EECS – Final Exam: Distributed Computation and Optimization for Electric Power Distribution Systems, Rabayet Sadnan
About the event
Student: Rabayet Sadnan
Advisor: Dr. Anamika Dubey
Degree: Electrical and Computer Engineering PhD
Thesis Title: Distributed Computation and Optimization for Electric Power Distribution Systems
Abstract: In order to operate a distribution grid efficiently, the optimal operation of electric power distribution systems has gained attention recently. However, as the number of controllable devices grows within the power distribution systems, a centrally-managed optimization paradigm faces augmented computational complexities and is susceptible to a single point of failure. These limitations of centralized optimization methods for power distribution systems operation have led to the distributed computing paradigm, where the overall problem is decomposed into several sub-problems and solved by different computing agents. In addition to that, from a network perspective, a distributed control structure establishes more resilient salience. It is expected that the future power distribution grid will adopt a distributed architecture with a systems-in-system structure. Thus, there has been a lot of focus on distributed computing and optimization. However, the existing techniques reported in recent literature for solving distributed optimization problems are not viable for power distribution systems applications. The essential drawback remains slower convergence for scaled systems. Specifically, a large number of communication rounds is required among the computing agents to solve one instance of the optimization problem; the typical number of iterations is in the order of 〖10〗^2~ 〖10〗^3. The focus of this dissertation is to develop a scalable Distributed Optimal Power Flow (D-OPF) method based on Equivalent Network Approximation (ENApp), to solve Optimal Power Flow (OPF) for power distribution networks. Specifically, the distribution system’s radial topology is leveraged to reduce the problem size of the decomposed sub-problems. The unique upstream and downstream relation among nodes in the radial system helps to properly reduce the size of the sub-problems. The proposed innovations reduce the required number of communication rounds for convergence by order of magnitudes. The proposed ENApp D-OPF method can be implemented for both a distributed control architecture and a distributed computation platform. In this dissertation, OPF problems for both balanced and three-phase unbalanced distribution systems with continuous and integer variables have been simulated and validated using an equivalent Centralized OPF (C-OPF). Besides, different optimization problem applications specific to the power distribution systems have been demonstrated in this dissertation using the proposed D-OPF method.