Optimization Under Uncertainty for Management of Renewables in Electricity Markets

Abstract

This thesis deals with the development and application of models for decision-making under uncertainty to support the participation of renewables in electricity markets. The output of most renewable sources, e.g., wind, is intermittent and, furthermore, it can only be predicted with a limited accuracy. As a result of their non-dispatchable and stochastic nature, the management of renewables poses new challenges as compared to conventional sources of electricity. Focusing in particular on short-term electricity markets, both the trading activities of market participants (producers, retailers and consumers) and the decision-making processes of system and market operators are challenged. As far as producers are concerned, participation in electricity markets imposes them to make their trading decisions with a certain advance in time as compared to energy delivery. Since their actual output is uncertain at the time of bidding, the trading problem for a renewable power producer translates into a stochastic optimization problem, whose objective is the maximization of the expected revenues. In this thesis, we consider the trading problem for a wind power producer both in markets with low penetration of renewables, where the producer is a price-taker, and in markets where the producer acts as a price-maker. Owing to the demand response initiatives to be undertaken in future power systems, the operation of electricity retailers and the behavior of consumers are also going to be influenced by renewable power production. Another focus of this thesis is on time-varying price mechanisms to make the most of end consumers' flexibility. In particular, the problem of managing optimally a virtual power plant equipped with renewable production facilities and flexible consumers is addressed through control-by-price. In a similar setup, the optimal trading (and pricing) problem for a retailer connected to flexible consumers is considered. Finally, market and system operators are challenged by the increasing penetration of renewables, which put stress on markets that were designed to accommodate a generation mix largely dominated by conventional sources. Indeed, the traditional market design, based on the sequential clearing of successive market floors and on deterministic rules and criteria, is characterized by higher and higher degrees of suboptimality and lower reliability as the penetration of renewables increases. This work contributes to the state-of-the-art by proposing new mechanisms for day-ahead dispatch and reserve determination in markets with high penetration of renewables, on the basis of stochastic criteria.