PNNL

Abstract

Dynamic electricity prices are increasingly being used to manage electricity availability and scarcity. Electric vehicle (EV) charging can take advantage of real-time electricity market price volatility. Presuming that an EV must be fully charged at a future target time, the EV should choose to charge using the lowest future electricity prices and thereby minimize electricity cost. This is straightforward to accomplish if forward prices are available; however, statistical methods must be used if forward prices are unavailable to the EV charger. In this case, historical prices and trends must be mined to anticipate which prices should and should not be used to charge the EV. Correlation models may be formulated to calibrate future electrical prices as a function of categorical variables like season, weekday, and hour. But price persistence, a tendency for electricity prices to inexplicably become and remain relatively high or low for extended durations, is a bias that is difficult to forecast and mitigate. This paper formulates and tests a pragmatic strategy for integrating conventional static statistical prices and the Bayesian propagation of price persistence from the current price to prices in the current and future hours. Simulations were conducted to test the cost effectiveness of charging strategy using real-time electricity prices. The effects of charge duration, initial state-of-charge, and time-of-day are evaluated and discussed. The results of 3600 scenarios are compared using this strategy, conventional charging, in which the first market prices are used until the EV is fully charged, and the best prices from perfectly accurate forecasts.