The Economics Of Demand Aggregators In Electricity Markets

Abstract

Distributed storage capacity at load centers (e.g. deferrable demand) can lower the total cost of generating electricity by smoothing out and flattening the daily dispatch profile of conventional generating units. The main savings come from shifting purchases from peak to off-peak periods and mitigating the variability of generation from renewable sources. Since it is unrealistic to assume that system operators can control large numbers of deferrable demand devices directly, aggregators will be responsible for managing these devices using instructions provided by a system operator. The objective of this thesis is to compare the performance of deferrable demand when 1) the aggregators treat deferrable demand as a grid-scale virtual battery and receive physical charge/discharge instructions for managing deferrable demand (i.e. centralized control), with 2) the aggregators follow their own interests and submit bids into the wholesale auction to minimize the expected cost of purchasing energy using projected prices provided by the system operator (i.e. hierarchical control). The analysis uses a stochastic form of multi-period Security Constrained Optimal Power Flow to simulate operations for representative days. This model treats potential wind generation and load as stochastic inputs and determines the optimum daily profiles of dispatch for different realizations of hourly wind generation and load. Ramping capacity is acquired to ensure that transitions from one hour to the next hour, as well as contingencies, can be supported. With stochastic price forecasts, the optimum strategy for submitting bids is to determine a quantity each hour and 1) a low threshold price for charging storage, and 2) a high threshold price for discharging storage. This strategy implicitly provides ramping services as well as shifting purchases to off-peak periods, and the results for centralized and hierarchical control are almost exactly the same. With hierarchical control, however, the physical limits of storage may be violated during the planning horizon. An effective solution to this problem is to operate the system using a rolling horizon and allow aggregators to modify their bids at each step.