The invention discloses a regional energy internet scheduling method based on random planning and model predictive control. The method comprises the steps of (1) analyzing the operation characteristics of all units and interconnection lines in a regional energy internet, and constructing a mathematical model of each unit; (2) analyzing the output of the wind and light renewable energy and the uncertainty of cold, heat and electric loads, and generating prediction scene data which are required for the operation of the optimization scheduling by adopting the scene generation and reduction technology; (3) establishing a day-ahead scheduling random planning model with the lowest total operating cost of the system as a target, and determining the on-off state of each unit; (4) establishing an intra-day scheduling random planning model with the lowest total operating cost of the system as a target, and determining the running power in each unit within a day; (5) establishing an intra-day real-time rolling correction model by adopting the model predictive control technology, and correcting the operation plan of each unit. According to the invention, the accurate tracking of loads and thesmooth unit output are achieved. The high dynamic performance and the strong robustness are also realized.