Public Utilities Reports

PUR Guide 2012 Fully Updated Version

Available NOW!
PUR Guide

This comprehensive self-study certification course is designed to teach the novice or pro everything they need to understand and succeed in every phase of the public utilities business.

Order Now

Smart Pricing, Smart Charging

Can time-of-use rates drive the behavior of electric vehicle owners?

Fortnightly Magazine - October 2011

Using this price elasticity with the High TOU rate shown in Figure 2, we find that the percent of customers charging during the peak period would drop from 60 percent to 55 percent. This isn’t likely to be a meaningful impact for grid operators who are trying to mitigate the adverse impact on the distribution system.

However, one could expect the price elasticity to be higher for a single, discretionary end use such as PEVs. But how much higher would it need to be to make a dent in the transformer overload problem? We have run simulations with a wide range of price elasticities to answer this question. 18 At the extreme, a value of -0.80 will be needed to effectively eliminate peak time charging. A value of -0.25 will be needed to eliminate half of the normal peak time charging load. The results of these simulations are shown in Figure 5.

Which of these price elasticities is realistic? That can only be resolved by conducting well-designed pricing experiments.

Designing the Study

The best way to estimate the parameters of a model for predicting charging behavior is to conduct a social experiment in which a large number, perhaps a thousand or more volunteers, are surveyed to study their charging behavior under alternative TOU rates. 19 Then these customers are randomly allocated to a control group and three treatment groups corresponding to the three rate types. Customers in the control group continue to drive their existing vehicles throughout the study while those in the treatment groups acquire or are provided a LEAF after a relatively short interval. Random allocation, similar to that carried out in medical clinical trials, ensures that treatment and control groups will be comparable both in observable and unobservable characteristics. The driving behavior and lifestyles of the experimental participants are then observed for several months before the LEAFs are delivered to them. This is necessary to establish a baseline since the best experimental designs feature before-and-after measurement as well as side-by-side measurement.

It would be a mistake to not include a control group in the design and also a mistake to not include a pre-treatment period. The temptation to only include a treatment group and measure their usage only after they have been given the LEAF should be resisted at all costs. Although it might yield results, those results could be subject to serious biases whose magnitudes might be incapable of being inferred from the experiment. The absence of a control group is also likely to lead to imprecise parameter estimates.

The design described above would yield both longitudinal and cross-participant data, i.e., constitute a cross-section of time series at the individual owner level. Such a panel data set would lend itself to econometric estimation by using either the fixed-effects or random-effects models that have successfully been used in the literature on dynamic pricing.

Because saving money is only one motive for charging at a particular time, we must examine other driver attributes before a predictive model of charging behavior can be developed. The most salient attributes are likely