Buildings in developed countries consume 20%–40% of the total final energy consumption [1
]; in 2017, residential and commercial buildings consumed 39% [2
] of all energy in the USA. Of this, about half [1
] is used to provide heating, ventilation, and cooling (HVAC). Likewise, in Sweden, in 2016, buildings consumed 39% of all energy [3
] and 55% of that quantity was used for space heating and domestic hot water (DHW) [4
]. An extensive amount of research, development, regulation, and investment has gone towards reducing energy consumption of building HVAC systems. Nevertheless, actual building energy consumption is often significantly higher than that planned at the design stage. This difference is sometimes referred to as the “performance gap” [5
The above studies rely on measurements of energy use intensity (EUI). Energy use intensity is commonly used for such studies because it is relatively easy to measure, needing as a minimum only building utility bills and floor area, and hence EUIs are widely available. The wide availability has facilitated compilation of large databases such as the US Department of Energy (DOE) Buildings Performance Database [14
], covering about 750,000 buildings, of which 44,000 are commercial and the remainder residential. Such a large data set allows a variety of “big data” approaches to analysis, including linear regression to estimate retrofit savings [15
], machine learning [16
], and clustering analysis [18
] to predict energy consumption. In many cases, submetering also allows EUIs for lighting and plug loads to be determined, further enhancing the value of these analyses.
Despite the many advantages and opportunities of using large data sets of energy use intensities, analyses of EUIs have inherent limitations. Specifically, it is difficult to differentiate between the effects of the building envelope loads and occupants and the performance of the HVAC system, e.g., if the building has high energy use, what is the cause? A poor building envelope? High internal heat gains or other occupant effects? Poor HVAC system performance?
This paper focuses on an approach that can help resolve this question for a specific building—field monitoring of the HVAC system performance—specifically for a ground-source heat pump system in a university building in Stockholm, Sweden. This is inherently more difficult and expensive, requiring additional sensors and data acquisition systems. Unfortunately, there are very limited comparison data available, so even after carefully measuring performance of a single building system, there are relatively few comparable measurements that have been reported in the literature. Nevertheless, the decreasing cost of sensors and data acquisition suggest that these analyses may be more widely available in the future. Thus, one contribution of this paper is providing a comparison benchmark for future studies.
One critical issue for representing field performance is the boundary schema—where should the boundaries be drawn when determining the system performance? Early work on defining boundary schema was done for residential systems, so we first review some of the work on field measurement of residential-building GSHP systems before covering the main focus of the review, measured-performance of commercial-building GSHP systems.
1.1. Literature Review—Residential Building GSHP System Performance Measurement
For residential buildings, field measurements of heat pump system performance have been reviewed by Gleeson and Lowe [21
]. The reviewed studies were undertaken in Austria, Denmark, Germany, Sweden, Switzerland, and the UK. The studies, which covered both ground-source and air-source heat pump systems, utilized sufficient instrumentation to measure system performance on a seasonal basis. Gleeson and Lowe identified two different boundary schema; both have multiple boundary definitions that may be applied. The boundary definitions differ as to which components are included—e.g., source-side circulating pump, load-side circulating pump, fans, etc.—and as to where heat delivered is measured, e.g., to the buffer tank or from the buffer tank to the heat distribution system. The authors attempted to harmonize the field trial results so that broad comparisons could be made between different field trials. Performance of over 600 heat pump installations was examined. One notable finding is an inconsistency in results that cannot readily be explained by the quality of the equipment or the building load profiles. For 216 GSHP installations, the seasonal performance factor that accounts for circulation pumps on both sides of the heat pumps and integrated backup heating ranged from 1.4 to 5.1. Gleeson and Lowe surmise that the issue is the quality of the design and installation.
A particular focus of the Gleeson and Lowe [21
] paper is harmonizing the field trial results computed with different boundary schema, each of which has different boundaries. The different boundaries are important for several reasons—these include comparison to conventional systems, benchmarking of GSHP system performance, identification of best practices, determining the causes of poor performance, and giving guidance as to how system performance may be improved. One such schema was defined by the EU project SEPEMO [22
], giving heating and cooling seasonal performance factors (SPFs) for a range of boundaries, as shown in Table 1
The SEPEMO boundary scheme was applied to 44 heat pump systems in six European countries (Sweden, Germany, Greece, the Netherlands, France, and Austria) with varying heat sources (air, ground) and heat distribution methods (panel radiators, floor heating). All but five of these heat pump systems were installed in residential buildings, mostly single-family houses. Nordman [22
] reports SPFH3
(heating system except the heat distribution inside the building) for the 44 heat pump systems, ranging from 1.3–7.3. Eight of the systems have SPFH3
below 2.6 and 15 systems (all GSHP systems) show SPFH3
above 4. Nordman [22
] notes that heat pump system performance depends not only on the heat pump, but also on the climate and quality of installation. The SEPEMO project guidelines work well for smaller residential GSHP systems, but do not fully address all of the features that may be found in GSHP systems serving larger and more complex GSHP systems, such as commercial, institutional, and multi-family buildings.
Note that the SEPEMO system boundaries for SPFH1 and SPFH2 correspond to SPFC1 and SPFC2, while the boundaries for SPFH3 do not correspond directly to those for SPFC3. SPFH3 includes auxiliary heating but not distribution pumps/fans, whereas SPFC3 includes distribution pumps/fans, but not supplementary cooling units. Hence if there is no auxiliary heating in the heating system, SPFH2 = SPFH3, while SPFC3 = SPFC4 for systems without supplementary cooling.
Miara et al. [23
] use a similar boundary scheme to the SEPEMO scheme, for residential heating only (Table 1
). System boundaries 0–2 correspond to the SEPEMO boundaries 1–3, while the fourth system boundary includes circulation pumps between back-up heater and storage tanks, but excludes the storage tanks and distribution system beyond the tanks. The field study by Miara et al. included 56 GSHP pump systems for small residential buildings over a period of three years. The reported average seasonal performance factors for the 56 GSHP systems were 4.19 for SPF0
, 3.93 for SPF1
, 3.88 for SPF2
, and 3.75 for SPF3
. The DHW share was 18% over the three-year period. SPF2
varied in the range of 3.75–3.9 over the three years when the measurements took place.
Another two system boundary schema suitable for residential GSHP heating systems are found in Koeningsdorff [24
] (pages 276–277) and in the German guidelines VDI 4650 [25
]. VDI 4650 (Table 1
) defines four system boundary levels, where the first only includes the heat pump unit (corresponds to the SEPEMO boundary H1), the second only includes the ground source with a load-side circulation pump, the third corresponds to the SEPEMO boundary H3, and the fourth boundary includes both the source-side and load-side. The Koenigsdorff scheme consists of seven system boundaries, of which the first three follow the VDI 4650 scheme, and the seventh corresponds to the fourth boundary in the VDI scheme.
1.2. Literature Review—Commercial Building GSHP System Performance Measurement
Commercial buildings often have considerably more complex GSHP systems than residential buildings. Beyond energy use intensity measurements, several authors have made measurements of system coefficient of performance (COP) for a single day [26
], several days [28
], or several weeks [29
Several authors [31
] have presented detailed performance measurements of building heating systems that incorporate GSHP and other renewable sources such as solar–thermal and solar–electric on site. The first paper focuses on other aspects of the system performance. Wemhoener et al. [32
] report monthly and annual performance factors, with an annual value of 3.92 for combined space heating and domestic hot water heating, accounting for the heat pump and source-side circulating pump. Lazzarin et al. [34
] describes a primary energy ratio metric that is the ratio of heating provided to primary energy consumed. Use of this metric helps get around the problem of how to treat electricity generated on site. The literature review in this paper focuses on GSHP systems that are characterized with SPF or COPs with the denominator being the electricity consumed without regard to whether it is generated on site or off site.
An aquifer thermal energy storage (ATES) system is a type of multi-well open-loop GSHP system, with hot and cold reservoirs maintained by seasonally switching the extraction and reinjection wells. Schmidt and Müller-Steinhagen [36
] described a rather complicated system used to provide heating to a 7000 m2
apartment building. This system combined solar collectors, a gas-fired boiler, ATES, and heat pumps. Seasonal heating COP for the heat pumps only was reported for three years, with an average value of 4.3.
Urchueguía et al. [37
] compared an air-source heat pump system and a ground-source heat pump system serving portions of a university building in Valencia, Spain. Daily and seasonal performance factors corresponding to the SEPEMO H2 and C2 boundaries were reported. For heating, SPFs of 3.5 ± 0.6 and 2.0 ± 0.3 were measured for the GSHP and ASHP systems, respectively. For cooling, SPFs of 4.3 ± 0.6 and 2.7 ± 0.4 were measured for the GSHP and ASHP systems, respectively. Montagud et al. [38
] present additional data for this system after five years of operation.
] utilizes the SEPEMO boundary schemes to determine daily performance factors (H2 and H4 boundaries) and seasonal performance factors (H1, H2, and H4 boundaries) for 19 GSHP systems serving non-residential buildings in the United Kingdom. All the buildings were part of the British Renewable Heat Incentive (RHI). SPFH1
values for the 19 buildings ranged from 2.36 to 4.64 with a mean value of 3.26, SPFH2
values ranged from 2.24 to 4.49 with a mean value of 2.95, and SPFH4
ranged from 1.21–4.12, with a mean value of 2.42, based on one year (July 2015–June 2016) of monitoring.
Zhai and Yang [44
] describe a large (280 boreholes) GSHP system serving a government building housing archives, and hence requiring close control of space air temperature and humidity. Average COPs of the heat pumps (presumably SEPEMO H1 and C1, but not explicitly stated) are given as 4.7 in the summer and 4.6 in the winter.
Vanhoudt et al. [45
] describe three years of performance monitoring results for an ATES system. The heat pumps have seasonal performance factors (SEPEMO boundaries H1 and C1) of 5.6 and 5.0. Seasonal system performance factors are also reported, though the boundaries are unclear—it is not clear whether “different circulation pumps” include both source-side and load-side pump energy consumption. Average values of 5.9 and 26.1 are reported. These are higher than the heat pump COPs because some heat is provided directly from the groundwater (presumably for tempering ventilation air) and much of the cooling is provided directly from the groundwater.
Michopoulos et al. [46
] monitored a ground-source heat pump system serving an office building in Greece. Weekly and seasonal performance factors corresponding to the SEPEMO H1 and C1 boundary conditions were presented. The seasonal performance factors vary from year-to-year, but are about 5–5.5 for heating and 4–4.5 for cooling.
Winiger et al. [42
] report on analysis of ten ground-source heat pump systems serving non-residential buildings in Germany. These systems were monitored as part of the broader Energy-Optimized Building (EnOB) project [47
]. The authors describe four system boundaries (see Table 1
) that have been evaluated, though none of the boundaries include only the heat pump. One case makes use of waste heating, allowing heating without heat pump operation—in this case, the heating SPF I was 61.7. Cooling SPF results are divided between systems using direct cooling (i.e., fluid circulated directly from the ground heat exchanger to the fan coil unit) and active cooling (i.e., using the heat pumps). When heating and cooling were being provided simultaneously, the electricity usage was allocated proportional to the thermal energy provided in each mode. Heating SPF II values (i.e., heat pump plus source-side circulating pump) ranged from 2.9 to 5.7. Heating and cooling SPF IV values (i.e., the entire system including the distribution system) ranged from 1.0–2.7 and 2.2–15.0, respectively, with higher cooling SPFs found for the direct cooling cases. SPF I and SPF II are given for six individual systems, with no other details provided.
Mermoud et al. [48
] used the SEPEMO boundary conditions to analyze a GSHP system serving a 2200 m2
office building in Geneva, Switzerland. The system provides heating with heat pumps and direct cooling using the ground heat exchanger fluid. Heat pump COPs are presented for five-minute interval data, showing a significant decrease from the manufacturer’s data. The system utilizes two heat pumps and the authors computed separate SPFH1
, for each heat pump, with values of SPFH1
of 3.5 and 3.1 and SPFH2
of 3.1 and 3.0. Overall values of SPFH3
are 3.1 and 3.0. As the system has favorable return fluid temperatures from the ground (exceeding 10 °C all year) and the system only provides space heating, the authors find the SPF values disappointing. The poor performance is attributed to the hydraulic design, which requires higher supply water temperatures than what is actually needed.
Monitoring studies of two distributed heat pump systems at the ASHRAE Headquarters building in Atlanta were presented by Southard et al. [49
] and Spitler et al. [51
]. The two systems were an air-source variable-refrigerant flow heat pump system and a ground-source heat pump system. Seasonal heating and cooling system coefficients of performance were calculated for the two systems. Because the systems were distributed and delivered heating and cooling with fans integrated in the heat pump units, and the electrical energy was measured for all heat pumps and the source-side circulating pump together, the coefficients of performance correspond to SEPEMO levels H4 and C4. The air-source heat pump system had seasonal heating and cooling COPs of 2.0 ± 0.1 and 2.5 ± 0.1, respectively. The ground-source heat pump system had seasonal heating and cooling COPs of 3.3 ± 0.2 and 4.3 ± 0.6, respectively.
Spitler et al. [51
] presented system heating and cooling COPs as a function of outdoor temperature and heat pump entering fluid temperature. For both the air-source and ground-source heat pump systems, cooling COP increased with increasing outdoor air temperature and heat pump entering fluid temperature over a substantial part of the operation range. The counterintuitive behavior of the GSHP system was explained by parasitic losses due to control boards that consume power all the time, fans that run even when all compressors are off, source-side circulating pump energy consumption, and cycling losses of the heat pumps. Recent research [52
] into cycling losses shows the decreased COP that occurs during the first minute of operation can have a variable but significant effect on heat pump performance.
Mendrinos and Karytsas [43
] report on annual heating and cooling SPF for eight GSHP systems in southern Europe monitored as part of the EU Ground-Med project. Additional details can be found in Carvalho et al. [54
] and Pardo and Michal [55
]. Four levels of system boundaries are defined. As shown in Table 1
, these are similar to the SEPEMO boundary conditions for heating, except they do not account for auxiliary heating, and the energy used by the fans in the fan-coil units is treated separately [55
] from the load-side circulating pumps. Heating and cooling SPFs for all buildings for all four levels are reported graphically; heating SPF II ranges from 3.6–5.9. Cooling SPF II results are divided between systems using direct cooling (as defined above)—values between 9.8 and 16.6 are achieved, and active cooling—values between 4.9 and 6.8 were achieved.
Schibuola and Scarpa [56
] made field measurements of system performance for a ground-source heat pump system providing heating to an 1840 m2
university building originally constructed in the 16th century. Annual heating COPs corresponding to SEPEMO H1 and H2 were measured at 3.96 and 3.65, respectively. Annual cooling energy efficiency ratios (EERs) corresponding to SEPEMO C1 and C2, were measured at 4.32 and 4.02, respectively. These results were compared to simulated results for an air-source heat pump system.
Pater and Ciesielczyk [57
] compare theoretical performance coefficients provided by the heat pump manufacturer with SPF obtained in real operating conditions for a GSHP installed in a 460 m2
mixed-use building near Krakow in Poland. Measurements were carried out over three heating seasons (September–May). The reported SPF3, according to the Miara et al. [23
] system boundary scheme, ranged between 3.4–3.8 over the three heating seasons, which fell within the range of expected SPF based on the manufacturer’s technical data for the heat pumps.
Garber-Slaght and Peterson [58
] describe a ground-source heat pump system utilizing horizontal ground heat exchangers in a sub-Arctic climate with permafrost present. The permafrost layer at the site was 3 to 7.3 m deep in 2006 and the slinky heat exchangers are installed at a depth of 2.7 m. The GSHP serves a 446 m2
office space. Monthly heating COPs were measured for three years; an approximate average value of 3.5 may be inferred, with some decrease from the first year to the third year.
Liu et al. [59
] summarized annual performance measurements of 10 ground-source heat pump systems serving commercial buildings. The ground-sources included both open loop and closed loop borehole systems, as well as a mine water system and a system utilizing municipal wastewater as the source. Most of the systems used distributed water-to-air heat pumps to provide both heating and cooling. Annual heating and cooling system coefficients of performance (SCOP) were determined for eight of the ten buildings. For the other two, difficulties in differentiating between energy used to provide heating and energy used to provide cooling led the authors to define an effective overall SCOP that combined heating and cooling. Annual heating SCOP varied between about 2.5 and 4.3; annual cooling SCOP varied between about 3 and 5.2. The most commonly identified problem was excessive pumping power caused by oversized circulation pumps, suboptimal controls, and excess hydraulic resistance. For the water-to-air heat pump systems, it was not possible to measure the heating and cooling provided, so a model-based approach was utilized. The resulting uncertainty was not characterized in the studies.
Naicker and Rees [60
] present seasonal system performance factors for a GSHP system serving a university building in Leicester, England. Seasonal performance factors are presented for SEPEMO levels C1 and H1 (3.19 and 4.06, respectively). Combined cooling and heating SPF are also defined corresponding to SEPEMO levels H1, H2, and H4 boundary conditions and are referred to as SPF1
, and SPF4
with values of 3.54, 2.97, 2.49, respectively. Hourly performance factors are also plotted against cycle time, demonstrating that cycling losses increase with decreasing cycle time. Daily combined performance factors are about 4 under high load conditions, but show considerable scatter at lower daily energy demands leading to an average SPF1
value of 3.54. SPF2 and SPF4 are affected by the pumping controls that start the circulating pumps three minutes before the compressors. Under low load conditions, with short cycles, the pumping energy consumption can be as high as 30% of the heat pump energy consumption. Several approaches to improving the system performance are identified, including incorporating buffer tanks, use of a smaller capacity “lead heat pump”, and use of variable speed compressors.
Gehlin et al. [61
] provide a preliminary analysis of the work described in this paper—analysis of a GSHP system serving a university building in Stockholm, Sweden.
1.3. Literature Review—Summary
Long-term measured (>1 year) SPF and COP values reported in the literature for 55 GSHP systems used in commercial or multifamily residential buildings are summarized in Table 2
, based on the SEPEMO definitions. In addition, there are three systems for which a combined cooling and heating SPF was given; these appear in the column “HC4”. In cases where values were given for multiple years, they have been averaged. The 55 systems shown in Table 2
are sparsely located on three continents. There is a notable inconsistency in which boundaries are used, making comparisons difficult. In a number of cases, the boundaries used are not clearly defined, again making comparisons difficult. Only about half the cases give the overall (H4 or C4) system SPF. Only three of the studies report any uncertainty analysis; only Hughes [41
] reports the methodology in detail. Yet this is critical to understanding the significance of the results, so a second contribution in this paper is a detailed description of the uncertainty analysis methodology.
Median values of annual SPF for heating are 4.1, 3.6, 3.1, and 2.9 for boundaries H1, H2, H3, and H4, respectively. Median values of SPF for cooling are 5.5, 6.4, and 4.2 for boundaries C1, C2, and C4, respectively. As noted by other authors [21
] covering residential GSHP systems, there is a significant range in SPF values that cannot be explained solely by different equipment or different climatic conditions.
Almost all of the past studies provide only seasonal or annual performance factors. But daily performance factors [59
] or binned average values [51
] have been useful in identifying causes for poor performance, such as heat pump cycling losses, pump controls, and equipment failures.
It is the objective of this paper to provide long-term measured system performance data including a detailed uncertainty analysis from a recently installed centralized GSHP system for a new-built office building in the Swedish capital Stockholm. The analysis includes seasonal performance factors and monthly, daily, and binned average values of coefficients of performance. These analyses show how the various system components and controls affect the performance and describe how some unanticipated consequences of the design may be ameliorated. (In the analysis of the Studenthuset GSHP system, we use the “Seasonal Performance Factor (SPF)” to refer to quantities calculated over a year, i.e., 365 sequential days, and “Coefficient of Performance (COP)” for performance factors calculated over shorter periods of time, such as monthly performance).