Whole Building Performance: Leveraging Utility Bills & Smart-Meters. Phillip N Price Scientist Lawrence Berkeley National Laboratory

Transcription

1 Whole Building Performance: Leveraging Utility Bills & Smart-Meters Phillip N Price Scientist Lawrence Berkeley National Laboratory

2 Colleagues Johanna Mathieu Sila Kiliccote Mary Ann Piette Research funders: Research funders: Department of Energy, California Energy Commission

3 AIA Quality Assurance Portland Energy Conservation, Inc is a registered provider with The American Institute of Architects Continuing Education Systems. Credit earned on completion of this program will be reported to CES Records for AIA members. Certificates of Completion for non-aia members are available on request. This program is registered with the AIA/CES for continuing professional education. As such, it does not include content that may be deemed or construed to be an approval or endorsement by the AIA of any material of construction or any method or manner of handling, using, distributing, or dealing in any material or product. Questions related to specific materials, methods, and services will be addressed at the conclusion of this presentation.

4 Session Objectives Learning Objectives 1. Gain perspective on how buildings react to the climate conditions, and how they don't. 2. Learn specific analysis methodologies for metered data that can be applied in field work to identify anomalous behavior. 3. Come away with a broader understanding of the strengths and limitations of remote analysis of buildings using metered data. 4. Learn how to use load data to measure the success (or not) of recommissioning or retrofits.

5 Building load shapes vary

6 You can see a lot just by watching Building stops shutting down at night around July 12. (But the building still hits the previous minimum load for 15 minutes each night!)

7 Graphical methods matter Same data as previous slide can t see a thing!

8 Comparing years to each other One month of interval data is about as much as you can usefully display at a time. These plots shows data from June 2007, 08, 09 Base load is lower in 2008 than Peak load looks lower too but maybe that s just because of weather. Dashed lines help compare years.

9 Comparing years to each other Furniture store, first two full weeks in June 2007, 08, 09.

10 Load shape features and parameters

11 Peak and Base Load can be informative We can display a whole year of peak and base load data; the interval data would be a forest of vertical lines.

12 Weather adjustment So far we have only looked at a single data stream: whole-building electric load versus time. But we also need to be able to adjust for weather. For example, the furniture store had lower peak load in June 2009 than in June Is that because of changes in the building, or just because June 2009 had milder weather? We need a statistical model. Simple model: Peak Load = (Day of week offset) + B*(Outdoor Temp 60 F) More complicated model fits the entire day, not just peak, and adjusts for time of day and day of week. Better model adjusts for time of week and allows non-linear temperature dependence.

13 Simple model: predict peak load from peak temp These are predictions, not forecasts: we know the peak temperature each day. Thick = model Thin = prediction

14 Weather adjustment Simple model: Peak Load = A + B*(Outdoor Temp 60 F) More complicated model first the entire day, not just peak, and adjusts for time of day and day of week. Better model adjusts for time of week and allows non-linear temperature dependence.

15 Better model for an office building Model is based on several months of data; just a few days shown

16 Nonlinear temperature dependence Expected behavior Estimated from a real building It is not possible to simply plot load versus temperature to determine this relationship! Temperature and load both change during the day (and peak around the same time), so time-of-day variation gets superimposed on temperature dependence.

17 Distinction between load and Temp-dependent load Above: load vs temp for occupied (black) and unoccupied (gray) periods. Right: temperature-dependent load vs temperature.

18 More real-world temperature dependences Temperature-dependent load doesn t always show clipping, even on hot days, perhaps because most cooling systems are at least adequately sized for even the hottest periods.

19 Model allows weather adjustment Bottom left plot shows that load for these three days in June was lower in 2009 than in Bottom right plot shows how the building would have performed in 2009 if the weather had been the same as in 2006 (middle curve). Model is based on several months of data; just a few days shown

20 Model allows quantifying effectiveness of retrofit or control change This building participated in Demand Response events in the periods shown with the dashed lines. Not effective 8/27; effective for the second half of the event on 8/28; effective for entire event on 8/29. The model is based on several months of data; just a few days shown

21 Plot interval data. Conclusions It can be helpful to plot derived parameters instead of raw data. Important to make good plots: Choose appropriate duration of data Overlay guide curves, previous year, or other. Statistical models can help a lot, especially comparing periods when weather differed. Models aren t trivial to fit. EIS companies are adding these capabilities; some already have them.

22 Do it yourself? These methods are not very complicated but they are probably beyond what most building energy managers are willing to do. To be widely used they need to be incorporated into EIS. Several EIS companies are working on automating these approaches. Some EIS say they already provide similar functionality; some of them may be right. Our paper (Price et al. in the 2011 NCBC proceedings) gives more detailed description of methods, and cites relevant papers including: Mathieu, J.L., P.N. Price, S. Kiliccote, and M.A. Piette, Quantifying Changes in Building Electricity Use, With Application to Demand Response, IEEE Transactions on Smart Grid, (in press). Price, P Methods for Quantifying Electric Load Shape and its Variability, Lawrence Berkeley National Laboratory, Berkeley, CA, LBNL- 3713E.

23 Co-workers include: Johanna Mathieu, Sila Kiliccote, Mary Ann Piette Contact: Phil Price at

24 Additional Slides

25 Sometimes restricting the dataset is helpful Even when looking only at peak load, complexity can hide some patterns or changes. Example: weekends differ from weekdays. You tend to focus on the weekly pattern, so it s hard to see what else is going on. Solution: Just look at weekdays.