Using Scatter Plots to Assess Building Performance–Part 1

Some of you may be wondering what happened to me and the string of posts I was doing about scoping a Dental Clinic.   The answer is that my summer became extremely busy between travel, projects, and little bit of personal time.  I hope to pick up where I left off in the next week or so.

Trying Out a New Idea

Meanwhile, I wanted to try something new;  specifically use a video clip to illustrate an technique vs. typing about it with a bunch of screen shots to illustrate key points.  Jay (Santos) one of the FDE principles actually set me up to do this years ago.  Bob Shultz showed him this program called Camtasia that did a video screen capture that he was using to show folks how to do things in control systems and Jay called me up and said “you should get that”.

So, I did and used it to convey stuff via illustration instead of words.  But given my perfectionist tendencies, I would turn the initial recording into a major editing session to make the video as perfect as possible;  if the phone rang in the middle of something I would edit it out; if Hobbes ….

Hobbs on pot

… (that’s him on pot) jumped up on the desk and flipped over on the keyboard for tummy rubs (causing all manner of things to happen depending on which keys he landed on), I would start over.

Recently a friend of mine (Sabastian St. John of St. John Consulting) suggested that a video of me illustrating, for instance, a spreadsheet technique I use, would not have to be perfect to be useful.  The ensuing discussion made sense to me, so what follows is me trying to embrace that concept in an effort to get some stuff out there as quickly as possible in an effort to help a number of people.

Looking for Shapes in the Clouds

I find myself using scatter plots more and more these days.   Basically, you are looking for shapes in the clouds …

image

image

… a past-time I enjoy quite a bit as a proud member of the Cloud Appreciation Society.

Of course, in the context of this blog, I am talking about this past-time in a less whimsical way, but even in that context, photos from the society members have provided some interesting insights, as illustrated in the string of pictures below.

image

image

image

image

Those photos are of real buildings down in Panama City Florida in the world’s biggest wind tunnel taken by J.R. Hott.

Looking for Shapes in Data Clouds

Unfortunately, for those of us working on buildings, the test set-up illustrated above is difficult to come by on an as needed basis.  But, what we do have available to us is data; lots and lots of data.  And our ability to gather it and target things for data gathering is constantly improving.

One of the very useful data sources that is becoming more common is interval data from “smart” utility meters.  Up until smart meters, the sampling interval for most of us for our utility data was approximately once a month, specifically on the day the meter was read.  While it was possible to do useful analysis with this type of data, the information was broad in context.

In other words, you could use the data to benchmark ..

image

.. or understand monthly patterns and correlate them to other data that was documented on a monthly basis like degree days.

image

But you could not use it to tell you if, for instance, the consumption pattern on Sunday was different from the other days of the week or the consumption pattern on a Wednesday in July was different from the pattern on a Wednesday in January.

Smart meters allow us analyze the details of how a building uses energy by providing data on a much more frequent basis;  typically once an hour for electricity and once a day for gas all though once every 15 minutes is not unusual for electricity and once an hour is not unheard of for gas.  Prior to smart meters, you had to have your own meter to get data like this or get a repeater signal from the utility meter that you then logged in your building automation system.

And smart meters have found their way to the consumer level.  For instance, here is a scatter plot of kW vs. time of day for my house last month that I just downloaded from my local utility’s web site, where I can access my smart meter data.

image

That’s a lot more granular than what I could do with the monthly bills using Bill Koran’s Utility Consumption Analysis tool (UCAT)

image

… especially if I use a little trick that lets me begin to see a shape in my meter data cloud.

image

By making the data points semi-transparent, it becomes clear that for our house, we seem to have a round-the-clock base load in the 0 to 1 kW range.  I am limited to 1 kW granularity by the utility data source.   I actually have a little Brultech meter on my panel that gives me even greater granularity and also lets me look at sub-circuits.  My point in bringing up my Brultech meter is to illustrate that the finer the sample size, the more your shape in the cloud can tell you. 

I have a little network problem right now so I can’t show you my Brultech data for August, but here is a data set I was looking at last winter, specifically from November 10-13th,  to try to understand the impact of the electric heat in my office and Kathy’s art studio, in particular to verify that the set-up/set-back features were working in the Art Studio since it is a separate structure.

image

Here is that same data set for those days in November from the perspective of my “Smart” utility meter.

image

Pretty different cloud.

A significant portion of the difference can simply be attributed to the number of samples.  My personal meter provided  23,862 data points  because it is logging data every 15 seconds or if anything changes by 60 watts or more.   In contrast, the utility meter data available to me is limited to once an hour, meaning a data set of about 96 data points for the four days in question.

As a result, I can see:

There Really Were No Hours  At 0 kW:  My guess is the utility rounds anything below 0.5 kW to 0 for the data set they furnish on line.

There Were Events Where the kW Were Above 6:  My guess is those events did not last long enough for them to impact what the utility presents to me when I log in.  For instance, most utilities bill demand based on a sliding 15 minute window.  That is why the in-rush from a motor start does not impact your demand.  So I suspect that the events my meter picked up that are over 6 kW did not last long and that the average for the hour they occurred in was at or below the peak of 6kW shown on the utility web site.  For instance, if I filter for events over 6 kW in my Brultech data set, there are about 1,200 (out of over 23,000) events and they last anywhere from 30 consecutive seconds to 4- 5 minutes, generally speaking.

My base load is less than 1 kW, probably floating between 0.5 and 0.8 kW:  Having watched my house for a while, I am pretty sure most of the base load (the load between midnight and 6 am) is our security lighting and some phantom loads I have not dealt with yet.  By midnight, Kathy and I are both asleep, the dishwater cycle has finished if we ran it, as have the laundry cycles if we are doing laundry. The base load variation is a result of some of the lights having motion detectors that trigger a bright cycle for a couple of minutes if motion is detected and then dimming after that while others are on steady.

Getting Started With Your Own Data Cloud

So at this point, if you are still reading this, you are probably interested in understanding how  you go about developing scatter plots like the ones I have been sharing.  That brings me to my video clips.  This first clip shows you how to start with a fairly basic data set, fix a few common problems (common in my experience) and develop a plot of energy consumption as a function of outdoor temperature.

Bear in mind that my goal here is to provide information but not necessarily perfection.  So, in this video, I would like to believe I get the point across, but there are a few blips here and there where I miss-click or a link doesn’t work or I forget a step and have to go back.  So bear with me (no visit from Hobbes in this one).

In the next post, I’ll discuss the advantages of applying filters to a scatter plot and then link you up with a video illustrating the technique I typically use.  Meanwhile, don’t forget to take a break from your data clouds to do something a bit more whimsical like looking for shapes in the real clouds.

For example on a break the other day, Kathy and I discovered Snoopy dancing in the sky over our house …

Snoopy in the Sky

… (or maybe Bullwinkle the Moose;  we weren’t quite sure).

David-Signature1_thumb1_thumb

David Sellers
Senior Engineer – Facility Dynamics Engineering
PowerPoint-Generated-White_thumb1_th

Click here for a recent index to previous posts

This entry was posted in Excel Techniques. Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s