Back to search

NAERINGSPH-Nærings-phd

Optimization of Tripping IN/OUT in Drilling Operations using Machine Learning

Alternative title: Optimalisering av tripping IN / OUT i boreoperasjoner ved bruk av maskinlæring

Awarded: NOK 2.4 mill.

Project Manager:

Project Number:

322794

Application Type:

Project Period:

2020 - 2024

Funding received from:

The project made an official request for the actual tripping in, tripping out, and drilling operation data to Equinor in March 2021. Equinor agreed to facilitate the research with the real-time data from the new source data while tripping tool (NOV DWT) and wired drill pipe telemetry system (NOV Wired Drill Pipe Telemetry Network). Equinor also agreed to provide the downhole memory data and the other relevant information needed for the project. The actual tripping in, tripping out, and drilling operations data are the basis of this research project. These field data belong to one of the fields in the Barents Sea, Norway. In this way, actual field problems relating to tripping in and tripping out encountered by Equinor can investigate, and potential data-driven solutions can be delivered. Moreover, this further develops a solid relationship between academics and industry by finding new solutions to drilling and well construction operations challenges. We have established another collaboration with NORCE, which is one of the largest research institutes in Norway. In Norce several projects are going on drilling automation. We presented the Ph.D. research project to several other companies for the research collaboration and we received positive responses from the companies. The possibilities of sharing knowledge and ideas with other researchers from the academics and E&P and drilling services industry will enhance learning and opportunities for further collaboration in other projects within drilling and well construction. The project aims to implement data-driven and Physics-based modelling techniques to learn the insight of downhole surge and swab pressures problems while utilizing the actual field data. Several researchers implemented data-driven methodologies to find new solutions within ROP modelling, reservoir management, drilling optimization, data processing and have already achieved promising results. The expected outcome of the ongoing PhD research will provide an insight into the swab and surge problem relating to tripping in and tripping out operations and new solutions to the problem. The aim results from the research will improve the performance of tripping in, tripping out process within drilling and well construction operations and will further reduce the risk related to swab and surge pressures. The project's current phase is developing the algorithms for the automated processing of tripping in, tripping out, and drilling data of an actual well. Wrote an algorithm in Python to segregate the five days data from the actual field into tripping in, tripping out, and drilling. The results obtained from the algorithm were tested against real-time logs from the well and found good agreements. During the tripping in and tripping out process in drilling operations, E&P companies are implementing the Data While Tripping tool to obtain internal, annular Pressure, and temperature while using Pressure While Drilling and Enhance Measurement System. However, Pressure While Drilling Tool and Enhance Measurement Sytems are from two different vendors, and the sensors mounted on these tools are of varying quality. Therefore, it is essential to determine the agreement between the measurements of the same variables from different methods. To investigate this as part of data processing, we examined different methodologies. We discussed this problem with Prof. of Statistic at the department of Mathematics and Statistics (Jan Terje Kvaløy). He suggested that we implement the Bland Altman Plot Analysis theory utilized in health sciences to analyze the measurements of the same variables from different methods. We implemented the Bland Altman Plot Analysis theory on the Pressure While Drilling and Enhance Measurement System measurements obtained via data while tripping tool through wired drill pipe telemetry system. The results obtained from the Bland Altman Plot showed that when the mean is higher, the difference between the measurements from the two sensors is lower, and when their mean is lower, the difference between the measurements is higher. Summary of meetings with Equinor, NORCE In March 2021 arranged an online meeting with Equinor about requesting the actual field data during tripping in, tripping out, and drilling data from the new source NOV Data While Tripping and wired drill pipe telemetry network. Presented the project proposal to Equinor, which received positive feedback, and Equinor showed great interest in the project. Therefore, they agreed to facilitate the research with the requested data and other relevant information. In August, an online meeting took place with NORCE, presenting the project proposal in this meeting. This meeting intended to establish a working relationship with NORCE, and utilize OpenLab (a drilling simulator) for the analytical modelling. The meeting resulted to establish a working relationship and sharing results and knowledge.

-

Method 1) Database Development: A database will be developed to manage the enormous data from high-frequency telemetry (wired drill pipe data) [M. Reeves et. al. 2006]. Today, information is sent from the near bit area of the drill string via mud-pulse telemetry. The signal rate for this form of telemetry is quite low. There are several drawbacks with this form of data transmission in addition to a low data rate: • Poor data quality • Limited number of sensors (a point in time and/or a point in the well) • Communication is unavailable under tripping/connection Wired drill pipe (WDP) [M. Reeves et. al. 2006], developed by NOV can greatly reduce the mentioned problems. Wired drill pipe breaks these constraints by having effectively continuous and multiple sensor measurements along the length of the wellbore [Sanna Z. et. al. 2016]. It has been long recognized that bad data quality is hampering our attempts to make use of the drilling data [Dan Sui, et. al. 2018]. Bad data obstruct integrated planning, burden collaborative environments and hamper workflows. Consistent data quality represents a challenge for the industry and represents a prerequisite for quality decision support. Both technological and systemic challenges must be addressed. If not, bad data quality will remain a barrier to safe and efficient drilling. Method 2) Validation of WDP Data: Develop a method for the validation of the quality of the data while tripping data. Method 3) Quality of the sensors and sensor error will be addressed. Machine learning enables identifying the key variables from thousands of attributes to minimize the noise and errors in the predictive models and reveal hidden relationships between dependent and independent variables. A data-driven control system will be developed while using deep learning & machine learning techniques to automate the tripping in and tripping out of hole process.

Funding scheme:

NAERINGSPH-Nærings-phd