HomeFeaturesExamplesTutorialsHelpTheoryVerificationPricingContact
BoreTherm / TutorialsTutorial 08 / 19
Workflow and theory · Ground data & uncertainty

Parametric study

How a parametric study varies inputs of one project, what it holds fixed, how each case is sized or simulated, and how to read the results. The numbers are those of the example project: a small office or school building of about 3000 m² with monthly loads of 200 MWh/year heating, 8 MWh/year hot water and 45 MWh/year cooling, a 4 × 4 field of 150 m boreholes at 6 m, simulated over 25 years. The study sizes 6 cases.

01 / The one idea

The one idea: one project, many copies, one answer each

A parametric study asks “what if?” for inputs of one project. It makes a copy of the project for every case, sets the case’s values in it, and sizes the borehole length (or simulates) as Size borehole length (menu Calculation › Size borehole length) or Calculate (menu Calculation › Calculate) would. Everything that is not varied stays at the project value. The study adds no physics: each case is an ordinary design calculation.

Projectλ = 2.6 W/(m K)B = 6 m, 4 × 4 fieldlimits −3 / +20 °Cloads, design, fluid …B = 5 mB = 7 mλ = 2.0H = 168.5 mrun 1H = 154.3 mrun 2λ = 2.6H = 150.8 mrun 3H = 138.3 mrun 4λ = 3.2H = 137.1 mrun 5H = 125.7 mrun 6Each run reportsH, total length,Tf min / max, Tm,Rb, Rb*, pump power,λ eff, limitingEvery run is a copy of the project with the run’s values. Everything else is held. Darker: longer boreholes.
The example: ground conductivity λ × borehole spacing B, a full grid of 3 × 2 = 6 cases, each sized.

Each run reports the borehole length H, the total length, the lowest and highest mean fluid temperature Tf,min / Tf,max over the period (Tf is the average of the fluid temperatures entering and leaving the boreholes), the undisturbed mean ground temperature Tm, the borehole thermal resistance Rb between fluid and borehole wall, the effective resistance Rb* (Rb plus the heat exchange between the pipes inside the borehole), the pump power, the effective ground conductivity λ eff and, when sized, which limit was limiting.

Values of one parametervk = min + k · step, k = 0 … ⌊(max − min)/step⌋, or a list
Full grid (1–3 parameters)N = n1 · n2 · n3
One at a time (any number)N = 1 + n1 + n2 + …
Size mode, every casesmallest H with Tf,min ≥ Tmin and Tf,max ≤ Tmax

The maximum is part of the range when it lies on the grid (within 10−9). An explicit list replaces min, max and step. The study can vary 29 inputs of this project; boretherm-cli params lists them with their units and bounds.

02 / The workflow

The workflow

Choose the tool by the question, then choose what each case reports.

What do you wantto know?Study: one at a timewhich inputs mattermost? → tornado chartStudy: full gridhow do 2–3 inputs acttogether? → lines, mapMonte Carlowhich length meets thelimits at 95 % confidence?Optimizewhich layout is theshortest or cheapest?Is the borehole lengthstill open?yesnoResult: Sizethe required length H of everyrun: the limits are just metResult: SimulateTf min / max at the currentlength: inside the limits?Monte Carlo has the same two results. The optimizer varies the layout itself; a study varies inputs of one layout.
One at a time and full grid are the two kinds of parametric study. Monte Carlo (menu Tools › Monte Carlo) and Optimize (menu Calculation › Optimize) are separate tools. Size is the default result; Simulate checks a length that is already fixed.
Input pagesthe project valuesof every parameterParametric study…Tools menu: kind, result,parameters, valuesRunall cases in parallel,progress, cancelParametric studyStudies & uncertainty:Chart and Table tabsReportsection ParametricstudySize borehole lengthor Calculate: size orcheck the designInput pagesthe run’s parametervalues (not the sized H)Apply run to projector double-click a run;one undo stepset up and runuse a result
The study is a tool around the design. Apply run to project sets the run’s parameter values on the input pages; it does not set the sized length. Run Size borehole length afterwards.
  • Parametric study…, menu Tools › Parametric study. Kind: full grid (every combination of 1–3 parameters) or one at a time (each parameter alone, the others at their project values). Result: size the borehole length, or simulate at the current length. The table has one row per parameter: project value, values (list), minimum, maximum, step and the number of runs. The dialog shows the total runs and, after a first study, an estimated time; above 2000 runs it warns.
  • Parametric study (Advanced analysis › Studies & uncertainty). Chart: one result against the first parameter, one line per value of the second (or a heat map); with a third parameter, one slice at a time. One at a time: a tornado chart. Table: every run with its values, results, limiting temperature and status; Export CSV gives the same columns as boretherm-cli study --csv.
03 / How the calculation works

How the calculation works

1 Check the studyat least one parameter; full grid ≤ 3; ids known, not twice2 Expand the valuesthe list, or min, min + step, … (max if on the grid)3 Build the casesfull grid: every combination, last parameter fastest;one at a time: the base run + each value alone4 For each case, in parallelcopy the project and set the case’s values;a value outside the parameter’s bounds fails that caseResult?SizeSimulateSize the length (Brent)as Size borehole length: H inH min … H max, g-function cacheSimulate at the current Has Calculate; cases of onegeometry share a g-function5 Metrics of each caseH, total length, Tf min / max, Tm, Rb, Rb*, pump power,λ eff; Size: limiting temperature, converged6 Parametric study pagechart and table; a failed case keeps its messageexample: 3 × 2 values6 cases0 failed6 sizings
The cases run in parallel. The result does not depend on the number of threads. Green: the example’s numbers.

Size runs the full sizing of Size borehole length in every case, with the project’s sizing options. Simulate runs the simulation of Calculate at the current length. Most inputs (loads, SPF, limits, fluid, pipes, grout, flow) do not change the g-function, the dimensionless temperature response of the field to a constant heat rate, so cases of one geometry share one g-function instead of computing it again. For homogeneous ground a conductivity sweep shares one too, rescaled through the time scale ts = H²/(9α) (α = λ/C, the thermal diffusivity); the difference is a few mK. The studies of the program use the analytic g-function.

What is varied and what is held

Each parameter has a setter. Most set one number; some do more:

ParameterWhat the setter changes
ground.conductivityλ of the ground; the volumetric heat capacity C stays, so the diffusivity changes too. Tm follows Ts + qgeo · H / (2 λ), with Ts the surface temperature and qgeo the geothermal heat flux.
layout.spacingB1 = B2 = B; the field is regenerated, each borehole keeps H, D and rb. Circle: B is the chord. Custom: scaled about the centroid.
borehole.Hthe mean length; equal lengths all get it, unequal ones are scaled in proportion.
loads.heatingScale, coolingScale, dhwScalemultiply that energy (monthly MWh, hourly kW); project value 1.
loads.peakScalemultiplies every monthly peak power (monthly loads only).
spf.*, limits.*, settings.yearsset the value (SPF: seasonal performance factor of the heat pump); years are rounded to whole years.
all others (ground, design, fluid, flow, …)set the one value; nothing else changes.

A value outside the parameter’s sanity bounds (for λ 0.1–10 W/(m K)) makes that case fail with a message; the other cases complete. The project value itself leaves the copy unchanged bit for bit, so the base run of a one-at-a-time study is the project.

04 / Reading the results

Reading the results: two modes, one physics

The same grid, once simulated at the current 150 m and once sized:

Simulate: Tf,min at H = 150 m1.62.02.42.83.23.6−7−6−5−4−3−2−10B = 5 mB = 7 mground conductivity λ [W/(m K)]Tf,min [°C]Size: required length H1.62.02.42.83.23.6110130150170190B = 5 mB = 7 mground conductivity λ [W/(m K)]H [m]limit −3 °Ccurrent 150 mLines: λ from 1.6 to 3.6 in steps of 0.2. Open circles: the six runs of the example.
Left: at a fixed length a poorer ground gives a colder fluid. Right: sized, it gives longer boreholes. Where a left line crosses the limit, the right line crosses the current length.
The length does not scale with 1/λ. From λ = 2.0 to 3.2 W/(m K) (+60 %) the required length at 5 m falls by only 19 %. Part of the resistance between fluid and ground hardly depends on the ground (Rb* = 0.121 m K/W here). And with a geothermal heat flux, a better conductor has a cooler ground: Tm = Ts + qgeo H /(2λ) is 11.53 °C at λ = 2.0 but 10.29 °C at 3.2. In a heating-dominated design that works against the better conductivity.

A one-at-a-time study ranks inputs by their swing. The tornado below varies six inputs of this project, each over a range an engineer might check, with the others at their project values (Size mode):

low value of the rangehigh value130140150160170Ground conductivity2.0 … 3.2 W/(m K)29.9 mSurface temperature8 … 10 °C20.8 mHeating energy0.9 … 1.1 ×20.7 mBorehole spacing5.0 … 7.0 m12.6 mGrout conductivity1.2 … 2.0 W/(m K)11.9 mFlow per borehole0.35 … 0.70 l/s3.7 mbase run: H = 143.7 m (the project)required borehole length H [m]
Bars: the change of H when one input goes to the low (blue) or high (orange) end of its range; the number is the swing. Sorted by swing, as the program sorts them.
A tornado ranks the ranges you enter, not the inputs. Here ±1 K of surface temperature (20.8 m) moves H as much as ±10 % of the heating energy (20.7 m), and the ground conductivity range 2.0 … 3.2 W/(m K) comes first with 29.9 m. A narrower conductivity range (after a thermal response test) would move it down. Give every input a range of the same likelihood, for example its uncertainty, and use Monte Carlo when you need a probability.
05 / How each result is derived

How each result is derived

The columns of the table, with run 1 of the example (λ = 2.0 W/(m K), B = 5 m, Size):

ResultHow BoreTherm derives itRun 1
HSize: the smallest length in H min … H max that keeps Tf within both limits (as Size borehole length). Simulate: the current mean length.168.47 m, limiting: Tf,min
Total lengthN · H (Σ Hi)16 × 168.47 = 2695.5 m
Tf,min / Tf,maxextremes of the mean fluid temperature over the period (or the last year)−3.00 °C / 16.69 °C
TmTs + qgeo · H / (2 λ)9 + 0.06 × 168.47 / (2 × 2.0) = 11.53 °C
Rbmultipole method: pipes, grout, shank spacing and the ground λ0.1155 m K/W
Rb*Rb + (H / ṁcp)² / (3 Ra), Ra the internal resistance between the pipes: grows with H0.1208 m K/W
Pump powerΔP · V̇ / ηpump (pressure drop × volume flow / pump efficiency): grows with H3407 W
λ effconductivity over the active length (layered ground: the effective value)2.0 W/(m K)
StatusOK; not converged (Size: limits not met within the range, H = H max); failed (message)OK

Rb changes a little with λ because the multipole method includes the ground around the borehole; Rb* and the pump power follow the sized length.

06 / The example in numbers

The example in numbers

λ [W/(m K)]H at B = 5 m (total)H at B = 7 m (total)5 → 7 m savesTm at 5 / 7 m [°C]
2.0168.5 m (2695 m)154.3 m (2469 m)14.2 m (8.4 %)11.53 / 11.31
2.6150.8 m (2414 m)138.3 m (2212 m)12.6 m (8.3 %)10.74 / 10.60
3.2137.1 m (2193 m)125.7 m (2012 m)11.4 m (8.3 %)10.29 / 10.18
  • All 6 cases converged and none failed. Every case is sized to Tf,min = −3 °C: heating decides; Tf,max stays below 17.4 °C.
  • Wider spacing saves 8.3–8.4 % of the length at every conductivity: the lines of the chart are nearly parallel. The two inputs hardly interact, so a one-at-a-time study would have told almost the same story with fewer runs.
  • The project itself (λ = 2.6, B = 6 m) is not one of the six cases. The mean of the 5 m and 7 m results is 144.6 m; sized directly it needs 143.7 m (the base run of the one-at-a-time study). The response is curved: add the project value to a range when you need it.
  • Apply run to project on run 6 sets λ = 3.2 W/(m K) and B = 7 m. The length stays 150 m until Size borehole length sizes it to 125.7 m.
07 / Limits to keep in mind

Limits to keep in mind

  • A study answers “what if” for the values you enter. It gives no probability; for a length at a confidence level use Monte Carlo.
  • One at a time misses interactions. A full grid shows them but takes at most three parameters, and the runs multiply: three parameters with ten values each are 1000 sizings.
  • In Size mode a case that cannot meet the limits within H min … H max is marked not converged; its H is the longest length of the range, not a required length. Widen the sizing range.
  • A varied spacing regenerates a generated layout; it does not move a custom field’s boreholes one by one (it scales the field about its centroid).
  • The program’s studies use the analytic g-function. With the FEM (finite-element) engine, check the chosen case with Size borehole length again.
  • Each case is as good as the design model: monthly or hourly loads, the borehole model and the ground model of the project apply unchanged.

Numbers calculated with BoreTherm. The one-at-a-time ranges are an illustration, not recommendations.