Research On Microseismic Data Processing Based On Spectral Multimanifold Cluster Method

With the increasing of oil and gas requirement in the world,they are gradually consumed as non renewable resources.Unconventional oil and gas become a new bright spot for development.Artificial seismological technology has been widely used in the fields of unconventional oil and gas exploitation.Microseism is a small vibration event caused by hydraulic fracturing technology,which injecting water at the bottom selected position and changing the bottom stress field through water injection.According to the difference of the main frequency,the microseismic signals are divided into ground microseismic and well microseismic.At present,microseismic monitoring is mainly based on three monitoring geophones with different components near the fractured well,and the monitoring results are obtained by continuous recording of three components of microseismic events.However,the acquisition of geophone is often accompanied by several problems,making the received signal difficult to distinguish.First,the surface is often uneven,and the different components of the same signal are not extended at the same time,and the adjacent return records on the same bottom are also different.Secondly,the vibration of hydraulic fracturing is very weak,so it is more likely to receive the interference of human noise and natural noise,so that the effective signal that can determine the bottom boundary is submerged in the noise.Third,the relative distance and angle of the geophone distribution may affect the received signal.Because of the above three points,it can be concluded that it is necessary to deal with the microseismic signal.It will help us to determine the location of the signal and determine whether the oil and gas resources are contained.In the processing of microseismic signals,there are two common processing directions.One is noise suppression and the other is P wave's time-picking.Therefore,the traditional method of suppressing noise,such as shearlet transform hard threshold,is not strong because of low signal to noise ratio(SNR)and close frequency to noise.The traditional methods such as AIC and STA/LTA are all processing signals in time domain.It is difficult to achieve satisfactory results when the quality of microseismic data is poor.In order to solve the above problem,we propose a new separation method based on the difference between the signal and the noise on the manifold.On this basis,the purpose of picking up and suppressing the noise is achieved.In this paper,a method of micro seismic signal processing based on spectral multimanifold clustering is proposed.First,esidual statics estimation by stack-power maximization method is carried out to compensate the uneven signal caused by the uneven surface of the surface.Then,the spectral multimanifold clustering method in the nonlinear dimensionality reduction method is selected and the corrected signal is used as a high-dimensional signal.Finally,the low dimensional representation can be known.This method separates the signal from the noise by the manifold difference between them.It successfully avoids the small difference in the amplitude,energy,frequency,and correlation between the effective signal and the noise of the microseismic signals and finally succeeded in separating effective signals and noises.In this paper,the algorithm is applied to the processing of microseismic signal,and the signal and noise are distinguished.After the location of the signal,noise suppression and time picking is also achieved.Compared with traditional noise suppression and time picking methods,we find that our algorithm achieves excellent performance in magnify diagram,accuracy rate line diagram,form,single channel contrast diagram and difference diagram.When the input signal to noise ratio is-6d B,the accuracy rate can reach 92.5%.Noise suppression can increase the signal to noise ratio from-5.1003 to 11.6525 d B.It can be proved that this algorithm has successfully applied the nonlinear dimensionality reduction theory to the field of seismic signal processing for the first time.By using the SMMC method,the method we proposed is better to solve the problem that the effective signal and noise are difficult to separate under the low signal to noise ratio.