| Nowadays, with the increasing demand for electricity from modern society andpursuit of economic benefits of power companies, power system is running towardsits stability limitation. When serious contingencies cause oscillation betweengenerators, some controls, such as generator tripping and load shedding, will be givento restore the synchronization of power system. Once these control measures is failure,power system should be separated into individual islands (i.e. islanding control) toavoid oscillation spreading to the entire network, which may cause widespread poweroutages. Traditionally, the islanding control strategy is given according to the offlinecalculation. And, it will be triggered when some condition is matched based on localmeasurement. This method has some shortcomings, such as high complexity in tuningdevices, hard to coordinate multiple control equipment, difficult to match the actualand pre-set conditions. With the development of wide area measurement system,online islanding control studies have been paid much more attentions. In this thesis,islanding control is deeply studied.A multilevel method for fast searching of the splitting surface is proposed in thispaper based on multilevel k-way partitioning method in graph theory. It combines thecharacteristics of graph theory and power system, and it searches the splitting surfacethrough three steps: grid graph coarsening, initial partitioning and refinement. In thecoarsening stage, according to the special problems of islanding identification, wepropose a high-efficiency tracking method based on the synchronizing machines.Then greed graph growth algorithm is adopted to obtain the initial partition result. Inthe refinement stage,0-1optimal planning method considering topology constraints isused to ensure high quality division and connectivity in subsystems. At last, IEEE39-bus and IEEE118-bus test system are used to validate the effectiveness of theproposed method. |
PDF Full Text Request