基于数据挖掘的储层地震属性优化研究-控制科学与工程·控制理论与控制工程专业论文.docxVIP

AbstradABSTRACT Abstrad ABSTRACT Seismic attributes refer to the parameters computed or calculated according to seismic records，such as amplitude，frequency,time，wave impedance，etc．Although reservoir character can be described by these seismic attributes from many aspects， yet the parameters show different due to the sensibility to different reservoir or the various liquids in different areas，which make the reservoir prediction by seismic attributes complicated．Therefore，how to select the best attributes of the reservoir information from the involved seismic attributes become the key point of improving the prediction accuracy for the complex oil and natural gas reservoirs． From the aspect of data mining，the present thesis is the research results involved with abstraction，optimization and prediction of seismic attributes in accordance with practical requirement．The relationship between optimization of seismic attributes and reservoir parameters is discussed for improving the accuracy of reservoir prediction and abstracting new seismic attribute，which is helpful to seismic interpretation． The thesis includes the following key points：Study on the normal abstraction skills of seismic attributes，comprising how to abstract instantaneous attribute based on the seismic section and how to abstract amplitude，time，frequency,absorbability attenuation based on seismic horizon；Study on the methods of optimization for seismic attributes，reducing dimension and correlation by K-L transform，selecting the best attributes by search algorithm；Study on the common ways to predict the reservoir parameters，including multiple linear regression and BP neural nets： Study on abstraction of new seismic attribute．Seismic attributes space is built according to the selected attributes，SO as to analyze the relationship among the attributes；common character of the selected attributes is described by inner projection attribute，abstracted by projection algorithm in attributes space．It is helpful to impr