Title : Prediction of rate decline by machine learning in bakken shale oil wells
Abstract:
Discrete mass-balance equations must be solved using commercial reservoir simulators. More grid blocks can be employed when the reservoir is heterogeneous and complex, which calls for precise reservoir data, like porosity and permeability that are usually unavailable in the field. It can consequently take hours or sometimes even days to predict the EUR (Estimated Ultimate Recovery) and rate decline for a single well, making them time-consuming and computationally expensive. In contrast, because decline curve models only need a few variables in the equation that can be easily obtained from the wells' most recent data, they are a simpler and faster choice. The publicly accessible databases of the Montana Board of Oil and Gas Conservation were used to collect the well data for this investigation. In a random oil field, well data set, the predictor parameters, and the SEDM (Stretched Exponential Decline Model) decline curve equation variables were correlated. The SEDM decline curve equation parameters were specifically created for unconventional reservoirs. The study looked at the relative weights of several well parameters. The original aspect of the study is the creation of a cutting-edge machine learning (ML) model based on a Support Vector Machine (SVM) for quick rate-decline and EUR prediction in Bakken Shale oil wells. The availability of a large, high-quality dataset is essential for the study's effective application.
Audience Take Away:
- The audience will learn how to use machine learning as an alternative to reservoir simulation and to overcome the complexities in traditional reservoir simulation problems also expand this study to the future to include more aspects and make the reservoir simulation process very computationally less expensive and rapid.
