The availability of reliable data is obviously a prerequisite to good production optimization. So far, many tools have been developed to increase the availability of data by using Smart Completions, Downhole Measurements and Remote control Sensors installed along the production chain. The progress in production models has also contributed to this improvement. Nevertheless, in term of reliability, the collected data are affected by uncertainties, which must be taken into account.

The process of Data Validation and Reconciliation (DVR) considers all the relevant information (measures/ models) of the production plant as one integral set of parameters. Furthermore, the originality of DVR as compared to equivalent systems is the use of a statistical model to control uncertainties associated with each parameter, and calculating the resulting error propagation. As results and thanks to the information redundancy, DVR allows an automatic real-time correction of each parameter based on its allocated uncertainty.

In this paper, we focused on a real time application with 16 Wells belonging to 3 offshore Platforms. The single-Well model is initially validated by using the Multiphase Flow Meter (MPFM) measurements.

The global results analysis is discussed by comparing the online application outputs with production tests using MPFM, in terms of Oil flow-rate and Water Liquid Ratio (WLR). Also, the monitoring of each information is highlighted in order to detect a model deviation or instruments failure. Furthermore, the error distribution is used to identify the impact of any deviation occurring in the application.

Conclusions are drawn on the performance of the virtual metering system in terms of online availability, results reliability and global consistency between data and results. The DVR application is considered as an operational monitoring tool thanks to alarms which are customized by operators, and as an analysis tool for investigation and diagnosis studies.

You can access this article if you purchase or spend a download.