Guest Editorial: Unlocking Unconventional Reservoirs With Data Analytics, Machine Learning, and Artificial Intelligence
- Dan Fu (BJ Services)
- Document ID
- Society of Petroleum Engineers
- Journal of Petroleum Technology
- Publication Date
- January 2019
- Document Type
- Journal Paper
- 14 - 15
- 2019. Copyright is retained by the author. This document is distributed by SPE with the permission of the author. Contact the author for permission to use material from this document.
- 13 in the last 30 days
- 120 since 2007
- Show more detail
- View rights & permissions
|SPE Member Price:||Free|
|SPE Non-Member Price:||USD 4.00|
The concept of the digital oil field has been around for many years and the industry is beginning to experience the benefits of data analytics, machine learning, and artificial intelligence (AI).
The upside of digital technology is it can improve operational efficiencies and reduce costs. The downside of digitization is that many companies have neither the resources nor the expertise to collect, analyze, and interpret a critical mass of data while maximizing its economic return.
Predictive Behavior for Unconventional Reservoirs
As North America shale plays are once again becoming economically viable, operators in all major basins have fast-adopted best practices to optimize drilling and completion processes to drive down the lifting costs.
Adoption of data-driven analytics to improve completion design, drive efficiency, and yield economic gains has been less swift. In years past, an operator would perform analysis on a single well completion and make decisions using publicly and privately available data, well logs, or sometimes nothing more than experience gained through the years. As the industry moves to more complex multipad, multiwell completion designs, intelligent completion optimization will require more sophisticated algorithms to improve decision making. The question remains as to how operators can use the vast amount of data readily available in the public domain and private data to develop and train models that enable them to realize these operational efficiencies and increase estimated ultimate recoveries.
This is where the use of analytics and AI can directly drive efficiencies to unlock the production potential of unconventional reservoirs specifically during hydraulic fracturing. The growing volume and availability of completion and production public data creates a revolutionary situation in which pad planning and completion design becomes possible with pure AI methods.
Benefits of AI-Driven Analysis
There are two main avenues where digital technologies are expected to make the largest impact: automation and optimization.
With predictive analysis, the industry will be able to automate and optimize the process of completion design to mitigate completion disfunctions such as frac hits or uneven cluster performance.
The inputs can come from quality data curated from wellsite records that include cluster spacing, pressure and rates, volumes pumped, fluids and sands used, diversion and flowback practices, production data, borehole assemblies, analyzed offset data, geology data, the plan of the well itself, and the offset wells that set up the pad development plan.
The most desired improvement of including AI into completion design is speed of the entire analysis and modeling process. The trained AI system does 6 to 8 weeks of manual research in seconds, identifying specific design hazards buried within amassed pages of completion and production reports. This accelerated tempo ultimately leads to reduction of nonproductive time.
Numerous completion parameters can be changed or improved using computational intelligence and mathematical optimization including stimulation volume, fluid types, flow rate, well later-al and cluster spacing, as well as numerous other considerations.
Optimization on these parameters must always be done coupled with economics. Long- or short-term goals will steer decisions to different completion designs. For example, with economics taken into account, the objective of avoiding frac hits at all cost may not be necessary. A trained AI model that predicts type curve and optimizes the cost function provided by economic objectives could discern that a certain level of risk of frac hits is acceptable and even beneficial.
|File Size||1 MB||Number of Pages||2|