Neural network based on a genetic algorithm for power system loading margin estimation

In this study, a new methodology based on the predictor-corrector method is introduced to trace the total system equilibrium of the power system model. This methodology considers the automatic voltage regulator voltage limit of all generation units and computes the loading margin associated with both saddle node bifurcation and saddle limit induced bifurcation points. Furthermore, a neural network based on the genetic algorithm approach is presented for fast voltage stability analysis. The neural network can establish a mapping between the operating conditions and the power system loading margin obtained by the above simultaneous equilibrium tracing technique. In order to improve the neural network efficiency, the genetic algorithm is used to generate the optimal feature subset and the neural network parameters at the same time. The method is examined on the New England 39-bus system and a mean absolute estimation error of 0.57% and a response time ≤0.1 s are obtained. By comparing the results obtained through different methods, it is concluded that the genetic algorithm-neural network approach provides a compromise between accuracy and speed of calculation. Hence, for operational purposes, for which having a fast response is vital, using the genetic algorithm-neural network approach is recommended.