جستجوی مقالات مرتبط با کلیدواژه « fault classification » در نشریات گروه « فنی و مهندسی »

The growth of exploitation of distributed generation sources (DGs) such as offshore wind farms makes DC networks an interesting alternative to conventional AC grids. But protection of DC lines is one of the main challenges of these grids especially in hybrid non-homogenous corridors including underground cables and overhead lines. In this paper, a new single-end time domain-based protection scheme for fault detection and classification is presented with remarkable features such as easy implementation, low computation burden, low sampling frequency, no setting parameters requirement, and also appropriate performance in noisy conditions. To validate the proper performance of the proposed scheme, several scenarios are simulated including internal and external DC, AC faults, and severe load variations in EMTDC/PSCAD software environment. Also, some hybrid line scenarios such as line length variation, OHL or Cable length changes, and increasing the number of line segments are investigated. The result shows desirable performance in various conditions.

In this paper, an effective method for fault detection and classification in a double-circuit transmission line compensated with TCSC is proposed. The mutual coupling of parallel transmission lines and presence of TCSC affect the frequency content of the input signal of a distance relay and hence fault detection and fault classification face some challenges. One of the most effective methods for fault detection and classification in a compensated line is pattern-recognition methods. Prerequisites for the optimal using of these methods are the extraction and selection of appropriate features to feed the classifier. In this paper, wavelet transform as a signal processing tool to extract features is used. Due to variety of mother wavelets, firstly the best mother wavelet is identified by using a new method and the feature vector is made by the coefficients obtained from the best mother wavelet. After this stage, decision tree, support vector machines, and k-nearest neighbor as the classifies are trained by feeding the feature vector. Then, their accuracies are evaluated against different simulation scenarios to select the best classifier which has the best performance among others. In this paper, the sample system and the proposed method is implemented in MATLAB environment.

In this paper, a new method for detection and fault location and classification in MTDC solar microgrid is presented. Some issues such as expanding renewable energy sources and DC loads and efforts to increase power quality and reduce the environmental impact of electricity generation have led to the expansion of solar networks. Identifying the types and locations of faults is important to ensure service continues and to prevent further breakdowns and the increasing the protection’s selectivity characteristic. In this method, an orbital kit is connected to the network. In the fault occurrence time in the network, the fault is detected by passing a current through the connected kits and measuring the traveling waves derived from the fault current, and applying it to a mathematical morphological filter The location of the error is determined using orbital equations and flow calculations. Mathematical morphology filter output and signal energy analysis were used to determine the type of faults. The method presented in an MTDC microgrid connected to energy storage and renewable sources was tested with many faults. The results indicate the accuracy of the proposed method. This method is resistant to changes in arcs resistance (up to 100 ohms), and has a very good performance in high impedance faults conditions(up to 1000 ohms).

In this paper, a new method for fault detection, location, and classification in multi-terminal DC microgrid (MTDC) is proposed. MTDC grids have expanded due to some issues such as the expansion of DC resources, loads, and aims to increase power quality. Diagnosing the types and location of faults is important to continue the service and prevent further outages. In this method, a circuit kit is connected to the grid. In fault time, the fault in the network is detected by passing the current through the connected kits and measuring the traveling waves derived from the fault current as well as applying it to a mathematical morphological filter. Determining the location of the fault is done using circuit equations and current calculations. Phaselet output and fuzzy inference systems are used to determine the type of faults. The presented method was tested in an MTDC microgrid connected to renewable and energy storages with many faults. The results show the ability of the proposed method. The error of the proposed fault location method is less than 7%. This method is resistant towards the change in sampling frequency (between 500 Hz and 50 kHz), fault resistance (up to 125 ohms), and loading (up to 120% of the nominal load); it acts very well in high impedance faults.

Two important issues in the modern transmission lines protection are the speed and accuracy of the fault type classification, which have a great impact on the duration of fault clearing time and the accuracy of fault detection by the distance relay. The purpose of this study was to use the phase space analysis and decision tree-learning algorithm to classify the fault type in single circuit transmission lines. Accordingly, an algorithm is developed in which the three-phase current and voltage signals are measured and sampled on one side of the transmission line, firstly. Then, after the phase space analyzing of the current and voltage samples, the statistical feature vector of the output of the analysis is calculated. In the end, the feature vector is fed to the pre-trained intelligent model, to determine the type of fault occurred. The proposed algorithm has been investigated and tested on the sample network in different fault conditions, including different values of fault resistance, fault inception time, the amount of the transferred power on the transmission line, and the fault location. The results show that the proposed algorithm can determine the fault type with a length of post-fault data window less than 2 ms and accuracy of 100 percent.

In this paper, using pattern recognition method all fault type is classified. Firstly, feature vectors obtained from sequence components of current and/or voltage signals are normalized by efficient technique. Afterwards, the proposed supervising function applies Kernel Naive Bayes classifier. The classification method through tuning of kernel function bandwidth s suitable for a complex and non-linear feature spaces. The signal processing procedures is done by using minimum sampling frequency hence the output of conventional current and voltage transformers can be utilized. Moreover, the performance of proposed pattern recognition methodology is evaluated from different point of views. The achieved results indicate that the proposed fault classifier has acceptable performance even in the noisy conditions.

Electrical Generator has a vital and important role in vehicles, especially the vehicles with Multiplex data transfer system. Developing electric malfunctions can cause catastrophic damages to other electric and electronic systems. Therefore alternator fault detection and monitoring has a significant role to avoid developing faults in other systems. In this research alternator fault detection and monitoring has been done with data extracted from vibration signals using Adaptive Neuro Fuzzy Inference System (ANFIS). To accomplish this task, certain faults are made on the alternator deliberately. Then vibrations from each specific fault are gathered and stored for subsequent analysis. The faults consist of: one phase disconnection, disconnection of positive voltage of regulator, burning of one and two diodes of rectifier set. The vibration signals of healthy alternator as well as different faulty states are gathered from two piezoelectric sensors mounted on alternator body for 30 seconds and 1000, 1500, and 2000 motor RPM. For analyzing vibration signals wavelet packet decomposition in level one was used. The mother wavelet with maximum energy to Shannon entropy was selected as the best choice. First and second energy bands were computed and used as the feature vector to the designed ANFIS. Results shows the proposed ANFIS model was effective and it could predict different faults with perfect match.

A traveling wave based algorithm for fault classification and faulted phases selection is presented in this paper. The proposed algorithm is composed of two major stages. In the first stage, after extraction of modal components from three phase signals and by means of Teager energy operator (TEO), traveling waves are extracted from modal signals. Then ratios of initial traveling wave in different modal components are used for the fault classification and faulted phases selection purpose. TEO is a very simple method which can extract traveling waves with very high speed and resolution. In the second stage, fault classification and faulted phases selection is performed based on the information extracted from the previous stage. Intelligent methods are the best choice to be used in the second stage and for the first time are utilized in this paper. Fuzzy system, which is used in this paper as a classifier, is able to classify the faulted signal with a high accuracy. Simulation results show that the proposed algorithm has an efficient and accurate performance in terms of close-in faults, high impedance faults and faults with very low inception angle.