Abstract: For the issue of predictive control of stochastic discrete linear systems, the stochastic predictive equation is transformed into the deterministive predictive-estimate equation based on the predication and estimation of Kalman filter. Accordingly, stochastic predictive control problem is transformed into deterministive online optimization based on the realtime observed value as the initial value. A stability condition for predictive control of stochastic system is deduced by the method of constructing feasible solution. Finally, using virtual LQG (linear quadratic Gaussian) feedback control law, a fast arithmetic with less on-line calculation cost and stability guarantee is presented, and the simulation results show the stability and the validity of the method.