Accurate Predictive Energy Modelling?

If I could predict the weather to the day and hour five years from now, or predict how many times a conference room would be full, at what days and what times, five years from now, or predict that the electrical subcontractor would wire the motor on one of nineteen air handlers backwards, or when the building operator would decide to leave all the lights on all night long (and then forget he had done that because he isn't ever in the building at night), I would no longer be in the energy modeling business.  I would be predicting the numbers in this week's PowerBall MegaMillions and be living on my own private island next week.   Modeling software has nothing to do with it.  It's all about the assumptions you have to make about what happens in the future.

See, it's times like these, when I thank my lucky stars that we invited Sustainable MEP Leaders to join our ranks.

Amen, Kim. It's not the software it's the assumptions. Short, relevant, recent story: I completed a model for a brand new office / lab building in California (~2018). During construction, we held the spec (envelope values, equipment efficiencies, lighting power densities), the building got commissioned, meters were set up to track everything you could think of. A few months after the building was occupied, COVID happened. The research areas went to 24/7 operation (COVID research and vaccine development), office AHUs were driven 100% OA, cleaning crews were in that building every few hours (even overnight!) which triggered occupancy sensors for lighting. To top all that, we had the hottest year in recent history recorded for 2020. When the model results were compared to the operations for 2020, our modeling team was asking about inaccuracies, inefficiencies, where we made mistakes.  I spent time 'calibrating' the model and we even purchased the weather file for that city for 2020 specifically. Wouldn't you know, updating those assumptions and weather file put us right back in line with the same model file that we used 3 years ago. It is a constant process to tracking energy, ensuring the building operates as designed and then, if using the model for current/future 'checks', it's engaging the same team to calibrate it and find where the assumptions changed.

To Kim's point (btw, he is obviously my paternal brother), it can be helpful to agree with all involved to use the concsus assumptions, and energy modeling to provide an "energy budget", that the team will need to abide.  Can you imagine on financial side the Proforma Statement is NOT followed by the development team?   Maybe similar?

Plug and process loads don’t get the attention they deserve. Either generic assumptions by space type are made or simulation defaults are left untouched. When appliances and equipment are properly accounted for, we notice that they never reach the 25% threshold that was previously expected by LEED energy models, at least for higher education and office buildings. Plug load energy use is likely similarly underpredicted in other building types. All things considered, predictive models are good for the set of assumptions they are based on. In other words, it is almost impossible for designers to foresee and account for all the possible operational variations. The work of Georgia Tech’s late Prof. Augenbroe (and others) has been insightful in this regard. They applied stress tests and risk analysis to design assumptions to learn how likely a building is to perform at net-zero in its first year and in its 25th year. The end goal is to account for some of this uncertainty in design decision making and modeling. Here’s an example paper: 046.pdf (researchgate.net) Ramana Koti Senior Associate, BEMP, LEED Fellow Responsive Design LordAeckSargent.com From:

I noticed that the hyperlink to the paper did not go through. Here’s the title of the paper to look up on researchgate.net: Modeling and Simulation of a Campus Living Building: A Case Study in Uncertainty Analysis and Stress Testing Ramana Koti Senior Associate, BEMP, LEED Fellow Responsive Design LordAeckSargent.com F

Occupancy is another overlooked input. It is very common to: (1) Assume default ASHRAE occupant densities (used by the MEP to size the systems), as opposed to actual anticipated occupancies. (2) Double-count occupants in a building (eg a school building where at noon the cafeteria –meant to fit all students- is at 90% occupancy, the classrooms are at 20%, to offices at 30% and the auditoria at 10%... ) (3) (related to the above) Assume that “no occupancy” in a space equals 5%-10% diversity. As modelers we rarely dare say “under regular occupancy there will be zero people in this room”. As a result, buildings end up being modeled as 10% occupied at 2am, which for a 1,000 peak occ building is equal to a full 100 people in it. While occupancy for some project typologies is not predictable at all (eg developer offices / labs, for others it is quite forecastable (eg elementary schools). I’ve seen models where the peak occupant count for a building gets to be four times the anticipated building count at any given time. Occupancy variations may have a large or a small impact on overall predicted energy use, though in my experience we’re always overestimating. For a modeler it shouldn’t be hard at all to provide total building occ counts on an hourly basis for a typical day. I highly recommend you ask for this information from your consultants, just to make sure you’re on the right ballpark. Ale. Alejandra Menchaca, PhD, LEED AP, WELL AP Vice President Direct 617.250.4177 | Mobile 617.999.0274 Looking to provide adequate ventilation and a comfortable indoors? Healthy Indoors Airflow Modeling. From: Ramana Koti