Research On Methods Of Optimal Filtering Based On Identified Model Of Microseismic Signal

Currently,microseismic monitoring technology has been widely used in fracturing of oil and gas well.Microseismic events generated in fracturing process usually have characteristics of weak energy,low signal-to-noise ratio(SNR)and sometimes valid signals are nearly submerged by environmental noises,which have greatly influenced the SNR.However,the SNR is the basis and key in the processing of microseismic data and has a significant effect on the accuracy of source location.If the microseismic data with low SNR is used for source location,the final results will appear serious deviation.Therefore,microseismic data must be preprocessed by various filtering methods to improve the SNR.As the major means to improve the SNR,the research on denoising methods has always been the focus of domestic and foreign experts and scholars.At present,various methods have been applied to microseismic data filtering and denoising,such as adaptive filtering,wavelet analysis and Kalman Filter(KF)based on state space model.Under the research background of microseismic monitoring technology and based on the research objects of microseismic signals in this paper,the identified models of the synthetic and practical microseismic signals are established successfully and the KF,UKF,SR-UKF filtering algorithms based on identified models have been carried on the exploratory research for the existing limitations of available denoising methods.The main work of this paper is divided into the following parts.(1)The identified modeling of the synthetic and practical microseismic signals is one of the main contents in this paper,which is also the necessary condition and basis of the optimal filtering algorithms.In this paper,microseismic signals are studied and analyzed.The synthetic microseismic signal is typically modeled as an exponentially decaying cyclic waveform and the practical microseismic signal is extracted from microseismic data with high SNR.In order to eliminate the errors caused by the assumptions and simplifications in mechanism modeling and to improve the accuracy of the mathematical model,the ARMA models of synthetic and practical microseismic signals are established by Prony algorithm.Moreover,the ARMA models are further converted into state-space models,which are suitable for KF framework.(2)In this paper,the KF algorithm is designed,which is based on identified models of microseismic signals.On the basis of the identified models,the KF algorithm is designed,which is suitable for microseismic data denoising.The filtering results of synthetic and practical microseismic data show that KF algorithm can suppress the noise mixed in synthetic microseismic data effectively.However,due to the factors such as complexity and variety of actual noise mixed in microseismic monitoring data,there are still some unsatisfactory issues that the filtering accuracy and the denoising results of KF is not ideal.(3)Based on the identified models of microseismic signals and unscented Kalman Filter(UKF),the square-root unscented Kalman Filter(SR-UKF)is introduced and designed,which is suitable for microseismic data denoising for the case of non-positive definite covariance matrix and consuming calculations in the iterative process of UKF algorithm.The filtering results of synthetic and practical microseismic data show that the SR-UKF algorithm based on identified models can suppress the noise effectively,and improve the filtering effect of KF algorithm and the numerical stability of UKF algorithm.(4)The paper concludes the limitations and shortcomings of KF,UKF and SR-UKF algorithm in the practical filtering process of microseismic data and discusses the improved filtering algorithm which can be implemented in the follow-up research.