Production Forecasting for the Duvernay Shale: Comparing Analytical and Machine Learning Methods

Date
2021-11
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
With the advance in drilling horizontal wells and hydraulic fracturing, shale production has increased dramatically in the past two decades. Shale formations are characterized by natural fractures, low-permeability rocks, and nanopores associated with organic matter, clay minerals, carbonate, and various silts. In shale, gas content is mainly composed of free gas in the nanopores, adsorbed gas in the rocks, and other components such as carbon dioxide, nitrogen, and hydrogen sulfide. The co-existence of these components with different sizes and properties has posed a significant challenge for investigating shale. Due to complex transport mechanisms and the existence of natural and hydraulic fracture networks, building reasonable reservoir simulation models and production forecasting is not entirely feasible with the existing simulators. Still, it is a challenging task for petroleum engineers. While simulation methods, including analytical models, semi-analytical models, and numerical simulation, could lead to reasonable production forecast, they are expensive and time-consuming and require many data such as well test data and well logs which are not always accessible. In contrast, Decline Curve Analysis (DCA) has the advantages of simplicity and efficiency for hydrocarbon production forecasting and mainly requires production data. In the past years, different DCA models have been proposed and benchmarked to meet industrial needs. DCA models have shown to be efficient for conventional reservoirs. When applied to shale wells, DCA has many shortcomings. Several authors have tried to overcome some of the well-known shortcomings of DCA. Nevertheless, many facts remain that make the use of DCA suboptimal. One of the main criticisms of DCA is its lack of sensitivity to major physical phenomena in shale wells that have to do with the fluid flow, the hydraulic fracture, and the reservoir characteristics. In cases like Duvernay shale reservoirs where short production periods are available, the use of DCA becomes increasingly problematic. In this study, different methods of production forecasting, including reservoir simulation, decline curve analysis, machine learning, and time series will be reviewed, and their limitations will be discussed. Then these methods have been used to predict and forecast gas production in Duvernay play in different scales of wells and fields.
Description
Keywords
Production Forecasting, Duvernay, Machine Learning, Time Series, Simulation
Citation
Malaieri, M. (2021). Production forecasting for the Duvernay shale: comparing analytical and machine learning methods (Master's thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca.