The key to automatic history matching for fractured shale oil reservoir is its precision. However, traditional widely applied data assimilation methods such as the ENKF is not applicable to a typical nonlinear and nonGaussian system of shale oil numerical simulation. In this paper, a history matching, based on particle filter is proposed to evaluate the state of the shale oil reservoir and to invert the fracture distribution parameters. The proposed method avoids the disadvantages of the traditional ENKF method. Results showed that the particle size was of great significance to the history matching results. For the established reference model with three discrete fractures, the reasonable particle size was about 400. Meanwhile, the inversion error of the fracture central points, the fracture half length, the fracture permeability and the matrix permeability had low average error of about 4.49%. Meanwhile, the accuracy of inversed fracture azimuth was up to 2.93%, illustrating that the rational application of geological information to decrease the uncertainty as much as possible is of essence. This work explores the feasibility of the application of the particle filter on history matching and aids in the development of shale reservoirs.
Keywords Shale oil, discrete fracture, history matching, particle filter