Based on complex geological form in Luning Mine, 3-d seismic exploration technology was used to explore. 4 seismic attributes were used in No.2 coal seam including average peak amplitude, slope reflection strength, percentage higher than the amplitude threshold and frequency sequence F1; 4 seismic attributes were selected in No.3 coal seam as well which included frequency sequence F3, amplitude of kurtosis, energy half a failure of the slope and frequency of the zero value. According to use BP neural network model for prediction technology, spectrum decomposition technique and reservoir prediction techniques on the basis of multi seismic features, thickness of coal seam and flush with coal seam in the middle river were explained, 3-d seismic data predicting thickness of coal seam and method model of erosion by ancient rivers which were appropriate for Luning Mine were summarized.