dc.description.abstract | Rapid evolution of technologies in petroleum industry in last decades has significantly improved our abilities in hydrocarbon reservoirs development. The number and complexity of tasks to be solved by reservoir engineers are gradually increasing, while the cost of field development projects is rising. In this conditions, optimal decision-making in reservoir management becomes critical since it might result in either significant benefit or financial loss to a production company. Although a significant improvement was made in project risk management to control project costs in the case of unfavorable outcome, reservoir evaluation still plays the important role and affect entire reservoir management and production process. Since the work of petroleum engineers actively involves reservoir simulation and target search for optimal solution of the particular reservoir assessment problems, selection of the most appropriate simulation approach in a timely manner is important. Successful search for suitable solution to a particular reservoir engineering problem is always a non-trivial task since it involves analysis and processing of large amounts of data and requires professional expertise in the subject area.
In this work we proposed an expert system, what provide flexible framework for the proper simulation approach selection and involves thorough data analysis, multiple constraints handling, expert knowledge utilization, and intelligent output requirements implementation. This expert system utilizes linguistic method of the pattern recognition theory for knowledge base design and inference engine implementation, what significantly simplifies procedures of the system design and provides it with tuning flexibility. This thesis elaborates on major aspects of the expert system design in close relation to data processing and recommended solution finding methods.
To validate the expert system’s applicability, several tests were designed based on the synthetic Brugge field case and real petroleum reservoir data. These tests demonstrate functionality of the major expert system elements and advantages of selected implementation methods. Based on obtained results we can conclude successful development of the expert system for appropriate simulation approach selection. | en |