A Dynamic Data-Driven Inversion Based Method for Multi-Layer Flow and Formation Properties Estimation

Abstract

Abstract Real-time monitoring of downhole temperature and pressure in injection and production wells equipped with permanent downhole gauges as well as distributed temperature and acoustic sensors plays an important role in production optimization, improved hydrocarbon recovery and daily well operations management decisions. To fully take advantage of the real-time data, there is currently a pervasive need in intelligent oilfield application areas for near real-time, high-fidelity, dynamic data-driven inversion methodologies combined with fast forward flow models comprising partial differential equations. Such methods facilitate dynamic data-driven interpretation of temperature and pressure measurements that can be used to provide continuous downhole production performance characteristics and forecasting, and are key to optimizing and controlling intelligent well production systems under closed-loop conditions. However, a central challenge is the construction of robust computational algorithms for near real-time solutions of the underlying data-driven inverse problem which is often more challenging to obtain than the corresponding forward simulation model. This paper presents the development and testing of a robust computational algorithm for the estimation of multi-layer reservoir flow in addition to static and dynamic formation properties. In this work, a new modeling framework which incorporates a forward simulation engine and an inverse model is developed. The forward simulation engine employs a fast and accurate mathematical model to predict flow and formation properties profiles in advanced completion wells. It also introduces specific measurement response functions for numerical simulation of permanent downhole gauge data in the advanced wells. The inversion model employs a modified Levenberg-Marquard algorithm and is used to interpret field-measured downhole pressure and temperature data to estimate permeabilities, transient reservoir pressure and temperature response, downhole multiphase flow rates and zonal production allocation to improve hydrocarbon recovery and mitigate water/gas breakthrough risk. Under the assumption of a subsurface transient thermal multiphase flow model, the inversion process yields estimated well rates, formation properties and transient reservoir pressure and temperature response specific to a given measurement domain by numerically reproducing the available measurements. Validation results show that the proposed computational workflow can improve the prediction results for the reservoir flow and properties in various layers, from the top of the reservoir to the bottom. The effects of calibration on performance and optimal parameter determination of the forward transient thermal multiphase flow model is assessed during model calibration periods. This result has implications for the effectiveness and efficiency of using limited measured data for non-continuous model calibration. Applicability of the system to various field cases has been demonstrated.