Abstract High Impedance Faults (HIFs) usually occur in distribution networks and, generally, cannot activate conventional protection relays because of high impedance at the fault point. These faults often occur when an overhead conductor breaks down and touches high impedance surfaces and/or where the conductors become in contact with a high impedance object such as a tree. The main purpose in HIF detection, in contrary to short circuit faults, is not to protect the system, but to protect the human lives and preventing fire hazards due to arcing phenomenon . In this thesis , for the first time a high impedance fault indicator (HIFI), to be mounted on the poles, is proposed to detect and track down the faulted branch. The main contribution of this thesis is the investigation of employing the magnetic field strength signal (MFSS), measured in the vicinity of the conductors of a feeder as a new signature, to detect HIFs based on a novel pattern recognition method. The proposed HIFI is capable of discriminating HIFs (with broken and unbroken conductor) from other similar phenomena, including capacitor switching, feeder switching, load switching, insulator leakage currents and harmonic load. Also , a new multi criteria pattern recognition based algorithm is presented to detect high impedance fault in distribution feeders. This method has three stages: feature generation, feature extraction and classification. The proposed feature sets are generated by applying the multi-resolution morphological gradient to each of the first three half cycles of the post-disturbance current and MFSS signatures . T hese feature sets are fed to support vector machines with r adial b asis f unction (RBF), as the kernel function, for distinguishing HIFs . Applying the data for HIF, insulator leakage current and harmonic load from field tests and for other similar phenomena from simulations has shown high security and dependability of the proposed method . |
