Inversion Of Fractures In Fractured Reservoirs Based On Assisted History Matching

Because the distribution of fractures in fractured reservoirs has a dominant influence on fluid flow behavior,effective prediction of the distribution of reservoir fractures is the foundation for the development of fractured reservoirs.Compared with the prediction of fracture distribution by analyzing logging data or in-situ stress information,the historical production data of oilfield contains important flow information affected by fracture distribution,and the information is the basis of applying assisted history matching for the inversion of fractures.Inaccurate fracture prediction makes the simulation value inconsistent with the historical production value,and therefore the process of automatic history matching is to minimize the difference between simulation value and historical production value to obtain accurate fracture distribution.Firstly,in order to guarantee that there is close connection between the fracture distribution and simulation value,equivalent continuous medium model and discrete fracture matrix(DFM)model of fractured reservoirs are used as simulation models respectively: the simulation value of equivalent continuous medium model is obtained by equivalently transforming fractured porous medium into anisotropic porous medium based on equivalent permeability tensor and then running ECLIPSE simulator;while the simulation value of DFM model is calculated by operating the DFM module of Matlab Reservoir Simulation Toolbox(MRST).Secondly,different characterization parameters of fracture distribution are studied,and these parameters are taken as inversion parameters.Then,based on Bayesian theory,the objective function of assisted history matching which represents the relationship between inversion parameters and production value is established,and inversion process is transformed into the process of minimizing the objective function.Finally,improved simultaneous perturbation stochastic approximation(SPSA)algorithm is adopted to minimize inversion objective function to obtain the optimal inversion results.However,the multiplicity of solutions of the inversion problem can reduce the accuracy of fracture prediction,or even cause the inversion results invalid.Therefore,at the early stage of inversion,the possible distribution ranges of fractures need to be estimated by logging data or in-situ stress information,and the ranges are used as the constraint conditions of objective function.In order to verify the feasibility of the inversion of fractures,the inversions of different flow models and different inversion parameters are all tested by theoretical cases.These cases demonstrate the effectiveness of the prediction achieved by inversion,and also show different prediction accuracy.