Browsing by Author "Sewilam, Tamer Sayed Abdelhamid Abdelgayed"

Now showing 1 - 1 of 1

Automated wavelet-based fault detection and diagnosis for smart distribution systems and microgrids
(2017-08-01) Sewilam, Tamer Sayed Abdelhamid Abdelgayed; Ibrahim, Walid M.; Sidhu, Tarlochan
The legacy electric power system is defined as a one-way power flow from a centralized power generation plant to customers (consumers). In the smart distribution systems, the customers are allowed to produce electricity through the distributed energy resources (e.g. solar photovoltaics), as well as to consume energy, hence, the smart distribution systems can be defined as a two-way power flow. The Micro-Grid system is defined as a part of the smart distribution system that may include distributed energy resources, energy storage systems and loads. In addition, the Micro-Grid system can operate in two modes, grid-connected or non-grid-connected (i.e., islanded mode). The protection of the Micro-Grid system represents one of the major operational challenges, in particular when considering the integration of distributed energy resources, which may result in different fault current levels, especially in islanding mode. However, the capability of protection system equipment to be more accurate and dependable for faults diagnostic in the Micro-Grid is considered a challenge until now. In this thesis, an automated wavelet-based fault detection and diagnosis technique based on a combination of Wavelet Transform, Harmony Search Algorithm, and Machine Learning approaches is developed for fault diagnosing in the Micro-Grid systems. The harmony search algorithm as an optimization technique is used to identify the optimum wavelet function(s) and the optimum wavelet decomposition level(s) to extract the most prominent features that are hidden in the current/voltage waveforms when applying the discrete wavelet transform. This is unlike previous works in which only one arbitrary wavelet function is used based on a trial and error process. In order to automate the fault classification process in Micro-Grid system, and to examine the effectiveness of the automated wavelet-based fault detection and diagnosis method against iii other approaches, two machine learning techniques (i.e. Decision Tree as an eager learner, and KNearest Neighbor as a lazy learner) are used. The performance of the two classifiers is estimated using the Monte Carlo stratified cross validation method. The Consortium for Electric Reliability Technology Solutions Micro-Grid is used as a test-bed system after modelling in Power Systems Computer Aided Design/Electromagnetic Transient Direct Current software package. The study also takes into consideration different operating modes, different fault types, different fault resistances, and also different fault locations. The results of implementing the proposed automated wavelet-based fault detection and diagnosis technique shows a significant improvement in the classification accuracy compared to other previous approaches reaching an overall accuracy of 95.63% in the Micro-Grid test-bed system. In addition, the proposed technique has been verified experimentally, and the results of the experimental set-up confirmed the validity/effectiveness of the proposed approach in real-time implementation.