Decision Support System for Tight Oil Fields Development AchimovDeposits and Their Analogues Using Machine Learning Algorithms

摘要

The aim of this work is to develop an approach to multivariate optimization of development systems for tightoil reservoirs of the Achimov formation,where large volumes of drilling of RN-Yuganskneftegaz LLC arecurrently concentrated on.The approach described in the paper is an integral part of the corporate module"Decision Support System for drilling out new sections of tight oil reservoirs",which allows making quickdesign decisions for new drilling sites of target objects.This work discusses the main parts of the integrated solution of this system that will be embedded intocorporate software.Also,the description of the global approach and obtained results are presented.The main idea of thisproject is based on automatic assignment of the prospective development zone to an existing cluster-analog,based on well logs response in exploration wells.Following this interpretation,the potential performanceof various development systems is evaluated and the optimal one is selected.Within the framework of these projects the following tasks were solved:1.Wells clustering in Achimov deposits and their analogs.The geological heterogeneity and reservoirconnectivity were characterized and a special algorithm for wells assignments to an existing clusterwas developed,that is done by:a.Wells clustering depending on their petrophysical properties derived from well logs interpretationvia k-means algorithm.b.Wells classification with a use of neural network.2.Multivariate 3D dynamic modeling and creation of surrogate models to provide predictions ofreservoir simulation results.3.Development of the software package with all mentioned functionality being implemented.