Feature Extraction for Soft Sensing and Process Monitoring in Steam Generators

  • Author / Creator
    Kwak, Seraphina Jinyeong
  • Data is becoming more valuable as there are still many uncertainties and hidden information that have yet to be discovered. For this reason, the application of data analysis and machine learning in the industry is becoming more popular. For example, SAGD (steam assisted gravity drainage) is a type of oil extraction process where high-pressure steam is used to heat the bitumen underground. Optimizing the steam generation is one of the ways to improve the SAGD process as steam is an important part of the SAGD process. One method that may be used to optimize this process is the feature extraction analysis.

    Feature extraction analysis is a method that tries to extract valuable information from a given dataset. Essentially, it projects the given dataset to another subspace such that a particular statistical property is amplified and noises are minimized. In this thesis, data analysis is explored to optimize the SAGD process. The first chapter defines the problem and feature extraction methods are introduced.

    In the second chapter, a grey box model is used to develop a soft sensor to predict the steam quality out of a steam generator in the real SAGD process. The core model structure is based on energy balance and data analytic methods is used to further improve the predictability strength by applying a Kalman filter and online bias updating technique. Later on, feature extraction methods are further explored to improve the developed soft sensor. Finally, cointegration analysis (CA), which is a type of feature extraction method, is modified to monitor fouling accumulation inside the steam generator tubes. The difficulty of predicting fouling buildup in a steam generator using process knowledge alone is addressed. Since fouling buildup involves complex chemical phenomena, a data analysis approach is proposed that can be easily applied in the industry. In the proposed method, PCA is paired with CA to develop a practical solution to predict fouling buildup.

  • Subjects / Keywords
  • Graduation date
    Fall 2019
  • Type of Item
  • Degree
    Master of Science
  • DOI
  • License
    Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. Where the thesis is converted to, or otherwise made available in digital form, the University of Alberta will advise potential users of the thesis of these terms. The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission.