large buildings. The findings also highlight the importance of equipment lifespan in decarbonisation planning, as it significantly influences life-cycle cost analyses and building owners are far more willing to replace ageing systems than newer ones, emphasising the need to proactively educate building owners about retrofits prior to equipment so that decarbonisation can be actively considered in capital planning. In lower grid intensity jurisdictions, the life-cycle carbon associated with premature replacement of fossil fuel heating equipment can be advantageous; however, building- and solution-specific analysis to quantify the carbon break-even point would be required to determine how far before equipment end of life this benefit exists.

4.4 Limitations

The characterisation of HVAC systems for archetype development requires generalisations of building consumption patterns. Factors such as changes to grid availability over time and costs of changes vary on a building-by-building basis. These factors cannot be generalised towards archetype development, regardless of how granular the archetypes are. Thus, this consideration was not included in the study.

It is important to note that the clusters and retrofits identified in this study are by no means exhaustive. Many more configurations could be developed for less-common cases; however, the objective was to characterise the most common HVAC systems by period. Moreover, this study only focuses on the HVAC characterisation aspect of archetype development and does not consider envelope characterisation or climate change modelling.

This study is geographically limited, focusing on Canada within a specific climate zone, using Toronto as a focus area. However, the methodology can be readily adapted to other regions with appropriate consideration of local building codes, policies, climate conditions, equipment availability and typology classifications. The Delphi technique could prove valuable in those contexts to obtain aggregate insights without requiring direct access to confidential data, engaging building science experts to ensure the applicability of available technologies.

Additional limitations of this work resulted from the use of DOE reference buildings files for energy modelling. First, these limited the building typologies considered to those with reference models, namely small, medium and large office and mid- and high-rise MURBs. Because no reference model exists for small MURBs, these were excluded from this study. Second, the EnergyPlus reference files developed by NREL have not been updated since 2013 and, when updated locally to current versions of the program, building features such as windows and HVAC were omitted, requiring these to be manually reinserted into the simulation file. Given that this study defined new HVAC systems, the latter was less critical but is important to note for future studies. Furthermore, EnergyPlus and OpenStudio provided unrealistic values for window ACs for archetype H5, resulting in lower electrical baseline emissions than expected. Any attempts to correct this resulted in system errors, so electrical values in H5 baseline emission calculations should be discounted. EnergyPlus Support has been contacted to flag this issue in the software.

4.5 Areas of future studies

Future studies should consider envelope and climate effects on different retrofit technology strategies. A recent study (Bagherzadeh et al. 2024) has explored the impact of future climate scenarios on envelope retrofit effectiveness. Extending this HVAC characterisation study with such scenario analysis will further inform more enriched archetype granularisation and the feasibility of future retrofit strategies and is recommended for future work.

Another study recommendation is to explore the development of surrogate models as an energy consumption prediction method in existing buildings and of various retrofit selections. Traditional calibration procedures are time-intensive and require expertise, making their large-scale application buildings restrictive (Ali et al. 2020). To develop a city-wide UBEM, a data-driven approximation technique to estimate the properties of unknown building parameters is necessary to limit the time and effort required. Overall, the HVAC system characterisation developed in this study is valuable to support large-scale retrofit planning, whether as inputs for bottom-up UBEMs using archetypes or to create data for surrogate models that can be used to rapidly evaluate the impact of HVAC retrofits and fuel-switching within a selected district, thus supporting urban decarbonisation efforts.