Power System Resilience Quantification and Prediction Using Machine Learning Techniques

Modern power system is prone to several types of disruptions and contingencies and power system resiliency or survivability is the inherent ability of a power system to withstand and recover quickly from these severe contingencies. To quantify the resiliency of a power system under a specific contingency, a system resilience index has been proposed here in this paper. This power system resilience index can be evaluated from the power flow results data of the power system under any contingency condition. The lower the value of this resilience index, the more will be the resiliency or survivability of the system. In this paper, a modified IEEE 30-bus power system has been considered for the quantification, analysis, and prediction of its resiliency under several contingency conditions. Two machine learning tools have been implemented here for the prediction of system resiliency. An Artificial Neural Network (ANN) model and an Adaptive Fuzzy-Neuro Inference System (ANFIS) model have been prepared separately for the proper prediction of the resilience of the system at a particular contingency condition. A training dataset has been prepared by performing the load flow simulations of an IEEE-30 bus power system in MATLAB/Simulink software.