A Storage pricing mechanism for learning agents in the smart electricity grid

Abstract

Masdar City is designed to be powered solely by renewable energy. The main of renewable energy generators to be used are solar photovoltaic panels and wind turbines. In the future, geothermal generators are planned to be added to the grid. However, the stochastic nature of renewable energy generators has remained a major challenge in their sole and large-scale deployment. Given that these renewable energy generators lie at the heart of the smart grid (a grid that enhances the existing grid by providing distributed communication and information systems), there is thus a need for computational solutions that will aid in the reliable deployment of renewable energy grids. The traditional way of dealing with the intermittency of renewable generators has been to over build generation capacity. This helps make up for shortfalls due to errors in prediction and also ensures that there is adequate supply to cover peak demand periods. Usually, peak demand do not coincide with peak generation periods. Another widely employed solution is the use of storage which also helps to smooth out the effects of prediction errors and provides electric energy during peak periods. These two approaches are sub-optimal in that they both require excess generation capacity which leads to wastage of renewable energy during off-peak hours. This thesis proposes a new approach to the problem by making demand respond to renewable generation patterns. Specifically, it couples electricity storage both at the generator and consumer ends and then uses real-time electricity pricing to incentivize consumers to respond to peak generation. The model consists of a green energy supplier (with utility-scale deep-cycle batteries) and a significant number of consumers that are subscribed to it for their total energy needs. The consumers are distributed on the grid each of whom has some storage capability. The novel storage-pricing mechanism presented in the thesis makes use of the storage information from the renewable supplier to generate daily, real-time electricity prices which are communicated to the consumers. This information is already available to the supplier and involves no extra communication overhead. The consumers are modeled as autonomous agents that are capable of communicating and reacting optimally to received signals. I empirically evaluate the system and show that, when all homes are equipped with storage devices, the supplier can significantly reduce its generation capacity (and capital cost) while satisfying peak consumer demands. By so doing, the supplier improves the ef?ciency of the system by up to 23% while the consumer reduces its costs by up to 35%. Finally, I evaluate the effect of varying the number of the consumers having storage and show that even with reduced storage presence, the system efficiency still improves.

Date of Award	2011
Original language	American English
Supervisor	Iyad Rahwan (Supervisor)