卷积神经网络微地震事件检测

Detection of microseismic events based on convolutional neural network

常规的微地震事件检测方法主要基于信号的特征计算,事件检测准确性依赖于算法的参数设置,受信号特征和信噪比变化的影响较大。为此,提出了一种基于卷积神经网络的微地震事件检测方法。该方法首先使用实际的油井水力压裂多站点微地震监测信号构建神经网络的样本数据集,样本数据包含有效事件信号和无效背景信号及其分类,然后用样本数据对神经网络进行训练和测试,得到微地震事件识别准确性最高的神经网络模型。使用训练好的神经网络模型对不同信噪比的合成微地震信号以及川渝地区多口油气井压裂微地震监测信号进行微地震事件检测。数据处理结果表明,训练好的卷积神经网络模型能有效自动检测微地震事件,且具有较好的抗噪性和泛化能力。

Conventional detection methods for microseismic events are almost based on calculating the features of signals.The accuracy of events detection depends on the parameters of algorithms,so it is greatly affected by the changes in the features of signals and signal-to-noise ratio.This paper proposes a method for microseismic event detection based on a convolutional neural network(CNN).To train and test the CNN,a sample set is constructed on the microseismic data monitored by multiple stations in an oil well that was hydraulically fractured.The data consist of effective event signals and ineffective background noises and their classifications.Then the CNN is trained and tested by the sample data set,and an optimal CNN model is obtained with best accuracy of event detection.To test the performance of the CNN model,synthesized microseismic signals with different signalto-noise ratios,and actual microseismic signals from several oil and gas wells are fed into the CNN model.The processing results demonstrate that the CNN model can automatically and effectively detect microseismic events.It has good abilities for noise suppressing and generalization.