Approximate Dynamic Programming Algorithms for the Control of Grid-Level Storage in the Presence of Renewable Generation

Abstract

In this dissertation, we present and benchmark an approximate dynamic programming algorithm that is capable of designing near-optimal control policies for time-dependent, finite-horizon energy storage problems, where wind supply, demand and electricity prices may evolve stochastically. We found that the algorithm was able to design storage policies that are within 0.08% of optimal in deterministic comparisons and within 0.86% in stochastic ones. We then use the algorithm to analyze a dual-storage system with different capacities and losses, and show that the policy properly uses the low-loss device (which is typically much more expensive) for high-frequency variations. We close this chapter by demonstrating the algorithm on a five-device system. We then use the algorithm to analyze the behavior of distributed storage across a model of the PJM electrical grid based on simulations with historical data, and we provide benchmarking for a deterministic version of this multidimensional problem. Finally, we study the integration of high levels of solar power in the presence of varying degrees of distributed storage and report on the effect of increased solar capacity on conventional generation levels, the value of storage at different levels of solar, the curtailment of solar power at high penetration levels, the impact of different levels of solar penetration on the storage control policy and the effect of combining distributed storage and solar power on load-weighted locational marginal prices of electricity.