Danomics + Claude: Selecting and Setting a Key Well

Your Claude (LLM) account can connect to your Danomics account. This is outlined in this article on how to set up a Danomics Connect for Claude. You will need to have completed this setup (or a similar one) to follow along with the tutorial below.

In this tutorial we will be using Claude to analyze our wells and stratigraphy to come up with a list of candidate key wells and to set one for us.

Continuing On

This tutorial continues on from using Claude + Danomics to make a new CPI. If you haven't yet done so, take a look at that video.

One common challenge we face is determining which wells will make for a good key well - does it cover the full stratigraphy, have complete logs, etc. Claude can help us with this. Let's give it the prompt:

In the new PRB.cpi can you go through and come up with 3-4 candidates for a key well?

Claude will analyze through the wells and come back with some candidates. It gave me the following choices (note, yours may be different).

All six wells I shortlisted carry a complete top set (Lewis → Teapot → Parkman → Sussex → Shannon → Niobrara → Carlile → Frontier → Mowry → Muddy → Dakota), so tops aren't the differentiator — log suite, penetration, and hole quality are. Here are the four strongest key-well candidates, best first.

1. API 49005618280000 — Devon "044672-1SXV" (spud 2013, TD ~11,760 ft)
The most complete suite in the dataset. Modern array induction (AF/AO/AT at five spacings + RXO), RHOZ density with HDRA/DRHO QC, TNPH/NPHI/NPOR neutron, PEFZ, DTCO sonic, spectral/enhanced GR (HGR/ECGR), plus HCAL caliper and SP for hole-quality flagging. Full quad-combo + sonic + PE + a proper caliper, centrally located (Campbell Co., ~43.99N/-105.51W), and it penetrates well below the Dakota. This is the cleanest well to anchor parameter picks on.

2. API 49005622740000 — Wesco "15462" (spud 2014, TD ~10,352 ft)
Essentially the same modern array-induction/RHOZ/TNPH/PEFZ/DTCO class as the Devon well, and it also carries computed RWA_HILT/AORT. Advantage: it's already one of the three wells in your existing PetroAI CPI crossplot set, so parameters would carry over cleanly. Trade-offs: it sits farther east and structurally shallower (Dakota at ~8,700 ft), so it images the section but with less overburden than the basin-center wells.

3. API 49005271630000 — Apache "34-9" (spud 1983, TD ~12,242 ft)
The best of the vintage logs and the deepest penetration of the group. Genuine PEF, full dual induction (ILD/ILM/SFLU), RHOB with CALD/DRHO, and — unusually — three porosity curves (NPHI_SS, DPHI_SS, SPHI_SS) plus DT sonic and GRD/GRR/GRS. A strong SW-quadrant anchor if you want a key well away from the two modern eastern/central wells.

4. API 49005315370000 — True Oil "1" (spud 1997, TD ~11,886 ft)
Your current key well in the tuned PetroAI CPI, so it's a proven, low-risk choice. Solid ZDEN/ZCOR density, NPHI_SS, AC sonic, focused resistivity (RILD/RILM/RFOC/SFLI), and — a real plus for QC — a logged repeat section (the _R curves). Main gap: no PE curve, which the two modern wells give you for lithology/MINV work.

My pick would be the Devon 49005618280000 for a fresh interpretation (most complete + central + PE + caliper), with 49005622740000 as the natural alternate if you'd rather stay consistent with the parameters you already tuned in PetroAI.

Want me to set one of these as key_well_api in the new PRB.cpi? I can also add all four to the crossplot/well-list selection so they're easy to compare side by side.

I then prompted it:

Okay, I'll take your recommendation, please set the key well to the Devon well you suggested.

Claude did as requested and let us know that was done.

Summary

Now Claude has helped us with a task that often takes significant manual inspection and analysis by shortlisting candidates for a key well. Once again, although this is a small step it shows how we could chain requests together and have Claude do significantly more.