Latin Hypercube Sampling and Spectral Clustering Based Typical Scenes Generation and Analysis for Effective Reserve Dispatch

This paper proposes a typical scenes generation method based on Latin Hypercube Sampling (LHS) and spectral clustering (SC). The proposed method can realize efficient typical scenes generation considering the stochastic fluctuation in renewable energy output and load demand. Then, the extreme scenes contained in these typical scenes are further analyzed, and the probability of the extreme scenes included in the typical scenes is defined as the membership degree of typical scenes in this paper. The results show that the membership degree of typical scenes is inversely proportional to the security margin of their branch power exceeding the limit, which is conducive to improving the security margin of the branch power for typical scenes by the reasonable allocation of reserve capacity, so that effective reserve dispatch of the system is guaranteed. Finally, the effectiveness of the proposed method is verified by experiments on the modified IEEE-14 bus system.