Research On The Key Techniques Of Parameterization And Intelligent Identification Of Seismic Facies For Clastic Rock Reservoir

With the constant exploration of the oilfield,the requirement of reservoir prediction accuracy and validity is higher and higher.Currently,traditional seismic facies analysis method and reservoir prediction method focus just on the local characteristics of seismic data and only use structural attributes which shows too much dependency on the expert experience and rock physics theory.Besides,the structure characteristics and all the elements controlling deposition are not taken into consideration.Faced with more and more hidden and complex exploration targets,the existing seismic reservoir prediction technology could not tackle the special problems such as weak reflection in calcareous mudstone and sandstones.In order to improve the accuracy of the seismic reservoir prediction,it needs innovation from the method.On the basis of conventional seismic facies analysis method,I studied the sedimentary facies model and build the seismic facies pattern library first.Afterwards,the relationship between the seismic facies and sedimentary facies was established.Also the specific geological meaning of seismic attributes was given in the paper and based on which,I did massive data analysis and intelligent identification test.Finally,under the constraints of the regularity of the sedimentary facies distribution and the relevance of the seismic attributes,I realized automatic recognition of seismic facies.The seismic reservoir prediction accuracy is further improved.The main research contents and innovation points are listed as follow:1.The lacustrine facies sand body genetic model is proposed based on key sedimentary transformation.The delta stage partition and genetic types of sand bodies deposition is put forward for the construction and destruction conversion level control of the delta.The deep water sedimentary bodies can be divided into four kinds of genetic sand body: sliding,slumping,clastic and turbidity.The sedimentary characteristics of all kinds of genetic sand body field outcrop are summarized through field history.The typical deep water sedimentary outcrop external geological model and the internal phase of lithology and lithofacies library are established.On this basis,the delta deep water sedimentary are divided from genetic types of delta phase and sand body genetic and statistics from sand body scale in Es3 middle sub-member of Dongying sag.A geology model libraries of delta-deep water sedimentary is formed in Dongying sag,including the library of internal structure and external form scale libraries.2.The technique of seismic facies digital representation is put forward.The seismic amplitude,frequency and continuity are characterized digitally by using the method of statistical sequence.The seismic internal reflection characteristics arecharacterized digitally according to the texture parameters.Through the chain code of seismic reflection event,the seismic external contours are characterized digitally with invariant moment parameter.3.A typical clastic sedimentary digital characterization library of seismic facies is established.The different sedimentary microfacies characteristics and space combination form is summarized and geological model is built,according to the lithologic facies and logging data analysis.The seismic pattern library combined different sedimentary subfacies(microfacies)is built with forward modeling and matching the actual seismic data;Based on seismic facies digital characterization techniques,two kinds of typical clastic sedimentary are established: delta–deep water sedimentary and digital characterization library of meandering channel sand body seismic facies model.4.The intelligent recognition method(reservoir prediction)to three kinds of seismic facies is researched by using seismic attributes,clustering particle swarm optimization and image training study.The several parameters reflected the strata sedimentary characteristic and seismic wave mathematical characteristic are put forward in the paper.The attributes are optimized alternative using genetic algorithm.The method of automatic identification of seismic facies is researched using many kinds of data mining algorithm such as support vector machine,BP neural network,naive Bayes,decision tree and so on.The seismic facies is predicted with depth learning method.The algorithm of clustering particle swarm optimization is proposed to realize integration of the BP neural network with particle swarm optimization algorithm.The stability of particle swarm neural network algorithm has been improved after the cluster analysis algorithm optimization.The image automatic recognition method of seismic facies structure is researched by experiment,including establishment of image seismic facies structure data set,convolution neural network model structure,the establishment of method and model training,and location identification method of the seismic facies structure.