Due to the gigantic power quality (PQ) demands for modern systems, power systems fault detection and diagnosis have become a significant issue. For this aim, it’s very important to detect the fault at early time and determine its location through any signal. Several methods and techniques are applied to solve this problem such as discrete wavelet transform (DWT). Although DWT has the ability of fast time detection of the fault, it has a problem to discriminate between faulty and noisy signals. DWT succeeds to extract features from altered transient disturbances, but it fails to differentiate between transient disturbances due to healthy or noisy signals. Fusion between DWT for its speed and radial basis function for its accuracy is done. The fusion technique used has a major disadvantage of its delay time as the fault can be detected after the exact location with several samples.
In this paper new technique will be proposed to overcome the DWT and data fusion method problems, to achieve the classification between noisy and faulty signals with high accuracy. The proposed method is executed to classify the signals premised on weights of them, complex tree classifier uses the energy of the signal as a feature. All simulations are achieved and done on IEEE standard 14 bus system to confirm the ability and capacity of new suggested technique. Simulation results show a better performance of the proposed system in comparison with other methods, and that it capable to differentiate between faulty and noisy signals and precisely locate the fault position. |
