Experimental Study On Water Detecting With Optical Fiber In Horizontal Wells

Horizontal well water production reduces oil recovery,resulting in poor development of oil fields.In order to achieve water seeking in horizontal wells,a horizontal well water finding method based on distributed fiber sound monitoring(DAS)was proposed.In this paper,single-phase fluid and two-phase fluid inflow experiments were carried out on a horizontal wellbore production profile monitoring simulation experimental device based on distributed fiber sound monitoring(DAS).The experimental results show that the sound signal energy in the sound waterfall diagram is related to the position of the fluid flowing into the horizontal wellbore.The larger the fluid flow rate is,the stronger the sound energy corresponding to the inflow position is and the brighter the color band is.The sound signal collected by the experiment is processed by the two-channel adaptive noise reduction method.The results show that the method can suppress the surrounding sound source,significantly reduce the noise in the signal and improve the signal-to-noise ratio of the signal.The frequency domain analysis of the fluid sound signal shows that water,oil and gas have their own frequency range,and some frequency ranges overlap.The frequency bands of sound generated by oil,water and airflow into the wellbore are concentrated at 800Hz-2300 Hz,Between 500Hz-1600 Hz and 1100Hz-3500 Hz.The results of the study on the voiceprint characteristics of the horizontal wellbore fluid inflow sound signal indicate that the linear predicted cepstral coefficient(LPCC)and the Mel frequency cepstral coefficient(MFCC)of the acoustic signals generated by three fluids of water,oil,gas are grouped around their respective categories,and the extracted timbre features can better differentiate the fluid sound types.Based on the above research,the method of support vector machine is used to study the fluid type identification method.The SVM classifier model of three fluid sounds of water,oil and gas is established.The test results using indoor experimental data show that the SVM The classifier model has a good sound recognition effect,and the recognition rate of water can reach 95%;the greater the water inflow speed,the better the recognition effect.The research results provide a technical reference for the engineering application of horizontal well water search.