New transient identification methods for automated pre-processing of pressure measurements with permanent well gauges

Abstract

Well monitoring with pressure and temperature gauges is a part of well and reservoir surveillance systems across various industries, including petroleum and geothermal energy production as well as carbon capture and storage. Installing permanent downhole gauges (PDG) becomes a standard in the industry, which in combination with flow rates, provides measurements for well and reservoir monitoring by using pressure transient analysis (PTA). The feasibility and accuracy of PTA are governed by proper identification of pressure transients. Transient identification is traditionally a heavy manual trial-and-error process. It often involves pre-processing PDG data by resampling, denoising or outlier removal. However, any pre-processing may have a risk of overlooking important information. In addition, lack of pressure-rate synchronization in raw data complicates further PTA applications.

This paper introduces a novel methodology for automated transient identification from raw gauge data. The methodology enables identification of both shut-in and multi-rate flowing transients by using pressure data only. Moreover, the new transient identification runs on the raw data without resampling, denoising or outlier removal, which ensures keeping all the information from the measurements. The methodology is a combination of two new independent methods: Topographic Prominence Max Rotation (TPMR) and Local Minimum in Rotation (LMIR). The TPMR method utilizes the concept of prominence to identify significant shut-in transients. The LMIR method detects multi-rate flowing transients by identifying local minima in transformed pressure data via proper rotation matrix. Together, these methods provide an automated solution for dividing a pressure history into sequential flowing and shut-in transients. The new methodology has been tested and verified using real PDG datasets from the Norwegian Continental Shelf. The testing confirmed stability and accuracy of the methods, providing fast results with minimal human intervention. Then, an automated data pre-processing framework is described integrating the transient identification methodology with pressure and rate synchronization, rate reconstruction, superposition time and Bourdet derivative calculations. Finally, an integration of the framework within an automated time-lapse PTA well monitoring workflow is demonstrated.