Aleena Swetapadma

A Novel Decision Tree Regression Based Fault Distance Estimation Scheme for Transmission Lines

In this paper, a decision tree regression (DTR)-based fault distance estimation scheme for double-circuit transmission lines is presented. Fault location is estimated using the information obtained from fault events data. The DTR was chosen because it requires less training time, offers greater accuracy with a large data set, and robustness than all other techniques like artificial neural networks, support vector machines, adaptive neurofuzzy inference systems, etc. Hitherto, DT has been used for fault detection/classification, but it has not been used for fault location. Three-phase current and voltage signals measured at one end of the line are used as inputs to a fault-location network. The proposed method does not require a communication link as it uses only one-end measurements. Signals are processed with two signal-processing techniques—discrete Fourier transforms and discrete wavelet transform. A comparative study of both techniques has been carried out to observe the effect of signal processing on the fault-location estimation method. The proposed method is tested on three test systems, namely: 1) the 2-bus; 2) the WSCC-9-bus; and 3) the IEEE 14-bus test systems. The test results confirm that the proposed DTR-based algorithm is not affected by the variation in fault type, fault location, fault inception angle, fault resistance, prefault load angle, SCC, load variation, and line parameters. The proposed scheme is relatively simple and easy in comparison with complex equation-based fault-location estimation methods.

Fuzzy inference system approach for locating series, shunt, and simultaneous series-shunt faults in double circuit transmission lines

Many schemes are reported for shunt fault location estimation, but fault location estimation of series or open conductor faults has not been dealt with so far. The existing numerical relays only detect the open conductor (series) fault and give the indication of the faulty phase(s), but they are unable to locate the series fault. The repair crew needs to patrol the complete line to find the location of series fault. In this paper fuzzy based fault detection/classification and location schemes in time domain are proposed for both series faults, shunt faults, and simultaneous series and shunt faults. The fault simulation studies and fault location algorithm have been developed using Matlab/Simulink. Synchronized phasors of voltage and current signals of both the ends of the line have been used as input to the proposed fuzzy based fault location scheme. Percentage of error in location of series fault is within 1% and shunt fault is 5% for all the tested fault cases. Validation of percentage of error in location estimation is done using Chi square test with both 1% and 5% level of significance.

Combined DWT and Naive Bayes based fault classifier for protection of double circuit transmission line

This paper describes combined discrete wavelet transform (DWT) and Naive Bayes (NB) fault classifier for protection of double circuit transmission line. Three phase currents of both circuits and zero sequence current are given as input to the NB network for classification of fault. Inputs are pre-processed using approximate coefficient of DWT. NB classifier uses Gaussian distribution function for classification. Seven classifiers are designed for fault classification for each phase A1, B1, C1, A2, B2, C2 and ground G. Advantage of using Naive Bayes classifier is that it take few seconds for training no matter how big the data is. Different cases of fault are studied like phase faults, phase to ground faults, inter-circuit faults, cross country faults, fault near boundaries, different fault location, different inception angle and different fault resistance with high fault resistance. Accuracy of the proposed method is 99% and reach setting is also 99% of line length.

A Finite-state Machine Based Approach for Fault Detection and Classification in Transmission Lines

Abstract In this article, a novel protective relaying scheme based on a finite-state machine is proposed to detect fault in transmission lines, classify the fault, and identify the faulty phase. The three-phase fundamental component of current and the zero-sequence current signals measured at one end of the double-circuit line are used as inputs. The finite-state machine based relaying scheme relies on time-series analysis of current signals only and is built upon the concepts from the finite-state automata theory. The finite-state machine works by transition of one state to another by following certain conditions. The proposed relay is tested during different shunt faults—inter-circuit and cross-country faults—with wide variations in fault parameters. The proposed method is adaptive to variation in fault type, fault resistance, fault inception angle, fault location, power flow angle, different line length, transient faults, current transformer (CT) saturation, and no-fault events. The relay performed correctly for 99.9% of test cases, proving the effectiveness of the proposed method. Furthermore, the proposed method can provide faster, more reliable protection against all shunt faults, inter-circuit and cross-country faults, with wide variations in parameters, and the protection range is effectively extended and greatly improved, which contributes to system safety and stability.

Fault analysis in three phase transmission lines using k-nearest neighbor algorithm

This paper proposes a method for fault detection and fault classification in time domain using k-NN algorithm as classifier. All the signals are generated using PSCAD software and processed using MATLAB software. Three phase currents generated are then processed to give fundamental component of currents and zero sequence component of current. Input given to the k-NN based fault detector is three phase fundamental component of currents. Input given to the k-NN based classifier is the fundamental component and zero sequence component of current. For each phase A, B, C and G separate classifier is designed. Proposed method is tested for different fault location, fault inception angle, fault resistance and fault type. Fault cases are tested against the trained k-NN modules and performance of the relay is found to be 99% accurate. Fault detection time is within half cycle for most cases and one cycle time for some cases.

A novel single-ended fault location scheme for parallel transmission lines using k-nearest neighbor algorithm

Abstract This paper proposes k-nearest neighbour (k-NN)-based method for fault location estimation of all types of fault in parallel lines using one-terminal measurement. Discrete Fourier Transform (DFT) is used for pre-processing the signals and then the standard deviation of one cycle of pre-fault and one cycle of post-fault samples are used as inputs to k-NN algorithm. The results obtained under various fault conditions demonstrate the high accuracy of the proposed scheme to estimate the fault location. The accuracy of the k-NN-based fault location scheme is not affected by alteration in fault type including inter-circuit faults, fault location, fault inception angle, fault resistance, and pre-fault load angle.

Intelligent computing based converter fault identification in line commutated high voltage direct current transmission line

Line commutated converter based high voltage transmission line suffers from converter faults where the converter experiences commutation failure of the device. This fault occurs when the ac voltage drops or the short circuit occurs and is more frequent in the inverter. The fast detection of a fault is essential because improper transferring of current from one device to another can cause stress on the device and interruption of transmitted power. In this paper, a fast scheme for protection against commutation failure due to decrease or faults in ac voltage is implemented using a fuzzy logic controller. The rectifier end data voltage and current signals are chosen as an input to detect the faults. This scheme has good selectivity, reliability, accuracy and robustness.

Fault detection in direct current transmission lines using discrete fourier transform from single terminal current signals

Line commutated converter based high voltage transmission line fault current rises instantly due to the absence of inductance as which is present in ac. In dc currents there is no natural current zero so interruption of current is difficult and therefore the focus of the paper is to identify the fault in less time so that trip command can be initiated to the dc breaker. The dc lines are economical for long length so at far end distance fault identification is essential. The converter control in dc transmission lines control the power and provides synchronous interconnection between two ac systems. The discrete Fourier transform is used to extract the dc current at rectifier end of the dc transmission line and the processed current is compared with the threshold value to identify the fault. The factors considering the effect of fault location and fault resistance are considered for the accuracy, reliability and selectivity.

An intelligent method for classification of normal and aggressive actions from electromyography signals

In this paper an intelligent method called adaptive neuro-fuzzy inference system (ANFIS) is proposed for discriminating normal actions from aggressive actions using the features extracted from electromyography (EMG) signals. Classification of normal and aggressive actions are essential for diseases and prosthetic arm controls. But accurate classification of physical actions are sometimes not possible using raw EMG signals. To enhance the classification accuracy feature extraction is an essential criterion. Hence in this work wavelet analysis is used for feature extraction from EMG signals to provide a suitable pattern to the ANFIS based classifier. The EMG signals are decomposed using DB-4 wavelet up to level 5 and approximate coefficients are extracted. Approximate coefficients from the signals are taken as input to the ANFIS module to classify the physical actions. The proposed method is validated using various test cases and it is observed that accuracy of the proposed method is up to 98% from all the tested cases.

Pediatric Seizure prediction from EEG signals based on unsupervised learning techniques using various distance measures

Epilepsy or recurrent seizures is one of the most common non communicable neurological disorder that is prevalent in todays world population are sudden outburst of excess electrical activity of the neurons. Epilepsy can be detected from Electroencephalogram (EEG) as EEG captures and presents the electrical activity of the brain. Non-invasive EEG or scalp EEG is generally used where electrodes are placed on the scalp in order to record the brain activity. In this work a unsupervised machine learning technique is explored which is used to cluster and extract features from EEG recordings (noninvasive) to detect seizures. A patient specific approach is adopted on an open dataset (Physionet database) from where 51 seizure and 51 non seizure recordings of pediatric subjects (age ranging from lyrs to 12yrs) are considered for the related work. Unsupervised algorithm used here is the k-means algorithm to cluster the recordings into two distinct clusters of seizure and non-seizure data. When the performance of the algorithm was tested the algorithm worked with 91.43% accuracy from nearly 18, 00, 000 data taken from various subject. In future scope of work the accuracy of the method can be enhanced using appropriate features for distinctly identifying different intractable seizures according to their characteristics that are prevalent among pediatric patients.