Optimizing the economic efficiency of wholesale electricity markets

This dissertation contributes design ideas and solution methodology to several policy and operational issues within the contemporary wholesale electricity markets in the United States. Mathematical programming, hierarchical optimization modeling and parallel computing techniques constitute its methodological basis. Chapter 2 interprets demand response as a trade of consuming rights among customers and provides a bi-level optimization model to simultaneously clear the energy and demand response markets. As an efficient implementation of FERC Order 745, a three-phase solution method involving joint use of nonlinear and mixed integer solvers as well as a bound-tightening technique is proposed for large-scale instances.;Chapter 3 analyzes the efficiency and equity issues of the uniform-price payment rule in bulk power markets, and justifies based on linear programming duality a pay-as-bid rule in the context of unit commitment. The effectiveness of this payment rule is validated via a simulation of market participants' bidding behaviors in a realistic setting.;Chapter 4 proposes a general bidding structure that clears obstacles for efficient demand-side participation. The structure promotes new bid types that allow deferrable, adjustable and storage-type loads to better express their value. These bid types are shown to be easily incorporated into the existing market and to preserve the efficiency and incentive-compatibility properties of the market.;Chapter 5 provides a Derand sampling method to approximate load forecast errors in a stochastic unit commitment model. Using only a few high-quality scenarios, Derand outperforms its deterministic counterpart and other conventional scenario reduction techniques by a significant margin.;Chapter 6 models a multi-stage security-constrained economic dispatch (SCED) problem. A series of algorithmic enhancements based on the Benders' decomposition and parallel computing are implemented to ameliorate the computational difficulty. The approach is able to process the N-1 contingency list in ten minutes for all large network cases available for experiments.;Chapter 7 proposes a novel solution approach for the SCED problem in a nonlinear AC setting. The approach approximates the nonconvex AC feasibility problem with its semidefinite programming (SDP) relaxation and uses the SDP model as a convex subproblem within a Benders' decomposition framework. Numerical experiments demonstrate the superior solution quality of the approach.