Application of machine learning for probabilistic production forecasting and ultimately recoverable reserves estimation

UDK: 622.276.1/.4.001.57
DOI: 10.24887/0028-2448-2020-9-109-113
Key words: machine learning, Markov Chain Monte Carlo simulation, Bayes theorem, risk assessment, uncertainty quantification, production forecasting, reserves estimation, displacement curve analysis, decline curve analysis
Authors: M.Yu. Nazarenko (Gubkin University, RF, Moscow; LUKOIL Mid-East Ltd. in Basrah, the Republic of Iraq, Basrah), A.B. Zolotukhin (Gubkin University, RF, Moscow; Northern (Arctic) Federal University named after M.V. Lomonosov, RF, Arkhangelsk)

Reservoir and production engineering processes involve a number of components applied to monitor and plan such parameters during oil and gas fields development as oil production profile calculation and ultimately recoverable reserves estimation. Field-statistical models such as displacement curve and decline curve analysis models are wildly used to forecast production performance and estimate reserves. As compared to decline curve analysis, displacement curve analysis provides a better result in production forecasting in waterflooded oil reservoirs since, apart from oil production, they also take into account liquid and water production. However, both methods are deterministic i.e. make no allowance for uncertainty in calculations that can result in failure to achieve planned production performance. In order to solve the problem of uncertainty quantification and risks assessment in production forecasts and ultimately recoverable reserves estimates, foreign scientists have applied methods of machine learning with traditional methods of oil and gas production forecasting.

This research for the first time applies such advanced machine learning methods as Bayes theorem and Markov Chain Monte Carlo simulation with integral displacement curve models for quantitative assessment of risks and uncertainty in the production forecasts and estimates of ultimately recoverable reserves. Developed methodology of probabilistic production forecasting yields to different probabilistic scenarios of production profiles and reserves estimates, accuracy improvement of the forecasts, and, as a consequent, increase in quality of decision making. In the framework of this study, hindcast has been carried out based on the history of 130 production wells of two waterflooded oil fields in order to assess reliability of the developed probabilistic methodology of production forecasting. In addition, the comparative study of developed method and foreign probabilistic methodology has been conducted on the basis of hindcast results.


1. Sazonov B.F., Sovershenstvovanie tekhnologii razrabotki neftyanykh mestorozhdeniy pri vodonapornom rezhime (Perfection of technology of development of oil deposits at a water-pressure mode), Moscow: Nedra Publ., 1973, 238 р.

2. Maksimov M.I., The method for calculating recoverable oil in the final stage of exploitation of oil reservoirs under oil displacement by water (In Russ.), Geologiya nefti i gaza, 1959, no. 3, pp. 42–47.

3. Burger J.G., Combarnous M., Thermal methods for hydrocarbon production, Journal Of Oil & Gas Science and Technology, IFP, 1975, V. 30, pp. 551–578.

4. Nazarov S.N., Sipachev N.V., Technique of forecasting technological indicators at the late stage of development of oil deposits (In Russ.), Izvestiya vuzov. Neft' i gaz, 1972, no. 10, pp. 41–46.

5. Sipachev N.V., Posevich A.G., Nazarov S.A. et al., Estimation of recoverable oil reserves based on integral curves of oil and water extraction (In Russ.), Azerbaydzhanskoe neftyanoe khozyaystvo, 1972, no. 5, pp. 20–21.

6. Gaysin D.K., Metod prognoza tekhnologicheskikh pokazateley i nefteotdachi plastov po promyslovym dannym v pozdney stadii razrabotki (Method for predicting technological indicators and oil recovery from reservoir data at a late stage of development), Proceedings of BashNIPIneft', 1986, V. 74, pp. 128–137.

7. Pirverdyan A.M., Nikitin P.I., Listengarten L.B., Danelyan M.G., On the forecast of oil and associated water production in the development of bedded heterogeneous reservoirs (In Russ.), Azerbaydzhanskoe neftyanoe khozyaystvo, 1970, no. 11.

8. Kambarov G.S., Almamedov D. G., Makhmudova T.Yu., Determining the initial recoverable reserves of oilfield (In Russ.), Azerbaydzhanskoe neftyanoe khozyaystvo, 1974, no. 3, pp. 22–24.

9. Abyzbaev I.I., Nasyrov G.T., On the factors affecting oil recovery in oil-water zones (In Russ.), Geologiya nefti i gaza, 1975, no. 2, pp. 60–63.

10. Gong X., Gonzalez R., McVay D., Bayesian probabilistic decline curve analysis quantifies shale gas reserves uncertainty, SPE-147588-MS, 2011, DOI: 10.2118/147588-ms.

11. Xie J., Efendiev Y., Datta-Gupta A., Uncertainty quantification in history matching of channelized reservoirs using Markov chain level set approaches,

SPE-141811-MS, 2011, DOI: 10.2118/141811-MS.

12. Liu C,  McVay D.A., Continuous reservoir simulation model updating and forecasting using a Markov chain Monte Carlo method, SPE-119197-MS, 2009.

13. Zolotukhin A.B., Frick T.P., A mobility driven fingering approach to the field-scale simulation of oil recovery, SPE-27017-MS, 1994, DOI: 10.2118/27017-MS.

14. Nazarenko M., Probabilistic production forecasting and reserves estimation in waterflooded oil reservoirs, SPE-192167-MS, 2018.

15. RD 153-39.0-110-01. Metodicheskie ukazaniya po geologo-promyslovomu analizu razrabotki neftyanykh i gazoneftyanykh mestorozhdeniy (Methodical instructions on geological and field analysis of oil and gas fields development), Moscow: Publ. of Ekspertneftegaz, 2002, 119 p.

To buy the complete text of article (a format - PDF) or to read the material which is in open access only the authorized visitors of the website can. .