A machine learning framework for rapid forecasting and history matching in unconventional reservoirs

Abstract

We present a novel workflow for forecasting production in unconventional reservoirs using reduced-order models and machine-learning. Our physics-informed machine-learning workflow addresses the challenges to real-time reservoir management in unconventionals, namely the lack of data (i.e., the time-frame for which the wells have been producing), and the significant computational expense of high-fidelity modeling. We do this by applying the machine-learning paradigm of transfer learning, where we combine fast, but less accurate reduced-order models with slow, but accurate high-fidelity models. We use the Patzek model (Proc Natl Acad Sci 11:19731–19736, 10.1073/pnas.1313380110, 2013) as the reduced-order model to generate synthetic production data and supplement this data with synthetic production data obtained from high-fidelity discrete fracture network simulations of the site of interest. Our results demonstrate that training with low-fidelity models is not sufficient for accurate forecasting, but transfer learning is able to augment the knowledge and perform well once trained with the small set of results from the high-fidelity model. Such a physics-informed machine-learning (PIML) workflow, grounded in physics, is a viable candidate for real-time history matching and production forecasting in a fractured shale gas reservoir.

A Geo‐Structurally Based Correction Factor for Apparent Dissolution Rates in Fractured Media

“…While this parameter cannot be easily measured in the field, estimates can be obtained through models of the fracture network that are informed by site characterization. DFN generation parameters for the fracture radius, family orientation, and hydraulic aperture distribution can be used to create an ensemble of realizations of the site using a geostatistical approach (Follin et al., 2014; Joyce et al., 2014; Srinivasan et al., 2021; Svensk Kärnbränslehantering AB, 2010). The geo‐structural parameters used in Equation can be directly measured in these ensembles without needing to run flow and transport simulations.…”

Physics-informed machine learning with differentiable programming for heterogeneous underground reservoir pressure management

“…The physics constraints needed for the training process are incorporated through the DPFEHM framework, which provides the physics information in our PIML framework. Our framework has similarities to the works of Harp et al 45 and Srinivasan et al 64 , that use physics-informed neural networks with physical constraints used during the NNM training. A significant difference in our approach is that instead of using a differentiable, simple analytical solution, we use a differentiable numerical physics model.…”

Physics-informed machine learning with differentiable programming for heterogeneous underground reservoir pressure management

“…The physics constraints needed for the training process are incorporated through the DPFEHM framework, which provides the physics information in our PIML framework. Our framework has similarities to the works of Harp et al and Srinivasan et al 44,47 , that use physics-informed neural networks with physical constraints used during the NNM training. A significant difference in our approach is that instead of using a differentiable, simple analytical solution, we use a differentiable numerical physics model.…”