Abstract: With the increasing permeability of renewable energy, a large number of uncertain devices pose a threat to the safe and efficient operation of the integrated energy system (IES). In general, solving the uncertainty optimization problem relies on a large amount of historical data, and assists some artificial intelligence technology to predict the analysis of renewable energy, but the usual process of separate prediction and decision making can produce serious deterioration of the model's optimization objective due to excessive prediction errors. Therefore, this paper proposes a prediction-decision method based on K-nearest neighbor (KNN) and robust optimization(RO) to improve the uncertainty optimization problem of IES. Then, the KMV ellipsoid set is constructed using the KNN + minimum volume (KMV) ellipsoid set method, and the two-stage robust model under the set is solved to obtain the optimal multi-energy flow solution. In order to balance the robustness and economy of IES, robust adjustable parameters are used to represent the appropriate level of the uncertainty set. Finally, through the simulation example,the change rule of the interval size and the adjustable parameters of the KMV ellipsoid set is proved, and the superiority of the set is proved.

    Key words: integrated energy system; machine learning; robust optimization; uncertain optimization