Abstract: The numerical simulation results of low Reynolds number turbine cascade flows are related to the computational boundary conditions and turbulence prediction capabilities. By applying the ensemble Kalman filtration method, a data assimilation approach was established, which is based on computational fluid dynamics (CFD) solutions obtained from a small amount of experimental measurement data and Reynolds-averaged Navier-Stokes (RANS) equations. After which, the proposed method was validated on two typical turbine cascades. The velocity distribution at the inlet of the cascade computational domain was corrected through using the measured values of the trail profile in the guide cascade. After assimilation, the average relative error of trail loss is reduced by 30%. For the PakB blade profile, eight preset constants of shear stress transport (SST) model were assimilated by using the measured values of static pressure on the blade surface. After assimilation, the average relative error of static pressure coefficient on the blade surface is reduced by 21%. With data assimilation based on experimental measurements, uncertainties in CFD simulations are reduced, and the credibility of analyzing low Reynolds number turbine flows by RANS equations is enhanced.