| Despite of the extensive applications in realtime geosteering and formation evaluation,Logging-While-Drilling(LWD)resistivity measurement is not capable of accurately detecting the geological structure around the borehole due to its limited detection depth of investigation(DOI)and non-azimuthal sensitivity to the bed boundaries.Therefore,it is necessary to further extend the deep detection capacity of LWD electromagnetic(EM)logging tool and to bridge the gap between the shallow DOI of logging tool and the low resolution of seismic measurement.In this paper,we conduct a thorough study on the theory,detection performance and classic responses of deep detection LWD multi-component EM logging(e.g.the LWD azimuthal electromagnetic measurements(AEM)and ultra-deep LWD electromagnetic measurements),which can provide strong support for the qualitive interpretation and the tool development.Then,the fast modeling and data processing of deep detection multi-component LWD electromagnetic measurement in complex formation structures are investigated.Finally,based on the multi-component LWD EM data,the realtime bed boundary extraction,determination of anisotropic resistivity and fast imaging of formation structure are estabilished,which are helpful to the realtime geosteering and optimized oil/gas production in horizontal wells.In the second chapter,the three-dimensional(3D)geometric factor of multi-component electromagnetic logging in anisotropic medium is first derived,and the micro and macor sensitivities of different electromagnetic components are analyzed.Then,the operation theory,signal synthesis method,detection modes and classic responses of LWD AEM and ultra-deep LWD electromagnetic measurements are introduced.To overcome the difficulty of the modeling of deep detection multi-component LWD EM logging in complex formation structure,a dimensionality reduction scheme which converts the 3D modeling of complex formation structure into 2.5-dimensional(2.5D)problem is proposed.Then,by using the sliding window technique,the 2.5D modeling is further reduced to a series of one-dimensional(1D)problems.Numerical example performed over classic folding structures with both 1D and 3D modeling algorithm shows that the dimensionality reduction algorithm whose speed is 70000 times of 3D modeling is accurate and reliable.In the third chapter,the detection performances of deep detection multi-component LWD EM logging tools(PeriScope and Geosphere)to bed boundaries are first analyzed.Then,the effects of relative dip,bed thickness and resistivity contrast to the responses of LWD AEM and ultra-deep LWD electromagnetic measurements is investigated.Numerical results show that the sensitivities of phase shift and attenuation geosignals from PeriScope to bed boundary,resistivity and relative dip reduce gradually,wherase the detection capability of Geosphere to bed boundary enlarges with the increasing spacing.When the operation frequencies of Geosphere increases,the maximum distances to boundary(DTB)available from Geosphere may exist non-linear variations.By using the combination of different modes,the Geosphere tool can reflect the relative dip and formation anisotropy accurately in arbitrary dip and formation anisotropy.Additionally,due to the the effect of severe variation of formation dip and high resistivity constrast,strong nonlinearity of Geosphere curves exists,which leads to the complicated logging responses.It should be noted that for the detection modes USDP,UADP and UHRP,discontinuous phase shift change may be observed,which further adds difficulity to the inversion process.In the forth chapter,the iteration formula of Levenberg-Marquardt algorithm,the automatic updating of regularization parameter and the uncertainty analysis method of gradient algorithm are derived first.Then,a modified simulated annealing-differential evoluation(SADE)and the Bayesian algorithm with a parallel-interacting Markov Chain Monte Carlo(PI-MCMC)sampling method are introduced to overcome the difficulity of inversion caused by the severe change of cost function and multiple local minima.Compared with conventional differential evoluation(DE),simulated annealing(SA)and MCMC algorithm,the modified SADE and PI-MCMC algorithms not only guarantee the global search to the parameters,but also increase the convergence speed.Therefore,the two methods are applicapable to the inversion problems where only litte prior information is available and there are multiple parameters to be inverted.In the fifth chapter,we estabilish a automation inversion algorithm to recover the bed boundaries from LWD AEM data and the effects of noise,dip and bed thickness to the inverted results are studied.The section criteria and its feasibility of inversion model are also investigated.Finally,the inversion performance and inversion speed of deterministic inversion and stochastic inversion are compared.Numerical results show that the simplified 1D inversion method is suitable for arbitrary well trajectory and the horizontal well with relative dip larger than 75°.The inverted 2D resistivity curtain not only shows the vertical resistivity distribution of formation nearby the borehole,the distribution of formation structure can also be derived from the lateral extension of the resistivity curtain.The smaller the relative dip between the tool and formation and the stronger the lateral heterogeneity of the formation structure,the worse the accuracy and stability of inversion results are.Meanwhile,the existence of resistivity transition zone may also affect the accuracy of the estimated bed boundaries.In general,the selection of inversion model is dominated by the thickness of targeted formation,while the selection of algorithm relies on the number of the inversion model.For the bed with thickness larger than 4.0m,one can use the single-boundary inversion model and gradient optimization.By contrast,when the bed thickness is between 1.0m and 4.0m,the nearby upper and lower boundaries can be derived by using the three-layered inversion model.For the processing of LWD AEM data acquired from thin beds,the multiple-layered inversion model and stochasitic optimization method are recommended.In the sixth chapter,a formation structure imaging method from the ultra-deep LWD reservoir imaing data is estabilished and the joint inversion of LWD electromagnetic measurements with different detection scales are performed.First,one inversion model with different formation anisotropy in each layer and one inversion model with all beds sharing a global anisotropy are estabilished.Then,the inversion speed and inversion performance using the two models are performed and compared.Finally,with the help of five-layered “global anisotropy” inversion model,the joint inversion of LWD electromagnetic measurements with different detection scales is discussed.Numerical results show that based on the five-layered “global anisotropy” inversion model,accurate formation structure nearby the borehole can be recovered by using 120000 random sampling.The joint inversion of different scale LWD EM data not only improves the accuracy and resolution of the inverted result,it also increases the inversion speed.Compared with the joint inversion with LWD electromagnetic data and LWD resistivity measurements,more accurate bed boundaries,better stability and continuity of the inverted 2D resistivity curtain can be recovered by joint inversion of LWD AEM data and LWD electromagnetic measurements. |
PDF Full Text Request