Abstract
Data analysis and event alarms are essential part of any production monitoring system. Most often standard event alarms in production monitoring are based on selected measurements. Oil rate, pump intake pressure and temperature, water cut are some of them. Less often alarms are based on calculated values, for example production increase potential. However, they always indicate something goes wrong but never indicate why. Alarms can indicate there are deviations from planned values but do not give any clue about deviation reasons. Production engineer should investigate deviation reasons himself.
From author's point of view, smart alarms are the tool to overcome these challenges. Smart alarms in production monitoring system should be able to determine reasons of parameters deviation and to predict deviations before they become critical.
Smart alarms should use all the information available in order to increase estimation and prediction accuracy. They should classify and rank problems and problem reasons identified from potential future problems to problems required immediate reactions. Smart alarms should give reasonable results based on partial data, and should be able to analyze thousands wells per day.
Primary goal for smart alarm tool is to automate business process for 80% of well stock working in a standard way and to allow engineers to concentrate on 20% of complicated wells.
Smart alarm tool based on Bayesian network framework is under development and pilot implementation in TNK BP of Samotlorskoe oil field. Smart alarm tool is a part of corporate production monitoring system. This paper discusses Bayesian network application for engineering tool development and results of pilot implementation including algorithm accuracy, lessons learned and development plans.