Qin He

Date of Graduation


Document Type


Degree Type



Statler College of Engineering and Mineral Resources


Lane Department of Computer Science and Electrical Engineering

Committee Chair

Shahab D Mohaghegh

Committee Co-Chair

Samuel Ameri

Committee Member

Ali Takbiri Borujeni

Committee Member

Tim Carr

Committee Member

Ebrahim Fathi


In oil and gas industry, quick decisions on reservoir management have a huge impact on business success. Typically, reservoir simulation is used to predict field performance and analyze uncertainties for assistance on decision making, while history matching is a key step of reservoir simulation, which is a process of model adjustment and simulation runs with different reservoir parameter settings until the difference between simulated data and historical data reaches minima. An efficient reservoir simulation model must be the one that can predict reservoir performance and update history matching results continuously by modifying reservoir model as long as new data becomes available. However, reservoir simulation can be very time consuming depending on the complexity of a reservoir model, and history matching is even more computational expensive since it requires lots of simulation runs. Nowadays, as intelligent technology advances in oil and gas industry, an initiation of a new era of real-time data acquisition begins. With the generation of high frequency data stream, how reservoir simulation should be performed in line with the real time data without compromise on the simulation time is a big concern for petroleum engineers.;In order to address this problem, lots of studies have been going on. Besides increasing computational power, varieties of research have focused on speeding up reservoir simulation process especially history matching by either implementing optimization algorithms or generating efficient proxy models. Nevertheless, there has not been a standard method recognized in reservoir simulation yet.;In this study, a novel methodology has been proposed as the first attempt to investigate the possibility of achieving continuously updated history matching by data-driven proxy model named Smart Proxy or Surrogate Reservoir Model (SRM). This research essentially involves detailing numerical reservoir modeling, continuously updated history matching and model modification process as new field data becomes available in a real case study. The objective is mainly focused on capability examination of Smart Proxy (SRM) in terms of continuously updated history matching.;According to the simulation results, the feasibility of Smart Proxy on a continuously updated history matching process has been proven successfully and efficiently. Importantly, tremendous time and efforts have been saved in the reservoir simulation process by using Smart Proxy compared with the traditional numerical reservoir simulation.