Abstract：Taking the volcanic in upper cycle of Yingcheng Formation in SP area of Songliao basin, the paper utilizes self-organization neural network approach to carry out waveform classification, uses such parameters as time-window, amplitude, frequency and facies to implement training and divides 15 kinds of model traces after 30 iterative computations. The seismic waveforms can be seen as pieces distribution or strips distribution along the faults on the resulted seismic facies map, which is coincident with geologic background. Then, following the lithofacies-naming principle, we calibrated volcanic facies at single well based on drilling lithofacies and conducted planar prediction of lithofacies. The distribution law of predicted volcanic facies is consistent with the statistical raw of drilling lothofacies. Application of the predicted lithofacies results to deploy the development well network of volcanic gas reservoir in SP area achieved good effects, showing the feasibility using the approach to predict the volcanic facies.
