Abstract: Taking prediction of member?3 mudstone with thin layer sand of Jurassic Badaowan formation in Bai?21 well area in Baikouquan oilfield in northwestern margin of Junggar basin as an example, the application of high?resolution logging constrained stochastic optimization seismic inversion technology to such a reservoir prediction is discussed, and good results are analyzed. As a further development of conventional seismic inversion, this technology takes advantage of the geostatistical law of the logging and seismic data and realizes the inversion of wave impedance and other physical properties using stochastic optimization theory, without complex modeling process. At the same time, this technology answered inverse problem by solving positive questions, thus making the constraint be relatively easy to achieve. In view of the partial overlap of sand?shale wave impedance in the study area and the corrected self?potential property that can do well in distinguishing the sandstone from mudstone, the wave impedance inversion and the self?potential inversion are all used for the target zone, in order to more properly predict the reservoir distribution in this area