Hello again. Last time, I talked about how I modeled PV production for the PV systems installed on the homes I studied. Today, I will introduce the discussion on how I modeled energy consumption for the homes. To save some time, I’m simply going to use some verbiage right out of my thesis (in italics).
Keep in mind that my research analyzed not just the energy performance of the homes I studied but also their economic performance, as well as the carbon dioxide emissions ramifications of these homes. I know, based on feedback I’ve received, that many readers are more concerned with these other topics than they are about the actual details on the modeling. Therefore, though I am going to give it substantial coverage in this blog-version of my research, I will try to keep things moving so that we can get to the other topics before the next Olympics…
I used the following paragraphs from my thesis to introduce the methodology I followed for energy consumption modeling:
The author created a “Custom Energy Model” (CEM) that uses as input basic building and system specifications, geographic location, and occupant behavior, and that yields estimates for a home’s monthly and annual energy consumption. These include energy consumed for domestic hot water (Edhw), lighting (Elight), appliances and miscellaneous electrical load (MEL) from cellphone chargers and modems (Eappl), and heating, ventilating and cooling (Ehvac).
Essentially, all the energy consumed in a home can be divided into these four categories. The very first time I used this method was when I was working on UMASS Lowell’s 2011 Solar Decathlon entry…see http://www.4dhome.us, though that effort was far less sophisticated than what I ultimately used in this research.
The initial homeowner survey contained about 50 questions that provided the necessary information to “run” the CEM. The equations in the following subsections are used in the CEM and generally follow methods outlined in the ASHRAE Fundamentals Handbook (ASHRAE, 2009), the DOE Building America Benchmark (Hendron C. E., 2010), and in Goswami (Goswami, Kreith, & Kreider, 2000). Note that these equations calculate energy loads for each hour of each day of the year (8760 values). Hourly calculations were undertaken in order to analyze the specific performance of various systems. They were summed over time to yield monthly and annual loads.
The above paragraph shows that I did not reinvent the wheel to come up with my energy consumption predictions, but rather simply applied existing principles, assumptions and techniques.
There were two reasons for developing an original energy model: necessity and reasonableness. It was necessary, as the more complex energy models currently available (Energy+, Energy10, EnergyGauge, etc.) require very in-depth specifications. However, the author did not have access to that kind of data for these homes, and typically, designers and homeowners did not either. Hence, a model was required that would work with only limited information available on each house.
Early into my PhD work, I had contemplated using an “off the shelf” model (there are literally dozens out there) to provide predictions of energy consumption against which I could compare measured consumption. However, as I stated in my thesis, I soon learned that I really did not have sufficient details on the homes’ specifications which all of these programs required. I thought about asking the homeowners for more details, but decided against that, in fear that several would decide this “volunteer” involvement in research was becoming too much of a hassle and drop out. I did not want anyone to suffer “volunteer fatigue”.
It was also a reasonable approach because occupant behavior and weather variation have an overwhelming impact on a home’s energy consumption. These completely mask the relatively minor deviations caused by using more detailed specifications in a model. Hence, since the purpose of this research was to measure energy performance relative to broad goals, a simpler model was more appropriate.
I was looking for big picture results, more concerned about accuracy than precision. If a home was designed to be net-zero, did it achieve net zero? If my model predicted a near net-zero home would consume 20,000 kWh in a year, did its actual energy consumption come close? If not, what were some reasons? Please keep in mind: just like with my own PV production models (see my previous post), I was not trying to prove my models were better than existing models.
The author used the CEM to produce an estimate of annual consumption for each house, and then compared this estimate to the one used in designing the home, if available. Finally, he compared these two estimates to the house’s actual consumption over the 12-month monitoring period.
I did ask each homeowner to provide the original designer’s energy consumption prediction for their home, though less than half were able to provide it to me. In subsequent posts that include energy consumption results, you’ll see the “homeowner provided” predictions for those houses displayed along with my model’s predictions and the measured consumption.
Ok, that’s the intro on energy modeling. I’ll take on the first category of load–DHW–in my next post.