A neural network is applied to high-quality 3-D seismic data during micro-seismic facies analysis to per- form the waveform analysis and training on single reflection events. Modeled seismic channels are estab- lished and the real seismic channels are classified. Thus, a distribution of micro-seismic facies having a high precision over a flat surface was acquired. This method applied to existing geological data allows the distribution of areas rich in coal bed methane to be clearly defined. A distribution map of the micro-seismic facies in the research area is shown. The data accord well with measured methane con- tents, indicating that the analysis using micro-seismic facies is reliable and effective. This method could be applied to coal bed methane exploration and is of great importance to future exploration work and to an increase in the drilling success rate. A neural network is applied to high-quality 3-D seismic data during micro-seismic facies analysis to per- form the waveform analysis and training on single reflection events. Modeled seismic channels are estabished and the real seismic channels are classified. , a distribution of micro-seismic facies having a high precision over a flat surface was acquired. This method applied to existing geological data allows the distribution of areas rich in coal bed methane to be defined defined. A distribution map of the micro-seismic facies in the research area is shown. The data accord well with measured methane con- tents, indicating that the analysis using micro-seismic facies is reliable and effective. This method could be applied to coal bed methane exploration and is of great importance to future exploration. work and to an increase in the drilling success rate.