Application Of S-Transform Based On Shearlet Correlation In Microseismic First Arrival Picking

With the large consumption of oil and gas resources,unconventional oil and gas reservoir exploitation has become a new research hotspot.Microseismic monitoring is the main technical means,which has great guidance on the development of low permeability oilfield.First arrival picking is a significant part of microseismic monitoring technology,and it is a basic step in downhole microseismic data processing.Picking arrival times automatically and accurately is an important precondition for high-precise microseismic hypocenter location.Due to the weak energy of valid signal and the low signal-to-noise ratio(SNR)of microseismic data,the first arrivals obtained by conventional methods are unsatisfactory,frequently be false or missing,which has a severe effect on the accuracy of hypocenter location.Therefore,it is of great significance to establish a novel first arrival picking method that applicable for low SNR microseismic data.This dissertation launches the research on first arrival picking for microseismic data,aiming to improve the picking accuracy in low SNR.There is no significant characteristic difference between the effective signal and noise in low SNR,which makes it difficult to complete arrival time picking effectively in time domain.Shearlet transform is introduced into first arrival picking in this dissertation,through which the microseismic data is mapped from time domain to shearlet domain.Based on the coefficient differences between the signal and noise at fine scales,the signal points can be preliminary identified from noise.Then,we use the scale correlation between adjacent scales to correct the selected signal points,which improves the accuracy of signal recognition.Finally,the first identified signal point is regarded as the first arrival.The performance of the proposed method was tested on synthetic and field microseismic data,experimental results indicated that compared to the short-term average and long-term average(STA/LTA)and the Akaike information criterion(AIC)algorithms,shearlet transform method improves the picking accuracy at low SNR effectively.The realization of first arrival picking method based on shearlet transform shows the effectiveness and feasibility of signal and noise distinction in shearlet domain.However,the parameters in method such as the threshold and correction value willaffect the final picking results and limit its practical applications.In order to further improve the picking performance in low SNR,we propose a first arrival picking method using S-transform based on shearlet correlation(SC-ST).Firstly,the correlation of signal coefficients between adjacent fine scales is used to enhance the signal coefficients,which makes the signal and noise have more obvious characteristic difference.Secondly,the coefficients are processed by S-transform,after which the signal energy can be effectively gathered.Finally,the first arrival picking is realized based on energy difference between the signal and noise.Due to the time-frequency clustering of S-transform,the effective signals can be identified more directly and accurately through the time-frequency representation,so that the proposed algorithm has better arrival time picking performance.In order to verify the validity of the proposed algorithm,we apply SC-ST method to noisy synthetic microseismic data with white Gaussian noise and real noise.The picking results are compared with those of STA/LTA,AIC and shearlet transform methods.The comparison results show that the SC-ST algorithm is superior to the other methods in the adaptability to noise and the accuracy of arrival time picking.In addition,the picking performance of SC-ST algorithm at different SNRs is analyzed by a large number of experiments.The experimental results on field microseismic data demonstrate the superiority of the proposed method,it is an effective and high accuracy first arrival picking method and it can provide accurate and reliable first arrivals for low SNR microseismic data.