Technical Paper Session 11 - Strategies to Improve Building Models and Operation

advertisement
2016 Winter Conference
Joshua New, Ph.D.
Oak Ridge National laboratory
newjr@ornl.gov
865-241-8783
Technical Paper Session 11 Strategies to Improve Building
Models and Operation
Paper #5 - Suitability of ASHRAE
Guideline 14 Metrics for Calibration
Orlando, Florida
Learning Objectives
Objective 1 - Describe the current state of testing
for building model calibration.
Objective 2 - Explain the major components of
the Trinity testing framework.
ASHRAE is a Registered Provider with The American Institute of Architects Continuing
Education Systems. Credit earned on completion of this program will be reported to ASHRAE
Records for AIA members. Certificates of Completion for non-AIA members are available on
request.
This program is registered with the AIA/ASHRAE 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.
Acknowledgments
• Aaron Garrett – JSU
• Amir Roth – DOE BTO
• Zheng O’Neill – UA
Outline/Agenda
• Publication edits
• Context
• Trinity Test
– Limitations
– Web service implementation
• Purpose of this paper
• Results of a large Calibration study
Publication Edits
• Trinity test – implementation of BESTEST-EX method
to evaluate calibration, whether manual or automatic
• ASHRAE Guideline 14 definition (b) of “calibration”
– “process of reducing the uncertainty of a model by
comparing the predicted output of the model under a
specific set of conditions to the actual measured data for
the same set of conditions.”
• We use the word “calibration” herein whether it is to
actual measured data or to simulation output (as a
surrogate to measured data)
Autotune
Automatic calibration of models to data
E+ Input
Model
.
.
.
6
Autotune Performance
DOE Office of Science
DOE-EERE: BTO
Industry and building owners
Results
Monthly
utility data
Hourly
utility data
CVR
NMBE
CVR
NMBE
ASHRAE
G14
Requires
Autotune
Results
15%
5%
30%
10%
1.20%
0.35%
3.65%
0.35%
Results of 20,000+ Autotune calibrations
(15 types, 47-282 tuned inputs each)
Features
• Calibrate any model to data
• Calibrates to the data you have
(monthly utility bills to submetering)
• Runs on a laptop and in the cloud
High Performance Computing
• Different calibration algorithms
• Machine learning – big data mining
• Large-scale calibration tests
• 30+ Publications:
http://bit.ly/autotune_papers
• Open source (GitHub):
Other error metrics
Residential
home
Tuned input
avg. error
Within
30¢/day (actual
Hourly – 8%
Monthly – 15%
use $4.97/day)
3 bldgs, 8-79 inputs
http://bit.ly/autotune_code
Leveraging HPC resources to calibrate models for optimized building efficiency decisions
Trinity Test – what is it?
• “True” model – defined by the user for a specific test case; the
answer key used to quantify accuracy of the calibrated model
• Calibration (edits) – simulation output as a surrogate for
measured data
Advantages
• Reproducibility!
– No specific, unique buildings of interest
– No faulty or unshared data used for calibration
– No variation in definitions or metrics
– No sole focus on simulation output
• Proliferation in calibration literature
– Necessarily unique
– Largely irreplicable
– Essentially incomparable
Limitations
• Cleanroom approach which has removed all realworld “noise” from the calibration process
– No: sensor drift, missing data, utility data measured at
different times, unaccounted for occupancy/behavior
changes, model/form uncertainty (but can allow study)
• Allows use of any weather file (TMY)
– For real-world application, you need AMY data
• No mapping of simulation output to measured data
– Temperature gradients: what point is “Temp. of N wall?”
• No sensor placement/material issues
• Test results equally weight all inputs/outputs, even
though some matter more than others
Results
IDF + CSV  XML
Thickness of metal siding?
Calibrator: Between 0 and 0.5
and less than 1-B
Oracle: 0.055
Website
http://bit.ly/trinity_test
XML
EPW
CSV
Website/service
Results
Metric
Input error average
Input error maximum
Input error minimum
Input error variance
CV(RMSE)
CH4:Facility [kg](Monthly)
CO2:Facility [kg](Monthly)
CO:Facility [kg](Monthly)
Carbon Equivalent:Facility [kg](Monthly)
Cooling:Electricity [J](Hourly)
Electricity:Facility [J](Hourly)
…
NMBE
CV(RMSE)<30%
CH4:Facility [kg](Monthly)
NMBE<10%CO2:Facility [kg](Monthly)
CO:Facility [kg](Monthly)
Exceeds G14!!!
Carbon Equivalent:Facility [kg](Monthly)
Cooling:Electricity [J](Hourly)
Electricity:Facility [J](Hourly)
Electricity:Facility [J](Monthly)
143+ outputs
Value
24.38
66.12
0.09
228.53
9.95
15.42
20.40
14.42
1577.96
10.48
-9.57
-14.78
-19.52
-13.83
592.77
-9.52
-9.52
Purpose of this Study
• Are CV(RMSE) and NMBE the best metrics to
use for calibration?
•
•
•
•
What about no-CV:
What about Mean Absolute Percent Error?
What about (non-normalized) Mean Bias Error?
What about Percent Absolute Error?
…maybe calibration using another metric would allow
a calibration algorithm to reach lower input-side error
(i.e. recover the “real” model of the building)
20,000 Building Calibration Study
#Inputs
#Groups
#Inputs
#Groups
Restaurant
Hospital
49
49
227
139
Large Hotel Large Office
110
67
85
43
Secondary
Stand-alone
Small Hotel Small Office
School
Retail
231
282
72
59
122
131
58
55
Medium
Office
81
36
Strip Mall
113
85
Midrise
Apartment
155
78
Super
Market
78
72
Primary
School
166
109
Quick
Service
54
54
Warehouse
TOTAL
47
44
1809
1142
Results
For the Strip Mall:
If you use MAPE to
minimize error to
“measured” data, then
you’ll have the closest
building match in terms of
CV(RMSE)
Is there anything better?
Output Variable
5
7
InteriorEquipment:Electricity [J](Hourly)
InteriorLights:Electricity [J](Hourly)
Number of Buildings
10
9
• Correlation to other properties showed that in most
buildings (out of 15), calibrating to electrical usage of
interior equipment and lights yielded better
calibration results than any other building properties
• ASHRAE G14 extension to allow a tier-2 calibration
using (increasingly feasible) submetering
requirements would allow more accurate and useful
models from calibration
Conclusions
• Trinity test allows replicable calibration studies
and quantifies calibration performance
• An unsupported website and web service:
http://bit.ly/trinity_test
• A calibration study was conducted
– 20,000 calibrations, 15 DOE commercial buildings,
each with 36-139 calibrated groups
• CV(RMSE) and NMBE are as good as any of the
proposed alternatives…which is to say, BAD.
– Calibration to important subsets is proposed
Bibliography
• Energy. 2015. Scalable tuning of building energy models
to hourly data. Energy 84, 493-502.
• ASHRAE. 2013. Evolutionary tuning of building models
to monthly electrical consumption. Transactions 119(2).
• Energy and Buildings. 2012. Evaluation of weather data
for building energy simulations. ENB 49(0), 109-18.
• IBPSA. 2012. Autotune E+ Building Models. IBPSA 37.
• NREL. 2011. Building energy simulation test for existing
homes (BESTEST-EX). NREL/TP-5500-52414.
• ASHRAE. 2002. ASHRAE Guideline 14, Measurement of
energy and demand savings.
Questions?
Joshua New
newjr@ornl.gov
Download