There is a lot of hype right now about the “Bloom Box.” Google installed one, Adobe installed one, and Bloom Energy has an active sales force hitting up the Fortune 500. So what is a Bloom Box and what does it mean for the future of the energy world?
100+
Lorem ipsum dolor sit amet, consectetur adipiscing elit.
2
Lorem ipsum dolor sit amet, consectetur adipiscing elit.
50%
Lorem ipsum dolor sit amet, consectetur adipiscing elit.
What’s a Rich Text element?
The rich text element allows you to create and format headings, paragraphs, blockquotes, images, and video all in one place instead of having to add and format them individually. Just double-click and easily create content.
Static and dynamic content editing
A rich text element can be used with static or dynamic content. For static content, just drop it into any page and begin editing. For dynamic content, add a rich text field to any collection and then connect a rich text element to that field in the settings panel. Voila!
How to customize formatting for each rich text
Headings, paragraphs, blockquotes, figures, images, and figure captions can all be styled after a class is added to the rich text element using the "When inside of" nested selector system.
There is a lot of hype right now about the “Bloom Box.” Google installed one, Adobe installed one, and Bloom Energy has an active sales force hitting up the Fortune 500. So what is a Bloom Box and what does it mean for the future of the energy world?
100+
Lorem ipsum dolor sit amet, consectetur adipiscing elit.
2
Lorem ipsum dolor sit amet, consectetur adipiscing elit.
50%
Lorem ipsum dolor sit amet, consectetur adipiscing elit.
We saw last time that commercial and industrial buildings are typically assessed demand charges based on the peak power demand they use. This power demand is measured in kW and tracked in 15-minute intervals. The interval with the greatest amount of energy usage in a month is used as the basis for the demand charge.
In theory, when we add a solar array to a building we should reduce these demand charges since most buildings use the most energy in the afternoons when it is hot outside and the sun is shining. Incidentally, when the sun is shining is also when a solar array generates the greatest amount of electricity.
Solar companies have a range of methods for calculating energy savings. Some are formal analyses, some are simple rules of thumb. One rule of thumb from our own experience is indicative: for a typical commercial building, with a solar array that meets 50% or more of the building’s peak energy needs, demand charge savings will be no more than 7%.
Seem small? What follows is a discussion of some of the factors that are critical to understanding the impact of the energy generated by a solar array on demand charges.
The Problem of Variability
Demand charge savings and solar are notoriously difficult to model because of the intersection of two unpredictable and partly independent factors: solar output and facility energy demand. The sun shines when it is hot, but may hide behind clouds at any moment for 15 minutes or more during which time facility energy needs will not change much. Buildings also need more energy when it is hot out, but sometimes require lots of energy at night as well. Energy demand differs by building: a university will have a different energy usage profile from a refrigerated warehouse which will have a different usage profile from an office building.
The demand charges will also differ based on the details of the utility’s rate structure. Some utilities have a single demand charge. Others have one demand charge for weekday peak energy demand between 11 am – 6 pm, another demand charge for weekday peak usage between 6 pm – 9 pm, and a third for peak usage on the weekend and between 9 pm and 11 am. There are many other variations. Peak demand rates can vary from winter to summer (e.g. $6/kW in the winter months and $15/kW in the summer months). Most rates are based on 15-minute demand intervals, but some use 10 minutes and others 30.
Resolving these issues requires the careful of collection of data about utility rate structures and facility energy demand.
The Problem of Average Solar Production
Using the National Renewable Energy Laboratory’s PV Watts tool, anyone can obtain hourly solar production data from key locations across the United States. PV Watts uses 40 years of historical data to model expected solar output.
What’s great about this hourly data for solar production is that if you also have hourly data for the energy demand of your building, you can deal with issues like the one we encountered at the end of the last post. As you may recall, we imagined having 500 kW building demand and 250 kW solar output at 1 pm and 475 kW demand and 0 kW solar output at 7 pm. It’s easy to use the hourly solar and demand data to calculate that in such a case, your demand charge would only be reduced from the previous maximum (500, which fell to 250 because of the solar) to the new highest amount (475 which does not fall at all because there is no solar). Calculating the corresponding changes in demand charges is just a matter of multiplying the kW savings (25) by the $/kW rate.
Problems emerge, however, with this hourly model when we model demand charges using the average solar hourly output from PV Watts.
For example, if the amount of solar energy produced by a solar array between 9 am and 10 am on:
January 1st was: 13 kWh = 13 kW (because it’s over one hour exactly)
January 2nd was: 11 kWh = 11 kW
…
January 31st was: 10 kWh = 10 kW
And you took the average of all the solar energy produced between 9 am and 10 am in the month of January to obtain say, 14 kW of power, that would give you the average solar power production at that hour. If you did it for every hour of every month, you would get a chart like this one (color added for clarity):
And if you knew the average facility demand between 9 am and 10 am across the month of January were 30 kW, then the average new facility demand would be:
30 kW – 14 kW = 16 kW
Then you could subtract the hourly solar production data from hourly building demand data for every hour of the year and multiply by the demand charge rates. The difference between that cost and the demand charges when there was no solar array is the demand charge savings.
While this is a good model, it has the following key shortcoming: the peak demand charges could very well occur when solar production is at its lowest not its average.
Energy consumption scales with temperature, which is correlated with solar irradiance, but a facility’s peak demand could overlap with minimal solar production. A temporary change in weather (cloud cover, brief storm) can cause a substantial drop in solar production with minimal change facility demand in a 15 minute period. Any drop in solar production for 15 minutes could throw off the demand savings for the entire month since demand savings are usually set by the highest facility demand during the entire month.
There is, fortunately, an easy way to make the hourly model of demand charge savings more accurate: select the minimum solar production for the month from the PV Watts data, rather than the average solar production. For example, if the amount of solar energy produced by a solar array between 9 am and 10 am on:
January 1st was: 13 kWh
January 2nd was: 11 kWh
….
January 31st was: 10 kWh
And you searched the month and took the minimum of the solar energy produced between 9 am and 10 am to obtain say, 6 kW of power, that would give you the minimum solar power production at that hour. And if you knew the average facility demand between 9 am and 10 am across the month of January were 30 kW, then the new net facility demand for that hour factoring in the solar array would be:
30 kW – 6 kW = 24 kW
You could then make a new chart of minimum hourly solar production and subtract off every hour from the facility demand as we did before. This would create a more conservative estimate of hourly solar production. You would be using the PV Watts historical solar irradiance (and hence weather-impacted) data to account for changes in weather. That would be preferable to using average hourly data.
The Problem with Hourly Data
Which brings us to an additional problem: we’re looking at hourly intervals but the utility invoices the building based on 15-minute demand intervals. Let’s say the hourly data show that the building’s peak hourly demand for the month of March happened at 3 pm and was 40 kW. Then we look at the 15-minute data and we see that between 3:15 and 3:30 the building really used 70 kW, it just happened to use 30 kW between 3 and 3:15, and 3:30 – 4 pm, so that the total averaged to 40 kW over the hour. That’s a real problem, and a likely one. The utility is going to bill the building based on having used 70 kW, but looking at the hourly data you would think you’re only being billed based on 40 kW of demand.
Hourly demand forecasts are a good proxy for how demand charges are calculated, but because they were calculated on the hour and not every 15 minutes, they are likely to be below actual peak demand given the 15-minute fluctuations within the data.
One thing you can do is compare the maximum 15-minute interval in every hour-long interval to the average hourly demand, and then analyze the differences across every hour of every month. Variation may exceed 100% within an hour (e.g. between 4-5 pm the facility needed an average of 100 kW, but at 4:45-5:00 it needed 200 kW). You could end up with a chart showing the percent greater that the max demand in the 15 minute intervals is compared to the average hourly demand:
You may note all the red around the hours of noon. That’s because this particular data is adapted from a facility that had a solar array. The solar production fluctuated from cloud cover and other weather-related changes causing substantial swings in the net energy demand of the facility within a given hour
A reasonable correction for this hourly/15-minute problem is to take the maximum of the 15 minute demands in every increment and update your hourly and monthly chart accordingly. Alternatively, one could come up with an hourly multiplication factor of say, 1.4, in which you assume that there will be 40% greater demand within a given hour than the hourly average. Then multiply all the hourly demand numbers by 40%. Both these solutions allow for more accurate modeling of the actual, 15-minute demand charges.
Final Improvements
It is certainly possible to conduct an even more comprehensive analysis of solar and demand charges than what we discussed above. One further step is to run a Monte Carlo simulation. You vary solar production based on weather data and vary facility energy demand based on historical energy demand data as well as weather data. Running this simulation a thousand or a hundred thousand times generates a range of likely savings – because of the range of overlaps of solar production and facility energy demand. Taking averages and ranges would give you an even better model of likely demand charge savings from solar.
But a model it remains. Because unless you can predict when the sun will shine, as well as when a building will use the most energy, in 15-minute intervals, exactly, for the next twenty years, without fail, then you cannot have certainty in demand charge savings.
We can do some very powerful modeling that approximates reality pretty closely. And now you know some of the complications that go into it.
