Joshua New Ph D Oak Ridge National Laboratory

Joshua New, Ph. D. Oak Ridge National Laboratory newjr@ornl. gov 865 -241 -8783 Technical Paper Session 3 Evolutionary Tuning of Building Models to Monthly Electrical Consumption Building Energy Modeling and Calculations 1

Learning Objectives • • • Describe reasons for and challenges involved with creation of an automated calibration methodology Explain how evolutionary computation works and how effectively it can create calibrated models Provide an overview of the Energy. Plus VRF Heat Pump Computer model Demonstrate the VRF computer model verification using manufacturer’s data Distinguish between five different existing methods for calculating distribution of absorbed direct and diffuse solar gains in perimeter building zones Understand the impact of solar energy distribution on heating and cooling loads as well as on free-floating room air temperatures for various climates and building envelope options 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. 2

Acknowledgements • Thanks go to: – – – Aaron Garrett, Ph. D. – Jacksonville State University Theodore Chandler – Jacksonville State University Amir Roth – DOE Building Technologies Office Oak Ridge Leadership Computing Facility Remote Data Analysis and Visualization Center 3

Objectives Q&A • What are two of ASHRAE’s primary sources for calibration, what is their purpose, and what performance metrics do they use? • What does SAE mean and what is its strength as a performance metric? • What is one of the acceleration methodologies used to speed up the calibration process and is it justifiable? • How well does envelope-only automated calibration currently do compared to human calibration? 4

Outline/Agenda • • • Context and calibration guidelines Evolutionary computation (EC) overview EC-based Autotune for building calibration Acceleration method Autotune calibration results 5

Context and Calibration Guidelines • Tool using BEM: retrofit optimization 6

Context and Calibration Guidelines • “All (building energy) models are wrong, but some are useful” – 22%-97% different from utility data for 3, 349 buildings • More accurate models are more useful – Error from inputs and algorithms for practical reasons – Useful for cost-effective energy efficiency (EE) at speed and scale • Calibration is required to be (legally) useful – ASHRAE G 14 (NMBE<5/10% and CV(RMSE)<15/30% monthly/hourly) • Manual calibration is risk/cost-prohibitive – Development costs 10 -45% of federal ESPC projects <$1 M – 114 of 119 US buildings are residential, 9% of ESCO market • Need robust and scalable automated calibration for market – Adjusts parameters in a physically realistic manner – Scales to any available data and model (audit) 7

Autotune E+ Input Model . . . 8

EC Overview • Evolutionary computation simulates natural selection – – – Genetic algorithms Evolution strategies Genetic programs Particle swarm optimization Ant colony optimization • EC approach to building calibration – Individual – a building (list of input parameters) – Fitness – error between simulation output and sensor data 9

EC Autotune What is an individual? • Defined by 108 real-valued parameters – Material • Thickness • Conductivity • Density • Specific Heat • Thermal Absorptance • Solar Absorptance • Visible Absorptance – Window. Material: Simple. Glazing. System • U-Factor • Solar Heat – Zone. Infiltration: Flow. Coefficient – Shadow Calculation Frequency 10

EC Autotune What is the fitness? Individual Fitness Model Error Actual Building Data 11

EC Autotune How do they evolve? Sister Mom Brother Dad 12

EC Autotune How are offspring produced? Thickness Conductivity Density Specific Heat Mom 0. 022 0. 031 29. 2 1647. 3 Dad 0. 027 0. 025 34. 3 1402. 5 Brother 0. 0229 0. 029 34. 13 1494. 7 Sister 0. 0262 0. 024 26. 72 1502. 9 • Average each component • Add Gaussian noise 13

EC Autotune • • • Population size 16 Tournament selection (tournament size 4) Generational replacement with weak elitism (1 elite) Gaussian mutation (mutation rate 10% of variable range) Heuristic crossover 14

Acceleration Method • Pick 1024 sub-atomic particles from the universe • Energy. Plus is slow – Full-year schedule – 2 minutes per simulation • Use abbreviated 4 -day schedule instead – Jan 1, Apr 1, Aug 1, Nov 1 – 10 – 20 seconds per simulation 15

Acceleration Method • 4 independent random trials • 1024 simulations per trial • Samples taken from high to low error r = 0. 94 Monthly Electrical Usage r = 0. 96 Hourly Electrical Usage 16

Acceleration Method Individual Fitness Model Error Actual Building Data 17

Acceleration Method Combining serially… Evolve 18

Acceleration Method Combining in parallel… Island Hopping 19

Autotune Calibration Results 25% reduction in error in 10 generations typical 20

Autotune Calibration Results What are you comparing to? Model Monthly SAE Hourly SAE (k. Wh) Hourly RMSE V 7 -A 2 1276. 340 6242. 036 1. 20594 28 July 2010 1623. 364 8113. 685 1. 62455 1800 1 623, 4 1600 1400 9000 8 113, 7 8000 1 276, 3 7000 1, 8 1, 4 6 242, 0 1200 6000 1, 2 1000 5000 1, 0 800 4000 0, 8 600 3000 0, 6 400 2000 0, 4 200 1000 0, 2 0 V 7 -A 2 28 July 2010 Monthly SAE 1, 6 0 1, 2 0, 0 V 7 -A 2 Hourly SAE 28 July 2010 V 7 -A 2 28 July 2010 Hourly RMSE 21

Autotune Calibration Results How well did Autotune do? • Autotune 108 envelope parameters 60% toward best manual model • Autotuned best model within $9. 46/month (actual use $152/month) 22

Bibliography • ASHRAE. 2013. Evolutionary Tuning of Building Models to Monthly Electrical Consumption. ASHRAE Transactions 119(1) (pending publication) • 22 Autotune-related publications: – 1 Ph. D dissertation, 9 accepted publications, 6 submitted publications, and 6 internal reports – Download data, view tuning dashboards, etc. 23

Questions? Joshua New newjr@ornl. gov 24