Abstract: An algorithm of real-time stochastic optimal control for large scale systems has been developd. The design problem of a new-type of Kalman filter with singularly perturbed form is investigated. A constant linear discrete time interconnected stochastic power system is represented in a singularly perturbed form. The computational time and required storage of Kalman filter per iteration can be significantly reduced, especially for the zeroth-order approximation. The control gains are calculated in the singularly perturbed form. The control effects and the efficiency of this form of Kalman filter are examined by simulation for the stochastic LFC of North China power system.