BackgroundPDF · 613 KB
Baghewala field guide
The problem, written for a reader with no oil and gas background. Why the oil will not flow, why steam fades, and why the pump breaks.
Updated 12 September 2026
SIH26120 · Baghewala heavy-oil field
After the steam, the well cools. The pump has to slow down with it.
Drag the day. The reservoir cools, the oil thickens, and the speed at which the rods still fall freely drops toward the speed the pump is set to. The twin sees that line coming. Nobody at the wellhead can, until now.
Production day
142
Day 0 · steam just endedDay 236
Reservoir temperatureEstimate
51.8°C
Down from 250 °C at day 0
Oil viscosityEstimate
9,637cP
2,005× thicker than at day 0
Safe pump speedEstimate
3.4SPM
Set speed 3.0 SPM · margin 1.13×
Production day · drag on the chart or the slider
Day 142: the oil is 2,005× as thick as when the steam ended. The rods can take 3.4 strokes a minute; the pump is set to 3.0.
Heavy oil, cold rock, and a pump that keeps breaking
Things cool down. Thick liquids resist movement. Metal breaks under repeated shock. No oil-field background needed.
The oil does not flow.
The crude is 14 to 19 degrees API, near 11,500 cP at 46 to 50 °C. Reservoir pressure is low. Without heat, the oil almost does not move.
Steam works, but not for long.
Steam thins the oil and the well produces. Over the following weeks the heat leaks into the cold rock around the well. The oil thickens again.
The pump breaks.
Thick oil slows the rod string's fall. It goes slack, then the next stroke snaps it tight. Enough shocks and the steel parts.
Oil production and pump safety are not two separate problems on the same well. They are two results of one variable: reservoir temperature. Optimise them apart, and the result is always worse than the best answer together.
A bigger heated zone has less surface against the cold rock, so it cools slower. Steam volume buys time, not only heat — that is the calculation behind “inject more steam now” or “wait.”
“CSS cycle design and SRP operation are optimised separately, using historical experience.”
Steam goes in
A bigger steam job heats a bigger zone. A bigger zone cools slower.
At Baghewala, one team picks the steam volume; a different team sets the pump speed. Both decisions turn on the same heated zone. Drag the volume and watch it.
m³ CWE
500Tier B4,000 m³ CWE
Field practice: 1,040–1,560 m³ CWE. The tinted part of the track is outside it.
- Heated radius
- 11.8 m
- Days of steam and soak
- 34 days
- Oil expected before the next steam job
- 2,946 bbl
Heated-zone temperature
50 °C250 °C
- Estimated zone
The pump has to slow down
Below the ceiling, the rods keep up. Above it, they float.
By production day 170, weeks of cooling have thickened the oil. The rod string can only fall so fast through it. Slide the set speed past that limit and the string cannot keep up: it goes slack, then slams tight on the next stroke.
SPM
0.93.4 SPM
Near the limitSafe speed today: 3.05 SPM
Heated-zone temperature
50 °C250 °C
- Estimated zone
A model of one well, running live
A digital twin is a live software model of one well, updated by that well's own sensors — not a general simulation. Physics does the calculation; the models learn where the physics is wrong, and never replace it.
- Step 1
Measure.
Read the live sensors: load, position, pressure, flow, temperature, steam.
- Step 2
Estimate.
Calculate what no sensor reports: reservoir temperature, viscosity, fillage, rod load.
- Step 3
Predict.
Run the physics forward: cooling, production decline, failure risk, days ahead.
- Step 4
Optimise.
Search the steam and pump settings for the lowest steam-oil ratio that keeps the speed safe.
- Step 5
Recommend.
Show the operator the setting, the reason, and the confidence. The operator approves it.
The operator stays in control.
- Step 6
Learn.
Compare the result against the prediction. Correct the model.
Objective: the most oil for each unit of steam. Constraint: keep the pump speed below the rod-floating limit.
Rod safety is a constraint, not the goal — a pump run at minimum speed forever would fail almost never and produce almost nothing. An accurate safe limit lets the pump run near it, not far below it.
The twin reads the well
One day's gap is noise. More days average it away.
The twin predicts what the wellhead thermometer will read before the day arrives. A sensor is noisy: one reading can sit several degrees off. Widen the window and the random noise cancels itself out.
days
5150 days
Average gap over the window: 0.08 °C (sensor noise is ±0.5 °C on any single day).
Wellhead temperature: the twin's prediction against the sensor
First 60 days of production
Wellhead temperature (°C)
- Sensor reading
- Twin's prediction
Data table
What the system does
Six functions. They match the six requirements in the problem statement.
- 1
Optimise the steam cycle.
Pick the steam volume, soak time and next steam date for the lowest steam-oil ratio, not past practice.
- 2
Predict heating, cooling and production.
Give a cooling curve and a production forecast for each well, updated daily from live data.
- 3
Set the pump speed continuously.
Recommend today's speed. Recalculate it when conditions change. Send it to a drive that can act on it.
- 4
Detect rod floating early.
Calculate the safe speed limit from the current viscosity. Slow the pump before the string goes slack, not after it fails.
- 5
Diagnose the pump condition.
Compare the predicted dynamometer card against the measured one. The gap separates underfill from valve, leakage and mechanical faults.
- 6
Cut steam and energy per barrel.
Steam is the operating cost of thermal recovery. Score every decision above against that one number.
Who uses it
One well's twin, read two ways
The operator
A daily check: today's recommended speed, why, and how sure the twin is. The operator decides, then acts. The system never acts alone.
The superadmin
The whole field, the field clock, and — only here — the reveal: the hidden true numbers behind the wall, next to the twin's estimate of them.
What changes
Oil India lists six expected results. Not one is a technical measurement — all six are operational or financial.
- 1More oil produced and recovered
- 2A lower steam-oil ratio
- 3Less energy for each barrel
- 4Fewer rod failures and pump unsettings
- 5Longer equipment life
- 6Decisions from data, not from habit
- Steam-oil ratio today
- 3.0 to 5.2
- Recovery factor
- below 20 percent
- Field production
- about 655 barrels per day in July 2025
Published field values. Source: the problem statement and Oil India field reports.
What we do not claim yet
Every parameter in our model has a value, a unit, a source and a confidence level. These four are still assumptions.
- 1.Our viscosity curve uses one measured point. A correct fit needs two — our highest-value data request.
- 2.Our rod-floating limit is a static force balance. It ignores rod stretch and wave effects, so it reads high. We will calibrate it against the real failure history.
- 3.We hold reservoir pressure constant in version one. It drives inflow directly, so a depletion term is the next addition.
- 4.The crude is non-Newtonian: its thickness depends on shear rate as well as temperature. We know this. We will not build it until the main chain works.
Project documentation
Technical referencePDF · 1.1 MB
Merged technical reference
The physics of the twin in one place. The governing equations, every parameter with its unit and source, and the objective function.
Updated 12 September 2026
Deep researchPDF · 5.7 MB
Full research document
The complete working research behind the twin, 350 pages with a dated errata page at the end. Model derivations, data sources, assumptions, and the open questions.
Updated 12 September 2026
The team
Pratham Jadhav
Shreyas Date
Vinit Limkar
Nidhee Rajesh Fukte
Sumit Sanjay Mate
Chitrangda Singh Parmar
Sarah Khambatta
Read the full technical documentation
A field guide, a physics reference, and the parameter tables. Written for a reader who has never worked in oil and gas.
SIH26120 · Oil India Limited