Description of ResFrac automated history matching and optimization workflow

TL;DR

This paper details an automated workflow integrating ResFrac for history matching and optimization in hydraulic fracturing, utilizing proxy models and Bayesian sampling to enhance calibration and decision-making efficiency.

Contribution

It introduces a novel automated workflow combining physics-based simulation, proxy modeling, and Bayesian sampling for reservoir history matching and optimization.

Findings

Workflow enables automated calibration to observed data.

Proxy models significantly reduce computational costs.

Application demonstrates improved reservoir management decisions.

Abstract

This document describes the functioning of the automated history matching and optimization workflow implemented with the ResFrac hydraulic fracturing and reservoir simulator. The purpose of the automated history matching and optimization workflow is enable automated calibration of simulator models to observed data and application to forward optimization. The automated workflow works by solving a formal mathematical optimization problem to minimize misfit with observations from any point in the lifecycle of a hydraulically fractured reservoir, or to maximize a quantity of interest associated such a reservoir, such as net present value. Objective function evaluations in the optimization problem consist of runs of the ResFrac simulator, which is a physics-based model of hydraulic fracturing and reservoir phenomena. The workflow employs a proxy model to improve computational speed and applies experimental design and Bayesian sampling techniques to generate points with which to train the proxy model. This document also provides an overview of the software infrastructure developed to support the automated workflow.