The invention discloses an intelligent
power grid deep learning scheduling method considering schedulable
electric vehicle fast / slow charging and discharging forms, belongs to the field of regional intelligent
power grid operation, and aims to perform power supply power optimal distribution by taking total operation cost as an objective function in a day-ahead scheduling stage. In the intra-day pre-scheduling stage, load fluctuation and a day-ahead scheduling plan are simulated to serve as input samples of the
deep learning network, prediction data generated through
simulation are input into aregional intelligent
power grid model in the intra-day pre-scheduling stage, and controllable unit scheduling data in the scheduling plan in the model training stage serve as output samples of the
deep learning network. An intra-day scheduling model of the regional
smart grid is trained based on the deep
learning network through the input sample and the output sample to acquire a predicted valueof the load at the next scheduling moment through ultra-short-term prediction. The predicted value and the day-ahead scheduling plan are inputted into an intra-day scheduling model of the regional
smart power grid to obtain an intra-day scheduling value of the controllable unit. The method solves the problems that errors exist in prediction of
distributed power supplies, electric vehicles and loads of a regional
smart power grid, and intra-day economic dispatching of the regional
smart power grid is difficult to achieve.