Abstract
The Litoral de Tabasco asset, in PEMEX's southwest marine region, was initially implemented to monitor and estimate the flow rate of naturally flowing wells, using daily operational conditions and well tests with the so-called Gilbert and Ashford and Pierce correlations.
The effectiveness of these correlations has now been reduced as a result of irregularities in operational conditions; lack of consistency in conducting the production well tests; limited availability of measurement and operational issues with mass, conventional, and multiphase flow meters; and ongoing problems with electronics communication, nonavailability of vessels, and high operational costs. If the operational conditions measured during the well tests were wrong, all the properties are suspect and may have affected the correlations developed in determining the estimated volume per well and associated monitoring workflow.
To address this problem, a new workflow using a set of predictive proxy models has been developed that combines artificial intelligence techniques such as neural networks with nodal analysis, and sporadic and high-frequency data to allow engineers to process and understand production behavior from the large amounts of information available, which are gathered according to the system being studied or evaluated.
This workflow enables validation of field well test data, which reduces the uncertainties in well production allocation, increases the accuracy of hydrocarbon accounting from the pumping process to the marine terminal, and implements an early detection system for anomalies that is published on the Internet for sharing with the entire asset management.
Now, when it is not possible to measure well production, the outputs of a proxy model and a nodal analysis are combined with sporadic and real-time data to reconstruct the historical and actual volume on a daily basis.