In partnership with PLMA, this group is for practitioners from energy utilities, solution providers, and trade allies to share load management expertise and explore innovative approaches to program delivery, pricing constructs, and technology adoption.


A peak coincidence visualization for quantifying ICAP risk

image credit: Advanced Energy Intelligence, LLC - All Rights Reserved
Carl Popolo's picture
CEO, Advanced Energy Intelligence, LLC

  • Member since 2018
  • 11 items added with 4,530 views
  • Jan 24, 2019

In a portfolio of dozens or hundreds of buildings, it can be a challenge to manage DR protocols and ICAP tags.  The challenge is a twelve-month challenge – avoiding new billing period peaks in January can be as cost-effective as avoiding a high ICAP tag in the summer.  But when focusing on ICAP specifically, we’ve tried to quantify the risk using a simple peak coincidence analytic that we developed to help assess this risk.

Statistically, if you take any month of the year, a building will typically peak around the same time each day, with some distribution around that central tendency.  Similarly, during the summer months, the ISO also peaks with some distribution around a central tendency.  The degree to which the two distributions overlap can be seen as a measure of risk coincidence.

Let’s look at a single building on a single day.  Specifically, this is a simple load plot for Boston City Hall on August 29th, 2018 and it so happens that the ISO New England peak hour for 2018 was at 5:30 pm on that day.

Advanced Energy Intelligence, LLC - All Rights Reserved

On this particular day, the building had a peak load of about 2,181 kW around noontime.  By the time of the grid peak at 17:30, the building load was closer to 2,031 kW.  In retrospect, knowing that ISO New England declared the ICAP hour at 5pm on this day, we can say that the natural tendency of the building helped save the building about 150 kW in its capacity charge basis for 2019.  In other words, if this day was typical then the natural tendency of the building is to peak earlier than the grid in a typical August.  That’s great in terms of this specific day, but is it statistically the case that the building is always likely to peak 5 hours before the grid?  And in fact, given that the building and the grid typically peak during hours of high occupancy, it just a matter of statistical musical chairs that the building might be peaking in the same hour as the grid.  Can this be quantified as a measure of the buildings peak coincidence risk?  We think it can, by statistically measuring the degree to which the distribution of peaks for the building and the grid overlap.  If the building always peaks at noon and the grid always peaks at 5pm, then we might calculate that the risk of a coincident peak is zero. But if the building peaks between 10am and 2pm while the grid peaks anywhere between 1pm and 8pm, then there is some non-zero risk that the two will peak together.

This visualization below attempts to quantify this statistical view of the data.  In this view, the y-axis is a time-of-day scale, and we are plotting the 31 daily peak times for the Boston Copley Library (blue) and the grid (purple).  This visualization is really nothing more than a tricked-out box and whisker plot which is a simple technique for determining the degree of similarity for two different distributions. 

Advanced Energy Intelligence, LLC - All Rights Reserved

The time of each daily peak load is plotted as a point on the vertical line.  The “box” in each plot contains the middle 50% of all the time values, and the median is shown as a black horizontal mark.  So for Boston Copley in August 2018, the building peaked before 3pm on 50% of the days, and for half of the days it peaked after 3pm. Conversely, the grid never peaked before 3pm in that month and had a predominant tendency to peak between 5pm and 6pm.  For those of you who remember your high school statistics, the degree to which these two distributions are coincident is a function of the median values and the interquartile ranges of the distributions.  Through a little transformation, we score this particular case with a value of 34 on a scale of 0 to 100.  The transformation isn’t important except that it converts the calculation into a scale that is easier to recognize as a risk factor when looking at a large population of buildings.  A score of zero implies no overlap at all and a score of 100 results when the medians and the interquartile ranges are the same.

Given that the ISO New England grid tends to peak late in the afternoon, it shouldn’t be a surprise that we rarely see buildings with significant risk factors – but in statistical terms, that risk is not zero so there is always the chance that a building will be peaking late on any given day at the same time as the grid.

Pragmatically, it’s obvious that a building with a low risk rating should not knowingly ignore a called ISO event.  But with this statistical method for assigning risk factors, the portfolio manager has a way to determine whether an aggressive DR position is worthwhile for certain buildings that have a high coincident risk.  The better analysis we are pursuing is to factor in the ratio of the building’s load at the ISO peak time as a fraction of the building’s load at its peak.  In other words, the building may peak 5 hours ahead of the grid, but that’s small consolation if its load at the time of the ISO peak is still relatively high.  Some buildings like Boston Copley exhibit square-wave behavior – the building comes on at 6am and basically ignores all outside influences until it closes at 9pm.  It may technically peak at noon, but its load at 6pm is almost as high.

Do you care enough about ICAP tags to employ some statistical method to assess risk, or is this just a paper chase to the obvious DR of flipping circuit breakers on hot days?

For a working demo of the visualization described above, visit our site at  Choose a building and then click on the Grid Coincidence tab.


Matt Chester's picture
Matt Chester on Jan 24, 2019

Carl-- in your analysis, are there any particular types of customers/buildings that tend to be higher risk and thus should be more interested in this type of analysis? Outside of just buildings that are vulnerable when there are power issues (hospitals, retirement homes), but ones that tend to line up and have riskier peak hour trends?

Carl Popolo's picture
Carl Popolo on Jan 24, 2019

Hi Matt,

We don't have a go-to metric for building types that are most susceptible, but as a class we think that Municipal buildings across the board tend to do better because the portfolios are significant and the towns are generally eager to find savings.  Within that, schools tend to peak earlier (dominant kitchen operations, school day ends earlier compared to a commercial building) - and generally overall risk is lower if they are closed for the summer.  But there's always the chance of afternoon activities that will light up a gymnasium or auditorium that can get a school in trouble.

Larger private operators usually have some sophisticated DR protocols or utility incentives precisely because their occupancy schedules are such that they must run with good building comfort well into the late afternoon.  So as a class, they are at risk but generally seem to be very proactive.

Bottom line is that if the building operator is paying attention, those buildings tend to do better.  They also tend to do a better job of scheduling equipment during unoccupied hours (70% of an average week), and recognize that demand charges are a twelve-month savings opportunity - they will tend to do better because they are managed more closely.

- Carl



Carl Popolo's picture
Carl Popolo on Feb 4, 2019

Hi Matt - I don't have permissions to contact you directly, but can you check why the images in my post have disappeared?  I was looking to check something I had written and just noticed that the 2 or 3 images in the article are no longer showing.  Thanks...

Carl Popolo's picture
Thank Carl for the Post!
Energy Central contributors share their experience and insights for the benefit of other Members (like you). Please show them your appreciation by leaving a comment, 'liking' this post, or following this Member.
More posts from this member

Get Published - Build a Following

The Energy Central Power Industry Network is based on one core idea - power industry professionals helping each other and advancing the industry by sharing and learning from each other.

If you have an experience or insight to share or have learned something from a conference or seminar, your peers and colleagues on Energy Central want to hear about it. It's also easy to share a link to an article you've liked or an industry resource that you think would be helpful.

                 Learn more about posting on Energy Central »