In the early hours of the morning, long before tenants start arriving, building engineers are already hard at work waking up their buildings for the day ahead.
Starting a building at the right time is one of the most important factors in improving its energy efficiency. Start too early, and you’ll waste energy heating, cooling, and lighting an empty building. But start too late, and you’ll risk frustrating your tenants — or incurring a peak load charge from the spike caused by starting all of your systems up at once.
That’s why we were so excited when we were asked to improve the system that Aquicore uses to recommend the optimal start time to our customers. It’s an interesting challenge that involves a lot of moving parts. The article will discuss the process and models Aquicore employs to save its customers the money and headaches that come with guessing at this crucial metric.
Why is this important?
In the case shown above, we see that a startup time of 3:15 AM led to a convergence (the point where the building reaches its preset indoor temperature and heating levels off) of 5:30 AM. That’s a full two-and-a-half hours before lease hours begin at 8:00 AM. Had the building been started at 5:15 AM instead of 3:15 AM, the resulting energy bill for this day would have been about 265 kWh lower.
It’s important to note that picking the proper start time each day is extremely complex. Building engineers know their buildings better than anyone, but the weather is hard to predict, and tenant comfort is typically the most important concern. As a result, there’s an incentive to err on the side of caution and start up the building early, as the engineer did in the example above.
By empowering engineers with an optimal start time tool, it equips them with the knowledge they need to start their buildings later, with confidence they will reach a comfortable temperature by the time tenants start filing through the door.
To get a sense of what this might look like, we took the actual start times for a real building and compared them to the optimal start times suggested by our model. In the figure above, the red dots represent actual start times, and the blue dots the optimal start times generated by our algorithm. As you can see, the biggest saving opportunities were concentrated in the winter and summer months, but there is a clear pattern of engineers playing it safe throughout the year.
In total, the estimated savings from using Aquicore to pick start times came to just over $8,800. In a building with a five percent cap rate, that’s the equivalent of adding $176,000 to the asset’s value, thanks to the increase in net operating income.
Let’s dig into the details of how our model generate these start times. The project required using anonymized data from hundreds of buildings in the Aquicore platform to generate two different models. We’ll start with the simplest first.
Finding Internal Temperature
How do you calculate what the temperature inside a building will be before you start it up? We determined that there are two basic factors that affect the temperature of a building at startup: the ambient temperature and the building’s level of insulation. (This last one is slightly more complicated because the size and shape of the building also has an effect on temperature, but the methodology that we used accounts for this naturally.)
Outside temperature is pulled from local weather predictions. Accuracy tends to be high for this type of data, because we pull the temperature data less than a day in advance of running the calculations.
To account for the impact of a building’s insulation (the other critical part of the equation) we had to go a bit deeper. We used anonymized data from a large set of buildings to group buildings into the four categories you see in the figure above. While very few buildings will fit this model exactly, the vast majority of buildings will fit one of them fairly well.
In the future, we plan to track temperature inside and outside our customers’ buildings, which will allow us to further enhance the fidelity of the data.
Between the insulation figure and the outside temperature at startup, we’re able to get an approximation of the temperature inside the building at startup. That gets us halfway there. Next, we have to calculate the time it will take for a building to get from its startup temperature to its target temperature. This requires a few more questions.
When calculating the time it takes for a building to reach its target temperature, one of the most important things to note is the thermostat’s global set point. Between that and the external temperature, we’ll be able to determine the amount of heating or cooling the building requires on any given day. We also ask how quickly a building’s HVAC system can heat or cool the building’s air when set to its maximum level.
Next, we want to know whether a building’s HVAC system pumps out a variable air volume (VAV) or a constant air volume (CAV). Generally, VAV is considered to be more energy efficient because it does a better job of hitting the temperature target without exceeding it. However, VAV systems accomplish this by slowly tapering off, meaning VAV buildings often take slightly longer to reach their target temperatures.
The last thing we ask is whether the building is air-cooled or water-cooled. Again, water-cooled buildings are usually more efficient, but air-cooled buildings are able to hit their target temperatures faster.
All of these data points give us a clear understanding of the mechanics of the building. The last step is just housekeeping. We need to know when your lease hours start, how much of a buffer you want to build in, and what your startup process looks like if you stagger your startup to avoid peak load charges.
Making an Impact
Put all of this together and you can calculate with a reasonably high degree of certainty how long it will take to bring your building to the right temperature (up or down) on any given day. Right now, we try to make predictions on 15-minute intervals. As more sensors find their way into buildings and we continue to hone our models, the level of accuracy will continue to increase.
We expect to see even more insights going forward. Our algorithm already uses the tariff schedules stored in the Aquicore platform to shift startup times earlier if doing so would save money based on demand times. It also requests input from users each day so we can better train the algorithm going forward.
Commercial buildings are often associated with inefficient energy practices. In fact, it’s been estimated that commercial buildings account for 20 percent of all energy used in the U.S. with as much as 30 percent of that energy being wasted. Now more than ever before it’s imperative for building operators, managers and owners to invest in the right tools to decrease operational costs, increase overall net operating income (NOI) and meet pre-selected energy targets if they want to remain competitive in today’s market.