When the low-permeability sandstone reservoirs enter into the middle-late stage of development，the relative high-permeability zones are easy to be formed between wells and the heterogeneity of reservoirs is enhanced because of the long-term water injection，which results in serious water coning and poor development effect. After washing in the relatively high-permeability zones，the physical properties of reservoirs are improved，and the remaining oil enrichment area can be easily formed in the external reservoirs due to the low degree of production. Accurate description on the state and distribution of the relatively high-permeability zones is the key of remaining oil prediction and tapping in the low-permeability sandstone reservoirs. Based on dynamic-static development data of the production wells，the reservoir parameter model can be established combined with the stochastic simulation. Ensemble Kalman filter algorithm was applied to achieving iteration and updating of the geological attribution model and the numerical simulation. The attribution model reflecting real geological conditions at various development stages can be obtained combined with the dynamic production data. The results show that the permeability model gained by this algorithm can point out the discrepancies in physical properties in the plane and identify the position and configuration of the relatively high-permeability zone in the low-permeability sandstone reservoir. It lays a foundation for the predicting and tapping of the residual oil distribution.