HomeFeaturesExamplesTutorialsHelpTheoryVerificationPricingContact
Theory manual

8. Mean fluid temperature simulation

8.1 Monthly base loads

The simulation covers M=12Y months (Y = 1 … 100 years). Month n=0…M−1 ends at tn=(n+1)tmonth and has the heat rate per metre qn of calendar month cn:

cn=(m0+n)mod12
(8.1)

where m0 is the start month (0 = January). Temporal superposition of (3.6) with g from Chapter 4 gives the wall temperature and the mean fluid temperature at the end of each month (Hellström & Sanner 1994):

Tb(tn)=Tm+12πλ∑i=0n(qi−qi−1)g(tn−itmonth),q−1=0
(8.2)
Tf(tn)=Tb(tn)+qnRb*
(8.3)

The classic monthly method prints calendar months with no load before the start month; BoreTherm's results start at the first operating month. The two agree after the index shift n=12(y−1)+m−m0. The monthly method of an open-source sizing tool is the same algorithm; given the same g-values it agrees with an independent implementation of (8.2)–(8.3) to 10−13 K.

8.2 Peak loads

The peak of month n is a rectangular pulse of duration tp at the end of the month that replaces the base heat rate; its energy is not added to the monthly balance, as in the classic monthly conventions. Its effect is the step from qn to qpeak during tp:

Tf,peak(tn)=Tf(tn)+(qpeak,n−qn)[gp(tp)2πλ+Rb*]
(8.4)

The peak is applied only if it is more extreme than the base in its direction: for extraction qpeak<0 and qpeak<qn, for injection qpeak>0 and qpeak>qn; otherwise Tf,peak=Tf. The peak response gp is an option:

gp(tp)=max(0,ln(H2rb)+12ln(tpts))
(8.5)

The use of (8.5) in the classic monthly method was established by back-calculation from the worked example of its published manual (residual ≤ 0.0024 in g over ten peak months). H and rb are the mean length and radius of the field. The two choices differ by 0.03–0.08 K in the verification cases and by at most 0.04 K on the peaks of the Linden worked example. For the example the overall minimum at 150 m moves from −2.41 to −2.39 °C. The option applies to monthly simulations only; in hourly mode the peaks are real hours of the load series.

The annual extremes and the overall extremes used by sizing are

Tmin,y=minn∈y{Tf,Tf,peak,ext},Tmax,y=maxn∈y{Tf,Tf,peak,inj}
(8.6)

For the example in January of year 1: the base rate is q = −13.28 W/m and the peak −30.70 W/m. With gp = 1.253 and Rb* = 0.1194 m K/W, (8.4) lowers the fluid temperature by 3.42 K, from 6.18 to 2.77 °C. With the asymptote (8.5) the drop would be 3.39 K.

Table 15. Year 1 of the example at H = 150 m (L = 2400 m): heat rates per metre from (7.3) and (7.4), and the month-end temperatures computed by BoreTherm. A dash marks a month without a peak in that direction; its peak column then equals the base value.
Monthq base [W/m]q peak ext. [W/m]q peak inj. [W/m]Tf [°C]Tf,peak,ext [°C]Tf,peak,inj [°C]
Jan−13.28−30.70–6.182.776.18
Feb−12.69−29.32–5.982.725.98
Mar−10.76−24.75–6.313.576.31
Apr−7.80−19.621.407.084.768.88
May−3.32−12.684.208.466.639.94
Jun5.92–11.2511.6211.6212.67
Jul12.08–22.5013.9813.9816.03
Aug7.46–14.0512.8112.8114.10
Sep−3.83−12.102.809.167.5410.46
Oct−7.56−17.22–7.755.867.75
Nov−10.09−23.18–6.674.106.67
Dec−12.36−28.52–5.632.465.63

8.3 Hourly simulation

In hourly mode the heat rate is constant within each hour. The wall temperature at the end of hour n is the discrete convolution of the load steps with the hourly step response:

Tb(n)=Tm+12πλ∑k=0nΔqkgn−k,Δqk=qk−qk−1
(8.7)
gj=g((j+1)·3600s),j=0…N−1
(8.8)

Index n is the end of hour n+1 and uses the load of that hour (the convention of the hourly method of an open-source sizing tool). The sum is evaluated in one of two ways, selected by an option.

8.3.1 FFT convolution

The first N terms of the linear convolution are computed with real FFTs (Marcotte & Pasquier 2008), zero-padded to a length P≥2N whose prime factors are all in {2, 3, 5, 7}, so that the circular convolution equals the linear one:

𝐒=ℱ−1[ℱ(Δ𝐪)·ℱ(𝐠)]0…N−1
(8.9)

The cost is O(NlogN) and the result is exact to round-off: 100 years (876 000 hours) take about 0.1 s.

8.3.2 Claesson–Javed load aggregation

Alternatively the past loads are aggregated into cells (Claesson & Javed 2012). Cell m=1,2,… is wm=2⌊(m−1)/5⌋ hours wide (five cells per level) and ends τm=w1+…+wm hours in the past. Cells cover the requested horizon plus five margin cells. After each hour the cell loads shift from the oldest cell down; the last cell absorbs incoming history without losing an outflow,

q¯M←q¯M+q¯M−1wM,q¯m←q¯m+q¯m−1−q¯mwm (2≤m<M),q¯1=qn
(8.10)

and the temperature change is

Tb(n)−Tm=12πλ∑mκmq¯m,κm=g(τm)−g(τm−1)
(8.11)

For a 25-year hourly simulation of the example project the two methods differ by at most 0.084 K. FFT is the default.

8.3.3 Inlet and outlet temperatures

With Q the field heat rate (+ injection) and ṁcp the heat-capacity flow of the whole field, the fluid enters warmer than the mean when heat is injected:

Tf=Tb+qRb*,Tin=Tf+Q2ṁcp,Tout=Tf−Q2ṁcp
(8.12)

The classic monthly method writes the same relation with its opposite sign. For monthly output from an hourly run, the base value of month n is the temperature of its last hour, and the peak values are the minimum and maximum over its 730 hours.

8.4 Warnings

Every result carries warnings when:

The monthly and hourly temperatures above use the steady borehole resistance. The short-time borehole model adds the heat capacity of fluid, pipes and grout to the hourly values and the peaks (Chapter 15); the fluid temperatures along strings of boreholes are in Chapter 14.