Research On Algorithm Of Dynamic Liquid Level Measurement In Oil Wells

The dynamic liquid level of a pumping unit well is the dynamic liquid level formed by the crude oil seeping from the annular space between the rod and the casing during the normal production operation of the pumping unit.Real-time monitoring of dynamic liquid level depth is an important parameter to ensure safe and efficient production operation of oil wells and to maximize production efficiency and output of oil wells.When using acoustic detection method to measure the depth of hydrodynamic surface,the acoustic signal is affected by the complex background noise and the attenuation of long-distance propagation,which makes the measured hydrodynamic surface echo signal curve more noisy and difficult to identify,so that the position of hydrodynamic surface can not be accurately identified by the measured echo signal,which affects the calculation of the depth of hydrodynamic surface.In order to solve this problem,this paper focuses on the analysis of the problems of the traditional Fourier transform,wavelet transform,spectral subtraction and other denoising algorithms,and studies the application of deep convolution neural network in signal processing and hydrodynamic position recognition.The deep convolution neural network with wide first layer core is used to extract the reflected wave signal from the mixed signal with complex background noise,and the position of the moving liquid level is accurately identified.By adding Gauss white noise to the ideal hydrodynamic level signal to simulate the real hydrodynamic level signal,the size of the first layer convolution core is constantly changed.According to the change of the recognition accuracy of hydrodynamic level signal,it is proved that it is necessary to increase the first layer convolution core to improve the recognition accuracy of hydrodynamic level.At the same time,the WDCNN model is compared horizontally with other denoising recognition algorithms.In the case of variable signal-to-noise ratio,the anti-noise performance of various algorithms is analyzed.Finally,the recognition and analysis of the measured dynamic liquid level signal are carried out.The original dynamic liquid level signals collected in real time from oil wells in Yanchang Oilfield are used as training samples to train and optimize the model,and model parameters suitable for dynamic liquid level signal position recognition are selected.Finally,the data collected by three oil production plants are tested and compared with other denoising methods such as band-pass filtering,spectralsubtraction,sparse decomposition and FFT-SVM.The final results show that the WDCNN dynamic liquid level recognition algorithm has the highest recognition accuracy and efficiency,and can meet the needs of oilfield production.Intelligent identification technology replaces the traditional time-consuming and unreliable manual analysis,reduces the production cost of oilfield production and improves economic benefits.