Optimal GSHP–ASHP hybridisation for large-scale tertiary buildings with imbalanced loads

2.1 Hybrid architectures and performance

The dominant hybridisation scheme couples a GSHP with an ASHP as auxiliary heat source. The switching logic between sources varies:

• Temperature-based switching: Grossi et al. (2018) introduced a fixed outdoor air switching temperature T_switch: when the outdoor air temperature is below T_switch, the DSHP operates in ground-source mode; above it, in air-source mode. This monovalent alternating strategy is the simplest to implement and constitutes the baseline for the present study. Pardo et al. (2010) used a similar priority-based approach with GSHP as primary source (COP 4.5 vs 2.5 for ASHP).

• Seasonal or calendar-based switching: Li et al. (2024) used ASHP alone in early/late season, GSHP alone in mid-winter, and a 2/3–1/3 split in the coldest months. You et al. (2015) operated an aerothermal heat compensator (AHC) to transfer heat from air to ground during off-season.

• COP-based real-time selection: Emmi et al. (2017) implemented a multi-source controller selecting at each timestep the source with the highest COP (air, ground, or PVT).

• Differential signal: Zheng et al. (2022) used a ΔT_soil–air > 1 K criterion to switch between GSHP and ASHP.

Several studies integrate photovoltaic-thermal (PVT) collectors to provide a triple benefit: electricity generation, thermal recharge of the ground, and PV cell cooling (Emmi et al., 2017; F. Wang et al., 2024; Yue et al., 2025). X. Wang et al. (2010) demonstrated seasonal thermal storage using solar collectors injecting heat into the ground from spring to autumn.

Table 2 summarises the main performance indicators. System COPs span a wide range, driven by climate, building type, and system complexity. The high values reported in solar-assisted or PV-coupled studies only count grid-supplied electricity in the denominator and are not directly comparable to a thermodynamic COP. The results provide a useful comparative basis but are difficult to synthesise given the diversity of metrics and cost assumptions: drilling costs range from EUR 10/m to EUR 32/m in Chinese studies to EUR 50/m in Grossi et al. (2018) (Italy) and EUR 80/m to EUR 120/m in France.

Ground temperature drift ranges from mild (about –4.5 K in 15 years at Bologna and stabilising thereafter (Grossi et al., 2018)) to catastrophic (–12.7 K in 10 years at Harbin (You et al., 2014)), with system shutdown reported by both You et al. (2015) (after 7 years) and F. Wang et al. (2024) (after 11 years) in the absence of mitigation. Most articles analyse the ground temperature decline, but the economic necessity of controlling it is rarely studied directly, except in extreme cold climates.

2.2 Identified gaps

Three gaps emerge from this review.

First, none of the ten studies formulates the borefield optimisation problem in terms of fluid temperature operating limits (–3°C to 40°C in France). In practice, this constraint is the binding criterion for borefield sizing: it determines the minimum number and spacing of boreholes required for a given extraction profile.

Second, the three European studies (Emmi et al., 2017; Grossi et al., 2018; Pardo et al., 2010) emphasise COP over costs. They report isolated economic indicators (a payback time in Pardo et al. (2010), a normalised borehole reduction in Grossi et al. (2018)), but none performs a detailed economic optimisation, and none treats borefield size reduction within a systematic framework. Paradoxically, it is in Europe—where drilling costs are highest—that this perspective is most critical. The Chinese studies (F. Wang et al., 2024; Yue et al., 2025; Zheng et al., 2022) do present formalised economic optimisations, but with cost assumptions that are not transferable to Europe.

Third, no existing study covers large-scale tertiary installations (500 kW to 2,000 kW) under French climatic conditions.

This study addresses these three gaps through a parametric LCOE analysis with French cost assumptions, explicit fluid temperature constraints, and building capacities of 500 kW to 2,000 kW.

3.1 Building models and climatic data

The case studies are well-insulated office buildings (new construction, envelope U-value 0.7 W/(m² · K)) modelled with the TRNSYS Type 56 multizone component. Three design heating capacities are considered: 500 kW, 1,000 kW, and 2,000 kW—the latter corresponding to the French regulatory threshold for géothermie de minime importance. Two representative French climates are selected via Typical Meteorological Year (TMY) data: Montpellier (Mediterranean, mild winters) and Langres (semi-continental, harsh winters). For each climate and building size, three cooling-to-heating demand ratios are simulated: 0%, 20%, and 60% (the latter excluded for Langres). The heating setpoint is 19°C; the cooling setpoint is adjusted to achieve the target ratio. The envelope assumes 50% glazing, 2 vol/h air renewal during occupancy, and standard internal gains.

3.2 TRNSYS simulation model

The system is implemented in TRNSYS 17 (Klein et al., 2010). The building uses Type 56 (multizone) with thermal demand computed from setpoint temperatures and weather data.

Ground heat exchanger Vertical BHEs are modelled with Type 557b (duct ground heat storage) (Hellström, 1991). Parameters include borehole depth (200 m reference), diameter, grout conductivity, U-tube geometry, soil thermal conductivity (2 W/(m · K)), volumetric heat capacity (2.16 MJ/(m³ · K)), and undisturbed ground temperature (mean annual air temperature from TMY). These soil thermal properties are reasonably homogeneous across metropolitan France, so no sensitivity analysis was performed on them. The hexagonal borehole layout has been verified against analytical g-functions from open-source Python library pygfunction (Cimmino, 2018), agreeing within 0.5 K on brine outlet temperature over 25 years.

Heat pumps The GSHP and ASHP are modelled with TRNSYS Type 927 and Type 941, normalised performance maps. The nominal heating COP is 5 at 10°C/35°C for the GSHP, and 4 at 7°C/35°C for the ASHP. Each unit operates on/off; capacity modulation is achieved by varying the number of units active at each timestep.

3.3 Hybridisation models

Two hybridisation control mechanisms are used in this work, both aimed at containing the brine temperature swing but acting on different timescales.

Alternating monovalent (T-switch), used in Sections 4 and 5. Inspired by Grossi et al. (2018): only one heat source is active at a time. The GSHP runs alone below T_switch, the ASHP runs alone above, and the GSHP also handles all cooling (dedicated cooling emitters are assumed installed). The GSHP is sized for the full peak heating load. This caps the annual ground extraction implicitly, which is well suited to conventionally sized or moderately reduced borefields, but ties borehole count to peak power rather than to annual energy.

Brine-temperature hysteresis controller (base-load), used in Section 6. Both heat pumps can operate, with the GSHP in base load and the ASHP in backup. The GSHP shuts off when the brine reaches the regulatory limits (–3/+40°C) and restarts after a hysteresis recovery. This controller acts only when the brine temperature actually leaves the allowed range, so it does not preventively limit the long-term ground temperature drift; it does, however, allow the GSHP to be sized well below peak load and the borefield to be deliberately undersized, in order to maximise the annual energy yield per metre of borehole and therefore the economic amortisation of the drilling investment.

3.4 Reference borefield sizing

The reference borefield for each building configuration is sized as follows:

• Borehole depth: 200 m;

• Borehole spacing: 10 m (5% of depth);

• Number of boreholes: minimum number ensuring that the brine temperature remains within –3°C to 40°C over a 25-year simulation;

• Hexagonal layout, confirmed by comparison with pygfunction as described above.

This reference case represents a conventionally sized geothermal installation: it is the minimum borefield that satisfies the regulatory fluid temperature limits over the 25-year horizon at a standard 10 m spacing, and it is the benchmark against which the hybrid configurations are compared.

3.5 Optimisation problem formulation

The fluid temperature constraint T_fluid ∈ [–3°C, 40°C] is the central feasibility condition. The brine temperature is governed by two interacting factors:

1. Short-term variation (dominant): swings of 10 K or more over a few hours, combining the temperature drop at the borehole wall under recent extraction intensity (strongest under sustained high-power extraction) with the instantaneous brine–wall ΔT required to drive the heat exchange;

2. Inter-annual ground temperature drift: a slow drift of the mean borefield temperature driven by the cumulative heating/cooling imbalance and borefield geometry.

Even without inter-annual drift, short-term variation alone imposes a minimum total borehole length; additional boreholes provide margin against the drift, spread over enough area to limit thermal interference.

Minimising the LCOE under the fluid-temperature constraint implicitly maximises the energy extracted per metre of borehole, since the drilling cost is amortised over the GSHP-delivered energy. This metric—the borehole energy yield—is used as an explicit indicator in Section 6 to show how the base-load model raises it well above conventional geothermal sizing. Two horizons are used: simulations run over 25 years (the French Fonds Chaleur (ADEME) eligibility horizon, on which fluid-temperature feasibility is verified), while the LCOE horizon is 10 years (typical investment-decision timeframe for tertiary owners). Each subsequent result indicates which horizon applies.

3.6 Parametric study design

For each building configuration and surface availability ratio (50%, 70%, 100% of the reference borefield surface), the sensitivity analysis explores two degrees of freedom:

• Borehole count ratio: 70%, 80%, 90%, or 100% of the reference number of boreholes;

• Switching temperature T_switch: six values from a climate-specific list (lower values for Langres than Montpellier).

The borehole spacing is derived from the number of boreholes and the available surface area, with a minimum spacing of 7.5 m. For each combination, a 25-year hourly TRNSYS simulation is run. The retained hybrid configuration for each (building, surface ratio) pair is the one achieving the lowest LCOE while satisfying the fluid temperature constraint. Approximately 1,000 simulations were executed in total.

3.7 Economic assumptions

The economic parameters are summarised in Table 3. The drilling cost of EUR 120/m is the fully equipped borehole cost (drilling, high-density polyethylene (HDPE) double-U probe, bentonite-cement grouting, head connection), consistent with French market data for deep boreholes (ADEME & BRGM, 2021). Heat-pump costs reflect large-scale tertiary installations and may be conservative for MW-scale systems. The electricity tariff includes taxes and network charges for tertiary consumers.

The LCOE combines capital expenditure (CAPEX) and operating expenditure (OPEX):

where H = 10 years is the LCOE horizon, r = 4 % is the WACC, Q_t is the annual thermal energy delivered, and OPEX_t includes electricity consumption and maintenance.

3.8 Simulation framework

A custom Python framework manages the study: TRNSYS deck generation, parallel execution (~5 min per 25-year simulation), and KPI extraction (COP, geothermal coverage, LCOE, fluid temperature extremes). Approximately 1,000 simulations were executed in total.

4.1 Simulation setup

Six values of the switching temperature are tested: T_switch ∈ {4,6,7,8,10} °C plus the pure-geothermal reference (T_switch = ∞); each is run as a 25-year hourly simulation. Two GSHP performance maps bracket the role of evaporator sensitivity:

• Default map: the standard TRNSYS map (rated COP = 5 at 10°C/35°C), which yields COP = 3.0 at the 0°C/50°C regime—a typical value for a brine/water heat pump and our realistic reference scenario;

• Steep-gradient map: same rated COP, but with a Carnot efficiency that drops more rapidly with the source/load spread, reaching COP = 1.7 at the 0°C/50°C regime. This map is not representative of any commercial heat pump; it serves as a sensitivity test for how steep the gradient would need to be for ground cooling to materially penalise GSHP operation.

4.2 Effect of switching temperature on system COP

Figure 1 shows, for each performance map, the annual GSHP COP, ASHP COP, and overall system COP as a function of the switching temperature. The dashed curves provide additional detail by plotting the GSHP COP per outdoor air temperature bin for each T_switch, illustrating how ground cooling affects performance at each temperature level.

Figure 1

With the default map (Figure 1a), the system COP increases monotonically with T_switch, from 3.15 at T_switch = 4°C (44% geothermal coverage) to 4.00 for pure geothermal. No optimum is observed: peak performance is reached without any hybridisation. The GSHP annual COP stays in 3.67–4.00, while the ASHP annual COP rises from 2.83 (T_switch = 4°C) to 3.71 (T_switch = 10°C). With a realistic GSHP map, ground cooling does not erode the GSHP COP enough to offset the GSHP’s inherent advantage over the ASHP.

With the steep-gradient map (Figure 1b), the GSHP annual COP now decreases with T_switch (3.10 at 4°C to 2.93 for pure geothermal), as ground cooling becomes much more penalising. A mild optimum appears around T_switch = 6–7°C (COP ≈ 3.05), a +4% gain over pure geothermal—the mechanism reported by Grossi et al. (2018).

Recovering such an optimum thus requires a steeper COP–evaporator gradient than commercial brine/water maps exhibit, and even then, the gain is marginal. Under realistic performance assumptions, alternating monovalent hybridisation does not optimise the system COP: the case for hybridisation must therefore be made on the other levers explored in the remainder of the paper, namely the reduction of the borefield surface area (Section 5) and the base-load hybrid operating model (Section 6).

4.3 Ground temperature evolution

Figure 2 shows the monthly-averaged borefield temperature over 25 years for each T_switch (default map). Pure-geothermal operation drifts by about 10 K (14 to 4°C), against 6 K at T_switch = 4°C. Despite this large difference, the impact on the system COP is modest: the brine inlet temperature during extraction is dominated by the instantaneous extraction rate rather than the average ground temperature, and the peak-extraction GSHP COP penalty is similar across switching temperatures.

Figure 2

The default TRNSYS normalised performance maps are retained for all subsequent analyses. The remaining sections explore the two other levers offered by hybridisation: the reduction of the borefield surface area (Section 5) and the base-load operating model (Section 6).

5.1 Methodology

For each building, the reference geothermal installation (Section 3.4) is deformed by reducing the available land area to 70% or 50%. Within each surface constraint, we sweep the borehole count ratio (70–100% of reference) and T_switch, retaining the lowest-LCOE configuration that satisfies T_fluid ∈ [–3°C, 40°C].

For Montpellier with 60% cooling, the binding constraint is the maximum fluid temperature, not the minimum. Since alternating monovalent hybridisation does not reduce cooling-season injection, those cases are excluded below (surface reduction remains feasible if the borehole count is preserved). In this model the GSHP covers the full peak heating load alone, which limits the borefield load factor; Section 6 explores an alternative logic operating the GSHP as the base-load source.

5.2 Results

Tables 4 and 5 consolidate the optimal hybrid configurations for each building and surface ratio. For each case, the table reports the borehole count ratio (n/n_ref), borehole spacing, switching temperature, system COP, and LCOE relative to the reference geothermal installation.

5.3 Analysis

Surface reduction with moderate COP trade-off. The central result is that the borefield surface can be reduced to 50% of the reference while maintaining a system COP above 3.1 in all cases. At 70% surface, the COP reduction is typically 0.1–0.7 units. A lower COP for a given building reflects a more hybridised configuration (lower T_switch, higher ASHP share) that trades modest COP for reduced drilling investment.

Climate effect on switching temperature. The optimal T_switch is systematically lower for Langres (4°C to 7°C) than for Montpellier (8°C to 11°C). This reflects the heating demand distribution by outdoor temperature: in Langres, a larger share of demand falls at very cold temperatures, so a low T_switch already captures most of the heating load on the geothermal side; a higher value would extend GSHP operation into mild hours that contribute little to annual demand. At 50% surface, T_switch converges to the lower end of each range to limit extraction intensity. Spacing drops to ~7.9 m at 50% surface, still above the practical minimum of 7.5 m.

LCOE. At full surface availability, hybrid configurations achieve LCOE reductions of 5–17% relative to the reference, with the best results at Langres (83.5% for 2,000 kW, 0% HL). At 50% surface, the LCOE is within 0–4% of the reference: hybridisation transforms an infeasible scenario into a viable one at comparable cost. LCOE gains are larger for Langres, where the greater heating demand amplifies the benefit of borehole count reduction.

6.1 Controller binding regimes

Whether the controller binds on the short-term or long-term timescale depends on the configuration. In mild climates with balanced loads the borefield recovers between peaks, so the binding constraint is the short-term brine temperature swing. In cold climates with heating-only loads, progressive cooling shifts more of the operation to the ASHP over the years, so the binding constraint becomes the inter-annual ground temperature drift.

6.2 Effect of GSHP extraction intensity

For the most favourable configuration (Montpellier, 1,000 kW, 60% cooling), Figure 3 shows the effect of the GSHP rated extraction intensity (50 W/m to 100 W/m) on the 10-year average yield and geothermal coverage, for a fixed borefield of 20 boreholes. Yield rises from 240 kWh/(m · yr) at 50 W/m to 286 kWh/(m · yr) at 80 W/m, where a clear plateau is reached. Beyond 80 W/m the controller increasingly cuts GSHP operation due to fluid-temperature violations. At the plateau, the geothermal system covers 68% of heating and 60% of cooling demand.

Figure 3

6.3 Ground cooling as the limiting factor

In the most unfavourable configuration (Langres, 1,000 kW, heating only), inter-annual ground cooling constrains the intensive extraction strategy. Figure 4 shows the annual evolution over 10 years. Year-1 yield is 208 kWh/(m · yr) with 55% geothermal coverage; the mean ground temperature drops from 11.8°C to 7.1°C (ΔT = –4.7 K), and by year 10 yield falls to 57 kWh/(m · yr) (16% coverage). The controller is self-protecting: as the ground cools, the GSHP is cut off more often, converging to a regime where geothermal only contributes marginally. Combining the controller with a switching temperature (prioritising GSHP when air is cold) was tested but added only +3% system COP (2.67 to 2.75); this modest gain likely reflects the default TRNSYS ASHP map underestimating defrost losses at low air temperatures.

Figure 4

6.4 Consolidated results across building configurations

Table 6 presents the 10-year averaged performance for all five building configurations; each tested with two borehole spacings (8 m and 12 m). All cases use 20 boreholes at 200 m. The GSHP is rated at 80 W/m (320 kW) for Montpellier (Section 6.2 plateau), and at 200 kW for Langres where ground cooling precludes higher intensities. Each configuration is compared against the GSHP-only reference of Section 5.

Three findings emerge from these results.

Substantial LCOE reductions vs. the GSHP-only reference. The base-load model reduces the LCOE by 37–52% relative to the reference (e.g. 352 → EUR 208/MWh for Montpellier 0% HL, 364 → EUR 176/MWh for Langres 0% HL). Most of these savings come from the borefield being sized at 20% of the reference, with the bulk of the load served by the cheaper ASHP. Unlike a standalone ASHP, however, the hybrid keeps geothermal in the energy mix—with the associated non-economic benefits (load diversification, lower peak grid demand, building real-estate value).

Cooling loads unlock geothermal economic value. With at least 20% cooling demand, the yield reaches 188 kWh/(m · yr) to 219 kWh/(m · yr), and up to 289 kWh/(m · yr) at 60% cooling (Montpellier)—three to four times the GSHP-only reference yield (42 kWh/(m · yr) to 92 kWh/(m · yr)). The cooling season partially recharges the ground. Heating-only configurations achieve much lower yields (93 kWh/(m · yr) to 149 kWh/(m · yr)), confirming that thermal imbalance is the fundamental barrier in heating-dominated climates.

Borehole spacing matters under intensive extraction. 12 m spacing consistently raises the yield by 15–40% over 8 m, as intensive extraction amplifies thermal interference. Since the borefield is sized at only 20% of the reference, the absolute footprint remains compact even at 12 m spacing.

Where ground cooling limits the base-load approach, active recharge strategies (solar thermal or ASHP-driven summer recharge) could stabilise the borefield temperature and extend the applicability of the base-load model to heating-dominated climates—a direction discussed further in Section 7.3.

7.1 Comparison with the literature

The borefield surface reductions of 30–50% achieved with the alternating monovalent model, with COP degradation limited to 0.1–0.7 units, are consistent with Grossi et al. (2018), who reported similar BHE length reductions for a residential DSHP in Bologna. The present study extends this finding to installations 50–200 × larger (500 kW to 2,000 kW vs. 8.6 kW), confirming the scalability of the hybrid approach. The LCOE improvements of 5–17% are comparable to the 8.1% ALCC reduction reported by Zheng et al. (2022), but achieved under European cost assumptions that are significantly less favourable to geothermal systems.

We do not, however, reproduce the hybridisation-driven gain on system performance reported by Grossi et al. (2018): under realistic performance assumptions the system COP is monotonic in T_switch and maximised without hybridisation; a mild optimum only reappears with our steep-gradient sensitivity map.

The base-load model shifts the focus from borefield size to energy yield. The borehole yields of 188 kWh/(m · yr) to 289 kWh/(m · yr) obtained under balanced loads far exceed conventional geothermal sizing, confirming that the hybridisation logic itself—not just the borefield geometry—is a key economic lever. The rapid degradation in heating-only Langres (208 to 57 kWh/(m · yr) in 10 years) shows that ground cooling remains the fundamental limitation under imbalanced loads.

A distinctive feature of this study is the explicit treatment of the fluid temperature constraint as the binding feasibility condition: no prior study formulates borefield optimisation in these terms (Section 2.2), yet these limits drive both the reference sizing and the base-load controller thresholds.

7.2 Practical implications

For tertiary developers and operators, three takeaways emerge. Land-constrained sites: the alternating monovalent model enables geothermal deployment with 50% of the reference borefield surface, COP above 3.1 and LCOE within 0–4% of the unconstrained reference. Economic optimisation: the base-load model profitably uses geothermal as soon as cooling demand reaches 20%, and clearly benefits the borefield at 60% cooling, with 12 m spacing further improving extraction by 15–40%. Climate sensitivity: cold heating-only climates benefit most from surface reduction (up to –17% LCOE) but are least favourable for the base-load approach; balanced loads and mild climates offer the best conditions for both strategies.

7.3 Limitations and future work

The main limitations are: (i) heat-pump performance maps are default TRNSYS data, possibly under-representing COP degradation in the critical –3°C to 5°C evaporator range; (ii) the analysis is purely simulation-based, as in 8 of 10 studies reviewed; (iii) the 10-year LCOE horizon understates the value of the borefield as a >50-year asset; (iv) no sensitivity on cost parameters has been performed—the LCOE is dominated by CAPEX (79–87%), so drilling cost is the most influential parameter.

The main future-work direction is active ground recharge (solar thermal or ASHP-driven summer injection) to stabilise the borefield temperature in heating-dominated climates and establish the conditions under which hybrid systems consistently outperform standalone ASHP solutions.