UC Berkeley

This thesis responds to the related imperatives of transitioning to low-carbon electricity systems and increasing global access to energy. It shows that microgrids and decentralized electricity systems are economically and technically capable of providing high levels of electricity access, and argues that incorporating active participation of electricity ``prosumers’’ into energy management systems enables more efficient electricity resource management.

Chapter 1 quantifies the tradeoffs between costs and electricity for autonomous solar and battery systems across sub-Saharan Africa, finding that on average these autonomous systems can achieve high levels of reliability at a cost of on the order of 10 USD cents per ‘9’ of reliability. Moreover, it shows that these costs could drop to as low as 3 cents per 9 as battery costs decline, and that decentralized systems are cost-competitive with legacy grids across much of the continent.

Chapter 2 proposes a load management system to manage electricity consumption in community microgrids with solar photovoltaics and battery storage while accounting for forecast uncertainty. It uses stochastic, model-predictive control techniques to set consumption limits during periods of low solar availability and high-demand. Simulation experiments show the management technique improves system reliability and consumer benefits from electricity through fewer interruptions and better electricity availability to high value uses.

Chapter 3 studies optimal pricing and peer-to-peer energy trading systems in microgrids with 100\% renewable energy sources. It promotes a utility-maximization framework from which prices arise from exchanging electricity under scarcity, in contrast to standard marginal-cost based pricing that breaks down in 100\% renewable systems. It further proposes a negotiation algorithm for peer-to-peer energy transactions and proves its convergence to optimal exchanges.

Chapter 4 extends the algorithm from Chapter 3 to a more broadly applicable system for optimizing power exchanges in microgrids and larger power grids in using forward markets and real-time controls. This approach, based on a decentralized optimization technique known as the Alternating Direction Method of Multipliers (ADMM), uses price-based coordination and independent agents in an iterative bidding procedure. Its equilibrium is a welfare-maximizing dispatch that solves the non-linear and non-convex power flow equations. This system preserves individual privacy, efficiently incorporates network congestion and voltage constraints, is highly scalable, and is robust in practice to model error. In addition to the forward market, the chapter introduces an agent-based feedback control system that continues to optimize power exchanges in real time.

The thesis concludes with a brief summary and directions for future research.