FastPrediction ofLoadabilityMargins by Constructing a Small-SignalStability Boundary Based on NeuralNetworks

Determining loadability margins tovarious security limits isofgreatimportance forthesecure operation ofapower system. A novelapproach isproposed inthispaperforfast prediction ofloadability marginswithrespect tosmall-signal stability basedon neuralnetworks. Small-signal stability boundaries areconstructed by meansofloading thepower systemuntil thestability limits arereached fromabaseoperating point alongvarious loading directions. Back-propagation neural networks(BPNN)fordifferent contingencies aretrained to approximate thesestability boundaries. A searchalgorithm is thenproposed topredict theloadability margins fromanystable operating pointalongarbitrary loading directions through an iterative technique basedonthetrained BPNNs.Thesimulation results fortheIEEEtwo-area benchmark systemdemonstrate the effectiveness oftheproposed methodforon-line prediction of loadability margins. IndexTerms--Loadability margins, neuralnetworks, oscillatory stability, stability limit prediction.