Applications of Intelligent Data Analysis Techniques for Modeling the Carbon Capture Process System and other Industrial Applications.

题目：Applications of Intelligent Data Analysis Techniques for Modeling the Carbon Capture Process System and other Industrial Applications.

报告人：Dr. Christine Chan，加拿大里贾纳大学

时间：2015年11月26日（周四）上午10：10

地点：国重 A403学术报告厅

   Abstract：This research work has the objective of improving efficiency of the carbon dioxide (CO2) capture process, but the approach differs from research work conducted in the field of carbon capture. Carbon capture studies have often focused on studying the features and performance of various aqueous amine solvents. Instead, our approach aims to achieve a good understanding of the intricate relationships among the key process parameters based on the assumption that a solid understanding of these relationships would enable operators to enhance efficiency and effectiveness of process operations. Therefore, we attempt to explicate the relationships among the key parameters of the process system by applying a variety of machine learning technologies. The study presented has two objectives: (1) to identify and determine the significances of the process parameters which have influence on the performance of the CO2 capture process, and (2) to model the relationships among the process parameters in order to explore the nature of their relationships. Our approach is to apply multiple data mining techniques to the operational data collected over three years from the amine-based post combustion CO2 capture process system at a pilot plant at Clean Energy Technology Research Institute (the former International Test Centre of CO2 Capture) located in Regina, Saskatchewan of Canada. The three data mining techniques adopted include: (1) statistical regression analysis, (2) artificial neural network (ANN) modeling combined with sensitivity analysis (SA), and (3) adaptive network based fuzzy inference system (ANFIS) modeling. Application of the three methods revealed the strengths and weaknesses of each method. Our study found that the ANFIS modeling method is the most satisfactory because it generated the interpretative models with high prediction accuracies. Some ongoing work in other industrial projects (in the areas of petroleum production, environmental engineering, process systems engineering, etc.) will also be briefly presented.

   Personal Introduction：Dr. Christine Chan, Ph.D., P.Eng. Canada Research Chair (Tier I) in Energy and Environmental Informatics, Professor, Software Systems Engineering.As Canada Research Chair Tier 1 in Energy and Environmental Informatics at U of Regina, Dr. Christine Chan’s research expertise lies in development and applications of artificial intelligence and knowledge-based system (KBS) technologies for energy and environmental systems analysis. She has led research initiatives in studying applications of advanced information technologies (IT) for problem solving in the areas of (i) solvent development and management and (ii) emissions control.

   In terms of emissions control, Dr. Chan contributed to the area of applications of artificial intelligence (AI) technologies for developing (a) monitoring and supervisory control systems, and (b) selection systems. She led the development of: an expert system for monitoring and supervisory control of a carbon dioxide capture process system, and developed an innovative intelligent system framework that integrates an ontological model with the monitoring and control functions. The framework supports a KBS development approach which has advantages over conventional approaches because it reflects the dynamic and multi-objective features of complex process systems while reducing the time and effort required for system construction. Based on this method, the system for monitoring, control, and diagnosis of the carbon dioxide capture process system was developed and implemented at the International Test Centre for Carbon Dioxide Capture (ITC) in Regina, Saskatchewan. The feedback from the operators of the pilot plant of ITC and the industrial sponsors has been highly favourable. As well, Chan tackled the research issue of developing an optimal approach for automated process control of complex processes, and produced three prototype process control systems which included: (i) a multi-objective optimization process control system, (ii) a model-based predictive control system, and (iii) a predictive control system based on a knowledge-based reasoning enhanced model. She published an internationally acclaimed paper in this area of AI techniques for monitoring and supervisory control of process systems, which won the “Top ten most cited articles 2005-2010” award of Engineering Applications of Artificial Intelligence (EAAI, published by Elsevier). Another paper on developing an automated monitoring and control system for the carbon dioxide capture process system won the “Best Paper Award” at International Conference on Ecological Information and Ecosystem Conservation 2010 held in Beijing, China. In terms of selection systems, Chan studied knowledge modeling and development of an intelligent system for selecting solvents for carbon dioxide separation processes. Based on this work, an architectural framework for developing intelligent systems for the selection task was completed. In the area of modeling and analysis of data, Chan led an international research initiative on modeling and analysis of data from the carbon dioxide capture process system in an effort to build correlation models that explicate relationships among key parameters of the system. The study adopted and compared diverse data modeling technologies, which included artificial neural networks, case-based reasoning, statistical analysis, neuro-fuzzy inference system (ANFIS), sensitivity analysis, rule-based system, visualization, and simulation. The approaches developed were also applied for modeling data of other industrial domains.

In general, Dr. Christine Chan has played a global leadership role in her work in research and development of applications of artificial intelligence technologies in the carbon dioxide capture systems area. Her work in development of knowledge-based systems and in data modeling and analysis for problem solving, whether it is related to operations of the carbon dioxide capture process system or to solvent selection for carbon dioxide separation, constitute first attempts in a systematic study of this complex area. Her studies have shed light on some specific issues in the area, which include (i) how to build a KBS system for monitoring and control of the carbon dioxide capture process system, (ii) how to automate the process of solvent selection for carbon dioxide separation, and (iii) investigating the intricate relationships among key parameters of the process system as a step towards realizing higher efficiency of the system.

油气藏地质及开发工程国家重点实验室

石油与天然气工程学院西南石油大学科研处

2015年11月23日