基于多模态神经网络的微地震事件检测

Microseismic event detection based on multi-modal neural network

1.东北石油大学计算机与信息技术学院,黑龙江大庆 163318;
2.东北石油大学物理与电子工程学院,黑龙江大庆 163318;
3.东北石油大学人工智能能源研究院,黑龙江大庆 163318

1. School of Computer and Information Technology,Northeast Petroleum University,Daqing 163318,China;
2. School of Physics and Electronic Engineering,Northeast Petroleum University,Daqing 163318,China;
3. Artificial Intelligence Energy Research Institute,Northeast Petroleum University,Daqing 163318,China

针对微地震有效信号时序特征存在的局限导致微地震事件识别准确率不高的问题,提出了一种基于多模态学习的神经网络微地震事件检测方法。首先,利用道集数据的相关性以目标道为轴对称制作多道时域模态,对目标道进行时频分析得到S域模态特征;然后,联合时域模态和S域模态设计微地震事件检测神经网络,综合多模态的特征进行训练学习,提高微地震事件识别的精度;最后,为验证方法的有效性,对合成微地震信号进行低信噪比数据分析、小幅值数据分析以及实际油井微地震监测信号事件分析。结果表明,该方法可以有效检测低信噪比及微弱的微地震事件;与支持向量机、卷积神经网络、基于监督机器学习方法的对比实验结果表明该方法具有更高的抗噪性与准确率。

A multimodal neural network-based microseismic event detection method is proposed to address the problem that the time series of effective microseismic signals has severe limitations.First,the multichannel time-domain mode with the target channel as the axis symmetry is established using gather data correlation,and the S-domain modal characteristics are obtained by using time-frequency analysis for the target channel.Then,the neural network for microseismic event detection is designed by combining the time-domain mode and S-domain mode.Multimodal features are synthesized for training and learning to improve the accuracy of detection.Finally,method validation is performed through the analyses of synthetic low-SNR and small-amplitude data and actual oil-well microseismic events.The results showed that our method could detect low-SNR and weak microseismic signals effectively.Compared with SVM,CNN,and supervised machine learning,our method has improved anti-noise performance and accuracy.

东北石油大学特色科研团队项目“智慧油田信息处理创新团队”(2023TSTD-04)资助。

10.12431/issn.1000-1441.2024.63.04.008