| Hydraulic fracturing diagnosis technology based on water hammer pressure signal has low cost and strong real-time performance,which has great development potential.In the actual process of fracturing pump shutdown,the collected signal may contain a lot of noise,which may be caused by various events such as fracture opening or pipeline vibration.The frequency spectrum of these noises is overlapped with part of the frequency spectrum of the water hammer pressure wave signal,which brings great difficulty to the spectrum analysis.At the same time,the response time on the pressure curve is presented as small pressure disturbance information,which may be drowned by noise.All these factors will affect the determination of response time and l reduce the accuracy of analysis.Therefore,an important problem of hydraulic fracturing diagnostics technology is to use appropriate filtering methods to remove all kinds of noise in the process of pump shutdown.This thesis analyzes the characteristics of water hammer pressure wave signals and noises,and builds a filtering method based on this signal.The main research contents are as follows:Firstly,we set up experimental equipment and collected laboratory and field signals.At the same time,the characteristics of water hammer pressure wave signal and noise are analyzed for the first time by using time-frequency domain and other methods.The frequency of water hammer pressure wave signal is low frequency within 2 Hz,including a lot of spike noise,random noise,fixed frequency noise which is close to the useful signal frequency or has a big difference from the useful signal frequency.In addition,the water hammer pressure wave pressure signals of some fracturing stages show the characteristics of square wave,and the frequency spectrum shows fundamental frequency and odd harmonic frequency.Feature analysis provides a basis for the research of subsequent filtering methods.Secondly,A new comprehensive filtering model for water hammer pressure wave signal is proposed.We analyze the filtering effect of analog signals without fractures and with two fractures.The SNR gain is about 6.8 d B,the mean square error can be reduced about 0.055,and the pressure response time with fractures can be retained and recognized well.The experimental results show that the model can retain useful signals well.At the same time,we also discuss the application scope of this method.The amplitude of low frequency noise is about 0.15,the amplitude of high frequency noise is about 0.2,and the variance of random noise is less than 0.15~2.Both random noise and high frequency noise cannot be greater than these two limits at the same time.The comprehensive filtering method provides the foundation for the subsequent intelligent filtering method.Finally,an adaptive noise cancellation system based on neural network for water hammer pressure wave signals is built.We use three different signals to do simulation analysis and study the filtering effect of two input noises with linear correlation and nonlinear correlation.When the two noise signals are linearly correlated,the SNR can be improved by about 46 d B and the mean square error can be reduced by about 1.99after filtering.When the two noise signals are nonlinearly correlated,the SNR can be increased by about 37 d B and the mean square error can be reduced by about 0.7.The adaptive noise canceller based on BP neural network has a good denoising ability.This paper studies the filtering method of water hammer pressure wave signal in fracturing stop pump,which can provide a good guide for the processing and filtering of water hammer pressure wave signal in site. |
PDF Full Text Request