A Fast Calculation Method of SCUC Based on Deep Learning

Under the background of the rapid change of power system and the wide application of artificial intelligence technology, it is of great significance to research how to combine artificial intelligence technology and Security Constrained Unit Commitment (SCUC) problem. This paper proposes a fast calculation method of SCUC based on deep learning (DL-SCUC). The proposed method uses off-line training convolution neural network (CNN) to extract features and learn rules of the time sequence information of unit startup and shutdown. In the on-line calculation, the integer variable reduction strategy is used, and the unit startup and shutdown scheme is arranged based on the trained CNN, which greatly reduces the difficulty of subsequent calculation of SCUC. The proposed method combines artificial intelligence with optimization methods, which not only improves the calculation speed but also ensures the feasibility of the solution. Case studies based on the modified XJTU-ROTS2017 testing system show that DL-SCUC can effectively improve the calculation speed of SCUC problem and has high accuracy.