Uncertainty Quantification And Economic Dispatch Models For The Power Grid

The modern power grid is constrained by several challenges, such as increased penetration of Distributed Energy Resources (DER), rising demand for Electric Vehicle (EV) integration, and the need to schedule resources in real-time accurately. To address the above challenges, this dissertation offers solutions through data-driven forecasting models, topology-aware economic dispatch models, and efficient optional power flow calculations for large scale grids. Particularly, in chapter 2, a novel microgrid decomposition scheme is proposed to divide the large scale power grids into smaller microgrids. Here, a two-stage Nearest-Generator Girvan-Newman (NGGN) algorithm, a graphicalclustering-based approach, followed by a distributed economic dispatch model, is deployed to yield a 12.64% cost savings. In chapter 3, a deep-learning-based scheduling scheme is intended for the EVs in a household community that uses forecasted demand, consumer preferences and Time-of-use (TOU) pricing scheme to reduce electricity costs for the consumers and peak shaving for the utilities. In chapter 4, a hybrid machine learning model using GLM with other methods was designed to forecast wind generation data. Finally, in chapter 5, multiple formulations for Alternating Current Optimal Power Flow (ACOPF) were designed for large scale grids in a high-performance computing environment. The ACOPF formulations, namely, power balance polar, power balance Cartesian, and current balance Cartesian, are tested on bus systems ranging from a 9-bus to 25,000. The current balance Cartesian formulation had an average of 23% faster computational time than two other formulations on a 25,000 bus system