Search

To increase the classification accuracy of a protection scheme for power transformer, an effective convolution neural network (CNN) extreme gradient boosting (XGBoost) combination is proposed in this work. Data generated from various test cases are fed to one‐dimensional CNN for high‐level feature extraction. After that, an efficient classifier tool XGBoost is used to properly discriminate different transformer internal faults against outside abnormalities. A portion of an Indian power system is considered and simulated in PSCAD software using the multi‐run feature to collect a large number of data for various fault/abnormal situations. The generated data are used in MATLAB software where the proposed algorithm is programmed. A high‐performance CPU is used for training and testing purpose of the projected artificial intelligent technique. The obtained results for classification accuracy as well as discrimination time shows that the proposed scheme is competent enough to properly discriminate transformer operational conditions. Further, the combined CNN‐XGBoost technique is compared with existing relevance vector machine and hierarchical ensemble of extreme learning machine classifier techniques. Moreover, a hardware experiment is performed in a laboratory prototype of 50 kVA, 440/220 V transformer to verify the authenticity of the developed protective scheme. After analyzing a variety of experiments, the authors note that the presented method provides promising classification accuracy within a short time period. [ABSTRACT FROM AUTHOR]

Various unwanted phenomena that are taken place in the transformer may occasionally mal‐operate selected fault classification based protective schemes. Hence, it is necessary to discriminate internal fault from external abnormal conditions for unit protection of power transformer. This study presents a new hierarchical ensemble extreme learning machine (HE‐ELM) based classifier technique to identify faults in & out of transformer. The component extreme learning machine (ELM) is structured hierarchically to improve its fault data classification accuracy. The developed algorithm is evaluated by simulating multiple disorders on 100 MVA, 132/220 kV transformer with the help of PSCAD software. DWT is used to extract features from acquired current signals from transformer. The feature vector formed after extraction process is fed to the HE‐ELM algorithm for data classification. The fault discrimination accuracy of HE‐ELM technique is 99.91%. This shows its effectiveness with respect to other classifier techniques. Moreover, the developed algorithm is successfully tested on hardware prototype in laboratory environment under various inrush and fault conditions using Cortex M4 microcontroller (STM32F407) with maximum identification time of 27 ms. The proposed HE‐ELM technique is compared with existing support vector machine, probabilistic neural network and ELM techniques for identical fault data. Results demonstrate that HE‐ELM outperforms than existing schemes in cross‐domain recognition task. [ABSTRACT FROM AUTHOR]

Sudden changes in loading or weak constitution of power network causes power swing which may aggravate missoperation of protective elements. Consequently, it becomes utmost essential to rapidly and accurately distinguish between fault and power swing conditions to prevent instability in smart power grid equipped with compensation. This study demonstrates an effective disturbance classifier scheme for series compensated transmission line (SCTL) for discrimination during disparity in power grid contexts using sequence-space-based support vector machine (SVM) classifier. The test data sets are generated by performing extensive simulations in PSCAD software by varying system and fault context. SVM architecture has been trained and tested by generating feature vector using modified full cycle discrete Fourier transform in MATLAB. After proper extraction of features of the interest at the time of disturbance, a decision about power swing or fault has been carried out using SVM classifiers. Regulation and kernel function parameter have been tuned using ten-fold cross-validation applied on training set. The developed scheme is also evaluated for symmetrical fault detection during power swing and shows remarkable improvement in accuracy and speed for protection of SCTL in comparison to existing schemes. [ABSTRACT FROM AUTHOR]

Discrimination between transient fault and permanent fault is contemporary problems for protection of transmission line. Auto-reclosure is one of the foremost solutions for the same. This study presents the development of auto-reclosure scheme using modified full-cycle discrete Fourier transform with adaptive dead time control. The discrimination between transient and permanent fault is done using differential voltage across the contacts of circuit breaker. Reclosing instance is identified by comparing absolute value of positive sequence voltage, phase difference and frequency difference with a respective preset threshold using synchronism check relay. The proposed scheme has been validated by simulating power system considering variation in fault location, fault clearance angle, fault resistance and load flow in PSCAD/EMTDC. The algorithm for the auto reclosing scheme is validated in MATLAB software. Real time emulation of the proposed algorithm is performed using AVR ATmega 328P controller on laboratory prototype. The test results of simulation and hardware emulation corroborate the efficacy of the proposed technique. [ABSTRACT FROM AUTHOR]

Identification of faults within power transformers is the means of ensuring unit transformer protection. Existing relay maloperates during abnormalities such as magnetising inrush, CT saturation and high resistance internal fault condition. Therefore, it is essential to categorise the internal fault and external abnormality/fault in case of transformer protection. This study presents a new scheme, based on relevance vector machine (RVM) as a fault classifier. The developed algorithm is assessed by simulating various disorders on 345 MVA, 400/220 kV transformer in PSCAD/EMTDC software and also on prototype model with 2 kVA, 230/110 V multi‐tapping transformer. One cycle post fault current signals are captured from primary and secondary to form feature vectors. These feature vectors are used as an input to RVM for classification of various test cases. Wide variation in system parameters and fault conditions are considered for test data generation and validation. The proposed scheme is compared with the support vector machine (SVM) and probabilistic neural network (PNN)‐based techniques. The proposed scheme successfully discriminates various types of internal faults and external abnormalities in power transformer within a short time. The fault classification accuracy obtained by proposed RVM technique is more than 99% in comparison to SVM and PNN‐based schemes. [ABSTRACT FROM AUTHOR]

This study presents a new current transformer (CT) saturation detection and compensation algorithm. The proposed algorithm depends on a saturation detection index (Dn) which is derived by using the derivatives of the current signals and Newton's backward difference formulae. The calculated index is continuously compared with an adaptive threshold (Th) to estimate the start and the end point of CT saturation. A low‐pass first‐order Butterworth filter is used to suppress noise and harmonics which may be present in CT secondary current. The proposed saturation detection algorithm has been tested by considering different values of remanent flux, fault type, fault inception angle, burden resistance, decaying DC component of fault current and noise. At the same time, modified discrete Fourier transforms‐based compensating algorithm has also been proposed to reconstruct the saturated samples. Validation of the proposed scheme is also performed on a developed laboratory prototype. A comparative evaluation of the proposed algorithm is also performed with the existing schemes. A series of test results from the simulation software and the laboratory prototype show the effectiveness of the proposed scheme. [ABSTRACT FROM AUTHOR]

This study presents a new support vector machine (SVM)‐based identification method which effectively discriminates between various types of fault and power swing conditions. Different power swing cases, fault cases and fault during power swing cases have been generated with varying fault and system parameters using PSCAD/EMTDC software package. The performance of the developed algorithm has been tested over 3510 testing dataset. Using three phase current samples for half cycle duration of each simulation case of post‐fault/power swing conditions, an overall classification accuracy of 98.71% is achieved. The proposed scheme is capable to detect faults during power swing condition accurately as the current waveform/signatures for both the cases are entirely different. On the other hand, the conventional scheme do not provide effective discrimination in the said situation as they compare magnitude/phase of the current signals (require pre/post‐processing of signals). In addition, conventional scheme gives poor accuracy during unknown system/unseen dataset whereas the proposed scheme provides promising accuracy (97.61%) in the said situation. [ABSTRACT FROM AUTHOR]