TODAY’S STUDY: WHAT SMART METER DATA SHOWS
Insights from Smart Meters: The Potential for Peak-Hour Savings from Behavior-Based Programs
Todd et al, March 2014 (Lawrence Berkeley National Lab)
Key Findings: Insights from the data
Previous to the rollout of smart meters, monthly utility billing data was used to estimate monthly and annual energy savings for BB programs. Without higher-frequency electricity consumption data, it was it was not possible to determine when during the day that these savings occurred. The analysis in this report is the one of the first to estimate the hourly profile of these savings.
New types of analysis enabled by investments in smart meters AMI allow us to examine hourly patterns of electricity usage and savings by customers participating in BB programs and perform statistical tests of whether savings during peak hours are higher than other hours. We employed a regression technique that compares the electricity use of the treatment group to the electricity use of the control group jointly for each hour of the day.
In addition, we used similar techniques to estimate the savings during all of the peak hours (which allows us to test whether or not the peak hours showed savings that were statistically significantly higher than savings during other hours), and the savings on the highest system peak days.
New kinds of results from the hour-by-hour electricity savings estimates are shown in Figure 1 (along with the 95% confidence intervals). The savings are shown with three different scales: first, kWh savings (left-hand y-axis on the top graph); second, normalized savings (right-hand y-axis on the top graph) as a percent of the total average energy usage of the control group across all hours (in order to give a sense as to how large the kWh savings are); and third, proportional savings (y-axis on the bottom graph) as a percentage of each hour’s average energy usage for the control group (in order to show the proportional savings relative to the energy consumed for each hour).
For analysis of the particular program rollout that we are using as our test-case (shown in Figure 1), we find:7
• Statistically significant electricity savings during every hour;
• Higher kWh savings during peak hours; and
• A higher percentage of savings during peak hours, relative to the energy usage in each hour.
These results show that this pilot program rollout resulted in savings that are higher during peak hours. It is particularly interesting because the savings disproportionately increase during the peak hours. Without hourly data, one assumption that was commonly used (based on anecdotal evidence) was that this was not the case; that either the savings are spread out evenly in proportion to the electricity usage, or that savings are actually harder to achieve during peak hours. Figure 2 displays hour-by-hour savings, but for only the ten highest and ten lowest system peak days included in our dataset. The X and Y-axis scales are similar to the previous graph: first, kWh savings; second, normalized savings as a percent of the total average energy usage of the control group across all hours; and third, proportional savings as a percentage of each hour’s average energy usage for the control group during the ten highest and ten lowest system peak days. For reference, Figure 2 also includes the savings during all days from Figure 1.
Figure 2 shows additional key findings:
• Higher peak savings during the ten highest system peak days and
• Slightly higher proportional peak savings during the ten highest system peak days
Together with the findings from Figure 1, this implies that BB programs have the potential to induce electricity savings exactly when they are most needed; the savings are disproportionately high during peak hours on peak days.
Key Finding 1: Proof-of-concept analytics tool
High-frequency data from smart meters enable new forms of analysis techniques that allow us to examine hourly usage patterns and determine when during the day households in BB programs are actually saving. This includes hour-by-hour savings estimates and rigorous peak versus off-peak statistical tests. Implication: This allows measurement of the effectiveness of BB programs in producing peak-hour savings and improves the prediction accuracy of load forecasts.
Key Finding 2: Novel result
Our results show an example of one rollout of a BB program that provides savings during every hour, with disproportionally high savings during peak hours and during high system peak days.
Implication: BB programs have the potential to provide peak-hour savings, and should be considered as a potential (non-dispatchable) resource for improving short-run reliability. If the peak hour energy savings can be maintained and accurately predicted over time, system planners can assess whether this type of program is treated as a planning capacity resource.
While we show that it is feasible for such BB programs to provide peak-hour savings, these results may be specific to this particular program in this specific situation. Because this is only one example of a BB program that provides peak-hour savings, this does not imply that these results can be generalized and that all BB programs can provide this kind of savings.
Until we have a better understanding of what is driving these savings levels and their differences across different populations and under different circumstances, it is not possible at this time to definitively conclude that all BB programs will produce peak hour savings.