Sagewell Blog

Back to Blog


Smart meter analytics has huge potential for improving the way utilities and program managers evaluate the performance of energy efficiency programs.

Traditional evaluation methods involve combining "deemed" savings from theoretical models - which may or may not be realistic - with occasional verification from a small sample of participating locations. Even these samples often do not actually measure that which is being targeted: changes in property energy consumption. 

In short, the performance metrics of most efficiency programs are essentially determined before a single measure has been implemented. We do not know how much energy has been saved by most energy efficiency programs simply because it has not been measured.

Luckily, this situation is entirely avoidable and the solution is clear: use hourly AMI meter data to actually determine how the energy consumption changes in properties that participate in programs.

Let's take a look at one example. Note - the below analysis is not intended to be a full-scale EM&V project, but is simply designed to demonstrate the power of AMI meter data. See the bottom of the post for a more thorough disclaimer!

Insulated-v-peerChart 1: Average summer load shape for weatherized homes vs comparable peer group.
Chart 1, above, shows the summertime average load shape for a group of recently weatherized homes (orange) vs a group of peer group properties of similar sizes (blue). We can clearly see that the weatherized homes used significantly less energy than their peers, by approximately 0.2 kWh/h, or around 17%. The difference is greatest during the evening hours (around 0.3 kWh/h), when overall consumption is highest. This is good news for peak avoidance!
But how - I hear you ask - can we be sure that the energy consumption difference was caused by the weatherization?
Chart 2, below, is a before (orange) and after (blue) analysis our recently weatherized sample homes. It shows the summertime average load shape in 2016 vs 2018 for homes that participated in an insulation and weatherization program during 2017. It clearly shows a drop in energy consumption between the two summers of around 0.15 kWh/h or 15% overall and 0.1 kWh/h on-peak (4pm to 9pm). This corroborates the data seen in the first chart. We're beginning to triangulate some results!
Insulation Before-After-2Chart 2: Average summer load shape for insulated homes before and after weatherization program completed.

But wait - what about weather effects? Couldn't they explain the difference between one summer and the next?

Absolutely - so let's try to get a rough measure of what those weather effects might be, so we can factor them into our analysis. Chart 3 below, shows the single family average load shape for the same two summers as used in the "before and after" chart shown above. Any differences in these shapes will be due, to a large degree, to changes in weather (see caveats at bottom of post). 

1-fam before-after-1Chart 3: Average summer load shape for all single family homes before and after weatherization program completed.

The chart shows a small decrease in energy consumption from one summer to the next. This average out to around 0.05 kWh/h. So we can factor this into the result we saw in chart 2 - around 0.05 kWh/h of the before-after difference might be due to the weather - leaving us with a 0.1 kWh/h net reduction.

To summarize, the results of method 1 (peer group) suggested a 17% energy saving while the results of method 2 (before-after) suggested a 9% energy saving. Not bad! 

The analysis presented here took the author less than a couple of hours to conceive, conduct and summarize in this post. High quality program evaluation need not be an expensive or difficult undertaking. Utilities already have the data they need and the tools are readily available. Let's get started!


Important caveats: we did not attempt to be scientific in our selection of peer group homes, we simply controlled for some sources of obvious error (e.g. homes with PV solar systems). A full EM&V project would need to select peer groups for each individual measured home, rather than comparing groups to one another. We also recommend that all EM&V projects use weather normalization to counter the confounds of different temperature profiles for different time ranges. This analysis used a simple way to infer weather effects, but would not be robust enough "in the real world".

Note - all analysis contained in this post was conducted using the SageSight analytics platform.


Leave a Reply.