Thursday, February 16 at 4:15pm Frank H.T. Rhodes Hall, 253 The large-scale integration of renewable energy sources is hindered by the fact that these resources are neither controllable nor accurately predictable. Our analysis focuses on quantifying the cost of balancing power system operations in the presence of renewable resources and on the amount of capital investment in operating and contingency reserves that is necessary for ensuring the reliable operation of the system. We also explore the extent to which demand-side flexibility can mitigate these impacts. We specifically focus on a contract that couples the operations of renewable energy resources with deferrable loads that can shift a fixed amount of energy demand over a given time window. Various flexible energy consumption tasks can be characterized in this way, including electric vehicle charging or agricultural pumping. We use a two-stage stochastic unit commitment model for our analysis. The use of this model is justified by the fact that it is capable of endogenously quantifying the amount of required reserve capacity by explicitly representing the uncertainty of daily operations in the model. We present a dual decomposition algorithm for solving the model, a parallel implementation of the algorithm, and various scenario selection algorithms for representing uncertainty. We present results for a reduced model of the California power system that consists of 124 generators, 225 buses and 375 lines. We validate the stochastic unit commitment policy that we derive from the stochastic optimization model by demonstrating that it outperforms deterministic unit commitment rules commonly used in practice. We demonstrate this superior performance for both a transmission-constrained as well as an unconstrained system for various types of uncertainty including network element failures as well as two levels of wind integration that roughly correspond to the 2012 and 2020 renewable energy integration targets of California. We present three fundamental approaches to modeling demand flexibility: central co-optimization of dispatchable generation and deferrable demand by the system operator, demand-side bidding and coupling renewable supply to deferrable demand.