Production Forecasting in Unconventional Reservoirs: A Workflow for Data-Driven Analysis
Date
2024-03-13
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The oil and gas industry has experienced a significant transformation in recent years, with the use of advanced data analytics and machine learning (ML) techniques. These innovations have opened new opportunities in reservoir engineering, allowing engineers to make data-driven decisions and optimize well completion parameters. The accuracy of forecasts in predicting hydrocarbon production from unconventional reservoirs has become very important, as it directly affects energy security, economic growth, and sustainable resource management. This thesis presents a novel and comprehensive automated multistep workflow that includes data collection, data preparation, feature selection, hyperparameter tuning, ML algorithm selection, and well completion parameter optimization. In the workflow, we also propose novel methods for outlier detection, feature selection, and a modified optimization algorithm. To verify the accuracy of the workflow, we applied it to various synthetic and field databases and considered different objective functions, such as oil and gas cumulative production and monthly/daily production rates. The proposed integrated workflow enhanced the accuracy and reliability of production forecasting, as well as the assessment and improvement of well performance in unconventional formations. It also helped us determine how geology, reservoir characteristics, and completion designs influence production, and how to adjust them to achieve better outcomes. Moreover, we used the workflow to find the best completion design and estimate the production loss due to sub-optimal completion practices for the wells in the unconventional plays. These results of the research offer valuable insights for the stakeholders in the energy sector who operate in unconventional resources to make informed decisions about well completions, field development, resource extraction, and operational costs.
Description
Keywords
Optimization, Machine Learning, Unconventional Reservoirs, Differential Evolution Algorithm, Data Analytics, Gas Production Forecast, Deep Learning, Time Series
Citation
Rahmanifard, H. (2024). Production forecasting in unconventional reservoirs: a workflow for data-driven analysis (Doctoral thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca.