Development of an Artificial Neural Network Based Model for Mimicking Combustion Tube Experiments for Heavy Oil Recovery

Abstract

Over the years, technologies for improved recovery of heavy oil have become an important part of the research efforts to meet the increasing demand for oil. Various methods are being developed for heavy oil recovery and among them in-situ combustion process has shown a good degree of potential in laboratory and pilot tests conducted in field. The in-situ combustion process needs to be studied further extensively because of the high degree of operational complexities involved in the process. Extensive laboratory studies have been conducted; however, a typical in-situ combustion tube experiment in the laboratory can be very costly in terms of its time, personnel and equipment requirements.

This work aims at reducing the number of laboratory experiments by developing an artificial neural network (ANN) that has the ability of emulating in-situ combustion tube experiments. An intelligent database was generated using commercial software to train the network within the parametric ranges adapted from previous experimental work. The ANN model was used to predict the cumulative production of oil, water and gas, peak temperature attained, location and velocity of the combustion front in the numerical combustion experiments mimicking the physical experiments. The proposed ANN model can be used to focus towards designing the experimental program within a rather small range of parameter variations resulting in more economical and focused analysis. Therefore, our methodology provides a unique and novel approach to understand in-situ combustion experiment.