Hidden Markov Model Based Islanding Prediction in Smart Grids

Abstract

Islanding is a recent issue for power system engineers. In every normal or faulty case, the power system undergoes a sequence of states. These states are not directly observable but have a particular pattern. The patterns can be reflected in the measurements. The nonobserving states are referred to as hidden states. This paper has developed a novel hidden Markov model (HMM) based algorithm to engender a probability related to the event of islanding, depending upon the phasor measurements obtained from the smart grid. After processing the phasor data, an artificial neural network is methodically trained to provide emission probability of the hidden states. This emission probability is useful for the evaluation of the HMM. An IEEE 9 bus system is selected to test the algorithm. Several case studies are performed to generate a statistical analysis for the parameters of the HMM. The posterior probability is responsible for final pronouncement of the occurrence of islanding. The accuracy of the algorithm is estimated on the basis of finishing result. Nondetection zone is also estimated for the HMM-based islanding detection method. It is observed that this method HMM can detect islanding with less nondetection zone proficiently in a short time.