1. Ensemble methods for process monitoring in oil and gas industry operations
- Author
-
Giulio Gola, Dan Sui, Davide Roverso, Mario Hoffmann, and Roar Nybø
- Subjects
Engineering ,Computer program ,Operating environment ,business.industry ,Process (computing) ,Energy Engineering and Power Technology ,Geotechnical Engineering and Engineering Geology ,computer.software_genre ,Ensemble learning ,Reliability engineering ,Fuel Technology ,Robustness (computer science) ,Data quality ,Probabilistic forecasting ,Data mining ,business ,computer ,Reliability (statistics) - Abstract
Complex operations carried out in the oil and gas industry such as drilling require constant and accurate real-time monitoring of the process. To this aim, a real-time model of the drilling operation is required. Such a model is used to estimate the state of the well when and where direct and reliable measurements are not available and it helps the driller gain an overview of the drilling process. Given the harsh operating environment, sensor reliability and sensor calibration are known problem areas, and bad data quality is a common problem, affecting the accuracy of the model. As a result, the driller may be misled about the downhole situation or receive conflicting claims about operating conditions. A way to reduce uncertainty and increase confidence is to aggregate the opinion of different experts. When the expert is a computer program, such aggregation is often referred to as an ensemble approach. The principle underlies techniques that have become popular in the oil industry in recent years, such as probabilistic forecasting and ensemble Kalman filters. In this paper, we discuss this trend and develop an ensemble system for predicting the bottom-hole pressure during a managed pressure drilling operation. The improved accuracy and robustness of the ensemble approach in situations with bad data quality is demonstrated.
- Published
- 2011