The article examines the issues of predicting the accumulated performance indicators for horizontal wells with multi-stage hydraulic fracturing. These wells penetrate the Bazhenov formation in the Palyanovskaya area of the Krasnoleninskoye field. This form has very low reservoir properties (permeability of the order of 0.001⋅10-3 μm2), according to the estimates obtained by specialists. This fact creates a number of difficulties in the operation of wells and in the modeling of the drainage processes of the target object and the calculation of forecasts.
The work carried out a comprehensive analysis of the accumulated data on well productivity indicators, technological indicators (features of construction and stimulation of wells) and geological indicators (geomechanical indicators of the target object). The current status of the development of the Palyanovskaya area of the Krasnoleninskoye field determines the type of task as data analysis on small samples. As a result of the analysis, statistical models were built to quantify performance indicators from the data. These data characterize the features of well construction and stimulation. This paper presents a methodology for preparing the initial data for analysis, selection of assessment tools and assessment of the significativity of the found dependencies. Finally, a probabilistic forecast of production rates for new wells was performed based on the constructed statistical models. The calculated forecast confirmed the choice of the direction of technological improvement of design. Cluster analysis of production indicators, correlation analysis of production indicators and technological indicators, analysis of regression models of production indicators and Monte Carlo sampling are used to build a forecast. The calculated indicators demonstrate a multiple increase of productivity compared to the average statistical indicators of previously drilled wells.
References
1. Plavinskiy S.L., Rabota s malymi vyborkami. Statisticheskiy analiz dannykh s propushchennymi nablyudeniyami (Working with small samples. Statistical analysis of missing observations), URL: https://www.youtube.com/watch?v=YeYs6GfNv-0
2. Dealing with very small datasets, URL: https://www.kaggle.com/rafjaa/dealing-with-very-small-datasets
3. Choosing the right estimator, URL: https://scikit-learn.org/stable/tutorial/machine_learning_map/index.html
4. Kobzar' A.I., Prikladnaya matematicheskaya statistika. Dlya inzhenerov i nauchnykh rabotnikov (Applied mathematical statistics. For engineers and scientists), Moscow: Fizmatlit Publ., 2012, 813 p.
5. Lapach S.N., Radchenko S.G., Regression analysis in conditions of heterogeneity of factor space (In Russ.), Matematichnі mashini і sistemi, 2016, no. 3, pp. 55–63.
6. Zubkov A.F., Derkachenko V.N., Barmin M.A., Cluster and discriminant analysis of regional insurance markets (In Russ.), Nauchno-tekhnicheskie vedomosti Sankt-Peterburgskogo gosudarstvennogo politekhnicheskogo universiteta. Informatika. Telekommunikatsii. Upravlenie, 2012, no. 1(140), pp. 113-118.
7. Sivogolovko E.V., Methods for assessing the quality of clear clustering (In Russ.), Komp'yuternye instrumenty v obrazovanii, 2011, no. 4, pp. 14-31.
8. Kim J.-O., Mueller C.W., Klekka W.R. et al., Faktornyy diskriminantnyy i klasternyy analiz (Factor, discriminant, and cluster analysis) [Russian translation], Moscow: Finansy i Statistika Publ., 1989, 215 p.
9. Tarantola A., Inverse problem theory and methods for model parameter estimation, Society for Industrial and Applied Mathematics, 2005, 352 r., https://doi.org/10.1137/1.9780898717921
