Thermal response test data

This step loads the measured thermal response test data and lets you isolate the phases of the test (circulation, heating and recovery) used by the analysis models.

To load a TRT data file, use the import button . The file must be a comma-separated values (.csv) file with 4 columns and with a full point as the decimal separator. The measurements can also be pasted straight from a spreadsheet with the paste button or Ctrl+V, in the same 4-column layout, and the loaded table can be copied back out with the copy button or Ctrl+C. A malformed file is refused outright rather than repaired: a short row or a non-numeric cell would otherwise enter the record as a 0 W heating rate or a 0 °C temperature, and dropping the row instead would shift every later timestamp. The message names the first offending line, which can then be corrected at the source.

Heating power — Q (W) - The first column of the file must be the heating power Q in watts.

Circulation flow rate — V (L/min) - The second column must be the flow rate V in liters per minute. The flow rate is only required for the Reynolds number calculation or for the short-term g-function. If the flow rate has not been measured and that model is not required, a dummy value can be used to populate the .csv file.

Fluid temperature — T (°C) - The last two columns are the temperatures, in degrees Celsius, measured in the fluid at the borehole inlet T(z=0 − ↓) and outlet T(z=0 − ↑), or at locations deemed appropriate to represent these measurement points. The notation names the point of measurement: z = 0 is the ground surface, where both probes sit, and the arrow gives the direction of flow in the pipe, ↓ for the fluid running down into the borehole and ↑ for the fluid coming back up. T(z=0 − ↓) is therefore the temperature entering the borehole and T(z=0 − ↑) the temperature returning from it.

Sampling period — dt (s) - Provide the time interval between two measurements, in seconds. It is assumed that the TRT data was recorded at a regular interval and at the end of the sampling period; thus, no data point is drawn at t = 0. Field records are usually sampled every 15 to 300 s. Correcting dt after an import is honoured: the time vector, the phase durations, the chart axes and the time base of both models are rebuilt over the same measurements, and the phase boundaries already set are preserved.

Low-pass filter length — n (-) - To remove white noise, a low-pass filter of length n can be used to smooth the data. A length n = 1 corresponds to no smoothing. For n > 1, a moving average centered at (n−1)/2+1 is applied over a window of length n. The same filter is applied to each column of the TRT data file. Four lengths are offered: 1, 3, 7 and 15.

Phase selection - Once the data is loaded, use the dual-range slider to select the heating phase and, if applicable, distinguish between the circulation and recovery phases. To refine your selection to the exact sample, press the Q (←) or E (→) keys to move the start of the heating phase and the A (←) or D (→) keys to move its end. The keys work as soon as this panel holds the keyboard focus, so clicking the slider, the data table or anywhere on the panel is enough; they do not act from the chart panel. The charts themselves can be zoomed with the mouse to inspect a boundary closely, which changes the view only and never the selection.

The three phases follow from the two handles: everything before the lower handle is the circulation phase, everything between the handles is the heating phase, and everything after the upper handle is the recovery phase. A test with no circulation, or none recorded after shut-off, simply leaves that phase empty, and every figure that depends on it is then reported as a dash rather than carried over from the previous dataset.

Heating power and flow rate - The heating power Q is plotted against the left axis and the circulation flow rate V against the right axis, each in the colour of its own axis title, over the whole duration of the record. This is the chart on which the quality of the test is judged: the heating power should form a flat plateau over the heating phase, and the flow rate should be steady. A drifting or stepped power trace means the constant-power assumption of the first-order approximation is not met, and the short-term g-function should be preferred. Vertical markers show the phase boundaries selected with the dual-range slider.

Inlet and outlet fluid temperature - Shown alongside the previous chart. The temperatures measured at the borehole inlet T(z=0 − ↓) and outlet T(z=0 − ↑) are drawn against time, with the same phase-boundary markers. The separation between the two curves is the temperature rise across the loop; a large separation, together with a low flow rate, is the visual signature of the near-laminar conditions that bias R_b^* high and λ_s low.

Selecting the Undisturbed temperature node does not replace these two charts: neither the circulation-phase view nor the vertical profile is drawn here (see Undisturbed temperature for both), so this pane keeps showing the heating power and the fluid temperatures over the whole test instead.

Summary & charts: the histograms and the statistics reported in the console

Data histogram - The panel to the right of the table shows the distribution of the heating power Q, or of the flow rate V, over each of the three phases, chosen with the drop-down above it. Its vertical axis is a count of samples, and hovering a bar names the class and its count. The distribution is what shows whether the heat injection and the circulation held steady: a narrow, single-peaked heating-phase distribution supports the constant-power assumption on which the first-order approximation rests, while a broad or multi-peaked one says the power drifted and the short-term g-function should be preferred.

Phase statistics - The numerical characterisation of each phase, duration, mean fluid temperature, mean heating power, the coefficient of variation of that power in percent, mean unit power, mean flow rate and Reynolds number, is reported on the first page of the console rather than on this panel, so that the raw-data table can use the full width. Selecting this node brings that page up, and the figures follow the handles as they move.

TRT in practice: what makes a dataset worth interpreting

A sound test, from the field. Everything the two models produce is conditioned by the record loaded here. A sound dataset comes from a test of sufficient duration, commonly 48 hours or more of heating, run at a stable, accurately metered power, with the flow held high enough to stay turbulent and the above-ground pipework insulated against day-to-night ambient swings. The fluid temperatures and the heat injection rate should be logged together, on the same clock, throughout circulation, heating and recovery. The conduct of the test itself is not covered by this manual; consult the IEA ECES Annex 21 final report for equipment, procedure and reporting practice.

The heating power itself. The heating power deserves as much scrutiny as the temperatures. The unit power q = Q/H enters the regression slope in direct proportion, so a wattmeter reading three percent high returns a conductivity three percent off, in one direction, with nothing in the residuals to reveal it. The calibration of the test unit itself is therefore a subject of its own: Eslami Nejad et al. (2018) and Karrer et al. (2024) both describe rigs built to check a TRT unit against a known response before it goes to site. Where the test hole can be instrumented over its depth rather than at its head, a distributed measurement returns conductivity as a function of depth instead of a single average, whether by heat extraction (Rolando et al., 2017) or by optical fibre (Galgaro et al., 2018). Dalla Santa et al. (2022) carry the fibre approach further, recovering a high-resolution conductivity profile from repeated tests in a stratified ground where a single test average would have flattened fifteen distinct units into one figure. Neither is what this module interprets, but both are worth knowing when a single figure for a layered ground looks too tidy.

Not a fix for an unstable test. The low-pass filter is there to remove measurement noise, not to rescue an unstable test. Real swings in the heating power or flow rate are physical signal, not noise, and smoothing them away merely hides a problem the first-order approximation is not equipped to handle. Keep the filter length n as small as the noise allows; if the power drifted substantially, do not filter it flat: turn instead to the short-term g-function, which is built to accept a varying heat-injection rate.

Phase boundaries. The phase boundaries deserve the same care. The first-order approximation in particular requires an accurate identification of the beginning and end of the heating phase, since its regression is fitted over exactly that window. Take the time to place those limits precisely, the fine-tuning keys move them one sample at a time, because a boundary off by even a few points shifts the slope and intercept from which λ_s and R_b^* are read. Beware in particular of short parasitic spikes in the Q column, before the test proper or after it: dragging a handle out to the apparent edge of the column can enclose them and drive the interpretation to an absurd conductivity and even a negative R_b^*.