The Research Of Data Assimilation In Fractured Shale Gas Reservoir Using Ensemble Kalman Filter

The accurate characterization of formation properties is crucial to numerical modeling of flow and transport in reservoir system.In the shale gas reservoir,since there existing a huge amount of natural fractures in the formation,the recognition of fracture system is a vital factor for shale gas production.But the ordinary fractures investigation methods,such as logging,coring,etc.,are usually error-prone,which could lead to the unreliable fractures characterization in reservoir model,and further,lead to the forecast errors in production period.The ensemble Kalman filter(EnKF)method,also known as automatic history matching method and geologic property inverse method,has been reported to be effective to use dynamics information to reduce the uncertainty within the geologic model.Through EnKF,the geologic features,which is extracted from the field observation as the constrain factor,can be used to update the model parameter,such as the static parameter,dynamic parameter,etc.In this research,to solve the difficulties of EnKF application in the fractures inversion problems in shale gas reservoir,the EnKF combined with dual porosity modeling method is proposed to build the fractured reservoir models.And the capability,sensitivity and application of the proposed method are analyzed in this thesis.Firstly,in the EnKF/DPDP capability demonstration part,the Sequential Gaussian Simulation and Boolean Stochastic Method are used to characterize the matrix and fracture system in shale gas reservoir model and quantitatively describe the uncertainty.Though Python programing language coupling the Eclipse reservoir simulator,The EnKF forecast and parameters updating steps are carried.With the observations assimilated with time,the parameter fields are updated and the uncertainties decrease accordingly.The simulation results shows that,the degree of uncertainty of posterior model is tremendously low compared with those of priori model.The parameter field,such as the fracture and matrix permeability pattern,can also perfectly fit those of reference model and the posterior model can also fit the history dynamic data very well.Secondly,in the research part of influence of priori data of fracture distribution to the data assimilation,3 different fracture models,deterministic fracture,global stochastic fracture and stochastic fracture cluster,are used to conduct the fracture modeling.Since the geologic data missing could improve the uncertainty degree of fracture modeling,the thesis investigated the assimilation quality under such condition by deducting the quality and quantity of the fracture statistics of fracture dataset.The research results show that,the EnKF/DPDP has powerful effect constraining the model parameters even without any fracture distribution statistics,in which condition the sequential Gaussian simulation are used to modeling the fracture parameter field.And with the fracture data added into the priori dataset,the quality of history matching and parameters estimation are improved accordingly.And within all 3 types fracture models,the stochastic fracture cluster has greater affects to the data assimilation procedure compared with all other models.Finally,in the study of EnKF/DPDP application part,such method is used to conduct data assimilation work in a 12-stage fractured horizontal well,updating the permeability distribution of fracturing area and pressure sensitive curve,in Haynesville shale formation.The research result show that,the proposed assimilation method can estimate the quality of horizontal well fracturing greatly and could describe the fracture close phenomenon caused by pressure drop in the gas production period.The history data also is fitted well in the posterior model,which demonstrate the application merit of such assimilation method.