Estimation with extreme learning algorithm of power system voltage stability index

Nowadays, electrical energy needs are increasing rapidly as a result of technological developments. In order to meet this growing demand, power plants have been built. The necessity of the building the production centers away from the consumption centers brought the electricity energy to be transmitted to the consumption centers with very high voltage and long transmission lines. Power systems have also grown rapidly and this situation formed a complex structure. Also, this case brought important operational and control problems. In this study, voltage stability in the the IEEE 14-bus power system was investigated by means of Extreme Learning Machine (ELM) and Artificial Neural Networks (ANN). For this purpose, the IEEE 14-bus power system model was built in Matlab environment and load flow analysis for this model was performed by using the Newton-Raphson method. In this power system, the voltage stability was evaluated by calculating the Line Stability Index (LSI). In the load flow analysis, the active and reactive powers of all bus were increased by 0.05 step (pu) and a total of 1000 active power, reactive power, voltage, and phase angle of the respective bus were obtained for each busbar. These values were used to calculate the values of LSI. The inputs of ELM and ANN are selected as active power, reactive power, voltage and phase angle of the respective bus. The output of the ELM and ANN was determined as LSI values. The test performance of the ELM is given by 5-fold cross-validation. In addition, the ELM's performance was investigated for the different number of hidden layer cell numbers and different types of activation functions. The activation function, which was determined during the creation of ANN, was chosen for the tangent sigmoid activation function for both inputs and the intermediate layer. Levenberg-Marquardt training algorithm was preferred in the training stage. From all the results, IEEE 14-bus power systems with the determination of the voltage stability of the ELM and ANN has been found to predict a very high performance in the LSI. However, since the estimated time and error values of the ANN are higher than the values of the ELM, it was evaluated that the ELM could be preferred.