Voltage instability analysis of electric power systems using artificial neural network based approach

Voltage instability analysis in electric power system is one of the most important factors in order to maintain the equilibrium of the electric power system. A power system is said to be experiencing voltage instability whenever the system is not able to maintain the voltage at all buses in the system remain the same after the system is being subjected to a disturbance. Voltage instability can lead to total system blackout. Therefore, it is important to implement voltage instability analysis in order to make sure that the voltage level at all buses is at stable state. Even though the research regarding voltage instability analysis has been carried out for decades, there is still room for improvement especially in terms of accuracy and time execution. The research work presented in this thesis is about the analysis of voltage instability of electric power system by using reactive power-voltage (QV) and real power-voltage (PV) curves. PV and QV curves are very important for calculating voltage instability indices. These voltage instability indices are voltage stability margin (VSM) and load power margin (LPM). VSM can be divided into two indices which are VSM for real and reactive power of load, VSM (P) and VSM (Q). Similarly, there are two categories of LPM which are LPM of real power and reactive power of load, LPM (P) and LPM (Q). Besides that, modal analysis technique is used in this research for determining the weakest load buses in the electrical power system. This research will explore the implementation of real power (P) modal analysis technique in addition to the reactive power (Q) modal analysis technique. It was found that reactive power (Q) of load gives more effects towards the stability of the system voltages than real power (P) of load. Subsequently, Artificial Neural Network (ANN) and Adaptive Neuro-Fuzzy Inference System (ANFIS) are used for providing the target values of VSM (P), VSM (Q), LPM (P) and LPM (Q). The accuracy and computing time of both ANN and ANFIS are being recorded and compared. The research showed that the accuracy and computing time of ANFIS are better than ANN’s. Finally, Probabilistic Neural Network is applied for classifying the voltage instability indices. IEEE 14, 30 and 39-Bus Test Power System were selected as the reference power systems in this research. The load flow analyses were simulated by using Power World Simulator software version 16. Both Q and P modal analysis were done by using MATLAB application software