Calibrating Energy Models

Sign Up for this course and learn how to match your energy models to existing utility rates.

How to take this course:

Watch the On-Demand video
Download the Spreadsheet
Enter Data an follow directions
Optional: Customize the spreadsheet


Matching Energy Models & METER Magic Overview

  • Importance of Matching to Existing Utilities
  • Getting started
  • 7 Step Process
  • Spreadsheet success

Why Match Energy-Models

  • Crucial cross-roads in the Building Simulation
  • Scalar savings?
  • They won’t believe you!
  • “Lingua Franca”
  • $$$

After this course

  • Understand the 7 step process to get BEM to match reality
  • Use the model matching spreadsheet
  • Simplify categories of energy consumption and peak demand
  • Use associated tools
  • Determine where to put your focus

Due Diligence

  • Before we can start
  • Verify Confirmable inputs (Cooling COP, LPD, etc)
  • Determine Climate Zone
  • Determine Building Fuel Types
  • Special Circumstances (Absorption chiller)

Sevem Step Process

  1. Verify Your Peak Loads
  2. Complete Energy Model and Obtain 12 months of Utility bills (or as much as possible)
  3. Determine Energy Usage for model and Utilities
  4. Plot Patterns, follow suggestion in tutorial
  5. Run METER, view correlation statistics
  6. Implement the Suggestions (add your own suggestions for future use)
  7. Consider swing months and user experience

The Trouble with Infinite Inputs

  • Cooling Efficiency
  • NET SEER
  • COP/EER
  • Economizers
  • Fluid Coolers
  • Evaporative coolers
  • Heating Efficiency
  • NET Heating Efficiency
  • Heating Efficiency
  • ERV Units
  • Hot Gas Bypass
  • Make note of special circumstances!

What is METER

  • Editable Spreadsheet tool
Transcript:
In this course, we're going to briefly discuss why it has become very important to match
your model to existing utilities.
We're going to cover new resources to help validate your model with existing data.
Our primary focus is going to be following a seven-step process which is a systematic
approach developed by the team at energy-models.com to get your existing model to match with a
given history of utility bills.
And we are going to be using a spreadsheet as a tool to get this job done.
The spreadsheet is customizable and you will get a copy of it later but the real key is
working with a defined system.
In this course, we are going to discuss in the next 90 minutes why it has become very
important to match our existing model to existing utilities.
We're going to cover new resources to help validate your model with existing data.
Our primary focus is going to be following this seven-step process which you've likely
heard about.
It's a systematic approach developed by the team at energy-models.com to get your existing
model to match with a given history of utility bills.
And we are going to be using a spreadsheet as a tool to get this job done.
Of course, this needs to be used with some user experience.
The spreadsheet is customizable and you will get a copy of it later but the real key is
working with a system and some of the early math that we're going to take on in this course.
The first question that we want to ask is why match energy models.
I probably think like a lot of you and I know that most of you want to jump straight into
the spreadsheet.
However, we do have to address this question.
Of course, for performance contractors, the answer to this is a no-primer and for those
of you who don't know what performance contracting is, essentially performance contractors bet
on their energy models so the logic is simple on why it is important.
They've essentially bet their money on the accuracy of their model.
Beyond that, as you may know, energy modeling is becoming more commonplace and subsequently
it is subject to more and more criticism.
I personally believe that we are at a crucial point in the history of building energy modeling
as a helpful tool in design and analysis.
If we cannot create models that are close to reality, there are going to be more and
more people who dismiss energy modeling as fuzzy math or the term that I've heard recently,
voodoo.
Given that I have done energy modeling support and energy models for several years, I am
no stranger to the concept of what you may have heard of called skill or savings where
you explain that basically if you run a model with one or two basic changes, the percent
savings scales with the utility bills.
For example, if you install high efficiency cooling equipment, that saves 10 percent.
If you double your energy, you still save 10 percent if you double your utility bills.
While I wish that more people understood this concept because it's a real concept that
I've seen proven time and time again, what it comes down to is that when you explain
skill or savings to someone who isn't well acquainted with modeling, they simply will
not believe you.
The question is, I guess, why is this?
We look at the issue of money and money is something that I essentially call a lingua
franca or to those of you who aren't familiar with that term, that's just a common language.
When you as an energy modeler or building simulator present to a decision maker, they
might not understand anything you say when it comes to technicalities, but they will
understand money.
It follows that someone becomes the decision maker because they are good with money.
Thus, if the model does not translate into the language of the decision maker, the model
will not be used by the decision maker and therefore the model is useless in the decision
making process.
Now I suppose I am preaching to the choir because you would likely not be watching this
if you did not understand the importance of matching the model.
That takes us to the course material.
After this course, you will be able to follow through the seven step process that we're
going to cover.
This process relies on the aforementioned spreadsheet that we will explain and walk
through.
We are going to simplify the categories of energy consumption and peak demand and look
at associated tools.
We're also going to try to illustrate a methodical way to focus on your models.
Here we can even dream of matching an energy model to existing rate structures.
We do have to assume that the modeler has done his or her due diligence, such as verifying
confirmable inputs, understanding the climate zone and the weather file, also determining
the fuel types.
This is highly important.
Determining the fuel types and how you are going to be charged for those fuel types.
All special circumstances should be noted up front.
For instance, if you have an absorption chiller or you use what's called cogeneration, such
inputs completely change the dynamic of your utility bill and they need to be identified
up front.
Other abnormalities should also be considered up front.
The most recent example actually made energy modelers look great that did heating updates.
The most recent example is that the year 2012 was abnormally warm and thus this would need
to be considered because this is probably the data that you're using if you're going
to be matching your model.
Most modeling programs have inputs that allow you to shift the weather and if you are matching
data to 2012 utility bills, for most of the U.S. at least, you'll want to consider that
it was an abnormally warm year.
Assuming you have done your due diligence, we can get to the seven step process that
I know that everyone wants to get to.
The first thing you want to do is verify your peak loads.
This step is essential.
Then you want to complete your energy model and obtain 12 months of utility bills or as
many utility bills as possible if you cannot obtain 12 months.
From the model and the bills, you want to determine energy usage and not just the cost.
You want to focus on energy usage underlined here for a purpose.
After that, we can plot the patterns and we're going to focus on the annual pattern, July,
January, and the annual totals.
From there, we're going to run meter, which is the name we have given to the spreadsheet
that you will get a copy of.
This spreadsheet provides suggestions and you can add to these, but you can weigh and
implement these suggestions based on what you know about your design.
Once we can look at the results visually, it's much easier to consider things like the
swing months, commonly April and October, and we can combine that with common sense and
user experience.
For anyone with user experience, you know that your energy model has essentially infinite
inputs.
Because of this complexity, we want to simplify some items.
For example, cooling efficiency.
Cooling efficiency is typically thought of as an input of COP or EER, but to match your
model to reality, it's best to lump a number of topics into one category and call it your
net SEER, your net SEER, seasonal EER.
This would include items that simply would not be predictable by anyone that does not
know the design, items such as economizers, fluid coolers, evaporative coolers.
These would all fall into the category of your net cooling efficiency.
Then probably the most common is the topic of unloading curves for your cooling equipment.
I want to mention right here, because a number of you may have questions on that, is it's
almost best to just stick with the defaults from your software to start with.
If that doesn't work, you may want to use, if you do not know the unloading curves, you
may want to use simply linear curves, which are actually used in heating efficiency.
The same set of circumstances applies to heating efficiency, but other items have a more direct
impact on the net heating efficiency, such as any type of heat recovery, like an ERV
or hot gas bypass.
As we mentioned, since only someone with the design documents will have an idea on some
of these inputs, we try to focus on the net efficiency within the spreadsheet.
But we did run a number of specific cases on some of these individual topics.
So what is Meter?
Meter is an editable spreadsheet tool that is under continuous development.
You will get a copy of it as it exists, and we are also going to add additional algorithms,
which we can now create with the release of Lawrence Berkeley Labs Energy IQ, which
is funny because that was publicly released while we were developing this.
So there's additional data that went into the development since we released this course.
We named it Meter as an acronym that stands for creatively matching energy models to existing
rates.
So with that background, we can start with the most important step.
If you do not get this step right, you're going to have to rely on your luck, and your
luck is going to run out.
This is our most complicated step, so please keep that in mind while we work through it.
The other steps will not be so math intensive.
In step one, the goal is to match our models to our documented peak electric demand, and
to a lesser extent, our peak fuel demand.
In fact, if you are not paying fuel demand charges, that may not be entirely important,
but we're going to walk through the key factors when doing this for a commercial building.
Some of the details about these factors are surprising.
For instance, once you start plotting the data, you note that actually some of the peak
electric profiles are essentially climate zone independent.
Climate zones 2 through 6 have very similar peak profile percentages, but also are actually
very similar in scale.
For instance, if you looked at Fargo versus Miami, the energy consumption is going to
be very different, but they both hit a little bit over 100 degrees in the summer for their
peaks.
So climate zone 1 is not even too different than climate zone 6.
Naturally, there is a large dependency on your cooling efficiency because that's a large
source of your kilowatt drop as we look at the next pie chart.
So if we consider a typical building, we mentioned that most buildings have a similar peak load
profile, especially between climate zones 1 through 6, which is most of the United States.
Climate zone 7 is not too far off, but to be fair, it is slightly different.
So we're focused on a median building between climate zones 1 through 6.
We gathered data from existing energy models and energy IQ, which we'll look at later,
and combined it with our models to develop this simple pie chart.
We've broken it into a few simple categories, and one of the best ways to approach matching
energy models to rates is to focus on the largest piece of the pie first, and it's clear
that cooling is the largest part of the peak, especially given that the peak electric demand
usually occurs on July around 3 p.m., sometimes in August, but subsequently, that's a time
of the year when there's a lot of cooling.
And as you may know, the cooling peak is largely dependent on a number of parameters.
So this is where we start our math.
We did a breakdown of the cooling peak based on existing models, and while there is a large
amount of variance, quite a bit of variance, and we do address that in the spreadsheet,
here we have a typical breakdown of a cooling peak on an office building.
Right away, we should note that the two biggest pieces of the pie are ventilation and the
solar heat gain.
The ventilation is the outdoor air coming into the building, including infiltration,
if you do have any in cooling mode.
Many buildings these days are pressurized, and we simply have a value of zero at our
peak design, and the solar heat gain is specifically the shading coefficient or the solar heat
gain coefficient on your windows.
Another important item to note here is that the miscellaneous loads and the lights are
two pieces of this pie, and they are marked in red because they were actually part of
the first pie chart.
And so if we start to do some math, we can see that these things are having a double
impact.
If we look at the second largest piece of the debt pie, which was the fan power, sometimes
this varies, but fans can also be broken down.
The largest piece here is the solar heat gain, and again we note that miscellaneous utilities
and lights contribute to the fan peak.
We also note that people, conduction, and solar heat were also pieces of the cooling
load.
So we did what's called cross multiplying, and we did the math for you, and we came up
with this new pie chart.
This new pie chart has our net kilowatt peaks, and as we mentioned before, it's best to
target the largest pieces first.
But to get an idea of where to start, through experience we've learned what variables we
can eliminate, because there is a certain pattern that most people go through, and if
you look at this pie chart that might not make entire sense, the typical pattern of
work.
One of the first things that people look at is the R value of walls, roofs, and windows.
The reason for this is that insulation has a huge impact on residential, and so we're
all well acquainted with it.
But in a typical office building, this is from data that we gathered, it really only
contributes to about 5% of the peak load as far as conduction goes.
And thus, as long as we are close on the R values, we can actually eliminate that variable.
Then assuming that we're close again, we can eliminate the smaller piece that is motor
heat.
This may be important for the net energy consumption, but for defining the peak, it's only 3%
and the fan efficiencies and pump efficiencies will probably be in the ballpark of 80-90%
efficient anyway, or possibly higher efficiency.
Then we can eliminate electric heat in the summer, which is the smallest piece of the
pie, and it may or may not be there at all, but in general there was some involved for
cooking or some sort of gas heat was used.
This leaves us with process loads, ventilation, windows, people, and lights, but lighting power
densities and ventilations are usually well known in the design comparatively.
So these numbers typically we have a pretty good idea of the lighting power density and
we have a pretty good idea of the ventilation, especially because the ventilation should
be in any design documents and it should be measurable at an air handling unit.
So this leaves us with process loads, the window solar heat gain coefficient, and the
people.
People is the smallest remaining piece of the pie, but there can be a large variance
with people.
For instance, if we modeled a school and we forgot to implement, get to the questions
actually in a minute all of them.
So if we look at a school and we modeled the peak people in every single room, a school
of 500 would show that it would have like 5,000 people at its peak because we'd have
the gym full, the auditorium full, the classrooms full, the hallways full, the restrooms full,
etc.
And so we want to make sure that we get the peak people correct.
So a building that has a peak population of 500, maybe if there's a safety factor built
in we model it as 600, but you can have a huge variance in the load from the people.
The other thing that the people will impact in your model oftentimes is the ventilation.
And so you may inadvertently change your ventilation by changing the people.
And so when you're matching your models, it's often a good idea to lock in your ventilation
and not have it based on the number of people, which may be a factor that you need to change.
The most common error is probably in process loads and or receptacles.
And someone is asking about infiltration.
We mentioned at cooling design that infiltration would be, it's often set to zero and as far
as loads go, it's essentially on the peak cooling.
It's essentially part of the, that it can be lumped in with the ventilation total.
So as far as this pie chart goes for simplicity sake, we have that lumped in here.
Now we will get to that for our heating peak.
So speaking of infiltration, it should be noted that heating design peaks are often overestimated
in the software by a factor of two or three versus reality.
But the typical breakdown is as such with a huge variance being allowed on the infiltration.
This is the one number that we may not know and of course this would impact our cooling
mode, but there's often a separate input for cooling infiltration versus heating infiltration
as the two numbers often differ.
And so we do consider that in our spreadsheet.
The other thing that's typically not taken into account in your energy modeling is that
you may have 24 seven plug loads, some lights on 24 seven that would actually lower your
peak directly.
Now this happens in energy mode, but for determining our heating peaks, it comes down to basically
these three categories.
It is actually a little bit harder to target the heating peaks.
We'll get to some of the variables in three slides from here.
And the only way that this becomes really important is when we start to pay if we're
paying a demand charge on fuel or we're paying a demand charge on electricity in the winter
as well.
So, we've got this narrowed down to four factors to figure first.
The first of these is the receptacles and when we talk about determining these factors,
we are also talking about schedules.
We want to get this down to the operating hours.
In a sense, the operating hours has a similar impact as increasing the total of receptacles.
So for instance, if we had energy wise, if we had one watt per square foot of receptacles
at 12 hours per day, if we doubled that, energy wise, it's not too different as making it
one watt per square foot 24 seven.
But that would have a tremendous impact on the peaks.
And so if we get the peaks correct here, we can focus on the schedules later.
And that's where, that's what makes it so important right here is to get the peaks correct
because then the only main variable is the schedule.
The other factor is the windows.
And the common mistake that people make in windows is they don't take the entire window
assembly for, especially for the solar heat gain coefficient.
The assembly will have an impact on the hour value, but the impact will likely be larger
on the solar heat gain coefficient.
The other factor is the design number of people or the number of people there at peak design.
And if you have demand control ventilation running at design, the impact that it would
have on ventilation.
But as we suggested earlier, hopefully you can lock in your ventilation in terms of CFM
to make these factors independent.
And a couple of you asked, the fourth factor that was not in the pie chart is the schedule.
So the first thing to discuss on schedules is schedule slip ups.
I would say, and this number is probably high, is that you have less than 0.1% chance of
a true 24-7 by 100% schedule for any sort of people load, lighting load, plug load.
Now a 100% schedule is different when you're talking about coils and fans, that's just
making it available.
But something that's actually in operation, this is actually, we do have an example on
this specific mistake and what it would do.
Another common mistake is a design day versus a typical day, missing your weekends and holidays
or simply lumping your weekends in with your typical days as well as your holidays.
If you do the math, that will come out as two-sevenths higher if you are unoccupied
on the weekends.
You're going to have two out of seven days more for operating cooling and heating equipment.
And so the next item on the schedule is multiple seasons, the classic example here is schools.
So you would now want to miss that because in schools, the reason that schools is such
a classic example is because the peak cooling load, if they don't have summer school, would
probably diminish quite a bit and might occur in September as opposed to July.
And so it would be very important to get the schedule correct for the typical day and the
design day.
The other schedule that's commonly mistaken is the hot water schedule.
The hot water schedule is not a 24-7 schedule.
There's default schedules for this in most software.
You don't want to make that mistake.
And so along the lines of hot water and other heating equipment, typically, are redundancies.
Most buildings have backup equipment.
For instance, buildings have extra boilers, buildings have extra cooling equipment, they
have extra hot water equipment, they have extra pumps.
But one of the mistakes that people make probably the most is when determining the heating and
especially the hot water capacity as well as the hot water schedule.
So one of the mistakes that people make is they look at, let's say they have three pieces
of hot water equipment and one of them is redundant and thus you really have two that
you're looking at for the peak size and they probably have a safety factor built in.
So instead of adding all three of them up, what you should do is actually try to determine
some sort of equation based on the number of people or it would probably be better to
let a program default if it's generated than adding up all three pieces of equipment.
And then on top of that, people make that available at 100%.
We have a slide coming up that would exploit a problem there.
Another case is extra pumps.
When you have large cooling plants for chilled water, it's common to have backup pumps for
multiple pieces of equipment.
So you want to make sure that you know which pumps are redundant and you look at which
pumps cannot operate simultaneously when determining your peak pump power.
So that is step one, as promised, that is the longest step.
Now move on to the next steps.
The last thing here is that it's important to note that peak loads do not equal the sum
of equipment capacities.
Most of you know that, especially if you're a designer, so you know that you have some
additional safety built into the building.
And so you would at least want to subtract out the safety factor.
But the common mistake that people make is in hot water and I would also say in heating
capacity just because it's easier to add a safety factor in heating capacity.
So that takes us to step two.
Step two is pretty simple.
With the peaks matched and the best guess of your schedules, you can run your energy
model and at that point you'd also want to make sure you obtained or you had already
obtained 12 consecutive months of utility bills.
In step three, we need to determine the energy usage from our energy simulation and 12 months
of utility bills.
We do not want to use the cost data from the model nor from the bills.
See cost data can be very complex, but what's most important is that cost delinearizes the
data.
It also effectively changes the units on heating and cooling, especially if you have fuel versus
electricity because a BTU of fuel costs a lot less than a BTU of electricity, at least
in most places in the country.
Speaking of heating and cooling, our example will work with electric only buildings, but
it is targeted towards electric and some sort of fuel such as natural gas or even steam,
which gives us additional data to check.
If you're modeling a building that is electric only and in climate zones two and above, it's
also good practice to isolate the heating equipment both in the model and in reality,
if possible.
In fact, anything that you have submitted in reality, it will help to match your rates.
The energy modeling software, of course, should be able to break out your various end
uses automatically, and we should do that and compare end use data relative to our climate
zone, which we'll do in a minute, which we can check for blatant errors, but if you did
in fact have your building submitted, that will make this infinitely easier because you
can identify your errors very quickly.
From our model and from our data, we can plot our typical end uses.
And here you can see we have two different charts, one's from the simulated building
and the other one is from a typical building.
The one on the left, we actually put together something that would auto generate this based
on your ash rate climate zone.
We got the data from energy IQ and it's important to note that energy IQ actually names the
climate zones differently, and it doesn't cover them all, so we did have to do some
intensive math, but you could get this data, we'll cover that.
We automatically generate this for a typical commercial building based on your climate
zone, and on the right is data that you would enter manually from your simulated building.
The important part is that this is dimensionless, it's in units of percentage.
Someone asked a really great question, talking about submetering, someone asked will utility
bills have a breakdown of submetering by default, they asked, well they had surmised that that
would be unlikely, and yes that is true, but computerized systems on buildings are submetering
more and more, so you might not get it from the utility bill, but you might be able to
get it from the building's computer, and so that would be ideal, I know that we're not
going to be dealing with an ideal case, and that's why we're starting with a typical building
and comparing it to the simulated building, and we can get typical building data from
the newly released energy IQ from Lawrence Berkeley Labs, and so we actually processed
a bunch of this and we broke it into ASHRAE climate zones, as opposed to the climate zones
that they have, which are quite different and a little confusing, and then we also extrapolated
some data because they left out climate zone 8 and actually climate zone 1 in terms of
ASHRAE climate zones.
Okay, so let's go ahead and see how this tool works.
This is the first step for your end uses.
Okay, this is the spreadsheet that you'll get, it just has the first page is just a copy
of the steps.
Really what we're starting with is the spreadsheet, and it's protected, you can unprotect it if
you want, I would suggest keeping it protected just so you don't accidentally change something.
The values to enter are highlighted in a dull green or an olive, if I get my colors correct
anyway, and so one of the things that we can enter here, and there's a lookup table in
the database that simply addresses, this was a lot of data gathering, so it's from offices
around the country we focused on.
You can get other building types from Energy IQ, but we focused on office buildings because
they're 5 days a week, 50 hours a week operation kind of deal, and so we focused on that there
is a large range of variants, which we'll explain.
So if we just, let's just enter something different here.
So if we enter 3 and hit enter, this will give us the average percent of typical building
based on this climate zone.
Energy IQ also gives high and low ranges, it goes from I believe the 5 percentile to
the 95th percentile, and instead of using that data we broke it down into standard deviations,
so this is within one standard deviation up and one standard deviation below.
So there's quite a variance on the amount of heating for climate zone 3, that's approximately
a factor of actually from low to high is 3.
If you add up all the numbers in the high percentile, you will not get 100, if you add
up all the numbers in the low percentile range, you will not get 100 percent, which should
be pretty obvious.
But these numbers are just the range of a typical building so that you could see right
here we had an outlier, and our pumps were 8 percent of the total energy, which sounds
like it's pretty high compared to an average of 2.9 percent, but depending on the way that
your chiller plant works, typical high could be as high as 11.5 percent, and you could
go above that number, these are just within approximately one standard deviation.
So the data we collected provided the 5 to 95 percentile, and here's some statistics we
could break out the standard deviations.
The data that we have here, and the key right here is the chiller BTUs, we could actually
make this any unit that we wanted, we could make it kill a lot of hours, but the key is
that all of these are in the same unit so that it can break out percentage.
If we entered heating and therms and everything else and kill a lot of hours, that would not
be fair because the therms are 1 therm equals 29.5 percent.
So the data that we enter is just in what I would call olive, I'm not a color expert,
but it's dull green, everything else is locked and auto generated.
And so we enter the climate zone, just look at climate zone 5, it will give us a typical
building with the high and low relative maximums that are within standard deviation.
And then we can enter our building data, and what this will do is because we've already
focused on the peaks in step one, this will help us determine if we have an outlier in
energy right away.
And so even though at this point we're just making up numbers because I've just entered
a false climate zone, but if we looked at this and we saw that heating was at 24 percent
here and it was at 35 percent for a typical building, what we'd want to check is what's
the low percentage, what's the typical low percent within one standard deviation.
There are, you can't do the comments on here as well.
And that would be 17.3 percent would be the typical low percent of heating energy within
one standard deviation.
So with a little knowledge about your building, it wouldn't be crazy to say that you were
far beyond that, depending on other things, and your heating could be as low as 5 percent
on climate zone 5.
There's actually, there's quite a bit of data available on energy IQ.
Someone asked a really good question, I had the same question.
We do not have the heat rejection broken out.
And that is falling into the cooling category here.
We don't have it broken out because they didn't break it out in energy IQ.
And so we're looking at essentially calling it a unitary piece of cooling equipment.
And typically you are correct, the heat rejection would be broken out in the energy model.
It's not broken out in energy IQ, at least not in the typical breakout data.
Along those lines, the lighting is also not broken out into interior and exterior lighting.
And so this would all be lumped into one as well.
And if you haven't noticed, that might do why it comes out a little bit higher than what
you would be used to for your interior lighting power density.
And so this is really just to give us a gauge of if we have any outliers.
If you did have an extreme outlier, it should be pretty noticeable.
Okay, we were in climate zone 4 to start with.
And so it would be pretty noticeable.
In fact, for instance, a real common outlier, it's not every model.
What really common is hot water.
And if you had hot water modeled incorrectly, you might see it, your hot water being 40%
of your total energy use.
A couple of questions rolling in.
Okay, first question.
Is a heat pump.
In the next tab, we do have some data on heat pumps addressed.
You can break out your heating and cooling energy separately for heat pumps in most software.
I typically use Trace 700, that's within the last few years they did add that feature.
You can do it in other software as well.
For instance, if you use E-Quest, you could put it on a separate meter, something like
that.
Another question is what's the definition of an outlier, beyond the minimum and maximum?
Yes, I think that would be if you were beyond the high and low percent maximum, that would
be an outlier.
Now, that does not mean that it would not be possible.
Questions rolling in here.
It does not mean, let's say for instance, the heating number here was 5%.
That does not mean that it would not be possible, but we would want to raise our bread flag
and say, why is heating so low in this building?
For instance, maybe this building has no ventilation.
Maybe it's hyper sealed, or maybe the lights are on 24-7 and it has high plug loads 24-7
that are knocking heating bills down.
Next question, we haven't got there yet, is about worldwide data and I knew this would
come up.
The way that they did it at Energy IQ is actually within ranges of heating degree days and cooling
degree days, and so the actual data comes from the US, but you could cross-reference
that with their heating degree days and their cooling degree days.
Based on that, they're accessing 2000-some buildings, you could likely come up with a
reasonable profile.
The other question is what we're ultimately working at is what do we do if they don't
match?
The first thing that we want to do here is if they don't match, if we assume that we
have the peaks matched up correctly, is we want to consider, first of all, did we make
a mistake in the peaks?
For instance, if we had the peak right, maybe we entered too high of the cooling efficiency,
so we entered a COP of, let's say, 7, something very high, a COP of 7, which knocked down
our total cooling power and we had entered our receptacles is too high, those two things
would offset.
Now, that's thinking pretty advanced.
I don't think in most scenarios when you have an outlier here, it's that advanced.
Typically, if you have an outlier here, you made a mistake in a schedule or you made a
mistake in your unit somewhere.
Really common mistake is if you're entering your plug loads in watts and you put in 200
watts, and then you accidentally changed it to watts per square foot for that room, now
your receptacles would end up being 60, 70% of your entire simulated building.
That might slip through the radar if you are not able to look at this number.
The other thing that I want to do with this spreadsheet, I just could not because the
data is just not available, I can't start making up data, is you could take this and
edit this if we unlocked this and broke this into further categories.
For instance, we could add right next to our cooling, we could add our heat rejection,
put them side by side and then just compare the two pieces of the pie.
By making this editable, ultimately, I want to make this open source.
There's a few other things that we can do, but this data, the additional data was just
released publicly, I would say, two weeks ago or less.
I'm pretty excited about it.
Basically, this tool here, we're not addressing anything major right now.
That's what we're addressing is everything major.
We're not going into finite details.
Here, we're looking for red flakes.
This is what this would tip us off to.
If we wanted to, let's say we had other data available.
For to determine the asteroid climate zone, we have this tool here called the Climate
Zoner, just energy-models.com slash tools climate zoner.
It only works for the US.
Again, it's one of those things where the data is only defined for the US, defined well.
We could add parts of Germany and parts of Canada, but it's just not added that well.
You can look that up in asteroid standard 90.1 if you need to know the exact climate
zone outside of the US, but this is a pretty popular tool that's used to look that up quickly.
Here's energyIQ.
A couple of you have asked about energyIQ, so it's a good time to look at it.
EnergyIQ, Lawrence Berkeley Labs, they're more respectable than any of us.
They basically collected data.
There's an article on this somewhere.
I'll make a note to cite that.
Essentially, I trust Lawrence Berkeley, but basically they ran this.
They determined a number of factors, but they gathered data on something like 2,000 buildings.
If we start and we just wanted to benchmark a building, there's more data in California.
They're out of California, so they have more data from California.
We typically would want to look at US buildings, so that does answer the question on somebody
asked about international data, but we'll get to that.
What we were looking at is we have this end-use breakout, we can just look at the total.
We can break it down to electricity, fuel, district heat, etc.
This is what the numbers that we were looking at.
There's a number of other options where you can isolate features, but our starting point,
again, I call it the shotgun approach.
We want to look at the largest area we can, broad spectrum, and notice any large pieces
of the pie are out of place.
That's the proverbial low-hanging fruit.
We have US buildings, and we select, then we'll look at the total.
Then you have to define your peer group, and this is where we just focused on offices.
Actually there was a number of outliers, so we got rid of some of these.
Look at the range of sizes.
We looked primarily at these three ranges, and then they cleverly, I think because they're
out of California, they called it the vintage, similar to wine, and it's basically, when
was this building made, or when was this class of buildings made?
By checking these, we're selecting from their pool of 2,000-some buildings.
If we knew when our building was built, we could really isolate this.
I would recommend choosing one or two groups because 2,000 buildings is a lot, but it's
not that much once we start to isolate things.
For instance, we could look at all education facilities.
Then they have census regions, as well as US climate.
This is all US data, but they did break this into cooling design days and heating design
days, and that data is available for international climates right out of the back of ASHRAE Standard
90.1.
You could look at an international climate, find the proper corresponding climate that
they have data from.
If we look at the map here, they have climate zone one, and this is more conventionally
six and seven.
It was a little bit tricky to extrapolate some data, and then instead of this being climate
zone two here, it's essentially climate zone five doesn't match up perfectly, but that's
why we had to plot a lot of data and make some charts.
We were able to do that, actually it was not too difficult.
The climate zones are numbered differently than ASHRAE, which can be confusing, but
if you don't look at that map and you simply just look at the cooling design days and the
heating design days, or heating degree days, I'm thinking in terms of the E-Quest here.
The cooling degree days and the heating degree days, we could select maybe two of these ranges,
and then from here we can go to draw our chart.
This might take a minute, because it's depending on how many buildings that this is pulling
from.
What's nice is they have several options here, and they actually do have this in a scalar
as opposed to a pie chart.
I think a scalar is nice, but it's harder to identify an outlier, because if this bottom
number is off, you don't have a side-by-side view of the surface hot water.
However, this number gives you some very good data here, and you could compare this to your
data as well to see if you're within the norm.
It tells us the median value is extreme, because we selected a lot of numbers.
This is in the 5 to 95th percentile.
We could spend an awful lot of time examining this data, but we clearly got 34 buildings.
We clearly got a large sample size, and we could go back to peer groups, and we could
eliminate a climate zone and draw the chart, still at 22 buildings.
If you notice, when we get the smaller sample size, the range gets a lot more precise.
It's a significant difference right here, just by narrowing down the climate zone.
One thing I remember is do not use your back button.
It will screw the energy IQ up.
Basically, you can get this data, and since they use kilobyte use, that's what we used
here.
You could copy this, and we can make another chart, and you could create this as the documented
building.
This is why we have this data that's editable.
We did not add a second building, because I believe that most people would just want
to start out by looking at a typical building from an existing database.
Again, we're not at an exact science here yet, but this will certainly help you find
what I referred to earlier as an outlier.
I find that most of us energy modelers are visual people, and so by making things visual,
we can find errors easier, more easily.
That takes us ... I can get all the questions answered.
The spreadsheet definitely does not have all of the energy IQ data built in.
We only looked at offices based on the five climate zones, and actually we removed ... Because
there was such a range, we started to remove medical offices and the miscellaneous classified
offices.
It only has the average data built in, and we can get to how you could edit that.
That would be editable, but at that point, I would suggest just going into energy IQ
and trying to run something specific on your building.
Someone asked about the pie chart.
No, it does not have a similar pie chart.
If you took their numbers and stuck it into the spreadsheet, it would immediately create
a pie chart for you.
Based on some of the questions that I'm getting, you don't have the spreadsheet yet.
I'm making a note that I'll add that to the pie chart.
I'll add an energy IQ input specifically to the spreadsheet.
You can have it generate a pie chart for your building and for the specific inputs on energy
IQ.
Given that, I was only made aware of energy IQ a few weeks ago, that certainly does not
have all the data built in, and in fact, that's actually not where I'm ultimately going with
this.
What I think we can do, I think we can do better than just a database.
I think from the database, we can develop formulas and hoping to develop something in
collaboration with some other people and make it open source.
But from existing data, we can calculate some equations that will predict how various things
interact.
We'll get to that, and that's part of the next step.
Right now, all we have is a lookup chart.
For step four, that's the next tab in our Excel spreadsheet.
What we want to do is we just plotted the totals, and this part is actually easier plotting
wise.
We want to plot the totals January through December for gas and electric if we have them.
If the gas is set to zero, we just leave it set to zero and we plot electric.
It is built into the spreadsheet.
We want to look at the general shape, because a similar shape is good news, and this is built
into our spreadsheet to detect this.
For example, one of the most common errors that people make by far is people ask, why
is my heating at either zero or why is my heating so low?
That's something that's pretty easily identifiable, especially if your model is showing as zero.
Usually, the corresponding question with that is, well, is your electric bill high?
The answer typically is yes.
This is probably one of the most common errors that I've seen.
One of the problems here is that people forget, and they make a 24-7 plug load schedule somewhere.
The nice part about having this all in the same units, notice that it puts the fuel in
kilowatt hours as well.
We can enter it in therms, again, the spreadsheet's editable, but we'll get into how it works
in a minute, but we enter it in therms and it plots it in kilowatt hours.
If we notice, we have so many extra kilowatt hours in January that also lowers the simulated
fuel by almost the exact amount for the laws of thermodynamics.
This, of course, is if you only had one mistake, but this would be a primary mistake.
What's built in is it calculates what's called the coefficient of variation.
Basically, we take the difference between this point, the actual electric, and the simulated
electric, so the difference between this point and this point, and we take the average of
all of those differences and we calculate the mean, and we also calculate the standard
deviation.
We take the standard deviation divided by the mean.
Someone raised their hand.
I can't seem to find it.
We'll get to this in the spreadsheet in a moment, but we calculate this coefficient
of variation.
The coefficient of variation is just an indication of how well your shape matches up to the
your simulated shape matches up with your actual electric bill.
This is a very common error.
Another common error that's easy to identify, and in this case, we actually added two errors.
In this case, we added too high of domestic hot water pumps, and we added too low of domestic
hot water.
You'll notice it's much closer, but what's interesting is numbers that are based on base
utilities have no other impact.
What we found is the consistent shapes are an indication that you're doing pretty well,
and this is done with the built-in test for the coefficient of variation.
What we found is in the scenario of a missing base utility, if that was the only mistake
you made, the coefficient of variation would probably be around .03, and very small, we
would say it's at least less than .1.
In the case of plug loads, it's quite low, so in this chart, the coefficient of variation,
if we took the difference between these two points, we can see that it's larger than these
two.
The reason being is, of course, that plug loads also increase fan energy and they also
increase the cooling load, and so they're going to change, there's going to be a slight
seasonal impact, and so the coefficient of variation of around .15 is an indication that
you have mismatched lighting power densities or plug loads.
Now, unfortunately, most of the time you're going to have more problems than this.
Another example of a problem where the coefficient would not be recognizable, but a visual plot
would really help you diagnose the problem is, again, we look at and we think in terms
of the coefficient of variation, if we notice that that variation is very, it would calculate
high over 12 months, but it's similar over these three months, we do not calculate that
because this is something that's really best identified as visually.
If your shear was too low, this is your net result, this is an exaggerated result, the
shear would have been pretty low, or you're inputted EER.
During the cooling months, so if you were in climate zones 3 through 7, during the cooling
months it would be highly exaggerated on your electric bill, so your simulated would be
much higher, or if it was too high, it would be much lower.
What we look at is the peak months, and we note that based on what we just said, the
exaggeration that we see for cooling and heating equipment, this happens in July and
January, so we can focus on those two, and then everything else can get lumped into the
annual totals. We consider permutations of July, January, and annual comparisons, and
from this data we can look at simulation versus existing bills and compare the two.
Most of the possible combinations are actually unlikely combinations. For instance, if electric
was high in January and electric was high in July, it's unlikely that your total electric
bill is going to come out as low, but it is possible.
Some of these questions I'll have to get back to privately, so we're down to 20 minutes
and we're running right on schedule here.
Let's look at how this next tab of the spreadsheet works.
If we enter our actual utility bills in terms of energy, so we use kilowatt hours and firms,
versus the simulation utilities, this will automatically decipher each month whether it's
high or low within 5%. This is editable. You can change the strength if you want. Again,
you'd have to be pretty self-heavy. This is where this calculates our coefficient of
variation, and this is in red here. Try to zoom in again. This value is in red here,
and if the number is less than the 0.1 and less than 0.17, it will give you a suggestion
telling you that the shape is consistent and what you should consider checking.
The first, actually the first three, up to three suggestions, is simply based on a combination
of data from July electric and gas, January electric and gas, and then your annual electric
and gas. Then it makes a string and we match up the two strings.
I've been using this method successfully for some time. If you notice, we don't have all
of the coefficients in the existing database or all of the strings, and some of them are
just extremely unlikely. What I have been doing is running numerous models against an
existing model and then logging this data over time. What I have noted is a lot of similar
combinations come up, and so this will give you up to three suggestions. One of them is
you can add in the olive areas, you can add your own suggestions. This is something that
I'll continue to update. I'm thinking something along the lines of the Google Doc to make
this updateable and some of you would like to collaborate. I'd be thrilled. That's something
that I think that is the way to start. From this data, we can definitely develop algorithms.
But from here, all we enter is this set of numbers and this set of numbers, and it will
give us suggestions for most combinations. In this case, the suggestions are a bit redundant
because the coefficient of variation is quite low. It would be indicative of our example
when we have mismatched base utilities. This would be, of course, an ideal scenario because
it's something that we can switch very quickly. Do you want it to enter this in different
units? At the bottom here, we do have, this is where we convert the data into all, into
kilowatt hours. We want to keep everything in the same units. Right here, we want to
convert the therms into kilowatt hours by multiplying by 29.3. If you wanted to change
this number, you just would have to remember to change this conversion factor. Again, that's
pretty complicated. It might just be easier to convert your default number into therms
and enter that here. Again, from your utility bill, you should get a dollar amount, but
you should also get a energy consumption amount. That is how the spreadsheet works. Now we
get to step six and seven, and I'll open it up for questions. Step six, when you change
your model, do a save as and change one variable at a time. There's no such thing as intuitive
in energy modeling. Subsequently, you change two variables. You won't know which one changed
what. You change one variable at a time, and if you iterate, you should be able to chip
away at patterns. If you plot this and you're completely all over the place, which is actually
not super likely, you probably want to, the first thing is consider the letter file. The
second thing is if you have your peak loads matched up, it's pretty unlikely that it would
be one of your main inputs that is making the plot scatter. The last thing to consider
is not built into the spreadsheet, but rather it would be something that you would see visually.
Two people actually asked about infiltration. Let's say we had infiltration, and this is
built into the spreadsheet, but let's say that our infiltration was too high. That would
make our January gas bill or electric, depending on what type of heat we had, would be too high.
If we had cooling, it would also be higher because of the infiltration in July. What
could happen in the swing months, especially if we had no airside economizer, the swing
months might come out perfect. The swing months, those are months like April, October, sometimes
September, November are swing months depending on the climate zones, but looking at March,
April, and May, I like to say April and October. Based on user experience, there's a number
of things that will make the swing months match up when other things do not. Common scenarios
when your infiltration is off, you just think about when you open your windows on your house.
That's typically a swing month, assuming that you open it, and you can only open it for
a set period of time, such as like a day, like it's done in the energy model. The swing
months are essentially the months that are comfortable naturally. Infiltration coming
in would not require any conditioning. In fact, because of those, we typically do have
an airside economizer built in. But many older buildings do not, and infiltration would
thus be a factor.
Once we have our energy matched, assuming we've iterated through those steps, once we
have our energy matched, the first thing that I would recommend is just implementing average
rates in the model. So if you have any idea on what the average rates are, implement them
in the model. And if that does not work, do not spend time, or I should say do not waste
time fudging the cost data.
The reason here is that most energy modeling software, you can implement the exact rate
structure in the model. And if the average rates don't work out, it's probably because
you have a demand charge or a ratchet, or you have some sort of stepped rates.
Where you get charged more or less, depending on the steps that you hit. And so the rate
structure, the exact rate structure is actually, it can be quite complicated. And I notice
that the complexity of it is almost directly corresponding with the population density.
And so for instance, New York will have a more complicated rate structure typically
than say San Francisco, but there's also power grid differences there as well that
make up for the difference. But New York is really complex, Chicago is complex, Dallas
is complex, LA is complex. And typically they have demand rates, and you would want to implement
the exact rate structure in the model.
Got some good recommendations, and we'll have a log of this. And so some of you that don't
get to. One of the questions is about what would be, to rephrase your question, is what
would be a reasonable goal to hit for consumption and demand on a monthly basis. And the suggested
number is 10%. And in the model, we have it at 5% would be the same, but that's just
to find subtle differences. I think that 10% within a month, and then on the year, you
should be less than 10%. So you'll have some minor variances, and that's based on actual
weather versus typical mean year weather. Most of you will not be able to obtain the actual
weather file, and that's often very expensive. And it's often actually not even accurate,
depending on when you get that precise. It can be different across town by a few degrees.
I did talk to someone who had spent much of his career performance contracting, and this
primarily for schools. And someone worked on this for, I would say, 30 to 40 years of
energy modeling performance contracting. And he said that the best that he could get was
5% with the school, and that he found out that this typical margin of error was within
the lights. And so I talked to him several years back, and I'd be interested to see what
happened now that more and more schools are implementing occupancy sensors. Because is
it the occupant that's making it inaccurate? And I would recommend not blaming the occupants.
There's articles published that we energy modelers always cop out and blame the occupants.
And so even though that's the hardest part to predict because it's randomness, it's something
that we think called chaos theory. And so essentially that these minor mistakes cancel
each other out. And if we're within 1% here and within 1% there, plus or minus, and we
do our due diligence, we should be able to match things up within 10% on a month-to-month
basis. And that should offset annually getting us within 10%.