Data processing: measuring – calibrating – or testing – Measurement system in a specific environment – Earth science
Reexamination Certificate
1999-02-09
2001-05-22
Patidar, Jay (Department: 2862)
Data processing: measuring, calibrating, or testing
Measurement system in a specific environment
Earth science
C702S014000
Reexamination Certificate
active
06236943
ABSTRACT:
FIELD OF THE INVENTION
This invention relates to data attributes and pattern recognition methods in data mining and analysis. More specifically, the invention is concerned with providing a method of analyzing patterns in one or more seismic signal attributes related to one or more properties of an underground reservoir.
BACKGROUND OF THE INVENTION
Analyzing and extracting the most information from indirect measurements of a property of matter, such as a subsurface formation or structure that might contain a natural resource, has long been an important goal for natural resource identification and recovery activities. For example, the information can be used to provide predictions of one or more structural and rock properties of a potential reservoir and enable the economic recovery of oil from the reservoir. One type of indirect measurement uses a compressional wave generator or other seismic source of seismic waves and a plurality of seismic transducers or sensors generating seismic data that represent the sensed seismic wave or disturbance. The generated wave is typically affected by at least one property of an underground formation of interest and the resulting sensed signals have long been used to predict at least one structural or lithological boundary location and other properties, referred to herein as a Primary Reservoir Property (PRP).
In a typical seismic analysis application, a firing or actuation of a seismic source (e.g., an airgun, explosive charge, or vibration device) generates one or more types of seismic waves such that the waves reach the formation of interest. A plurality of seismic sensors is located so as to be influenced by the subsurface seismic waves after they reach the formation of interest and produce seismic data. The seismic data are typically comprised of events resulting from reflections of the seismic waves from formation boundaries, such as shale breaks or fault lines. The locations of the reflecting boundaries can be determined from an analysis of the attributes of the seismic data. For example, measuring the time delay from the firing source to the first detection of the reflected signal gives some indication of the distance traveled using an estimate of the speed of the underground seismic wave. The speed of the seismic wave through different formations is typically estimated or determined from other available information.
In order to predict other formation or reservoir properties, different methods are used to determine a relationship between one or more attributes of the measured seismic signals and the reservoir property. Two methods typically used are categorized as a physical method (PM) and a statistical method (SM). A physical method (PM) seeks to correlate one or more formation properties with seismic or other indirect measurements based on physics or other known physical relationships. An example of a physical method is the bright spot analysis technique linking the amplitude responses of stacked seismic data to the presence of hydrocarbons in the formation, e.g., see Exploration Seismology, Sheriff and Geldart, Cambridge Press, Vol. 1, 1982 and Vol. 2, 1983. Another example of a physical method of analyzing seismic data is a conventional AVO analysis that under certain conditions can be related to the behavior of seismic amplitude as a function of offset to lithology and/or fluid properties of the reservoir using Biot-Gassman theory, e.g., see Plane Wave Reflection Coefficients for Gas Sands at Non-Normal Angles of Incidence, Ostrander, Geophysics, Vol. 49, No. 10, 1984, pp. 1637-1647.
Other analysis methods, such as statistical methods (SM), are also used since not all physical effects of various formation properties on seismic signals or other indirect measurements are known. One conventional statistical method is based on correlating an unknown property and a seismic signal attribute using more direct measurements of the property at one or more specific locations, such as well log data. Conventional linear statistical methods are used to develop a correlation for other locations, including regression analysis (e.g., see “The Application of Pattern Recognition in Seismic Signal Interpretation,” Z. Bian, Y. Li, P. Yan, and T. Chang, in Advances in Geophysical Data Processing, Vol. 3. entitled Artificial Intelligence and Expert Systems in Petroleum Exploration, M. Smaan & F. Aminzadeh editors, JAI Press, 1989, pp. 175-199), clustering (e.g., see “Application of Clustering in Exploration Seismology,” F. Aminzadeh and S. L. Chatterjee, Geoexploration, Vol. 23, 1984, pp. 147-159), linear discriminent analysis (e.g., see “Seismic Indicators of Stratigraphy,” A. Sinhal, K. N. Khattri, H. Sinvhahal, and A. K. Awasthi, in Pattern Recognition and Image Processing, Geophysical Press, 1987, pp. 225-262), and cokriging (e.g., see “Integrating Seismic Data in Reservoir Modeling, the Colocated Cokriging Alternative,” W. Xu, T. T. Tran, R. M. Srivastava, and A. G. Journal, SPE Paper #24742, 1992). Examples of non-linear statistical analysis methods include neural networks (e.g., see “Reservoir Architecture and Porosity Distribution, Pegasus Field, West Texas, an Integrated Sequence Stratigraphic Seismic Attribute Study Using Neural Networks,” J. S. Schuelke, et. al., in the Proceedings of the Society of Exploration Geophysicists Meeting, Dallas, 1977, pp. 668-671), fuzzy logic techniques (e.g., see “Fuzzy Classification with Application to Geophysical Data,” by B. Lashgari, in Expert Systems in Exploration, F. Aminzadeh and M. Sinaan, Editors, SEG Press, 1991, pp. 161-178, “Applications of Fuzzy Expert Systems in Integrated Oil Exploration,” F. Aminzadeh, Computers and Electrical Engineering, Pergamon Press, 1994, pp. 89-97, and “Applications of Fuzzy Expert Systems in Integrated Oil Exploration,” F. Aminzadeh, Soft Computing, Prentice Hall, 1994, pp. 29-43), and generic algorithms that try to identify correlations where there is no well defined or well known relationship between one or more seismic attributes and one or more formation properties. Besides log data from wells penetrating the formation of interest at a known location, signal-correlatable formation properties can also be derived from other sources including laboratory or scaled down physical models of the reservoir, measurements and analysis of core samples, or well log data from similar formations or nearby locations.
Both physical and statistical methods have been successfully used to accurately predict at least one primary reservoir property (PRP) in environments such as a shallow reservoir, but have experienced problems when applied to seismic data for complex underground environments. A complex environment is defined herein as having one or more of the following characteristics: a reservoir that is deeper than about 10,000 feet and more especially a reservoir that is deeper than about 15,000 feet, a reservoir having a near surface weathering layer more than 500 feet thick, a reservoir having a shallow gas layer over 200 feet thick, seismic data having a low signal to noise ratio, and data having multiple reflections and/or static problems. A very complex environment is defined herein as having two or more of the above-mentioned characteristics, and an extremely complex environment is defined herein as having three or more of the above-mentioned characteristics. Other conventional seismic attributes may be used to predict a PRP in some complex environments, such as frequency and AVO, but accurate predictions using seismic attributes in very complex environments and extremely complex environments have been a problem.
More recently, seismic data have been used to predict other unknown or inaccurately known properties of a potential fluid-containing reservoir, each referred to as a secondary reservoir property (SRP), e.g., formation fluid saturation, porosity, and pore pressure. These SRP predictions are especially desirable when direct measurements are very costly, e.g., predicting the fluid-related properties of an offshore underground formation that may contain hydrocarbo
Aminzadeh Fred
Xu Wenlong
Jacobson William O.
Patidar Jay
Taylor Victor J.
Union Oil Company of California
Wirzbicki Gregory F.
LandOfFree
Hybrid reservoir characterization method does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Hybrid reservoir characterization method, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Hybrid reservoir characterization method will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2506689