Author ORCID Identifier

Date Available


Year of Publication


Degree Name

Doctor of Philosophy (PhD)

Document Type

Doctoral Dissertation




Electrical and Computer Engineering

First Advisor

Dr. Yuan Liao


Transmission lines are designed to transport large amounts of electrical power from the point of generation to the point of consumption. Since transmission lines are built to span over long distances, they are frequently exposed to many different situations that can cause abnormal conditions known as electrical faults. Electrical faults, when isolated, can cripple the transmission system as power flows are directed around these faults therefore leading to other numerous potential issues such as thermal and voltage violations, customer interruptions, or cascading events. When faults occur, protection systems installed near the faulted transmission lines will isolate these faults from the transmission system as quickly as possible. Accurate fault location is essential in reducing outage times and enhancing system reliability. Repairing these faulted elements and restoring the transmission lines to service quickly is highly important since outages can create congestion in other parts of the transmission grid, therefore making them more vulnerable to additional outages. Therefore, identifying the classification and location of these faults as quickly and accurately as possible is crucial.

Diverse fault location methods exist and have different strengths and weaknesses. This research aims to investigate the use of an intelligent technique based on artificial neural networks. The neural networks will attempt to determine the fault classification and precise fault location. Different fault cases are analyzed on multiple transmission line configurations using various phasor measurement arrangements from the two substations connecting the transmission line. These phasor measurements will be used as inputs into the artificial neural network.

The transmission system configurations studied in this research are the two-terminal single and parallel transmission lines. Power flows studied in this work are left static, but multiple sets of fault resistances will be tested at many points along the transmission line. Since any fault that occurs on the transmission system may never experience the same fault resistance or fault location, fault data was collected that relates to different scenarios of fault resistances and fault locations. In order to analyze how many different fault resistance and fault location scenarios need to be collected to allow accurate neural network predictions, multiple sets of fault data were collected. The multiple sets of fault data contain phasor measurements with different sets of fault resistance and fault location combinations. Having the multiple sets of fault data help determine how well the neural networks can predict the fault identification based on more training data.

There has been a lack of guidelines on designing the architecture for artificial neural network structures including the number of hidden layers and the number of neurons in each hidden layer. This research will fill this gap by providing insights on choosing effective neural network structures for fault classification and location applications.

Digital Object Identifier (DOI)