Date of Graduation


Document Type


Degree Type



Statler College of Engineering and Mineral Resources


Lane Department of Computer Science and Electrical Engineering

Committee Chair

Muhammad A Choudhry

Committee Co-Chair

Ali Feliachi

Committee Member

Natalia A Schmid


The state of the art of power distribution systems is to demand a more accurate response. It also provides more reliability for fault location and restoration respectively. A multi-agent system design for power distribution has been developed using the change of current methodology to detect and locate any type of faults. Employing the artificial intelligence for restoration process is the most important contribution to this study. Since feed-forward neural networks are weight training based back propagation concept, radial basis neural networks showed more efficiency by using the minimum error method to optimize the decision. A Probabilistic radial basis Neural Network (PNN) is designated at each feeder agent to implement the reconfiguration by analyzing the impedance and current values for each zone. The appropriate decision for the optimal reconfiguration case is a vector of activation signals associated with each switch to restore the power to the un-faulted zones of distribution feeder.;This study examines the role of Universal Asynchronous Receiver Transmitter (UART) buffer circuits in the laboratory experiment demonstration of the multi-agent system design. The main approach of a self-healing concept is the protection system. A recloser has been developed and improved for more sensitivity and faster response to detecting a fault where ever it occurs and lead the process of isolating and re-configuration. An electronic buffer circuit using digital microcontroller has been associated with the recloser and agents switches in order to offer a satisfying feedback for the proposed approach. Simulation studies, using MATLAB SimPowerSystems and, Neural Network toolboxes, for the proposed power distribution system showed improved results for fault location and restoration using Radbas neural networks. Hardware implementation with high accurate software data scoping of results has been employed to show the difference in time response using Universal Asynchronous Receiver Transmitter buffers at each switching relay in the design.