Applied Mechanics and Materials Vol. 818 (2016) pp 47-51
© (2016) Trans Tech Publications, Switzerland
doi:10.4028/www.scientific.net/AMM.818.47
Submitted: 2014-08-27
Accepted: 2015-01-22
Single-Line to Ground-Fault Detection for Unit Generator-Transformerbased on Wavelet Transform and Neural Networks
Ahmad Rizal Sultan1,a *, Mohd Wazir Mustafa2,b, Makmur Saini3,c
1,3.Politeknik Negeri Ujung Pandang, Makassar, South Sulawesi, 90245 Indonesia
1,2,3.Faculty of Electrical Engineering, Universiti Teknologi Malaysia, Johor, 81300 Malaysia
a.rizal.sultan@fkegraduate.utm.my, b.wazir@fke.utm.my, c.makmur.saini@fkegraduate.utm.my
Keywords:
ground-fault detection; unit generator-transformer; dispersion factor of the wavelet
Abstract.
This paper proposes an approach for the detection of the single line to ground fault on a unit generator-transformer, based on the extraction of statistical parameters from wavelet transform based neural network. In the simulation, the current and voltage signals were found decomposed over wavelet analysis into several approximations and details. The simulation of the unit generatortransformer was carried out using the Sim-PowerSystem Blockset of MATLAB. The statistical parameters analysis involved measurement of the dispersion factors (range and standard deviation)of wavelet coefficients. Regarding the pattern recognition of neural networks performance, the accuracy of SLG-fault detection of neural networks was 97.45 %. The results indicated that dispersion factor feature of wavelet transforms was accurate enough in distinguishing a single line to ground-fault and normal condition for a unit generator-transformer.
Introduction
Small current Ground-Fault (GF) detection has been a major concern in protective relaying for a long time. Relaying engineers and researchers often faces the challenge of developing the most suitable technique that can detect faults with reasonable reliability to secure the run of a power system [1]. In general, a step up transformer at an electric power station can be categorized either as a unit generator-transformer configuration, a unit generator-transformer configuration with generator breaker, a cross-compound generator or a generator involving a unit transformer [2, 3]. A GF on the transmission line or busbar can disturb the system configuration of the generator. Numerous feature extraction methods based on Wavelet Transform (WT) have been used for the detection and classification of faults. Reference [4] proposes a new technique for arcing fault location by using Discrete Wavelet Transform (DWT) and wavelet networks. Fault classification procedure based on wavelet in transmission is suggested in [5]. Reference [6, 7, 8] describes the feature extraction technique based on fast WT, a fault index and wavelet power for use to detect the stator faults in the synchronous generator and transformer. Abstraction of a statistical parameter as fault detection has been used for fault detection in previous studies, but they only consisted of standard deviation, kurtosis and skewness [9]. Meanwhile, the statistical feature parameters include kurtosis, skewness, crest factor, clearance factor, shape factor, impulse factor, variance, square root amplitude value and absolute mean amplitude value to fault diagnosis in rotating machine, as defined in reference [10]. The new approach as proposed in this paper uses a dispersion factors of statistical parameters for single-line to ground (SLG) detection. In the analysis, the GF signals were calculated by using DWT. The GF detection was carried out including the analysis on significance of Range (R) and Standard Deviation (STD) of the current and voltage wavelet coefficients, as well as the detail andapproximate of wavelet coefficients to distinguish SLG-fault.A WT is an instrument which functions for the extraction of the transient signals. The best select of the mother wavelet plays an important part for detecting different types of transient signals. The finest wavelet for extracting signal information is that which can produce as many coefficients as possible to characterize the individual signals.
Conclusion
Concerning the ANN performance, the accuracy of the SLG-fault detection was 97.45 %. In this paper, analysis of a dispersion factor of a statistical parameters on wavelet coefficients as ANN input successfully detection of SLG-fault at the unit generator-transformer.