油井动液面位置智能识别算法研究

Research on intelligent recognition algorithms for dynamic fluid level position of oil wells

针对采用声波探测法测得油井动液面数据时,采集到的信号由于受到长距离衰减和复杂背景噪声的影响,动液面回波位置容易淹没在复杂噪声之中不易识别的问题,本文采用一种新的带有宽第一层核的深度卷积神经网络(WDCNN)的方法。即使用采集到的原始声波信号作为输入,并使用第一卷积层中的宽内核来提取特征和抑制高频噪声;卷积层中的小卷积核用于多层非线性映射,池化层用来减少特征的空间大小和网络的参数;在输出层使用softmax函数转化识别的不同液面深度值。现场试验结果表明,构建的WDCNN模型提高了动液面位置识别的准确性与识别效率,智能识别技术取代了传统的耗时且不可靠的人工分析,降低了油田开采生产成本,提高了经济效益。

Aiming at the problem that the echo position of the dynamic liquid level is easily submerged in the complex noise due to the influence of long-distance attenuation and complex background noise when the acoustic detection method is used to measure the dynamic liquid level data of oil wells,this paper uses a new method of Deep Convolutional Neural Network (WDCNN) with a wide first-layer kernel.That is,the original acquired acoustic signal is used as an input,and a wide kernel in the first convolutional layer is used to extract features and suppress high frequency noise.The small convolution kernel in the front layer is used for multi-layer nonlinear mapping,the pooling layer is used to reduce the spatial size of the feature and the parameters of the network;the softmax function is used in the output layer to convert the identified liquid surface depth value.Field test results show that the WDCNN model improves the accuracy and efficiency of dynamic liquid level position recognition.Intelligent recognition technology replaces the traditional time-consuming and unreliable manual analysis,reduces the cost of oil field production and improves economic benefits.