Production Well Placement and History Matching by Hyperparametric Optimization and Machine Learning
摘要
Abstract
This article explores methods for resolving two practical issues in reservoir engineering: discovering an optimal position for the production well on a grid, and recovering unknown fracture position based on the historical production data. To speed up the conventional approach (numerical simulation and the full grid search), we assess application of optimization techniques from the Hyperopt package as well as several machine learning models.