Date of Graduation


Document Type


Degree Type



Statler College of Engineering and Mineral Resources


Lane Department of Computer Science and Electrical Engineering

Committee Chair

Ali Feliachi

Committee Co-Chair

Muhammad A Choudhry

Committee Member

Powsiri Klinkhachorn

Committee Member

Hong-Jian Lai

Committee Member

Daryl S Reynolds

Committee Member

Sarika K Solanki


The proliferation of Smart Grid technologies such as distributed energy resources (DER) and microgrids has resulted in new opportunities and challenges to existing Distribution Automation (DA) solutions. A novel hybrid Multi-Agent System (MAS) Fault Location, Isolation, and Restoration (FLIR) solution is proposed for electric distribution systems with different types of distributed generation (DG) resources and microgrids. The main goal of the MAS is to locate and isolate the fault and then to automatically reenergize fault-free portions of the network to restore power to as many customers as possible while maintaining system constraints like voltage and thermal capacity limits.;Determination of the optimum number and placement of automated switches is an important and a daunting task in the economic feasibility process of any DA project. An innovative algorithm based on relative reduction in the normalized customer interruption costs is presented for the switch optimization problem. The proposed approach isolates the impacts of varying switch investment and customer interruption costs that are usually based on outdated surveys.;A distributed Fault Detection algorithm is presented to support sectionalizing switch agents (SSA) to autonomously detect the fault condition and type of the fault by using the local data such as voltage and current phasor measurements. Fault characteristics of various DG systems including inverter based DG, synchronous generator, and induction generator are considered in the proposed algorithm. Fault isolation and service restoration functionalities are achieved through coordinated communications among the agents using both centralized and decentralized control strategies. Feeder Agent (FA) uses the proposed "Tie-Switch Ranking Algorithm" and "Zone Priority Algorithm" to solve the service restoration problem. A new reinforcement agent learning mechanism based on Q-learning is introduced to support FAs with service restoration task.;The proposed MAS is designed to be demonstrated on Mon Power, a FirstEnergy company, distribution system as part of the DOE project, West Virginia Super Circuit (WVSC). The actual distribution network is simulated using CYME RTM, MatlabRTM/SimpowerRTM, and PSCADRTM software, whereas the MAS and its communications are simulated using MatlabRTM S-functions. The switch optimization results show that the proposed iterative algorithm can drastically reduce the search space, and can find optimal number and placement of the switches with minimum computational effort. The FLIR simulation results show promising advantages of using the proposed MAS solution with the agent learning capabilities for fault location and service restoration tasks.