A Machine Learning Approach For Virtual Flow Metering In Geothermal Fields: A Case Study In Salihli, Turkey

For proper field management in geothermal fields, it is vital to know the flow rates of each well. Increasing the accuracy of the flow rate provides more control over the field monitoring and management; thus, it enhances the decision-making process in optimizing the fields’ production. A Virtual Flow Meter (VFM) is a cost-effective solution for increasing the accuracy of flow rate measurements that uses the data collected from existing equipment and infrastructure in the field. In this study, we developed a method for VFM with a data-driven approach using machine learning algorithms and tested the method with existing physical flow meters. In this method, for every producing well, a machine learning model is created from the data of various sensors and measurements of the field, such as the pressure and temperature readings from wellheads, separators, and pipelines, flow rate information from power plants’ inlet and outlet sensors, physical flow meters and also meteorological conditions. The developed VFM algorithm is created and tested using the Caferbeyli Geothermal Field data in Salihli, Turkey, in which Sanko Enerji operates. The developed method minimizes the difference between the estimated and measured flow rates. It can be used as an independent flow metering system with periodical tuning from field sensors and can also be used as a backup system to validate the physical flow meters.