Power system closed-loop prediction-decision scheduling method based on data driving

The invention provides a power system closed-loop prediction-decision scheduling method based on data driving, which comprises the following steps: in a data processing stage, performing feature extraction by adopting a standard regression coefficient to determine a most relevant feature category, and selecting a training sample for a next stage based on a Wasserstein distance; in the training stage, constructing an ERM model taking the unit commitment economy as a prediction evaluation index based on the determined feature category and the training sample, and solving by adopting a Lagrange decomposition algorithm to obtain a renewable energy predictor taking the unit commitment economy as a guide; in the prediction-decision-making stage, a renewable energy predictor is embedded into a traditional unit commitment model, and a prediction-decision-making model capable of performing renewable energy prediction and unit commitment decision-making at the same time is obtained. According to the method, the influen