Application of neural network seismic waveform classification technology in thin sand body prediction
Wang Lei*, Yuan Lichuan, Yingchun He, Ying Zhang, Ke Qin, Ding Chaolong · 2020
Due to the resolution of seismic data, there is uncertainty in the prediction of thin sand bodies. A seismic waveform is mostly formed by two or more stages of sand body superimposed interference. Therefore, it is difficult to accurately describe single-stage sand bodies with conventional properties such as amplitude and frequency. In this paper, the neural network seismic waveform classification technology is used to study the variation law of seismic waveforms under different sand body superposition modes, and the typical sand body superposition pattern in the study area is established, and the appropriate time window is selected for self-organizing learning of seismic waveforms, according to the time window. The spatial distribution of different seismic waveforms, and the plane heterogeneity characteristics of different sand body superposition modes are described in detail. The method was applied in the prediction of thin sand body of Qingyi III sand group in the Daqingzijing area in the southern Songliao Basin and achieved good results.