摘要
This paper reveals that the existing techniques have some deficiencies in the proper estimation of voltage stability margin(VSM) when applied to a power system with different load change scenarios.The problem gets worse when credible contingencies occur.This paper proposes a real-time wide-area approach to estimate VSM of power systems with different possible load change scenarios under normal and contingency operating conditions.The new method is based on an artificial neural network(ANN) whose inputs are bus voltage phasors captured by phasor measurement units(PMUs) and rates of change of active power loads.A new input feature is also accommodated to overcome the inability of trained ANN in prediction of VSM under N-1 and N-2 contingencies.With a new algorithm, the number of contingencies is reduced for the effective training of ANN.Robustness of the proposed technique is assured through adding a random noise to input variables.To deal with systems with a limited number of PMUs, a search algorithm is accomplished to identify the optimal placement of PMUs.The proposed method is examined on the IEEE 6-bus and the New England 39-bus test system.Results show that the VSM could be predicted with less than 1% error.
This paper reveals that the existing techniques have some deficiencies in the proper estimation of voltage stability margin(VSM) when applied to a power system with different load change scenarios.The problem gets worse when credible contingencies occur.This paper proposes a real-time wide-area approach to estimate VSM of power systems with different possible load change scenarios under normal and contingency operating conditions.The new method is based on an artificial neural network(ANN) whose inputs are bus voltage phasors captured by phasor measurement units(PMUs) and rates of change of active power loads.A new input feature is also accommodated to overcome the inability of trained ANN in prediction of VSM under N-1 and N-2 contingencies.With a new algorithm, the number of contingencies is reduced for the effective training of ANN.Robustness of the proposed technique is assured through adding a random noise to input variables.To deal with systems with a limited number of PMUs, a search algorithm is accomplished to identify the optimal placement of PMUs.The proposed method is examined on the IEEE 6-bus and the New England 39-bus test system.Results show that the VSM could be predicted with less than 1% error.
