Foreword — What to do with an interrupted test?

A thermal response test is the only in-situ measurement that ties the design of a borehole field to the ground it will be drilled into. The ground thermal conductivity and the borehole resistance it yields set the required borehole length, the undisturbed ground temperature it also measures sets the boundary condition that length is sized against, and together they set a large share of the capital cost of a ground-source heat pump installation. Yet the test is a field operation rather than a laboratory one: it runs for several days on a construction site, usually unattended, with a generator, a heater and a pump all expected to hold a steady output from start to finish. The conventional interpretation asks for exactly that constancy, and the field does not always grant it. A generator trips, the grid fails, the heat injection stops for a few hours, and the record no longer meets the assumptions the straight-line methods rest on. Such a test is then either repeated at full cost, or interpreted as though nothing had happened, with a bias nobody has quantified.

This software is rooted in my own experience. Before moving to academia, I worked as a consultant in the geothermal industry, where I designed, performed and interpreted dozens of thermal response tests myself, and lost a few of them to an interruption. I know what it costs to tell a client that a TRT will have to be restarted because no recognized method can exploit an interrupted one, and I know the quieter temptation of fitting a line through it anyway. TRT Analysis began there: a desktop application for the interpretation of the thermal response tests carried out ahead of any large ground-source heat pump project, and in particular of the records that a field incident occasionally leaves behind.

To read those records as well as the well-behaved ones, I paired two complementary analysis methods. The first-order approximation provides the standards-aligned, slope-and-intercept reading of the heating phase, and of the recovery phase when the test carries one, that practitioners know and trust, while clearly marking the point in time before which the line-source assumption does not yet hold. It keeps the conventional assumptions in exchange: a constant heating power over the window being fitted, and no injection at all once the recovery has started. The short-term g-function goes further. A pre-trained neural network reconstructs the borehole's early-time response, reproducing a detailed transient model, and that response is convolved with the power actually measured, so the interpretation uses the entire test, including the early hours and the recovery, and extends naturally to the pulsed and interrupted tests where the simple model breaks down.

TRT Analysis is developed and distributed by P³ Geothermal, a spin-off company originating from Polytechnique Montréal. Over the years, many graduate students and colleagues have contributed to scientific publications on the models the software rests on: the p-linear averaging of the fluid temperature, fast spectral convolution, and the neural-network approach to transient borehole response. These research contributions have been extensively incorporated into the software (see the References section). Whenever applicable, a portion of the revenues generated by TRT Analysis is shared with students whose research has contributed to its development. This model reflects my commitment to ensuring that academic contributions are recognized and rewarded beyond the publication of scientific work.

My hope is that TRT Analysis helps the specialists who perform these tests extract everything their data has to offer, and, just as importantly, recognize when a test does not support a confident answer.

Professor Philippe Pasquier, eng., Ph.D.
Lead Programmer, P³ Geothermal