Research And Application Of Seismic Waveform Classification Technology

The seismic waveform is a comprehensive reflection of seismic attributes such as amplitude,frequency,and phase.Seismic waveform classification is a commonly used reservoir prediction method.Seismic waveform clustering is a kind of seismic waveform classification.It only depends on the seismic data itself and does not rely on a priori information,so that the seismic data of the same classification phase is clustered under the same cluster of interpretation.K-Means(K-means)is a commonly used seismic waveline clustering method based on Euclidean distance.K represents the number of classifications.This paper proposes a new method of K-Medoids(K center point)waveform clustering based on DTW(Dynamic time warping)algorithm.The innovation of this method includes the following three aspects:(1)For the optimization problem of the classification number K,use the elbow discrimination method,pre-select and optimize the initial K value to determine the classification number(the elbow discrimination method is named because its intersection graph is elbow-shaped);(2)adopt the DTW algorithm Accurately and automatically pick up the target horizon to be classified,aiming at the horizon error caused by manual interpretation;(3)The K-Medoids algorithm is used to solve the meaningless centroid problem,and it has better noise resistance than the K-Means algorithm.This paper first uses numerical forward modeling data and seismic physics model data based on the Longgang area in central Sichuan to test the new method,and proves that the method can better characterize fault edges and smaller faults than the traditional K-Means algorithm based on Euclidean distance.Geological features such as fracture groups.Finally,the seismic waveform clustering method is applied to the target layer of actual data in Daniudi gas field.By comparing and analyzing traditional seismic attribute(such as variance and amplitude envelope)extraction methods,K-Medoids waveform clustering method based on DTW algorithm can The edge is more accurately portrayed,which is more advantageous in describing details.