Currently there is a trend in the oil and gas industry towards the deterioration of reserves due to which the number of fields which are significantly dominated by an unstable well stock is growing. Basically such assets include fields with high gas-oil ratios and fields with oil rims. The complicated fund can also include more complex operation of mechanized equipment, for example, in intermitted mode. Such regimes are used when it is inefficient to produce in a stable regime, for example, in fields with low permeability marginal sections.
In most fields, wells are well equipped with telemetry sensors. On wells with electrical submersible pumps (ESPs) dozens of parameters are recorded (electrical parameters of ESP operation, pressure and temperature in key system nodes). Wellhead parameters ( gas-lift gas injection parameters, pressure, temperature) at the gas-lift well stock are measured along the wellbore a little less often. The discreteness of measurements for most of the sensors reaches a minute. All this leads to a large daily accumulation of information about the work of the well. Often not all of the information is used in daily work. This is primarily due to the inability to process such an amount of information manually, without the use of digital approaches. The analysis is carried out on averaged data for the day, missing useful information about the daily operation of the well, reducing the efficiency of decisions made.
The task of an unstable stock analysis is most critical since it is impossible to get a qualitative understanding of the operation of such a stock without looking at the level of current discretization of key well performance parameters. Therefore it is impossible to solve the problem using standard methods based on adapting models to daily measurements and even more so to monthly data.
This article presents the experience of solving the problem for an unstable artificial lift well which provides a compromise between the accuracy of the results and the complexity of modeling non-stationary processes. This approach allows getting the result relatively quickly, while maintaining the consistency of all parameters with each other.
References
1. Хабибуллин Р.А., Бурцев Я.А., New approach for gas lift optimization calculations (In Russ.), SPE-176668-RU, 2015, DOI: http://doi.org/10.2118/176668-MS
2. Moitra S.K., Chand S., Barua S. et al., A fieldwide integrated production model and asset management system for the Mumbai High field, Proceedings of Offshore Technology Conference, 30 April-3 May, 2007, Houston, Texas, DOI: http:// doi.org/10.4043/18678-MS
3. Bishop C.M., Pattern recognition and machine learning. - New York: Springer, 2006.
4. Bingham E., Chen J.P., Jankowiak M. et al., Pyro: Deep universal probabilistic programming, Journal of Machine Learning Research, 2018, DOI: https://doi.org/10.48550/arXiv.1810.09538
5. Андрианова А.М., Маргарит А.С., Перец Д.С., Симонов М.В., Hierarchy of data verification approaches for production and development control (In Russ.), Нефтяное хозяйство = Oil Industry, 2017, no. 12, pp. 75-77, DOI: https://doi.org/10.24887/0028-2448-2017-12-75-77
6. Lake L.W., Petroleum engineering handbook - Texas; Austin, 2010.
7. Vazquez M., Beggs, H.D., Correlations for fluid physical property prediction, SPE-6719-MS, 1977, DOI: https://doi.org/10.2118/6719-MS
8. Andriasov R.S., Mishchenko I.T., Petrov A.I. et al., Spravochnoe rukovodstvo po proektirovaniyu razrabotki i ekspluatatsii neftyanykh mestorozhdeniy. Dobycha nefti (Reference guide for the design, development and operation of oil fields. Oil production): edited by Gimatudinov Sh.K., Moscow: Nedra Publ., 1983, 455 p.
