Preprocessing And Research Of Denoising Methods For Marine Controlled Source Electromagnetic Data

With the increasing demand of the energy, it is becoming more important thatpetroleum and gas resource is explored in deep sea areas. Marine Controlled-SourceElectroMagnetic (MCSEM) method is a relatively young geophysical method forseafloor oil and gas prospecting, it has achieved remarkable results through thetheoretical development in decades and practical applications of seafloor oil-gasreservoirs exploration in recent years. At the moment, the main way of theprospecting method is deploying receivers on the seafloor, survey vessel tows theunderwater transmitting system to transmit the MCSEM signal, one need study theproperties of the target reservoirs, identify the high resistance oil-gas layers, and thenimprove the drilling success ratio by processing and analyzing the received electricand magnetic field signal. The MCSEM method research just starts in China, and it islack of systematic data pre-processing. Therefore, this study designs a MCSEM datapre-processing flow and implements a corresponding MCSEM data pre-processingsystem. In addtion, under the seafloor conditions with selected source-receiverdistance and transmit frequencies, the MCSEM surveying results will be affected bydifferent kinds of noise, which influence the accuracy of interpretation and follow-upinversion results. This study specially aims at analysis the characteristics of the noisesin MCSEM signal and presents the corresponding denoising methods especially forthe low SNR (signal-to-noise ratio) problem at the middle-far and far offset area.After testing synthetic models and field data, the results show the signal-to-noise ratiocan be effectively improved by the methods that presented in this dissertation, and itis significant to improve the interpretation accuracy and inversion data quality. Because of the different data format between receiver and navigation data files inthe actual exploration, it is necessary to study the time matching relation of differentdata files during the pre-processing and extract the correct MCSEM data for furtherprocessing. Decompose time domain data of electric field or magnetic fieldcomponent into consecutive bins based on the sampling period of fundamentalfrequency, Fourier transform, can obtain the MVO (Magnitude Versus Offset) andPVO (Phase Versus Offset) curves of corresponding frequencies by extracting theamplitude of corresponding frequency in each bin and merging with navigation dataof source. In this basis, this study focuses on the frequency spectrum leakage whenthe sub-bin of the time domain data is non-periodic. Through the comparative analysisof several different window functions, the solution for the low frequency spectrumresolution is presented. After systematically studying the data pre-processing methods,this research achieves corresponding MCSEM data pre-processing system, and thesoftware with independent intellectual property is formed.Through the preliminary analysis to the noise in the measured data, results showthat the interference noise is mainly caused by nature low frequency electromagneticfield and seafloor ocean current noise, and when the offset increased, the receivedenergy of effective signal is decreased, the effective signal in the middle-far and faroffset area will be contaminated by low frequency noise. According to thecharacteristics of the SNR in MCSEM signals change with the offset, the frequencydomain stacking method, time-varying smoothing filtering method and time-varyingbilateral filtering method are presented to suppress noise. Among them, frequencydomain stacking method stacks frequency spectrum in two times bins or four timesbins base on the length of the data, the stacked spectrum replaces the raw spectrum.When time-varying smoothing filtering and time-varying bilateral filtering mthodwork, one need to break the low SNR part of MCSEM data into fixed-length bins.Along SNR of the fixed-length bins decreases the increased smoothing radius offilters will suppress MCSEM noise. The bilateral filter has a better edge preserveingability for square wave signal with random noise, compared with smoothing filter.To futher analyze the methods, three different one-dimensional seafloor models with homogeneous layered media are established: the first one is the model withoutoil and gas high resistance layer, the second one is low sampling rate model with oiland gas high resistance layer, the third one is high sampling rate model with oil andgas high resistance layer. The first two models discusses the MCSEM response signalcharacteristics with different emission frequencies, different field component, withand without oil and gas high resistance layer. The relationship between media andwave field response is given.The frequency domain stacking method and two kinds ofnonlinear time-varying filtering methods are applied to the third model, the ability ofdifferent methods for handling noise are analyzed. The dissertation disscusses theprinciple for choosing parameters of different methods, a comprehensive denoisingscheme is given. For further verifying the validity of the proposed time-varyingfiltering methods, the one-dimensional OCCAM inversion method can compare theresults before denoising and after denoising inversion, respectively, and gives theapplicability analysis.In the last part, the proposed denoising methods are applied to the field data, andthe denoising results are analyzed. The results show that the processed MVO curvesare high fidelity in the near offset, and display more reasonable attenuationrelationship in middle-far and far offset area, meanwhile, the denosing methodsimprove the quality of MCSEM data for follow-up processing. The dissertationprovides new techniques for MCSEM noises suppression methods, but also gives areference for other relevant research fields.