Machine Learning Application to Predict the Efficiency of Water Coning Prevention Techniques Implementation

The increase in number of the mature fields is accompanied by an increase in the water cut of the produced fluids. One of the most common causes of this phenomenon is the process of water coning, that is, the breakthrough of the bottom water to the wellbore, in which water flows form a figure similar to a cone. The paper proposes a ranking mechanism based on machine learning methods that allow to significantly reduce the resource intensity of existing prediction models. In order to preserve the simplicity of presentation, the proposed mechanism is considered on the example of one technology - DWL. Obtained results show about 10% smaller deviation values when using the least squares support vector machine in comparison with the ANN. Both developed models demonstrated acceptable results for practical application. (C) 2021 "OilGasScientificResearchProject" Institute. All rights reserved.