The Improvement Of Sand Body Superimposed Pattern Recognition Method And Its Application In Seismic Drive Modeling

With the exploiting of the most oil and gas fields in China,remaining potentials has gradually been from a large area of connected reservoirs to small scattering areas but with local relatively rich area.Therefore,it is very important to characterize the reservoir more carefully and to establish a reservoir model which can describe the internal architecture and spacial distribution of the reservoir in order to optimize the production process of the oil field.The fluvial sands are important oil and gas reservoirs in the continental basins of China.At present,most of the researches on fluvial sand bodies focus on modern sedimentation and field outcrops.But less focus on the internal architecture and superposition patterns of sand bodies.It is very important to study the superposition model of fluvial sands and to assist reservoir modeling with the recognition of sedimentary facies and the actual seismic data.The seismic driver modeling method uses both log data and seismic data as hard data to construct the reservoir.Through the combination of log data and seismic data to calculate the transform function to estimate the unsampled data.By fully using the geological spatial information contained in seismic data,reasonable modeling results can be obtained in areas where the log is relatively sparse.On the basis of previous research,this work improves the recognition algorithm of sand superposition model.By matching the waveform information of sand body superposition template with actual seismic data and seismic multi-attributes,the algorithm can adapt to observed data for pattern recognition.The sparse auto encoding method is explored in machine learning for the recognition of sand superposition-pattern.Finally,the improved recognition method is applied to seismic driver modeling process.