Aiming at the packet loss problem that may occur under the structure of Cyber-Physical Systems, an analysis and design method of stochastic distributed model predictive control is proposed. The packet dropouts from the controller to actuator transmission channel is modeled with the Markov process. By augmenting the linear model of Markov jump, a distributed model predictive control method is designed for the stochastic uncertain system with packet loss. The system is decomposed into multiple subsystems, a stochastic distributed predictive control is designed based on the minimum maximization optimization problem. An iterative algorithm is proposed for the coordination of the distributed predictive controllers. The proposed distributed predictive control algorithm is tested on a DC motor system, and is verified to be effective.