Automated Workflow Uses Pressure, Rate Measurements for Well Monitoring

_ This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper SPE 218470, “A New Automated Workflow for Well Monitoring Using Permanent Pressure and Rate Measurements,” by Anton Shchipanov, SPE, NORCE; Boyu Cui, SPE, University of Stavanger; and Vitaly Starikov, SPE, Heriot-Watt University, et al. The paper has not been peer reviewed. _ The complete paper describes an integrated workflow for automated well monitoring using pressure and rate measurements obtained with permanent gauges and flowmeters. The workflow is based on time-lapse pressure transient analysis (PTA) and integrates the following components: virtual flowmetering (VFM), transient identification, feature extraction and pattern recognition in transient pressure responses, and assessment of well performance based on PTA metrics. The methodology behind the workflow combines different physics-informed and data-driven methods described in the paper. A New Workflow for Automated Well Monitoring The concept of the automated workflow described in the complete paper has been suggested in the literature. The workflow consists of four main steps (Fig. 1), and integrates the following components: - VFM - Pressure transient identification - PTA feature extraction and time-lapse pattern recognition - PTA metrics to obtain the well-performance profile VFM. The well flow rate is often measured more sparsely than its bottomhole or wellhead pressure for many reasons. However, the permanent downhole or wellhead gauges, when installed in a well, provide high-frequency real-time pressure and temperature measurements. This disparity in the pressure- and rate-measurement sampling rates is a problem for PTA, which requires accurate and synchronized input for flow rates in line with the pressure time series. VFM helps reconstruct missing flow-rate measurements using available pressure measurements. The automated well-monitoring workflow described in this paper would benefit from integration of a relatively simple but robust and scalable VFM model having modest input-data requirements. The steady-state vertical-lift-performance-based VFM model was chosen for use by the authors after testing. The model uses only the bottomhole and wellhead pressure difference, which fits well with the automated workflow requirements. Put simply, the VFM-developed algorithm identifies the suitable pressure and flow-rate samples to train the VFM model; estimates, filters, and verifies the evolution of the model coefficients; and predicts the missing flow-rate values using the pressure measurements where available. This prediction applies only to the time when the well is flowing; as for shut-in periods, the algorithm assigns zero rate values. Transient Identification. Proper transient identification is crucial for PTA because accurate identification of the starting break point of a transient and sequential synchronization with corresponding rate governs the reliability of PTA interpretation results. This is important not only for the transient in question but also for the well history preceding the transient. In the latter case, proper transient identification may help in reliable rate time allocation throughout the well history for common cases of uncertain rates. The workflow of the extended method consists of two main stages. First, the breakpoints indicating start and end of shut-in transients are detected using the topographic prominence max rotation (TPMR) method. This automatically provides detection of flowing transients between the shut-ins identified. Second, the breakpoints of multirate periods (assuming stepwise rate changes) are detected within the long flowing transients using the local minimum in rotation (LMIR) method. As a result, the well pressure history is divided into sequential flowing and shut-in transients, also differentiating flowing transients governed by stepwise rate changes. Neither TPMR nor LMIR methods require denoising, so raw data from permanent gauges may be used without any preprocessing.