The ANN Dynamic Security Regions of Power System Dynamic Security Analysis

摘要

PDSR imitated by ANN to judge the security of the power system is described in this paper. The two-kind and three-kind classifier are constructed by ANN to judge the stability of running points exactly, and the remains imitator built by ANN, the output of which is shown by CCT t_(cr), is used to decide the stability remains of running points of the power system. The emulation on the three-machine and nine-bus test system demonstrates that the DSR imitated by ANN is more accurate than the hyper-plane method. However, with the expansion of the injection elements, the 'dimensions misfortune' problem is caused easily, so optimal feature subset of decision tables is used, combined with element discretization of rough-set, to select a group of effective feature elements from larger dimensions and rebuild the ANN. The simulation performed on EPRI-36 system shows a good result.