Independend g-function and variable mass flow for optimization/calibration

[fuma01]

With reference to issue #215, I encountered a similar question in my current project regarding a variable mass flow rate (laminar–turbulent) in an hourly time‑resolved simulation with prescribed load and mass flow, including load aggregation. In addition, I attempted to implement the core idea of Eskilson/Claesson of a material‑independent g‑function.
Objective: Simulation of borehole fields that have been in operation for several years, using measurement data (heat meter and forward/return temperature measurements). The ground parameters (T_s, k_s, cv_s, power_start) and grout conductivity (k_g) are to be “calibrated” through optimization, and the sensitivity of the parameters is to be analyzed. The power_start parameter was introduced because no measurement data are available for the first few months of operation. Subsequently, forecasts and load‑profile optimizations are to be carried out using the optimized parameters.
I will send you an invitation to my example repository (unfortunately, I do not yet have approval from my company to make it publicly accessible; i sent an invitation to privat repo) and would greatly appreciate a review. I am interested in your assessment of the implementation and whether variable mass flow and material‑independent g‑functions are already achievable in other ways within pygfunction.
Given constructive feedback, I am open to making the repository publicly accessible or to pursuing a joint publication.
@MassimoCimmino
@j-c-cook

[j-c-cook]

Thank you for sharing this -- having real operational data to validate against is extremely valuable, and sounds like a great opportunity for both pygfunction and practicing GHE designers in general.

I'd be interested in exploring a two-pronged approach if @MassimoCimmino is also on board:

1. Near-term publication: Using the data you have now, we could validate aspects of the model that your dataset can meaningfully exercise (and possibly resolve NewFeature Analytical Short-term g-function #149 as apart of the effort)
2. Long-term follow-up: Some of the more interesting questions - long-term ground temperature drift, borehole-to-borehole thermal interference and whether the g-functions hold over extended operation - likely require more data than a few years can provide. If measurements are going to continue on this system, a follow-up study revisiting the same model against a longer record may be a natural second piece of this work.

[MassimoCimmino]

@fuma01 I missed the invitation. I am curious to take a look.

[fuma01]

@MassimoCimmino : send another invitaion: I'd appreciate it if you took a look. I like the basic idea of ​​a geometry-only dependent g-function and the pre-calculation of the borehole resistance for multiple flow rates due to the speed of the time simulation. Perhaps these ideas could be integrated into the PygFunction more intuitively, but I don't have enough insight into that. As the examples show, this approach can be used for optimizations where the g-function is only calculated once for the geometry of the probe field despite changes in the soil parameters—or where multiple changes in the flow rate don't lead to constant recalculation of the network in the time loop.

[fuma01]

@j-c-cook: Thanks for your reply. Indeed, interesting points. I've added the folder "real_data_otp" to the repository, containing the input and output of a simulation with optimized parameters of a BHE field using real measurement data. For me, the results are very useful, especially for predictions. The measurement data isn't of optimal quality: there are unrealistic power peaks, very inaccurate volumetric flow measurements, and the sensors aren't optimally positioned and have gaps (as is typical). Additionally, 10% of the length is in groundwater, and it borders directly on another BHE field. Multiple influences...