Detection of microseismic events based on time-frequency analysis and convolutional neural network

Aiming at the problem that traditional microseismic event detection methods have cumbersome pretreatment steps and severe manual intervention, a novel detection method is proposed based on the time-frequency analysis and convolutional neural network (CNN). The actual microseismic signals of oil and gas well hydraulic fracturing is treated as the original data. Then, the sample data set is constructed by using the spectrum extracted by S-transform. Finally, the CNN is constructed to realize the feature extraction and classification recognition of the time-spectrum samples. In order to verify the feasibility of the proposed method, both synthetic microseismic signals with low signal-to-noise ratio (SNR) and different types of surface microseismic signals of oil wells are tested, respectively. The results show that the method can effectively detect multiple types of microseismic events, including low SNR signals and weak signals. Compared with the algorithms combining CNN with other time-frequency analysis methods such as short-time Fourier transform and wavelet transform, the detection method based on S-transform and CNN has higher recognition accuracy and stability. © 2021, Editorial Office of Journal of China University of Petroleum(Edition of Natural Science). All right reserved.