Wind-solar power consumption optimization method based on data driving

The invention discloses a wind-solar power consumption optimization method based on data driving. The method comprises the following steps: S1, acquiring historical data and carrying out data preprocessing; s2, establishing a day-ahead load prediction model based on LSTM; s3, generating and reducing a renewable energy output scene set based on k-means; s4, formulating a wind and light power supply optimal consumption scheme; according to the invention, for a load prediction problem, on the basis of a long-short-term memory learning model, preprocessed system historical load data are trained, and hyper-parameters of the model are optimized, so that more accurate day-ahead load prediction data are obtained; aiming at the problem of wind and light power supply output prediction, probability distribution of prediction deviation is counted based on historical measurement data and historical prediction data, a large number of initial deviation scene sets considering prediction deviation time-varying characteristics