| In the oil and gas industry,exploration is crucial to guarantee the energy supply for the coming decades.The success of the hydrocarbon exploration projects is the key factor in identifying potential hydrocarbon resources and to turn potential resources into producing reservoirs.Hydrocarbon exploration mainly relies on highly sophisticated exploration techniques.The main goal of exploration geophysics is to extract structural information and physical properties of the Earth's subsurface from measurements that have been carried out on the basis of physical phenomena like acoustics,electrodynamics,gravity,etc.Seismic reflection surveying is one of the most common geophysical methods that are used in hydrocarbon exploration because of its depth of penetration and resolution.Although recently tremendous efforts have been made to generate more energy from renewable sources,fossil fuels(mainly hydrocarbons)are still providing more than 80 percent of the world's energy.At present,the global demand for energy is huge.In some developing countries and emerging economies,the demand for world energy is in a gradually rising stage.Specially,as a large ocean country,China's offshore petroleum reserves are abundant,and offshore petroleum exploration has gradually begun to take over the task of land petroleum exploration.However,since offshore petroleum exploration costs are much higher than the land,it has greatly restricted the development of offshore exploration technology.In addition,because of the geopolitical factors,offshore petroleum resources,especially in the South China Sea,are increasingly intensifying,and therefore,there is an urgent need to develop efficient,high-density acquisition technologies.With the exploration of the offshore petroleum resources gradually deepening into the complex structural oil and gas fields,the deep seabed reservoirs and the concealed structural reservoirs,traditional marine seismic acquisition technology can no longer meet the needs of today's oil and gas exploration situation.At present,the seismic technology commonly used in the petroleum exploration industry,whether in two dimensions or three dimensions,is a conventional single source shooting pattern.The shooting between different single shots requires a certain time interval to avoid mutual interference,which directly leads to the inefficiency of field acquisition.Especially in 3D seismic exploration,field acquisition and its seismic data processing are very time consuming.Even in the case of continuous improvement of computer performance,the demand can not be met,which seriously restricts the development of seismic exploration technology.The development of high-efficiency and high-density acquisition technology is imminent.In addition,the quality requirements for seismic images are becoming stronger and stronger over time.This is because the reservoirs to be discovered are increasingly complex,meaning that a higher resolution is needed as well as more accurate layer properties.With the rapid development of simultaneous source acquisition technology in the world,oil exploration industry has gradually realized the huge advantages in its potential data acquisition and processing.Compared with traditional single source acquisition method,the advantages of simultaneous source acquisition are mainly existed in two aspects: 1? simultaneous source acquisition greatly improves the seismic data acquisition efficiency under the same data quality as the traditional single source acquisition method,reducing the exploration cycle,thereby reducing the exploration cost;2? In the case of the same acquisition time,simultaneous source acquisition can obtain high-density sampling data,so the data quality is higher,and it is conducive to the acquisition of wide-azimuth seismic data,and effectively improve the data quality of deep water seismic exploration.Therefore,there is an urgent need for research on simultaneous source acquisition.However,due to factors such as higher cost of the acquisition,higher requirements for acquisition equipments,and more complicated acquisition processing,the development of simultaneous source acquisition technology in marine exploration is far slower than that on the land acquisition.In order to accelerate the development of blended marine ac-quisition technology in China,this thesis systematically studies the simultaneous source acquisition technology.In blending,the source space can be densely sampled for a given survey time.Alternatively,a traditional survey design can be carried out more efficiently.In addition,the signal-to-noise ratio of a blended survey is higher and finally,blending is better from a health,safety and environment(HSE)viewpoint.This is why today blending is considered better,cheaper,faster and safer.Once a blended survey has been carried out,there are several options for the processing.A first option would be to carry out the data processing directly on the blended data,i.e.,without a deblending process.However,we realize that such a 'direct' approach has the disadvantage that new processing algorithms and workflows are required.Current algorithms and workflows expect 'unblended' data.Therefore,a second option for processing blended data is to carry out a 'deblending' procedure.After such a procedure conventional data processing can be carried out.The deblending problem is underdetermined,i.e.,there are many solutions.Therefore,additional knowledge must be provided to get(close to)the desired solution.An example of such knowledge is the fact that seismic data is laterally coherent,whereas blending noise is laterally incoherent in certain domains.Many of the deblending algorithms that are currently used in the industry are based on this knowledge.Another approach is based on the fact that out of the many possible solutions to the deblending problem,the 'sparsest'-defined in a suitable domain-is the most likely.An inversion procedure based on this property was implemented using an ?1-norm type solution,as known from 'compressive sensing'.This thesis consists of following investigations:(1)We investigated the focal deblending methods.Double focal transformation is an extended version of the concept of the single-side focal transform in which a focal transform operator was used,which aims at projecting all of the seismic data into the source points.The main difference with respect to the original focal domain is that it does not use two-way focal operators,with a need to accurately include propagation and reflectioninformation,but one-way operators,containing only propagation information.However,these one-way operators are applied at both source and receiver side,thus representing a full redatuming.Therefore,the transform domain was called 'double focal domain'.The nice advantage is that it fits very well with the CFP-technology,as it requires the same type of focusing operators.Actually,by considering focal operators for one reflecting level,the bi-focal domain is a fully redatumed dataset.However,if the situation requires this,the original focal transform based on two-way operators can still be used.This could be the case when the sampling in one of the dimensions is not dense enough to warrant a complete transformation axis(think of 3D OBN data,where the receivers are very sparsely positioned).In addition,the double focal domain is required to be sparse.This makes sense,as reflection information will be focused to a small area(ideally a band-limited spike).We utilize the sparse inversion to solve this problem such that the focal domain is minimum in the ?1sense.Because only the reflection information around the focal point is really sparse,we extend the double focal domain to a multi-level double focal domain,where the data is simultaneously redatumed to several depth levels.Each focal operator at each level carries its own focal domain.The sparseness constraint will find the most effective focal operator to describe each seismic event in the input data.(2)The forward model describing conventional and blended seismic data is given by using a matrix notation.This is an essential step in solving inverse problems.In addition,it is investigated what is the influence of the accuracy of the focal operator,as usually we choose simple,effective levels using NMO velocities.When more accurate focal operators are chosen,seismic events can better compress and the deblending capabilities should improve.It turns out that this has only small effects on the deblending quality for some examples at hand.(3)It is investigated how more efficient solvers can be used to replace the thorough,but expensive SPGL1 solver.In particular,Chapter 3 focused on the use of a so called greedy solver.The greedy inversion introduces a coherence-oriented mechanism to enhance focusing of the significant model space,leading to a sparse model space and fastconvergence rate.It first defines a subspace in the transform domain by a thresholding procedure,after which this subspace is estimated with an ?2norm inversion.(4)Because the focal transform can be used for both reconstruction of under-sampled(aliased)data as well as blended data,it is investigated if both problems can be solved simultaneously.From the examples in Chapter 4 it is shown that indeed both problems can be handled simultaneously although there is of course a limit to what the methodology can achieve,as we need enough information to steer towards the right solution.Adding noise to the input data will create also some noise in the output,so therefore,a denoising process applied in advance can help to improve results.In Chapter 5,especially the case of irregularly sub-sampled seismic data is considered.A surprising outcome of this study is that the reconstruction capabilities of this joint deblending/reconstruction process is not dependent on the fact of data is regularly or irregularly subsampled,whereas the experience within the field of compressive sensing seems otherwise.(5)It is investigated if other transform domains can also assist in deblending.Particularly,the seislet transform is considered as a flexible and effective transform domain to apply deblending under sparsity constraint.Some first examples on field data are treated. |
PDF Full Text Request