Study on micro-seismic detection in coalbed methane based on multiscale neural network

Abstract

In the exploration of coalbed methane, it is of great significance for safe production to master the direction of artificial fractures and adjust the position and orientation of coal mining roadway through micro-seismic monitoring technology. In this paper, a micro-seismic detection method based on a multi-scale convolutional neural network is proposed, which better solves the problem of weak signal recognition in a strong interference environment. The results show that the detection precision and recall rate of the proposed method is better than those of wavelet analysis, BP network, and convolutional neural network under different intensities of noise interference, and have good noise immunity. When the signal-to-noise ratio is greater than 6 dB, the model detection recall rate can reach more than 90%, the accuracy rate can reach more than 92.1%, and the model has good performance through the verification of the actual coal-bed methane micro-seismic monitoring data in a certain area of Heilongjiang Province.

Keywords

• coalbed methane
• micro-seismic detection
• neural network
• certainty factor
• multiscale detection