Voltage stability assessment of power systems using voltage stability indices and artificial intelligence techniques

The research presented in this thesis uses the Artificial Intelligence (AI) techniques to assess the voltage stability condition in power systems. Voltage stability index is a feature for evaluating the voltage stability condition. It is generated from the basic power flow equations and/or energy functions. The research is very timely and current and would be a substantial contribution to the present body of knowledge in programming and voltage stability assessment. The methods developed in this research would be faster than presently available voltage stability indices. In this study, five voltage stability indices previously developed, namely, Fast Voltage Stability Index (FVSI), On Line Voltage Stability Index (LVSI), Line Stability Index (Lmn), Line Stability Factor (LQP), and Power Transfer Stability Index (PTSI) were utilized by using MATLAB software. All the indices were subjected to various contingencies including variable load increase and line outage. The range of all the indices was found to be falling between 0 and 1. When the voltage stability indices are near to 1, the system became unstable, thus the system went to instability with increasing the load change or line outage and increasing voltage stability indices depend on the bus type. That is, the transmission line connected to generation bus or reference bus is more stable because it is near to source. The results obtained from the indices were compared with each other,and the conclusions on the performance of the indices were discussed. Two Artificial Neural Networks (ANNs), namely Radial Basis Function Neural Network (RBFNN) and Multi Layer Perception Neural Network (MLPNN) were considered in fitting all the indices for voltage stability assessment of power systems. The data generated from the contingency analysis of all indices were used for training and testing the ANN. Suitable power system features were selected for the ANN which include voltage, active power, reactive power and load angle. Using the mentioned approach, for a given operating conditions, the most critical transmission lines and buses of the systems have been identified. Moreover, the voltage stability assessment by using ANNs was monitored throughout the generalization test. It is appeared that difference between the prediction computed by ANNs, and conventional methods of voltage stability indices tests is considered almost negligible. The analysis of features sensitivity of the ANN has been investigated and found out that the selection of features affect the performance of the ANN. In conclusion, using ANN for fitting the voltage stability indices shows a lot of potential in assessing voltage stability problems. In this research, the first objective was to implement several existing voltage stability indices and compare it with each others. The second one was to apply Radial Basis Function Neural Network and Multi-Layer Perceptron Neural Network for all the indices in order to improve the indices performance in terms of computational time and accuracy. While the third one was touse feature selection on the input features of artificial neural network to decide the most important features