IMPROVING THE QUALITY OF DRILLING DATA. PhD thesis, King Fahd University of Petroleum and Minerals.

PDF (Salem AlGharbi -Dissertation) Salem AlGharbi -Dissertation_v_1000_FINAL.pdf - Accepted Version Restricted to Repository staff only until 25 October 2023. Download (17MB)

Arabic Abstract

رفع كفاءة البيانات المباشرة لعمليات الحفر بسبب الزيادة في الطلب العالمي على النفط و الطاقة, قامت شركات البترول بالعديد من المبادرات لتلبية هذا الطلب, منها الحفر في مناطق نائية, تحت ظروف غير إعتيادية. مما أدي ظهور تحديات جديدة أمام عمليات الحفر و اللتى في أصلها عمليات مليئة بالتحديات. للمساعدة في هذا الأمر, تم إستخدام تطبيقات الذكاء الإصطناعي و تعلم الآلة لمراقبة بيانات الحفر و توفير ملاحظات دقيقة على آلية العمل. للأسف بسبب طبيعة و بيئة العمل القاسية, فإن آلات رصد هذة العمليات قد تصاب بالعطب و تتنتج بيانات غير دقيقة, فلا تستطيع تطبيقات الذكاء الإصطناعي و تعلم الآلة العمل بصورة جيدة. فكانت البيانات عائقا أمام الإستفادة المثلي من هذه التطبيقات. لحل هذه المعضلة, تم طرح العديد من الحلول, كإستخدام أجهزة أحدث, قوائم الآداء, بناء مركز مراقبة البيانات. و هي كلها حلول منطقية, لكنها تستلزم الكثير من الوقت و الجهد و المال, و تعاون عدة جهات مختلفة لكي تتحول إلي خانة التطبيق. رسالة الدكتوراه التي بين يديك تطرح نظره مختلفة لحل هذه المعضلة, بإيجاد نموذج عمل مستخدما تقنيات الذكاء الاصطناعي لزيادة كفاءة البيانات المباشرة لعمليات الحفر. هذا النموذج سيفتح المجال لزيادة عدد تطبيقات الذكاء الإصطناعي و تعلم الآلة في مجال الحفر, و الذي بدورة سيؤدي إلي زيادة سلامة عمليات حفر و إرتفاع كفاءتها.

English Abstract

To ensure a sustainable level of energy supply, the oil and gas industry has made multiple efforts to increase production. One of these drilling in remote areas and in unconventional conditions. This has introduced additional challenges to the drilling operations, which are, by their nature, full of complexity and uncontrollable factors. Due to this complexity level, it become important to keep close monitoring of the live activities. It worth to highlight that these activities produced big column of many data types in a very fast rate it become difficult to human capabilities to follow. Since artificial intelligence and machine learning has a capability to process massive amount of data converting raw data to information, it become essential to capitalize on this technology to convert the drilling raw data to decision-like support systems. Bring attention to the harsh environment of drilling operation, and how it negatively affected the sensors collecting data, and the transmission systems, add to that, the site location (offshore, or remote area) limited the maintenance and calibration services to such sensors and systems, such situation lead to dropping, adding or altering data points from the dataset, introducing a drilling data quality issue. This situation is note new, drilling engineers and operation crew were used to facing such drilling data quality challenges, but based on their experience and analytical skills, they were able to do on-the-fly processing, remove all the noise, and provide the needed decision. Unfortunately, this is not the case with artificial intelligence and machine learning. Such techniques base their function on the input data. If the data quality is low, the output quality will be low, based on the garbage-in garbage-out principle, leading to inefficient decisions. Such decisions could even jeapordise the entire drilling operation. As a result, data quality is an important factor that needs to be addressed. Improving data quality, in general, is a challenging task. It requires sector specific business knowledge, as well as statistical knowledge. Moreover, development skills are needed to carry out the needed fixes in a timely manner. Despite the great need for such skills in the drilling domain, they are still limited, leading to many vacancies in this area. There are some efforts to fill this gap, but they mostly need time and effort, are relevant to future data points only, and will not address data already collected. This work addresses the challenge and solves it by introducing a process to improve the drilling data quality. It starts by introducing the challenges in drilling and the important of utilizing the AI technology in this challenging domain. Then explore previous work toward improving the quality of the drilling data, including their limitation and gaps. After that more insight about the drilling data will be listed, followed by the current practice in the area of data quality improvement. The work will show the important of having high quality drilling data on drilling-AI models. Then will evaluate multiple methods of data quality improvement techniques, mainly in the statistical and AI fields. Based on that, the work propos a new approach, and ways to increase the efficiency. This work succeeded in reducing the drilling data quality from several days to few minutes with accuracy reaching 95%. This approach will help the industry to improve drilling data quality, setting the floor for rabid integration for more artificial intelligence and machine learning models to achieve safer, faster and more cost effective operations. It is worth emphasizing, that this work is one of very limited number researches targeting the challenging domain.

Item Type:	Thesis (PhD)
Subjects:	Computer Research > Petroleum Petroleum > Drilling Engineering
Department:	College of Petroleum Engineering and Geosciences > Petroleum Engineering
Committee Advisor:	Abdulraheem, Abdulazeez
Committee Members:	Patil, Shirish and Ahmed, Moataz and Al-Nuaim, Sami
Depositing User:	SALEM SALEM AL-GHARBI (g199819730)
Date Deposited:	25 Oct 2022 07:54
Last Modified:	25 Oct 2022 07:54
URI:	http://eprints.kfupm.edu.sa/id/eprint/142222