Abstract
For years, the oil and gas industry has been utilizing real-time sensors to monitor multiple parameters. In drilling operations, these sensors are highly important in achieving the well objectives. They provide data to guide the well execution and aid during the evaluation phase. The downside is the limited capability to identify operational anomalies. The latest popular strategy to mitigate this downside is using a Decision Support Center to obtain the most benefit from this data. Decision Support Center should be available for 24-hour operations and consists of multiple personnel to review data, provide path forward recommendation, and alert the field supervisor.
During these past years, most companies have optimized the number personnel to keep competitive in a low oil price environment. However, some knowledge has been lost during the process. On the other hand, the industry worldwide is entering an era of digital transformation. This triggers the need for artificial intelligence systems that can translate analog knowledge from personnel to digital data, analyze for execution anomalies from sensor readings, and provide feedback and recommendations to the end-users.
In one of largest mature onshore Indonesian blocks, a pilot implementation of artificial intelligence assisted Decision Support Center was developed. With limited time and cost, the project needed to be developed as fit-for-purpose. Thus, the system focused on the significant problem which historically caused million dollars in lost opportunity in the form of stuck pipe. A single stuck pipe event could result in a US$ 2.5 MM loss, which is more than 2 times the typical well AFE in this block. Beside surface and downhole sensors readings, subsurface prognosis, torque and drag model; operational best practices were also included in the system. Anomaly alarms were developed, including ECD limit by depth, pick up weight limit by depth, torque limit by depth, potential loss circulation zone, maximum allowable static time, and maximum allowable pumps off time. Whenever the sensor reading exceeds these limits, an automatic alarm would be generated and sent to the field supervisor, drilling superintendent and drilling engineer. As result, it only required 1 personnel per 4 rigs to maintain the process compare to 3-4 personnel per rig required with the conventional Decision Support Center process. This process successfully eliminated stuck pipe events though to the remaining of the drilling campaign of 56 wells, where the historical frequency had been 1 stuck pipe event out of 20 wells.
Future improvement in this system will utilize machine learning to develop anomaly alarms by gathering data from offset well operations. An additional benefit of this system is to implement it in workover operations.
