Seismic Horizon Automatic Extracting And Hydroearbon Detection Technology Based On Higher Order Statisties Study

As the oil and gas exploration level deepens, the oil and gas exploration difficulties are increasing. Owing to the complicated conditions of exploration, the SNR and resolution of data are very low. thereby seriously impact on effect of processing and interpretation. On the condition that weak signal, there is particularly important to propose a new method of signal detection for the new exploration object. As a mathematical tool describe random signal HOS theory has been initial applied in the exploration area; based on this theory, two kinds of methods were discussed by this article with the view of inhibit coherent gauss noise to interfere with the seismic horizon extracting and mend frequency-divided technique and improve the time-frequency resolution.As the base of oil and gas detection, seismic horizon extracting appears more and more important. To improve the efficiency, people propose many ways, such as artificial neural network, pattern recognition and cross-correlation, etc. The cross-correlation way has the best stability and applies the most widely. But when the noise of the data has strong relevance, cross-correlation way pick up the time delay of noise, bring about the error of seismic horizon extracting. This paper is the full use of the transverse wave comparability and one-dimensional slice of cross-fourth-order cumulants is immune to Gaussian noise, and propose a method of automatic extracting seismic horizon based on one-dimensional slice of fourth-order cumulants, there are the following specific innovation points:1, in 1993 tugnait gave the time delay formula based on one-dimensional slice of fourth-order cumulants, the formula uses of variable length analyzing window. When data processing uses of this variable length analyzing window, there is a problem of producing great influence by bad trace, the instability calculation results, and inadequate time-delay. To avoid above problems, this paper uses the more robust and fixed length analysis window, and according to the actual need to change the location of delay the parameter.2, when the underground media is even, the seismic signal is a kind random signal of symmetry distribution of bordering on generalized Gaussian distribution, skewness is zero. The three order cumulant of symmetrical distribution random signal snooze is zero, so it is not fit for processing such signals. But the time delay formula based on one-dimensional slice of fourth-order cumulants doesn't has this limitation.3. in process of extracting seismic horizon needs to reasonably setup parameters, these parameters concern the accuracy of horizon extracting, and the choice of referenced trace is especially important. Based on the study of traditional choice pattern of referenced trace, a new choice pattern of referenced trace has been devised; this way overcome the problem that the fixed referenced trace pattern leads calculation error because of transverse wave change:and also correct the shortcoming of floating referenced trace leads cumulative error.Frequency-divided technique deriving from B P Amoco is a resolution technique to transforms seismic data to frequency domain or time-frequency domain through mathematical transform, through it. cut frequency axis into slices, make use of slices to predicate reservoirs finally. At present sorts of document considered, existing frequency-divided technique include:The information from various sources, the existing businesses of technology are mainly:Fourier transformation, linear frequency analysis, the frequency distribution.etc. This paper in order to mend frequency-divided technique and improve the time-frequency resolution, make use of the high time-frequency resolution specialty of Wigner bispectrum diagonal slice, first propose the frequency-divided technique based on the Wigner bispectrum diagonal slice and apply it on the oil and gas detection, achieved following results:1, in order to solve the problem of cross-term interference in Wigner bispectrum diagonal slice, through fuzzy fields'kernel wave-filtering method for restraining cross-term interference and derivate formula of Wigner bispectrum diagonal slice fuzzy functions.2, based on fully studying cone-shaped kernel and exponential kernel, through combine exponential function in cone-shaped kernel with exponential kernel and give a new kernel. Numerical simulations indicated that the new kernel function has the advantages of above two kernels, and has a strong inhibition to cross-term interference in delay axis and beyond axis.3, by way of moving coordinate to superpose the center point between fuzzy functions signal terms and new kernel. Thereby solve a problem that the new kernel couldn't be capable of restraining cross-term interference because of filtering in fuzzy fields directly.4, in order to solve the problem of program error messages:"memory overflow' because that datasize of 3D data volume is too massive. So the frequency-divided technique procedure is improved, avoiding generating Wigner bispectrum diagonal slice3D data volume.In accordance with research base from home and abroad as well as this field knowledge of myself, two kinds of methods were discussed by this article need to be intensive studied from following three areas:1, one-dimensional slice of fourth-order cumulants of this paper is based on non-parametric estimation method. This method is simple to implement, but its estimation variance is excessive, in the case of finite length of data. Estimation variance directly impact on the accuracy of the results, therefore it can make us of little variance parametric estimation method to give seismic signal delay estimation operator.2, it can be seen from Wigner bispectrum diagonal slice fuzzy functions that space between effective signal and cross-term is closer, the delay is the less. This goes against resolving the thin bed. The formula of Wigner bispectrum diagonal slice fuzzy functions, and increases the space between effective signal and cross-term, improving ability of resolving the thin bed.3, while effectively restraining cross-term, fuzzy field's wave-filtering method reduces the time-frequency resolution. So may make us of rearrangement theory or adaptive optimal-kernel and so on to restrain cross-term, and further improve time-frequency aggregation.