Decentralized scheduling of EV energy and regulation reserve services in distribution network markets
Yanikara, Fatma Selin
MetadataShow full item record
The electricity transmission and distribution (T&D) grid is undergoing a paradigm shift as renewable generation explodes while flexible, storage-like loads are being massively adopted. We address the intermittency and volatility issues of renewable resources in connection with spatiotemporal distribution location-specific marginal-cost-based prices (DLMPs) that guide flexible loads to utilize their significant degrees of freedom for the purpose of providing valuable storage-like services to the grid including demand response, energy charge/discharge arbitrage and regulation reserve services. Dynamic DLMPs can induce socially optimal energy and reserve schedules to be adopted by flexible load. To this end, existing transmission wholesale markets must be extended to include distribution network connected participants. Since the inclusion of the complex preferences of many flexible loads renders familiar centralized transmission market designs intractable, we propose and investigate tractable decentralized market designs with Electric Vehicle (EV) battery charging selected as the representative flexible load. We address the equilibrium existence, uniqueness, and efficiency issues that arise with decentralized market designs, using game theory techniques. We investigate various multi-hour and multi-commodity (energy and reserves) market designs including EV self-scheduling under distribution network information aware/unaware conditions, and single or multiple load aggregator(s) scheduling groups of EVs. We investigate the role of network related information in enabling partially price anticipating EVs to acquire market power and self-schedule to achieve individual benefits at the expense of social welfare. Our contribution is the proof of existence and uniqueness of decentralized market equilibria, as well as analytical and numerical comparative analysis. Secondly, we depart from the usual ideal battery assumption, employing instead a realistic two bucket model. We then develop a novel Markovian Decision Process (MDP) application to estimate the regulation tracking cost incurred over an hour by an EV charger employing an optimal controller to respond to the regulation signal which is reset every two seconds by the system operator. The hourly tracking error increases when the EV promises higher regulation reserves while at the same time demanding an achievable albeit high average charging rate. We solve the MDP repeatedly, in fact off line, to capture the impact of the average charging rate and the regulation reserves promised at the beginning of an hour to the resulting hourly regulation tracking error. We then estimate a convex closed form relationship mapping hourly charging rate and regulation reserve offerings to the expected hourly tracking error cost. These convex tracking cost functions provide crucial input to the day ahead hourly energy bids and regulation reserve offers made by individual EVs to the Day Ahead market in response to spatiotemporal DLMPs.