Integration of Real-time Data and Past Experiences for Reducing Operational Problems

摘要

Oil well drilling operation is a complex process, in which there are always new lessons to be learned during drilling operation. A case-based reasoning (CBR) system is used to provide an intelligent advisory system based on previous experiences. Whenever the process is running smoothly, or is failing, the experiences gained during such episodes are valuable and should be stored for later re-use and prediction. Contributing features are selected for characterizing an episode (case) that is compared to other cases in a case database. In this way, cases are classifed into case classes according to the similarity between a new case and other cases. As most problems during drilling operation are depth dependant, the system keeps all the cases and experiences in each defined depth interval to compose sequences of cases. Our aim is to implement the sequence building for the purpose of predicting problems before they occur. To meet this goal, each sequence is composed of previous, present and next case. The paper presents a methodology of a semi-automatic case building and case discrimination process to make a robust sequence-based reasoning system. Our approach addresses the task of developing an intelligent system for prediction through sequeces. A CBR platform, developped in our group, was used and employed in the oil well drilling domain. The methodology was applied in one well section in the North Sea and the methodology clearly showed its ability through the good results which were obtained.