A well is in natural flowing state when its bottom-hole pressure is enough to produce to the surface. Natural flowing well’s production is regulated by using surface restrictions to regulate the production rate in such a way that the overall well performance is a function of several variables. Examples of these variables are tubing size, choke size, wellhead pressure, flow line size, and perforation density. This implies that changing any of these variables will modify well performance. One of the techniques for the analysis of production performsnce is studying the wellhead pressure declination, since, in critical flow conditions, flow is a function of wellhead pressure. From wellhead pressure trends you can identify the behavior of each well and determine some issues, such as: choke erosion due to sand production, choke o tubing paraffin plugging or choke obstruction. In order to achieve an effective real-time monitoring of this type of wells, and in this way reduce the production losses, the challenge was to create online tools that could detect those mentioned issues.
The present work performs the analysis of wellhead pressure curves using data science, with the purpose of predicting real time anomalies that could occur for timely correction. The data correspond to 130 flowing wells from the Loma Campana Field. The study began with a filtering process of the pressure curve, with two specific objectives: first, eliminate atypical values from the time series, and second, smooth the curve in such a way that future predictions can be performed. Next, the Prophet methodology was applied with the purpose of predicting values of the curve. This is based on historicsl values of the time series to predict future values; the trend characteristic of the curve was used to apply this methodology. Then, to identify the anomaly a model was designed based on the declination of the curve. The pressure declination curve is a descending exponential function, so the first and second derivatives indicate the trend (ascending - descending) and curvature (concave or convex) of it. Once these values are available, they are classified according to the anomaly: paraffin, encrustation or obstruction. Finally, the model is being tested in the Loma Campana control room, delivering a probability of occurrence of any anomalies every hour.
Gilbert, W.E., (1954). "Flowing and Gas-lift Well Performance". N.V. De Bataafsche Petroleum Maatschippj. 54-126 API Conference Paper.Grubbs, F. E. (1969). "Procedures for detecting outlying observations in samples". Technometrics. 11 (1): 1–21. doi:10.1080/00401706.1969.10490657Hedong, S (2015). Advanced Production Decline Analysis and Application. doi: https://doi.org/10.1016/C2014-0-01693-0 Loader, C. (1999). Local Regression and Likelihood. Sringer: New York Shumway, Robert. H. and Stoffer, D. (1991). Time Series Analysis and Its Applications with R Examples, Third edition. Springer New York Dordrecht Heidelberg London.Taylor SJ, Letham B. (2017). Forecasting at scale. PeerJ Preprints 5: e3190v2 https://doi.org/10.7287/peerj.preprints.3190v2 Tukey, John W (1977). Exploratory Data Analysis. Addison-Wesley. ISBN 978-0-201-07616-5. OCLC 3058187 Desarrollo Operativo Loma Campana YPF. S.A. Carga de Controles con Fórmula (March 2015).J. I. Alvarez Claramunt, P. E. Bizzotto, F. Sapag, E. Ferrigno, J. L. Barros, and G. Martinez, YPF. ( August 2018). "Cultural change in ALS Real Time Operation and Optimization in Vaca Muerta Through CWE". SPE-190963-MS.
