HomeFeaturesExamplesTutorialsHelpTheoryVerificationPricingContact
Theory manual

18. The dynamic borehole model

The usual design model treats the inside of a borehole as one thermal resistance. The fluid, the pipes and the grout store no heat, so a change of the load reaches the fluid temperature in full at once. That is accurate when the load changes slowly compared with the thermal time constant of the borehole, which is a few hours. With hourly loads, short peaks, a pump that switches on and off, or an inlet temperature that is prescribed over time (a thermal response test, a heat pump model, a thermal store), the heat stored in the borehole and the time the fluid needs to pass through it matter. The dynamic borehole model resolves the inside of the borehole for these cases. The ground outside the borehole wall stays the analytic g-function response of the field.

This chapter describes the three borehole models, the network of the dynamic model, its coupling to the ground, how hourly loads and sizing work with it, and what it changes against the steady model in a worked example.

18.1 Three borehole models

In the steady model the mean fluid temperature follows from the borehole wall temperature Tb and the heat rate per metre q (positive into the ground) through the effective borehole resistance Rb*, at every time:

Tf=Tb+qRb*
(18.1)

Rb* contains the borehole resistance Rb between the fluid and the wall and the thermal short-circuit between the legs through the internal resistance Ra (Hellström 1991). Two further models add the heat capacity of the borehole interior; the model is chosen on the Design limits & simulation page.

Table 27. The borehole models.
ModelInside the boreholeLoadsTypical use
Steadyone resistance Rb*, no storagemonthly, hourly, time seriesdesign of heating and cooling fields
Short-time capacitythe steady model plus one combined short-time response of fluid, pipes and groutmonthly, hourly, time serieshourly designs with short peaks; energy piles
Dynamicfluid cells that move with the flow, grout nodes on depth levels, transit time, flow schedulehourly or time seriesverification of an hourly design, pump on/off, prescribed inlet temperature, thermal stores

18.1.1 The short-time capacity model

The short-time model keeps everything the steady model computes and adds a correction for the heat capacity of the borehole. The pipes of a U-tube borehole are replaced by one concentric equivalent pipe that keeps the fluid and pipe-wall volumes and the film and wall resistances of the legs in parallel (Xu & Spitler 2006); its grout annulus gets the conductivity that makes the total resistance equal to Rb*, so the long-time end of the model is the steady model. A one-dimensional radial finite-volume model of fluid, pipe wall, grout and ground is run for a unit step load twice, with and without the capacities of fluid, pipes and grout. The difference is the capacity correction

δg(t)=2πλ[Tf(t; capacities)−Tf(t; no capacities)]w(t)
(18.2)

with a blend weight w(t) that is 1 up to a quarter of the blend time tblend (100 h for boreholes) and falls smoothly in lnt to 0 at tblend. The combined response of a unit step is

gc(t)=g(t)+2πλRb*+δg(t),Tf(t)−Tm=qgc(t)2πλ
(18.3)

where g is the g-function of the field (Eskilson 1987), λ the ground conductivity and Tm the undisturbed ground temperature. In the hourly simulation the correction enters as a short convolution of the load steps with gc−g−2πλRb* over less than 100 hours, so the steady part of the calculation is untouched. At 100 h the capacity effect is below 0.3 % of the response, and from tblend on the model is the steady one. With all capacities set to zero the correction is exactly zero.

18.2 The cross-section network on depth levels

The dynamic model cuts the active length H into K=max(1,round(H/Δz)) levels of equal height (default Δz = 2 m). Every level carries the thermal resistance–capacity model (TRCM) of Bauer et al. (2011): every leg i of a U-tube has a fluid node fi and a grout node gi; the fluid exchanges heat with its grout node, the grout node with the borehole wall, and the grout nodes with each other:

Rfg,i=Rp,i+xiRg,i,Rgb,i=(1−xi)Rg,i
(18.4)

with Rp,i the film and wall resistance of the pipe, Rg,i the grout resistance of the leg and xi the position of the grout node. Bauer et al. derive Rg and the grout–grout resistance from two-pipe line-source formulas. Here they are derived from the multipole resistance matrix 𝐑 of the cross-section, so that the steady network reproduces it exactly. With 𝐊=𝐑−1 (the heat from the legs is 𝐊(𝐓f−Tb1)), 𝐀=diag(1/Rfg,i), 𝐁=diag(1/Rgb,i) and the Laplacian 𝐋 of the grout–grout conductances, eliminating the grout nodes gives 𝐊=𝐀−𝐀(𝐀+𝐁+𝐋)−1𝐀, hence

𝐌=𝐀(𝐀−𝐊)−1𝐀,𝐁=diag(𝐌1)−𝐀,𝐋=𝐌−𝐀−𝐁
(18.5)

and Rg,i=1/∑jKij−Rp,i. For a single U-tube this is Bauer's Rg=2Rb−Rp. The grout node sits at Bauer's position, the radius that halves the grout area of the equivalent annulus (db the borehole diameter, do the outer pipe diameter, n = 2, 4, 6 legs):

x=ln(db2+ndo2/(2ndo))ln(db/(ndo))
(18.6)

The network must respect the second law: 𝐀−𝐊 and 𝐁+𝐋 must be positive definite and every Rgb>0. Otherwise x is reduced to 2x/3 (Diersch et al. 2011) until they are. A grout–grout conductance may be negative (legs close to the wall) as long as 𝐁+𝐋 stays positive definite. Capacities per metre: the fluid ρcpA of every channel and the grout CgroutAgrout/n of every grout node; the pipe walls carry none, as in Bauer et al. A coaxial borehole has the chain inner fluid – annulus fluid – grout node – wall, with an inner-pipe-wall node without capacity that splits Ra in half.

Borehole sections along the depth (other diameter, grout or insulation) act per level: the network of every construction is averaged onto the levels it overlaps, weighted by the overlap length, and the heat capacity of a pipe sleeve or casing is added to the grout nodes it sits with. The film resistances depend on the flow; the level parameters are computed at relative flows 2k/16 of the nominal flow and cached, and with the pump off the film resistance of a laminar flow of 10−3 of the nominal flow is used.

18.3 Fluid transport: plug flow without numerical diffusion

Every flow path (a U-tube: down one leg, through the U-bend and up the other) carries fluid cells of equal volume ΔV that move with the fluid: cell j occupies the path volume [φ+(j−1)ΔV, φ+jΔV] with 0≤φ<ΔV. Advecting a volume moves only φ; when φ passes ΔV the outlet cell leaves the path (its heat is the outflow) and a new inlet cell enters. The inflow is mixed only into the partly filled inlet cell. The transport is therefore exact plug flow: a temperature front keeps its shape and arrives after the transit time ttransit=Vpath/V˙, with no numerical diffusion beyond one cell. A cell exchanges heat with the nodes of the levels it overlaps, in proportion to the overlap.

A sub-step of length δt=min(h, ttransit/10), with h the control step (default 5 min), advects half the volume, exchanges heat for δt and advects the other half (Strang splitting). With the pump off δt=min(h,300 s). The run time is therefore set by the transit time and the length of the period, not by the control step.

18.4 Heat exchange between fluid and grout

A fluid cell of capacity C coupled by conductances Gl to node temperatures Tl is advanced from T to T′ by a time-centred, exponentially fitted step:

CT′−Tδt=∑lG~l[Tl−T+T′2],G~=G2tanh(ξ/2)ξ,ξ=δt∑lGlC
(18.7)

For constant node temperatures this is exactly T′=Tn+(T−Tn)e−ξ for every ξ, with Tn the conductance-weighted node mean, so the step never oscillates. With the Strang splitting the quasi-steady axial profile is that of the continuous equations (Hellström's exponential leg temperatures) independently of the sub-step. The grout nodes and the wall coupling are integrated with backward Euler. The fluid cells that touch only fixed nodes are eliminated onto those nodes, and the remaining banded system (grout nodes in depth order) is factorised once per sub-step, so the cost per sub-step grows linearly with the number of levels.

18.5 Coupling to the ground

The borehole wall temperature is uniform along the depth and equal in all boreholes, consistent with the g-function for a uniform borehole wall temperature (UBWT), which is used here with the cylindrical-source correction because the wall is resolved. On the control grid h the wall temperature of step n is

Tw(n)=Tm+Hhist(n)+g(h)q¯w(n)2πλ
(18.8)

with q¯w(n) the mean heat flux through the walls of the field over the step, taken from the grout nodes. It differs from the fluid load while the borehole stores or releases heat; once a load has been steady for a few days the two agree and the dynamic model gives the steady result. Hhist is the response to the earlier steps, kept by the load aggregation of Claesson & Javed (2012): the history is held in cells m of growing width wm (in steps) with mean loads q¯m, and every step shifts a fraction 1/wm of each cell's load on to the next, older cell:

q¯m←q¯m+q¯m−1−q¯mwm,Hhist=∑m≥1q¯m[g(τmh)−g(τm−1h)]
(18.9)

where τm is the age (in steps) of the far end of cell m. The cells cover the run plus one more level of cells, and the last cell keeps all the load it receives (no outflow), so every step conserves the energy of the load history exactly; what remains is the smearing error of the aggregation itself, a few 10−3 of the response for a constant load.

With a prescribed load, the field inlet temperature Tin and the wall temperature Tw are the unknowns that couple all boreholes. Every node temperature is carried as an affine form c0+cinTin+cwTw, and the three parts are solved with one factorisation. In every sub-step the fluid balance of the field,

Qδt=ρcp(VinTin−∫TdVout)
(18.10)

is affine in Tin and Tw and gives Tin exactly, without iterating on a lagged outlet. At the end of the step the scalar equation (18.8), with q¯w affine in Tw, gives Tw, and every node is resolved. With a prescribed inlet temperature the load Q follows instead.

Boreholes in series and parallel groups of the connection network are chained: the outflow of a stage, mixed over its boreholes, is the inflow of the next. Boreholes with equal length, depth, radius and flow are simulated once and weighted by their number, so an all-parallel field of equal boreholes costs one borehole.

18.6 Flow control, pump off and fluid properties

18.7 Energy balance and outputs

Over the dynamic periods the program sums the energy delivered by the fluid, the energy through the walls and the change of the heat stored in fluid and grout, and reports the largest relative step error of the balance:

e=|(Ein−Eout)−Ewall−ΔS||Ein|+|Eout|+|Ewall|+|ΔS|+10−9|S|+10−3 J
(18.11)

with the enthalpies and the stored heat S taken relative to 0 °C. The Dynamic model page shows the inlet, outlet, mean fluid and wall temperatures, the wall flux, the load and the flow at the output interval, depth profiles of the down-going and up-going legs and of the grout at chosen times, and the flow direction. The hourly arrays (hour means) and a monthly view (base = the last hour of each 730-hour month, peaks = its extremes) are filled as by the other models, so the Results summary, the Mean fluid temperature page, the Annual page and the limit check work unchanged.

18.8 Hourly loads and sizing with the dynamic model

The dynamic model needs hourly or time-series loads. Hourly loads may be generated from the monthly table: the generator keeps the energy and the peak power of every month and spreads the rest by the outdoor temperature and a use pattern (for an office: office hours on working days). The limits are checked on the hour means of all years (or of the last year).

Size borehole length works with every borehole model: with the dynamic model it runs the dynamic model at every trial length, on the same cached g-functions as the steady sizing. That is accurate but slow. The faster route is to size with the steady model and then run Verify with dynamic model on the Required borehole length page: it copies the project, sets the boreholes to the sized length, switches to the dynamic model with the project's dynamic settings and runs it together with the steady model at the same length. The project itself is not changed. The result shows the hour-mean range against the limits, the range at the output interval (more extreme when the control step is shorter than an hour) and the steady range.

18.9 Example: an office field with hourly loads

A small office with a ground-source heat pump: 200 MWh/year space heating (peak 100 kW), 8 MWh/year domestic hot water and 45 MWh/year cooling on 16 boreholes of 150 m in a 4 × 4 grid, ground conductivity 2.6 W/(m K), over 25 years; the limits are −3 / 20 °C. The monthly loads are turned into hourly loads with the office pattern. Table 28 lists the runs.

Table 28. The office example with the three borehole models.
Loads and runBorehole modelTf,min [°C]Length H [m]
Monthly loads, 150 msteady, monthly−2.41–
Monthly loads, Size borehole lengthsteady, monthly−3.00143.7
Hourly loads, 150 msteady, hourly−3.06–
Hourly loads, Size borehole lengthsteady, hourly−3.00150.7
Verify at 150.7 mshort-time capacity−2.92–
Verify at 150.7 mdynamic−2.89limits met
Hourly loads, Size borehole lengthdynamic−3.00149.5

Hourly loads make the field 7.0 m per borehole longer than the monthly sizing (112 m in total): the office draws its heating peak for a whole working day, day after day, while the monthly method assumes a 6-hour peak on top of the monthly mean. The dynamic model at the hourly length keeps Tf,min 0.11 K above the limit, so the steady length is confirmed with a small reserve. Sizing with the dynamic model gives 149.5 m, 1.2 m shorter (0.8 %), from 10 dynamic runs of 25 years each.

Figure 17 shows why the difference is small. In the first hour of the coldest 12-hour load block of the last year the steady model drops by the full qRb*, to −1.68 °C; the dynamic model reaches only −0.87 °C, because the fluid and the grout give off stored heat first. The coldest hour, however, is the last one of the block, when that storage is spent: there the models differ by only 0.11 K (−3.00 against −2.89 °C). The dynamic model matters most for short peaks; for long load blocks the steady length is close. The depth profiles show the fluid warming on its way down and up, with the grout between fluid and wall; the difference between the legs is the thermal short-circuit that Rb* accounts for in the steady model.

(a) Hour means of the mean fluid temperature over three days of the last year, around the coldest load block, at 150.7 m, for the three borehole models; (b) depth profiles of the dynamic model (down-going leg, up-going leg, grout) in the first and the last hour of the block. The steps in the profiles are the distance the fluid moves in one sub-step.
Figure 17. (a) Hour means of the mean fluid temperature over three days of the last year, around the coldest load block, at 150.7 m, for the three borehole models; (b) depth profiles of the dynamic model (down-going leg, up-going leg, grout) in the first and the last hour of the block. The steps in the profiles are the distance the fluid moves in one sub-step.

Discretisation of the run: 75 levels of 2.01 m, control step 300 s, sub-step 32.5 s. Over 25 years the fluid took 2463 MWh from the ground and 2463 MWh passed through the walls; the largest step error of the energy balance (18.11) is 1.7·10−14. The lowest inlet temperature is −4.18 °C with the steady model and −4.09 °C with the dynamic model.

18.10 Verification

The network itself goes back to Bauer et al. (2011) and Pasquier & Marcotte (2012); the multipole-derived parameters reproduce Hellström's steady leg temperatures and the effective resistance Rb* along the depth.

18.11 Limits