botminds.aiRequest a demo
SearchNew Ventures, Exploration & Reserves

Reservoir Study Search

Reservoir study search for the engineer who has to defend a forecast: where each input began, page by page, and who has ruled on it.

See the demoBook a demo
Follow one model input back to its first page ↓
The problem

The model input everyone copies and nobody has opened since it was typed

A forecast comes from a model, the model takes its inputs from the last study, and that study took them from the one before. Most of the older ones are scans by people who retired long ago. Following one number back by hand means days in the records store, so the engineer stops at the latest model note and takes the value as given.

That is how a guess survives. Each report cites the one before it, so a copied number reads like a fact. A history match fits pressure and water cut and says nothing about the inputs it never varied. Meanwhile a core test that measured the property sits in a well file, where nobody looking for studies thinks to look.

Built forReservoir engineersSenior reservoir engineersSimulation engineersSubsurface data managers

The Problem Today

A number in the current model can rest on an old guess, copied from report to report.

24 hours
to trace one number through six reports by hand our estimate

Read by hand today

  • Studies
  • Core data
  • Model notes
  • Well files
  • Measured, then filed awayA core result kept with one well did not reach a study
  • The same study, started againNew work begins without the earlier work on the question

The Solution

  1. Ask in plain wordsWhere a value came from, or whether someone has looked already
  2. Back to the first pageEach copy shown on its own page, scans included
  3. How each author got itMeasured, inherited, assumed or fitted, and what was missed
  4. A senior engineer rulesThe engineer flags; the paper waits for the ruling and the re-run

Built-in controls

  • Agents never change or run a model
  • Whoever flags a number cannot rule on it
  • Questions about working a well reach the person in charge

What You Get

Engineers see what a costly decision rests on before the money is committed.

34%
less flow from a lateral at 0.02 than at 0.1, by hand check our estimate
  • Earlier work found firstWhat was concluded, or a plain reply that nothing turned up
  • Rulings come back in answersThe next person to ask sees who decided, and why
  • Evidence for the reviewEach step of the tree and the ruling, in one file
One input, one infill paper

A reservoir study chain of five copies under a $174 million infill decision

A reservoir engineer is writing the paper for six horizontal infill wells. Before it goes to investment review, he asks one question about the model the forecast comes from.

demoThe field, its wells and studies, the engineers and every figure here are sample data.

  1. 01Before the review

    The kv/kh in the model runs back through five studies to a guess

    Reservoir engineer

    He asks where the vertical permeability in the N-1 model comes from, and the answer grows as a family tree. The model note copied it, and so did each study before it, back to the oldest plan: the one page that says the value was assumed. Below sits a core test that measured 0.02. At the right stands what it all carries: $174 million.

  2. 02Minutes later

    No core had been cut, and the author wrote that down

    Reservoir engineer

    Opening the root shows the typed page itself, with the line lit: no core had been cut in the sand, so a ratio of 0.1 was assumed. The page also says the value should be reviewed once core is available. Beside it sit the author, how the scan was read and how sure the reading is. The guess was in plain sight all along, buried under copies.

  3. 03Minutes later

    A median of 0.02 from 38 plug pairs, filed with one well

    Reservoir engineer

    The side branch opens on a core lab table: 38 pairs of plugs, a median ratio of 0.02 and thin clay laminae crossing the interval. A caution follows on the page itself: plugs are an inch across, so a model cell may behave differently. Because the handover list had filed the report with its well, no later study found it.

  4. 04Before flagging

    At 0.02 a 2,000 ft lateral gives about 34% less, by hand

    Reservoir engineer

    He wants to know whether the difference matters before raising it. The sum behind the bars uses a published formula for a horizontal well, and each input cites its page: lateral length, sand thickness, permeability, fluid properties and a 250 psi drop. The screen calls it a hand check of oil alone. Only a run of the model can weigh the two values properly.

  5. 0509:23

    Send stays locked until both pages are open

    Reservoir engineer

    To flag the value he must open the assumed line and the core table himself, here 2 of 2 opened, and choose why: a measured value exists and was not used. He can send it to the senior engineer. Ruling on it is closed to him, because he raised it. Once it goes, the infill paper is held.

  6. 0609:29

    She opens all four pages herself, after the flag arrives

    Senior reservoir engineer

    The evidence panel quotes each author: the assumed line, the core median, the history match that held the ratio from an earlier description, and the model note that left it unchanged. Her signature stays locked until the counter reads 4 of 4 opened since the flag was sent. Pages she read before it reached her do not count.

  7. 0709:41

    Re-run as a range of 0.02, 0.05 and 0.1 before the decision

    Senior reservoir engineer

    She keeps neither number. A plug is an inch across and a model cell is tens of feet, so the core value cannot simply replace the old one. Her ruling asks for three runs and states the reason: a measured value exists that was not tried. The ruling records that she opened all four pages before signing.

  8. 08When the runs are in

    Each case needs a run file whose input file holds that ratio

    Simulation engineer

    The simulation engineer owns the model and makes the runs in the simulator. Here one file per case is picked from the simulation file store, which is read only. Typed numbers are not accepted, and a file only fits a case when its input holds that kv/kh. The paper beside the table still reads held: re-run owed.

  9. 09When the runs are in

    The hold lifts only once she has opened all three run files

    Senior reservoir engineer

    She opens each of the three run files before the countersign button wakes. Then the hold lifts, and the paper sets the signed runs beside the forecast it started with. These are sample run files, and the screen says no simulator ran. What the page shows is the rule: results trace to a file a second engineer opened.

  10. 10Next time anyone asks

    The ruling now rides on the answer to the same question

    Any reservoir engineer

    Asked again, the tree gives the same chain and adds who ruled, what she ruled and what the signed runs gave at each value. An engineer who joins after both of them retire still reads the reason, which is the part a hallway decision loses.

What six sidetracks carry at each kv/kh in the signed sample runs

From the paper once the runs were countersigned. These are authored sample run files and no simulator ran, so they show how a ruling reaches the paper.

kv/khNet a wellSix wellsPer barrel
0.02, as measured0.73 million4.4 million$40
0.05, between0.88 million5.3 million$33
0.1, as in the model0.95 million5.7 million$31

demoSample data, from the working solution. Barrels, net at the 60% share. Each well costs $29 million net.

Forty-one held files sit behind each answer that says no study cites the core

A gap is called a gap. Held files are counted, so a missing citation is a statement about what was read.

Water damaged boxes from an older store23 files
Password protected consultant files11 files
Old file formats from the tape era7 files
Held, unread, and named in each answer41 files

demoSample data, from the working solution.

Around the tree

Earlier work, the model's nine inputs and the reservoir study library

Other screens the subsurface team works from, on the same sample library.

Has anyone done this already?Before a study starts: 3 found in the 22 of 6,120 open in full, what each concluded, and the numbers copied without a new test.
Model inputs and where they beganNine key inputs of the model: 5 measured, 2 assumed, 2 fitted to history, each with its oldest source and how many copies back.
The studies6,120 studies on 34 reservoirs: 3,806 scans and 2,314 typed. Held files are listed with their reason and who is chasing them.
Questions sent to peopleA question about working the well goes to the installation manager with no data back. A judgment call goes to the senior engineer.
RulesThe records manager proposes and the senior engineer signs. Five rules are fixed, among them that agents do not change or run a model.
Audit trailEach page opened, flag, ruling, run file and countersign, by person or agent, with the time. Exports for the trail and an evidence file.

demoSample data, from the working solution.

Who decides

Flag, rule, re-run: three engineers and the page each must open first

No one signs their own work, and openings count only after the item reached the signer.

Reservoir engineer
Sees
The tree, the author's own words and the hand check
Signs
Raises a flag with a reason, and proposes release of the paper
Senior reservoir engineer
Sees
Each quoted page, opened after the flag was sent
Signs
The ruling, the countersign on the runs and the release
Reservoir simulation engineer
Sees
The cases owed and the run files in the store
Signs
Registers one run file per case; can flag, cannot rule
Subsurface records manager
Sees
Hard readings, label lots and held files
Signs
Confirms readings and proposes settings for a second signature
Reservoir manager
Sees
Flags past their clock
Signs
May rule on an overdue flag, but not on one he raised
Technical assurance reviewer
Sees
Everything, read only
Signs
Signs nothing; downloads the evidence file for a review
The agents

Seven agents follow what each author wrote, and none picks the value

Judgment stays with engineers.

Question planner

Works out what is asked and answers records questions.

  • Sends operations and safety questions to a person
  • Never decides which study is right
Study reader

Reads scans, carbons, typed reports and handwriting, with a confidence.

  • A reading under the line waits for a person
Value tracer

Follows stated sources: a citation, a table note, the author's words.

  • Never links two values on the number alone
  • Never fills a gap
Origin labeler

Labels each appearance measured, inherited, assumed or fitted.

  • Never upgrades inherited to measured
  • Says when it cannot tell
Earlier work finder

Finds earlier work on the same question and what it copied.

  • Never reports nothing found as proof
Effect explainer

Sets out what a value does, from a published formula with cited inputs.

  • Never runs or changes a model
  • Gives no forecast
Ruling keeper

Keeps flags, clocks and holds, and repeats signed rulings in later answers.

  • Never rules
  • Never clears a hold without a signature

What reservoir engineers ask before trusting a search over their studies

What is reservoir study search?

Reservoir study search answers questions from decades of reservoir study reports, scans included, with each answer cited to its page. Ask where a model input came from and you get a family tree back to the first page that states it, with how each author got the value: measured, inherited, assumed or fitted.

How does it trace reservoir simulation model inputs?

The value tracer follows what each author wrote about the source: a citation, a table note, a sentence such as unchanged from the history match. Two reports are not joined just because the same number appears in both. Where the trail breaks, the tree shows a gap.

Does a good history match prove the inputs?

No. A history match fits what it was tuned on. In the sample the match changed aquifer size and one fault, and its author wrote that pressure does not respond to kv/kh and that it was not varied. The tree quotes that page, so the reader sees what the match left untested.

What happens when it finds nothing?

It says so, and names what it searched and which files it could not read. A missing citation or an empty search does not count as proof that no test was made.

Does it change our models or run the simulator?

No. A re-run is made by the simulation engineer, who registers a run file from the store for each case. Typed results are refused, and a senior engineer opens each file before countersigning. The model stays with its owner.

Can it read our legacy reservoir reports?

It reads scanned, typed and handwritten pages and gives each reading a confidence. Readings under the line go to the records manager before anything rests on them. Files that cannot be read are counted, with the reason, and named in every answer.

Where does it sit beside reservoir engineering knowledge management?

Most knowledge tools find documents. This one follows a single value through the documents that carried it and shows how each author got it, then holds the decision until a senior engineer rules. The ruling is repeated to the next person who asks.

New Ventures, Exploration & Reserves

More reserves, studies and subsurface records solutions

All 9 solutions →
Upstream Due Diligence

From the seller’s price to your bid, each step shown well by well and signed.

1-min demoPreferential Rights Review

Flip the deal: see every right that fires, in the governing words.

1-min demoDecommissioning Liability Due Diligence

The seller’s one number tested well by well, every gap tied to a page.

1-min demo

Bring one model input
and the studies behind it.

Send us one input and its study chain, and we’ll walk the family tree on your own reports.

Book a demoBrowse solutions