Qualified Two-Hybrid Techniques by DWT Output to Predict Fault Location

Abstract

The power transmission system is essential for the power scheme to transfer the energy from generators to consumers. The short circuit problem repeatedly occurs in the transmission system, and the main problem is to separate the sources from users. This research has applied two hybrid techniques to predict fault location. The first hybrid technique has involved the Discrete Wavelet Transformation (DWT) and Adaptive Neuro-Fuzzy Inference System (ANFIS), while the second hybrid technique is for DWT grouping and Support Vector Machine (SVM). These hybrid techniques are intended to estimate the fault location of each fault category in a transmission system. The DWT was applied to both D8 and D9 level at the 50 kHz sample frequency. The root mean square (RMS) values of the D8 and D9 coefficients were used for training using ANFIS and SVM techniques. After that, ANFIS and SVM were utilised to detect faults in the phase and ground lines. Several types of fault have been simulated, i.e. fault location, fault resistance, and original point of view. The RMS results from the two hybrid techniques were compared to find the best results. The tests of error estimation were performed for the three bus systems. The comparison of error estimation of the two methods shows that both hybrid techniques can be applied to predict fault locations.