Can Electricity Demand Management Drive the Transition to Clean and Affordable Energy in Poor Economies?

The Government of India has a target of installing 250 million smart meters by 2025. This rapid deployment of smart meters has the potential to accelerate the clean energy transition by enabling dynamic retail pricing of electricity, which could incentivise consumers to use power when it is generated from clean sources and thereby reduce the need for costly supply-side solutions such as energy storage to compensate for times when the supply of renewable energy is insufficient to meet the demand. Furthermore, to the extent that dynamic pricing induces peak-shaving, regulations that enable dynamic pricing could lower the cost of supply, which is especially important in the Indian context where distribution companies often contract a large amount of generation capacity to meet the anticipated demand of only a few hours of the year. However, the effectiveness of dynamic pricing depends on (a) consumers’ awareness of the retail price of electricity, and (b) their ability and willingness to respond to changes in the retail price of electricity. We will conduct a randomised control trial in partnership with an electricity distribution company in India, where participants will be offered simple IoT-enabled automation devices that generate automated switch-off events. Participants will be offered rewards for each kWh of energy saved during the switch-off events. By allowing the rewards to vary over the hours of the day, the trial can be used to simulate a dynamic price, although we also expect to uncover alternative ways of generating incentives for load balancing.

To recruit participants for the study, we will work with a large Indian electricity distribution company to target distribution feeders with a high proportion of domestic smart meter users. A random sample of 5,000 domestic smart meter users will be invited every three months over the course of one year to participate in the study. The invitation will describe the purpose of the study, the study procedures, and the benefits to participants as individuals and to society. The first 1,000 customers that express an interest will be invited to enroll in the study on a web platform. Participants will be sent a Wi-Fi enabled smart switch, which will be used to remotely operate one large appliance, such as a washing machine, an air conditioner or an electric geyser. The IoT algorithm will generate brief switch-off events one or more times a day, which participants can override by turning the smart switch on manually or via a smartphone app. At the time of registration, participants will have the opportunity to state all the times during each day of the week when they would prefer not to have a switch-off event. They can change their preferences on the app at any time during the three months for which they will be enrolled in the study. All participants will be offered guaranteed monetary rewards in proportion to the units of electricity saved during the switch-off event. If they override the event and turn on the smart switch, they will only be rewarded for the amount of time that the appliance was switched off before the override.

2023-07-01

2024-06-30

The primary outcomes are (a) participation, and (b) meter- and switch-level electricity consumption, which will be monitored automatically through an integrated web platform and smartphone app that have been developed for this project.

In each three-month iteration of the study, 1,000 participants will be randomly assigned to one of two cross-cutting treatment groups: (1) high and low reward per kWh of electricity saved during switch-off event and (2) fixed vs variable reward rate schedule over the hours of the day. Depending on the number of switch-off events administered per customer-day, the research team may also (a) periodically vary the fixed and variable rate schedules for a subset of participants to more precisely estimate the willingness to pay to avoid a switch-off at each hour of the day, and (b) vary the amount of notice time given to participants before each switch-off event to estimate how the scope of demand flexibility varies with the amount of advance notice given to consumers.

After collecting one month of baseline plug usage data, we will assign prospective participants into 24 groups based on the hour of the day when the first switch-off event will occur. We will randomly rotate the group assignment at the start of each day in order to generate within-customer variation in the timing of the switch-off events, which will allow us to develop hourly estimates of the welfare cost of a power supply interruption to the selected appliance. Participants will be informed prior to take-up that a switch-off event may occur during a one hour window at least once a day. We will stratify the sample based on baseline electricity consumption as a proxy for wealth. Stratified randomisation is important as more affluent households are likely to be less responsive to switch-off events perhaps because they may have access to multiple appliances of the same type. Conversely, poorer households may be oversensitive to switch-off events as their appliance stocks may be smaller. We will run the intervention in four phases, enrolling four cohorts of 1,000 participants each in October 2023, January 2024, April 2024, and July 2024 respectively, to study how seasonality affects take-up. Each participant will remain in the study for three months unless they choose to exit earlier. By observing users’ tendencies to override the switch-off events, we will generate insights into their true electricity demand flexibility; in other words, we will be able to observe the times of the day and days of the week when individuals may or may not be willing to turn off the selected appliance. We will also be able to explore how the potential for demand flexibility differs by location, season, weather, and baseline electricity consumption.

While IoT technologies provide no intrinsic utility to the user, they generate a positive externality in terms of increased efficiency of the power grid through the demand flexibility that they make possible, implying that users may need to be compensated (Richter and Pollitt, 2018). Therefore, we will offer households the possibility of earning monetary rewards commensurate with plug usage. Participants will earn rewards for each unit of electricity saved as a result of the switch-off event triggered through our web platform. Participants in each cohort will be randomly assigned to one of four reward treatments at the start of the experiment:

A: Fixed and Low Reward Rate: Participants in this group will receive a fixed payout rate INR X per kWh of electricity saved during switch-off events at any time of the day
B: Fixed and High Reward Rate: Participants in this group will receive a fixed payout rate INR Y per kWh of electricity saved during switch-off events at any time of the day, where Y is greater than X.
C: Variable and Low Reward Rate: Participants in this group will receive a variable payout rate for the electricity saved during switch-off events at different time of the day that averages to INR X per day
D: Variable and High Reward Rate: Participants in this group will receive a variable payout rate for the electricity saved during switch-off events at different time of the day that averages to INR Y per day, where Y is greater than X.

If the participant overrides the switch-off event (i.e., by turning the switch on manually or through the app), they will only be rewarded for the amount of time that the plug was switched off before the override. Depending on the number of switch-off events we can administer per customer-day, we will introduce two additional cross-cutting interventions: (a) vary the fixed and variable reward rate schedules every week for a subset of participants, which will allow us to develop precise estimates of the willingness to pay to avoid a switch-off event, and (b) vary the amount of notice time given to customers prior to switch-off events, which will allow us to study how the scope for electricity demand flexibility varies by notice time, an essential parameter in the design of demand response programmes. The exact reward rates will be determined after we have conducted a pilot, but it will be in the INR 20-50 per kWh range.

Randomization done in office by a computer

Residential smart-meter user x 30-minute interval

No

N/A

4,000 participants

1,000 participants in one of four reward treatment arms