Do Time-of-Use Prices Deliver Energy Savings at the Right Time?
Yes
This article is a little different from my usual fare. It focuses on a recent research paper about electricity prices that my UC Davis colleague Kevin Novan, a graduate student Zheng Fu, and I wrote. This article is co-authored with Kevin and cross posted on the EI Blog.
In 2018 and 2019, the Sacramento Municipal Utility District (SMUD) blanketed local media with information about electricity prices. They hit local TV, radio, newspapers, and the internet. They sent out mailers. It was like being in a swing state during election season.
SMUD’s message was that electricity prices were about to change to a time-of-use (TOU) system. Customers would pay a higher price in the early evening hours on non-holiday weekdays in the summer. As the figure below shows, instead of paying a constant rate of 13c per kilowatt hour (kwh) for all consumption, customers would pay 28c/kwh from 5-8pm, 16c/kwh from noon-5pm and 8pm-midnight, and 12c/kwh from midnight to noon. On weekends and holidays customers continue to pay a constant rate. This change was the culmination of a long process, including pricing trials that Catherine Wolfram summarized in this blog more than a decade ago.
SMUD charges more in the early evening because that is when electricity use is highest just as production from solar is declining. Reducing peak electricity use can have huge benefits. The grid needs to have enough capacity to meet peak demand, so higher peak load requires more generators that are only used sporadically. These “peaker” plants are expensive because they are rarely used, less efficient and more polluting.
By charging high prices during peak periods, the utility hopes to reduce peak load.
Peak electricity demand is much higher on hot days than cooler days because of air conditioning (AC). However, TOU prices apply every day. If TOU pricing is to reduce peak load, then it needs to reduce air AC use on the hottest days.
In a new EI working paper, the two of us, along with UC Davis PhD graduate Zheng Fu, evaluated the effect of SMUD’s switch to TOU pricing on residential AC. We used data from Ecobee smart thermostats. Ecobee has a program called Donate your Data in which owners of Ecobee thermostats can elect to make their anonymized data available for research. The data includes information on the number of minutes that the AC unit is running each hour and the thermostat set point.
We compare the number of minutes that Ecobee customers used their AC before and after the TOU switch. As a control group we used data from Ecobee customers outside SMUD in Climate Zones 12 and 13, which includes Fresno, Merced, Davis, and other cities. Customers in these non-SMUD areas did not experience a price change between 2018 and 2019.
Between 2018 and 2019, AC use in peak hours in the SMUD service area did not change, as shown in the figure below. However, AC use in non-peak hours increased. In the non-SMUD areas, AC use increased in both peak and non-peak hours. Assuming that, in the absence of TOU pricing, peak AC use in SMUD would have increased similarly to the non-SMUD areas, we estimate that TOU pricing caused peak-hour AC use to decrease by 3.5 minutes per hour, a 15% reduction. In additional results in the paper, we also show that there is no effect on AC use on weekends. This is reassuring because during weekends there are no TOU prices.
Using the data from Ecobee we can also look directly at thermostat settings. In 2019, SMUD users turned up their thermostats by nearly one degree during peak hours from an average of 77.6° to 78.4°. Non-SMUD customers lowered their peak-hour set points from 77.8° to 77.5° between the two years. Outside of peak hours, SMUD customers kept their set points at 77.8° across the two years, whereas non-SMUD customers lowered theirs by about 0.4° on average.
So, relative to the control group, SMUD customers increased their set point in peak hours by 1.1°, as shown below.
This evidence suggests that TOU pricing reduces peak energy use, but there are (at least) two reasons we should hesitate to draw that conclusion yet. First, Ecobee users reduce AC use on an average summer day, but do they do it on the hottest days? In Sacramento, an average daily temperature of 62° implies a high of about 75° and a typical day with an average of 75° will exceed 90°. Focusing only on days when the daily average temperature exceeds 75°, we see slightly larger reductions in AC use (4.5 minutes) and similar increases in set point (one degree) after the introduction of TOU prices.
A second potential issue with our analysis is that Ecobee customers may be different from the average customer. If you get a smart thermostat, you may be someone who pays more than average attention to AC use. On the other hand, Ecobee users may be richer and therefore less sensitive to price than the average customer.
To address this point, we examine total load in SMUD’s service territory. Controlling for temperature, we estimate that SMUD’s average hourly peak period load fell by 187 MWh on summer weekdays in 2019 relative to 2018 (about 10%). It did not fall significantly on weekends, as illustrated below.
This number is probably an overestimate of the effect of TOU pricing because consumption in non-peak hours also fell between the two years, although not by as much. Here’s another way to estimate how much of a reduction in peak load came from TOU prices. SMUD has about 575,000 residential customers. If each has a 4kW air conditioning unit and it runs 3.5 fewer minutes in an hour, that is a savings of 0.233 kW per customer/hour, which adds up to 134 MW or about 7% of average hourly peak period load.
To put these number in context, on September 6, 2022, Sacramento hit a record high temperature of 116º amid a statewide heat wave. At 4:55pm on that day, electricity demand reached its highest ever level in the state. To avoid the need for widespread rolling blackouts, an emergency text message was sent to millions of Californians urging them to “conserve energy now to protect public health and safety.” This message was credited with reducing the system load by 4%, likely averting costly outages.
A common concern with TOU rates is that they don’t provide strong enough signals to reduce consumption on the hottest days. The importance of reducing consumption on the hottest days will only increase as the climate warms. Although larger efficiency gains could be achieved by adopting prices that vary with the marginal cost across days, our results reveal that, even without varying the price signal across days, TOU rates can induce sizable reductions in peak period cooling, and thus achieve peak period energy savings that are concentrated on the highest demand days.