HomeFeaturesExamplesTutorialsHelpTheoryVerificationPricingContact
Theory manual

11. Layered ground: segmented model, borehole sections and uniform fluid temperature

The effective model of Section 3.3 gives every layer its share of the active length in three mean properties and leaves the g-function that of a homogeneous ground. This chapter describes the models that resolve the depth: the segmented layer model, in which every segment of the finite line source responds with the properties of its own layer, the borehole sections that describe the construction along the depth (diameter, grout, finite insulation), and the uniform mean fluid temperature boundary condition (UFT) that lets the section resistances act on the distribution of the heat.

11.1 Segmentation at layer interfaces and section boundaries

The segments of a borehole [D,D+H] are those of the homogeneous engine ((4.14)), split at every layer interface and every section boundary inside (D,D+H):

Without an interface inside the active length the segments, and therefore the g-function, are bit for bit those of the homogeneous engine.

11.2 Pair kernel

The response of receiving segment i to emitting segment j is the finite line source (4.4) of a homogeneous ground with the properties of the pair:

hij(t)=λrefλijF(t;αij)2Hi
(11.1)
λij=λi+λj2,Cij=Ci+Cj2,αij=λijCij
(11.2)

with λij=λk, Cij=Ck when both segments lie in layer k. The g-function stays defined against the reference values of the effective model (λref=λeff, αref=αeff, ts=H2/(9αref)), so the superposition (3.6) and everything after it is unchanged; the factor λref/λij converts the physical temperature qF/(2πλij2Hi) into g-function units. F uses s=1/4αijt and the same axial function I(s) with its mirror image, so the one-pass evaluation of Section 4.3 carries over with one lower limit per αij.

Why the arithmetic mean. A point source on the plane interface of two half-spaces has the exact steady temperature q/(4πλ¯r) with λ¯=(λ1+λ2)/2. The mean also keeps F symmetric in the two segments, so the reciprocity (4.9) and the symmetric job structure (and cost) of the homogeneous engine are kept. The obvious alternative is the emitter's layer (Abdelaziz et al. 2014; Erol & François 2018). Against the finite-volume reference of Section 11.5 the emitter's layer deviates by 1.25 % (two layers) and 3.64 % (three layers, strong contrast) at 50 years, the pair mean by 0.58 % and 2.23 %; the harmonic mean gives 1.35 / 3.82 %, the geometric mean 0.96 / 3.00 %.

Limits. Heat that crosses an intermediate layer between two segments sees only the two end layers, and the long-time response of a thin poor layer between good ones is overestimated (in the kernel it cannot spread vertically into its neighbours): the remaining deviation grows with time (2.2 % at 50 years for 3.0 / 1.2 / 2.8 W/(m K)).

Short times. The thresholds tlin and tsolve of (4.18) use the smallest α and the largest rb of all segments, and the self response of (4.19) becomes the length-weighted

S(t)=∑sHs(λref/λs)Ss(t)∑sHs,Ss(t)=12E1(rb,s24αst)
(11.3)

The cylindrical correction (4.22) of a segment uses its own α, rb and λref/λ.

11.3 Borehole sections

A borehole design may carry sections that describe the construction along the depth, independently of the ground layers: each section has its own borehole diameter, grout conductivity and grout heat capacity, and optionally an insulation; gaps between sections use the design values, and sections outside [D,D+H] have no effect. Insulation is always a finite resistance:

Rp,i′=Rp,i+ln((ro+t)/ro)2πλins
(11.4)
Rcas=ln(rb/(rb−t))2πλcas,𝐑′=𝐑+Rcas11T,Rb′=Rb+Rcas
(11.5)

Ra is unchanged by a casing for symmetric leg layouts.

11.3.1 Two-channel conductances and the sectioned effective resistance

With 𝐊=𝐑′−1 summed over the legs of the down-going channel D and the up-going channel U, the conductances per metre of the two channels to the wall and between them are

g1=KDD+KDU,g2=KUU+KUD,g12=−KDU
(11.6)
Rb=1g1+g2,1Ra=g12+g1g2g1+g2
(11.7)

(g12 is the direct conductance of the delta circuit; it is negative when the legs lie close to the borehole wall.) With the wall at Tb=0 the leg temperatures of a uniform-wall-temperature borehole follow

ṁcpdTDdz=−g1TD−g12(TD−TU),ṁcpdTUdz=g2TU+g12(TU−TD)
(11.8)

with TD(0)=Tin and TD(H)=TU(H). Section by section the solution is the 2×2 matrix exponential exp(𝐌kLk), propagated upward from the bottom (the stable direction); it gives Tout/Tin and

Rb*=H(Tin+Tout)2ṁcp(Tin−Tout)
(11.9)

One symmetric piece reproduces (5.22) to 10−9. For a uniform heat flux (the classic form) the wall temperature follows from g1(TD−Tb)+g2(TU−Tb)=q, so TD−TU=q(H−z)/(ṁcp) and TD is quadratic in z on each piece; Rb*=(T¯f−T¯b)/q is evaluated exactly (Simpson's rule per piece) and one piece reproduces (5.21) to 10−9 for every pipe type.

Coaxial exchangers. (5.22) assumes two channels with the same resistance to the wall. The exact coaxial solution (11.8) lies 20 % above it at 0.05 l/s and 0.3 % at 1 l/s. A coaxial design without sections keeps (5.22) (unchanged results); with sections the exact solution is used.

Results by section. The reported resistances are Rb and Ra as length-weighted parallel means H/∑k(Lk/Rk) and Rb* from (11.9) (or the fixed value); the results by section list every section and gap of the active length with Rb, Ra and its share of the heat (from the segment heat rates of the g-function solution at its last time). Insulated return leg: insulating the up-going leg in the top 30 % lowers Rb* where the thermal short-circuit dominates, always for the inner pipe of a coaxial exchanger (300 m, 0.5 l/s: 0.195 → 0.150 m K/W) and for a single-U only at low flow in a deep borehole (300 m, 0.1 l/s: 0.513 → 0.344 m K/W). At ordinary flows (150 m, 0.5 l/s) it raises Rb*, because the return leg also loses its own exchange with the ground.

11.4 Uniform mean fluid temperature (UFT)

Under UBWT all segments share one wall temperature, so a section resistance cannot change where the heat goes. Under the UFT condition every segment exchanges heat with one mean fluid temperature Tf through its own local resistance Ri (the Rb of its section), and the coupling row of (4.16) becomes

Tb,i+qiRi=Tf,∑iHiniqi=Htotal
(11.10)

where ni is the number of boreholes the segment represents. One diagonal term 2πλrefRi is added to the UBWT matrix; its size and the solver stay the same. The result is an effective-wall g-function:

g(t)=2πλrefTf(t)−Tmq−2πλrefRpar,Rpar=Htotal∑iHini/Ri
(11.11)

It tends to 0 at short times like any wall g-function, so the short-time handling, the evaluation rules and the sizing cache apply unchanged (the g-function is marked as an effective-wall response that already contains Rpar). The simulation uses Tf=Tb*+qRb* with Tb* from (3.6) and the sectioned Rb*; without short-circuit (large flow) this is the Tf of the UFT solution exactly. Bounds (verified): R→0 gives the UBWT g-function, R→∞ the UHTR one, and g grows monotonically with R in between. An insulated section receives a reduced, never zero, heat rate, and its share tends smoothly to 0 as λins→0 (PU foam sleeves of 20 mm on both legs between 50 and 80 m: 0.197 of the heat per metre of the bare part).

Radius along the depth. A section with its own diameter gives its segments that radius: the self-term distance and the short-time response use the receiver's rb, so F is no longer symmetric and every ordered segment pair gets its own entry (the Equivalent method then falls back to Similarities). This happens whenever the g-function is segmented at the sections (segmented ground or UFT); otherwise the g-function uses the length-weighted mean radius over the active length and the simulation warns.

11.5 Examples and verification

Table 18. The office example (4 × 4, B = 6 m) on the layered example grounds (simulated at H = 150 m, then sized). Five layers: clay, sand, claystone, limestone and granite (VDI 4640 values). Clay over rock: 70 m at 1.5 W/(m K) over 3.5 W/(m K), the same mean conductivity as the homogeneous example.
Ground and layer modelTm modelTm [°C]λeff [W/(m K)]Tf,min / Tf,max [°C]Sized H [m]
Homogeneous (λ = 2.6 W/(m K))–10.732.600−2.41 / 16.03143.7
Five layers, effective modelactive-length average11.072.621−1.92 / 16.34140.3
Five layers, effective modelmid-depth10.722.621−2.27 / 15.99143.1
Five layers, segmented modelactive-length average11.072.621−2.00 / 16.34140.9
Five layers, segmented modelmid-depth10.722.621−2.35 / 15.99143.7
Clay over rock, effective modelactive-length average11.622.620−1.38 / 16.89136.1
Clay over rock, effective modelmid-depth10.722.620−2.28 / 15.99143.3
Clay over rock, segmented modelactive-length average11.622.620−1.65 / 16.91138.3
Clay over rock, segmented modelmid-depth10.722.620−2.56 / 16.01145.8

Table 18 and Figure 10 show what the layering changes. With the same mean conductivity the effective model gives the homogeneous g-function; the segmented model lets the rock below 70 m take more than its thickness share of the heat, and the g-function of the clay-over-rock ground is up to 5.6 % higher at long times. The segmented field therefore needs 138.3 m instead of 136.1 m (active-length average Tm). The larger effect comes from the undisturbed temperature: the poorly conducting clay raises Tm under the active-length average by 0.90 K, which shortens the heating-dominated design by 7.2 m. On the five-layer ground both effects are small.

(a) g-functions of the office field on the homogeneous ground and on the clay-over-rock ground with the effective and the segmented model; (b) relative difference segmented / effective.
Figure 10. (a) g-functions of the office field on the homogeneous ground and on the clay-over-rock ground with the effective and the segmented model; (b) relative difference segmented / effective.

Verification. An independent axisymmetric finite-volume solution of one borehole (H = 100 m, D = 4 m, uniform flux on the wall, 26 000 cells) is the reference. BoreTherm's segmented UHTR g-function with the cylindrical correction agrees with it within 0.02 % for a homogeneous ground, 0.58 % for two layers and 2.23 % for three layers with a strong contrast (all at 50 years; 0.12 % and 0.30 % at 9 months). The sectioned resistances reproduce Hellström's closed forms to 10−9, a casing the analytic shell resistance to 10−12; sections that repeat the design values give bit-identical results.