Short and long term production forecasts are important for assessing the performance and value of unconventional oil and gas wells, setting production targets and predicting cash flows. Often the best way to try to estimate future production of a new completion is by analogy with production from other nearby completions with similar treatment histories. Forecasting production using type curves combines knowledge with experience using physics-based decline equations, type curve coefficients and an engineer's judgment. Modern algorithms for predictive analytics can be based on a generative model (GM) which does the same thing. A GM is a stochastic model which can be constructed to capture physics (solutions to differential fluid-flow equations) with experience (probability distributions that are calibrated using historic data). The result is a model which is similar to a type curve in that it can be used to give a reasonable forecast of the future of a completion based on only a small amount of actual information about that particular well; the rest it draws from what it has seen before. By combining a GM with an algorithm for statistical inference, it is possible to make accurate forecasts that include quantified uncertainty bounds. The combination of a forecast and uncertainty leads to powerful new ways of looking at future production including Forecast Certainty Maps.
Generative Models for Production Forecasting in Unconventional Oil and Gas Plays
- Share Icon Share
- Search Site
Kuzma, Heidi Anderson, Arora, Nimar S., and Kemal Farid. "Generative Models for Production Forecasting in Unconventional Oil and Gas Plays." Paper presented at the SPE/AAPG/SEG Unconventional Resources Technology Conference, Denver, Colorado, USA, August 2014. doi: https://doi.org/10.15530/URTEC-2014-1928595
Download citation file: