Master of Science in Electrical Engineering (MSEE)
Electrical and Computer Engineering
In this thesis we propose a distributed algorithm, based on diffusion, to balance loads on an electrical power grid, while maintaining stable operation (system’s ability to maintain bus voltages within preset bounds). This algorithm, called the Diffusion-driven Distributed Load Balancing (DDLB) algorithm, is implemented on the OMNET++ Discrete Event Simulator and the response of the physical grid is simulated on a load flow program, which together simulate a deployment of the DDLB algorithm on the grid. The electrical grid is represented as a graph whose nodes are buses and whose edges are power lines connecting buses. Each node (except the slack bus) has a load (positive if power consumed, negative if power generated) associated with it. The slack bus is a special bus that covers any power surplus or deficit due to a load assignment. A given load assignment, when applied to the grid affects bus voltages and system stability. The problem we address is as follows. Given a preferred load for each node and a load cost (a measure of deviation from this preferred load), the ideal solution is a load assignment with lowest cost that results in a stable system. We measure the performance of our algorithm (DDLB) against the "one-shot" algorithm, a naive distributed solution in which each node uses its preferred load directly for a load assignment, without any regard for system stability. Through extensive simulations with 1.6 million test cases, we show that the DDLB algorithm vastly outperforms one-shot. Specifically, the one-shot algorithm causes instability in over 57% of the cases tested albeit with zero load cost. For the same cases when applied to the DDLB algorithm only 0.65% were unstable; the average load cost was less than 2%. Our simulations included a study of several scenarios that a grid could be subjected to, including balanced load, overloaded, underloaded grids, local generator failures, and a sparser communication network for the DDLB algorithm; in this context one could view the one-shot algorithm as a distributed algorithm with no communication network. In all these scenarios studied the DDLB algorithm outperforms the one-shot algorithm.
Document Availability at the Time of Submission
Secure the entire work for patent and/or proprietary purposes for a period of one year. Student has submitted appropriate documentation which states: During this period the copyright owner also agrees not to exercise her/his ownership rights, including public use in works, without prior authorization from LSU. At the end of the one year period, either we or LSU may request an automatic extension for one additional year. At the end of the one year secure period (or its extension, if such is requested), the work will be released for access worldwide.
Ayyala, Isaac Abhilash, "A Distributed Diffusion-Driven Algorithm for Load Balancing in an Electrical Power Grid" (2015). LSU Master's Theses. 1688.