SR_Nicholson
Mechanical
- Sep 30, 2021
- 1
I've seen references to a common equation using the "trumpet curve" to approximate soil temperature below depth across time, however, all of these require a known thermal diffusivity value.
From my experience working and researching in the field of geothermal energy (mec background), I know that thermal diffusivity (or thermal conductivity, which it is a function of) is extremely hard to calculate without tests of each soil sample. Using approximate values for this from tables of generalized soil types isn't accurate enough for global soil variations, and for variations by depth. In geo-exchange applications, the accuracy of the ground temperature (and of the soil's properties) have serious effects on the performance of a system.
So far, I've found this tool: [URL unfurl="true"]https://groundtemperatures.com/[/url]
That uses a machine learning model to predict ground temperature (with traditional physics calculation like the "trumpet curve" combined with measured datasets from across the globe). However, unlike the traditional method (or apps like cableizer), this tool automatically finds the soil thermal properties - and only needs the user to input date, location, and depth. It also works fairly quickly. For every location I've tested, from the arctic, to deserts, to urban areas, to jungles and farmland have produced accurate temperature predictions within a reasonable error.
I'm wondering how others in the industry typically deal with the difficulty of calculating ground temperatures using the "trumpet curve" method - and if the Ground Temperature Predictor tool I found would be useful for feasibility, or as a sanity check for on-site tests?
This is a continuation of this thread: thread261-228349
From my experience working and researching in the field of geothermal energy (mec background), I know that thermal diffusivity (or thermal conductivity, which it is a function of) is extremely hard to calculate without tests of each soil sample. Using approximate values for this from tables of generalized soil types isn't accurate enough for global soil variations, and for variations by depth. In geo-exchange applications, the accuracy of the ground temperature (and of the soil's properties) have serious effects on the performance of a system.
So far, I've found this tool: [URL unfurl="true"]https://groundtemperatures.com/[/url]
That uses a machine learning model to predict ground temperature (with traditional physics calculation like the "trumpet curve" combined with measured datasets from across the globe). However, unlike the traditional method (or apps like cableizer), this tool automatically finds the soil thermal properties - and only needs the user to input date, location, and depth. It also works fairly quickly. For every location I've tested, from the arctic, to deserts, to urban areas, to jungles and farmland have produced accurate temperature predictions within a reasonable error.
I'm wondering how others in the industry typically deal with the difficulty of calculating ground temperatures using the "trumpet curve" method - and if the Ground Temperature Predictor tool I found would be useful for feasibility, or as a sanity check for on-site tests?
This is a continuation of this thread: thread261-228349
