Font Size: a A A

Reservoir Prediction Research Based On Curvelet Transform And Bayesian Theory

Posted on:2012-09-16Degree:DoctorType:Dissertation
Country:ChinaCandidate:J J ZhengFull Text:PDF
GTID:1100330338493195Subject:Geological Resources and Geological Engineering
Abstract/Summary:PDF Full Text Request
With the demand growing of oil and gas and the improving of seismic exploration technology, the exploration and development field of oil and gas become complex (such as complex fractures type, concealment type and deep hydrocarbon reservoirs, etc.), so requirements for reservoir prediction is getting higher. Therefore, more effective methods and techniques are needed for reservoir prediction. First, this thesis researched set of Curvelet-based methods to predict the fracture strength and fracture zone strike to ook favorable oil and gas accumulation area. And then, methods based on Bayesian theory were used to optimize a large number of attributes from the perspective of probability analysis, and main aim is to predict hydrocarbon reservoir and its scope by the post-stack attribute analysis approach. At last, the new methods and technology based on prestack data were researched and were used to distinct the nature of the fluid in reservoir. In general, the idea of the thesis is prediction of favorable oil and gas accumulation zone-prediction of the range of oil and gas reservoirs-the identification of reservoir fluid properties. Starting from the source of reservoir prediction, apply the progressive methods to predict the reservoir range and the fluid properties.Fractures and cracks are common in the strata. This paper deep studies reservoir fractures and proposes edge preserving and sharpen filter, multi-scale, multi- direction coherence technology and multi-spectral body curvature analysis based on the Curvelet transform. First, since the seismic reflection data near-fault /fractures is complex, and contains relatively strong noise, an edge preserving and sharpen filter is proposed based on edge preserving smooth (EPS) techniques and the lower-upper-middle (LUM) filters. This filter preserves small linear features, while removing noise by adjusting the parameters. After that,combining Curvelet transform with coherence and curvature technology, we develop a new and effective multiscale, multidirectional coherence cube and multi-spectral volumetric curvature method of predicting different scale fracture zones and their strike. Multiscale and multidirectional coherence cube methods gives different reconstruction coefficient in Curvelet domain, gains seismic data that bursts characteristic of the different frequency bands and different directions, and then obtains multi-scale and multi-directional coherence at last. By applying Curvelet transform to separate different wavelengths wave motion information and using multi-spectral derivative operator to finely tune scale coefficients, multi-spectral volumetric curvature method reconstructs wave motion information matching with the scale of target reservoir and gains seismic data that bursts characteristic of the different wave-number. The three methods predict the fracture strength and fracture zone strike from the waveform and structure, respective and improve the reliability of prediction results. Real data were used to test the sets of methods. The results showed they can improve the signal to noise ratio (SNR) of data, protect small fracture signals and burst seismic data of the different frequency bands and different directions and better represent complex and variable geologic body details. These methods provide a new way to predict lithology.In reservoir prediction, seismic attribute analysis has been an important way to obtain reservoir parameters. According to analysis results, accurate reservoir information can be got. Based on Bayesian theory and principal component analysis, PKPCA method was proposed. This method use probability model constraints kernel principal component analysis and overcome the two shortcomings of principal component analysis that lack probabilistic model and higher order statistics information. Then, a mix model of PKPCA was developed-mix probabilistic principal component analysis (MPKPCA). It mixed a variety of probability models to characterize the data. At last, EM algorithm is used to get the best probability model. The application of the actual data showed the attribute probability optimization method based on Bayesian theory improves the accuracy of attribute optimization, while increasing the accuracy of reservoir prediction.Reservoir fluid recognition is one of the final targets of the seismic exploration, so we carried out the research in the method of fluid recognition based on the reservoir prediction. First, the focus of our research is the incidence angle AVO approximation equation, which is constructed based on aki, Shuey approximate equation and snell law, and expresses the inner relationship between the reflection coefficients and the upper incident and petrophysic parameters more clearly. The practical seismic CRP gathers transferring offset into angles is also the function of upper incident angles, so the theory formulas are more consistent with the objective conditions of actual seismic data. After that, based on the incident angle AVO approximation equations, we studied the more accurate AVO attribute extraction method, proposed the correction formulas of the prestack AVO attributes (G, Rs, etc.). Then, using the advantages of stack data in the angle part , this article proposes three effective methods and techniques of fluid recognition and reservoir prediction, and combining it with the prestack AVO attributes correction formula, realizes the purpose of improving reservoir prediction accuracy and reliability. Finally, after some theoretical researches on the common fluid factor relations were carried out, we discovered the inner relation between them, and respectively established the new fluid factor formula in the reflection coefficients domain and impedance domain according to their essence. The application results of actual data show that the methods in this the paper can not only be used to accurately determine the position and scope of reservoirs, but also better distinguish gas, water reservoirs.
Keywords/Search Tags:Curvelet transform, multi-scale and multi-direction, fracture zone and its strike, attribute optimization analysis, Bayesian theory, the incidence angle, AVO approximation equation, reservoir prediction, fluid recognition
PDF Full Text Request
Related items