RESERVOIR
Decline models fitted to production history, ranked by AIC, with P90/P50/P10 EUR from a bootstrap. Rate-transient analysis then reads permeability, fracture length and oil in place from the same rates and flowing pressures.
WHAT IT DOES
Decline curve analysis fits an empirical model to a well's rate history and extends it to an economic limit. Converge fits four models to the same data, by least squares on log rate, with outliers down-weighted and downtime excluded, and ranks them by the Akaike information criterion so a better score cannot come from extra parameters alone.
A single EUR hides the uncertainty, so Converge adds a block bootstrap: the residuals are resampled in blocks that keep month-to-month correlation, every resample is refitted, and the terminal decline is drawn around your input. The result is a P90, P50 and P10 EUR, and the spread between models is shown separately.
Rate-transient analysis goes further: with flowing pressures, rate-normalised pressure, Blasingame and Agarwal-Gardner plots, a type-curve match, linear-flow analysis and flowing material balance turn production data into reservoir properties.
KEY CAPABILITIES
Four model families, all fitted to the same data so the comparison is fair.
A fit that holds up on real, noisy production data.
EUR to the limits you set, with the uncertainty stated.
Work across a field, not one well at a time.
Production data read as a long, variable-rate well test.
In-place volume from flowing data, without shutting the well in.
WORKFLOW
Paste rate history, or send it from the Data module with downtime already flagged.
Fit all four models, review the AIC ranking and click out points that do not belong.
Enter the economic limit, terminal decline and time cap for the EUR.
Get the P90, P50 and P10 EUR in the background while the page stays usable.
Build type wells and run batch decline over every well, with QC flags.
Add flowing pressures for rate-transient analysis and flowing material balance.
VALIDATION
The decline cases were generated from known parameters, so the truth is exact, and the rate and cumulative forms are checked against an independent open-source decline library. One coverage case is marginal, and the benchmarks page says so.
| Check | Reference | Converge | Tolerance |
|---|---|---|---|
| Arps fit to exact data, qi 3,500 STB/d, Di 60%/yr, b 0.70 (V15) | The generating parameters | Recovered exactly | 3 decimals |
| Arps fit with 3% noise (V15) | The generating parameters | Recovered | 5% on qi |
| Arps EUR to a 30-year cap (V16) | 4.78 MMSTB | 4.78 MMSTB | 0.005 MMSTB |
| Exponential and harmonic EUR (V17) | 1,980 and 9,210 STB | Closed forms matched | 0.5 and 5 STB |
| Rate and cumulative, all four models | Independent decline library | Matches | 10⁻⁶ relative |
| Oil flowing material balance (V14) | Known synthetic truth | Within tolerance | 1% |
| Bootstrap P90–P10 contains the truth, 12 synthetic wells (V18) | At least 80% coverage | 10 of 12, marginal | 80% coverage |
Not built yet: type-curve matching for gas wells and Fetkovich type curves. Gas wells get diagnostics and flowing material balance.
INPUTS AND OUTPUTS
CONNECTED MODULES
Rates arrive checked from Data, and the forecast and its range flow on to money and reserves.
FAQ
It fits Arps, modified hyperbolic, Duong and stretched exponential to the same data and stars the one with the lowest AIC. You can use any of them.
No. Downtime is detected and excluded, outliers are down-weighted with a Tukey biweight, and you can click any point in or out.
By a moving-block bootstrap of the log-rate residuals, 200 resamples by default, each refitted with the same model, with the terminal decline drawn around your input. The spread between different models is reported separately.
Yes. The batch tab fits every well, flags problems in a per-well QC table and exports the results as CSV. Type wells average aligned wells for new-well forecasts.
Diagnostics and gas flowing material balance (Mattar-McNeil), yes. The type-curve match is for oil wells in this release.