Abstract: The ouput uncertainty of renewable generation and the fluctuation of load require integrated risk evaluation and optimal scheduling based on the real-time operation state of the system. Through the timely warning of the operational risk and adjusting the current control strategy accordingly, the safety and economy of power grid operation is coordinated and ensured. This paper presents a multi-time scale dynamic reliable optimal dispatch method for high proportion renewable energy power systems based on Model Predictive Control. The reliable optimal dispatch of electric system with high share of renewables is carried out through day ahead predictive optimal dispatching, day rolling optimal regulation, operation risk assessment and feedback correction of renewable generations and adjustable resources in the system. In the intra-day stage, based on the day ahead scheduling plan and real-time operation status, regression prediction algorithm is used to adaptively select key variables to predict the future operation status of the system. The probability density of representative variable is obtained through Gibbs sampling, which can quickly quantify the operation risk of the system at the next moment. The risk is then feedback to the day rolling optimal regulation to perform repeated reliable optimal scheduling. The simulation results verify the adaptability and feasibility of the proposed method.