Manohar Mishra

Study the performance of S-transform based extreme learning Machine for islanding detection in distributed generation

With the recent development of Microgrid system, the penetration of distributed generation (DG) into the distribution network is bringing various protection and operational problem. Among them islanding is a major issue of Microgrid system which leads to equipment hazard as well as risk of life. Therefore, the detection of islanding has become a must. The islanding detection scheme proposed in this paper constitutes of S-Transform based feature extraction method and evaluation of feature vector by the use of extreme learning machine (ELM) classifier. In this manuscript, a system containing two DGs, one synchronous wind generator and one asynchronous hydro generator are taken into consideration. The effectiveness of ELM is shown in this paper by testing and training the feature vector with different noisy environment and comparing them with other existing classifiers like artificial neural network (ANN), support vector machine (SVM), clustering etc. The whole system was simulated by using MATLAB software.

Study of performance of pattern recognition techniques based on wavelet features for Islanding detection in distributed generation

In this rapidly growing world, the need of distributed generation (DG) is increasing. But there are some problems related to it, one of them is Islanding. In this manuscript, a comparative study of different pattern recognition techniques using wavelet based features for islanding detection in distributed generation is presented. A system containing two DGs, one synchronous wind generator and an asynchronous hydro generator are taken into consideration. The model is simulated and wavelet transform is applied to extract the features like energy, entropy and standard deviation. Pattern recognition techniques including decision tree (DT), support vector machine (SVM), artificial neural network (ANN), extreme learning machine (ELM) and differential evolution (DE)-clustering are applied to segregate the islanding condition from non-islanding events like single line to ground fault, double line to ground fault, triple line to ground fault, load switching, capacitive switching, voltage sag, voltage swell and DG tripping. The comparative results show the effectiveness of each individual classifier for islanding detection in both noiseless and noisy environments. The whole system is simulated by using MATLAB software.

High impedance fault detection in radial distribution system using wavelet transform

This paper presents a novel wavelet transform based approach to detect High impedance fault (HIF) in the distribution line. Due to the limitation of distance relays and over current relays like insensitive to detect very low value fault current, it is unreliable to apply as a fault detector for HIF in distribution line. As wavelet transform (WT) is a very useful tool for analyzing transient fault signal which also provides both time and frequency information, the same has been considered for High impedance fault detection. This method based on various features like the sum of energy contents, standard deviation and entropy of coefficients in multiresolution analysis (MRA) based on wavelet transform. Artificial Neural Network (ANN) and Support Vector Machines (SVM) are used as a machine learning technique to discriminate the HIF from other transient phenomenon (Load switching, capacitor Switching) and normal fault. The proposed schemes are fully analyzed by extensive MATLAB simulation studies that clearly reveal that the proposed method can detect HIF in high voltage distribution line with high accuracy.

Power quality disturbance recognition using hybrid signal processing and machine intelligence techniques

This paper presents two novel approaches for power quality (PQ) event classification. It is a two stage system in which optimal features that characterise the complete signal behaviour are extracted in the first stage and in second stage, based on these features various disturbance waveforms are classified. In the first classifier, a hybrid approach between S-transform and decision tree (DT) is presented. In the second classifier, the S-transform (ST) technique is integrated with neural network (NN) model with multilayer perceptron. Power system suffers from different PQ events such as sag, swell, momentary interruptions, impulsive transients, notch, spike, harmonics and also combination of the above with noise. The above-mentioned events comprise high-frequency and low-frequency components. Thus, it is difficult to classify these PQ events using traditional approaches. Both the classification methods derive various statistical parameters of eight types of single power disturbance and two types of complex...