Optimization of the Reservoir Management System and the ESP Operation Control Process by Means of Machine Learning on the Oilfields of Salym Petroleum Development N.V.
{"title":"Optimization of the Reservoir Management System and the ESP Operation Control Process by Means of Machine Learning on the Oilfields of Salym Petroleum Development N.V.","authors":"A. Musorina, Grigory Sergeyevich Ishimbayev","doi":"10.2118/206518-ms","DOIUrl":null,"url":null,"abstract":"\n Under the present conditions of oil and gas production, which are characterized by mature production fields and the focus shifted towards digitalization of production processes and use of machine learning (ML) models, the issues related to the improvement of accuracy and consistency of the well operation control data are becoming increasingly important. As a result, SPD has successfully implemented the project of using annular pressure sensors in combination with machine learning models to control the well annular pressure as part of the field development program compliance.\n Under the field development program, echosounder and telemetry system readings are typically used to control the annular pressure and the dynamic flowing level. Echosounders, however, are not designed as measuring instruments, the accuracy of their readings being low and making it impossible to reliably evaluate the well's dynamic flowing level and annular pressure, as well as to achieve the well's maximum potential, and the telemetry systems used to measure the pump intake pressure may go wrong.\n This manuscript describes the approach to the producer well annular pressure assessment based on the machine learning model data. The machine learning (ML) model is a function of the target variable (bottom-hole pressure), which is predicted on the basis of the actual data: static parameters (well schematic, pump design) and dynamic parameters (annular and line pressures, flowrate). The input parameter interpretation results in the most probable value of the target variable based on the historic data.","PeriodicalId":11017,"journal":{"name":"Day 2 Wed, October 13, 2021","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2021-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Day 2 Wed, October 13, 2021","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.2118/206518-ms","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Under the present conditions of oil and gas production, which are characterized by mature production fields and the focus shifted towards digitalization of production processes and use of machine learning (ML) models, the issues related to the improvement of accuracy and consistency of the well operation control data are becoming increasingly important. As a result, SPD has successfully implemented the project of using annular pressure sensors in combination with machine learning models to control the well annular pressure as part of the field development program compliance.
Under the field development program, echosounder and telemetry system readings are typically used to control the annular pressure and the dynamic flowing level. Echosounders, however, are not designed as measuring instruments, the accuracy of their readings being low and making it impossible to reliably evaluate the well's dynamic flowing level and annular pressure, as well as to achieve the well's maximum potential, and the telemetry systems used to measure the pump intake pressure may go wrong.
This manuscript describes the approach to the producer well annular pressure assessment based on the machine learning model data. The machine learning (ML) model is a function of the target variable (bottom-hole pressure), which is predicted on the basis of the actual data: static parameters (well schematic, pump design) and dynamic parameters (annular and line pressures, flowrate). The input parameter interpretation results in the most probable value of the target variable based on the historic data.