Decomposition of uncertainty propagation through networks of - PowerPoint PPT Presentation

Decomposition of uncertainty propagation through networks of heterogeneous energy systems Bryan Eisenhower Associate Director Center for Energy Efficient Design UCSB Focus Period on Dynamics, Control and Pricing in Power Systems Lund, Sweden: May 2 - May 27, 2011

Motivation – On Average 40% For Buildings 60% Wasted 15% Renewables

Motivation – On Average End Use 2008 Annual Energy Use (QBTU) Residential & 18.75 Commercial Buildings Lighting 2.01 Transportation 21.63 Cars 8.83  ~30% reduction can be achieved by occupancy based lighting control (0.8 QBTU) DoD Spends ~3.4Billion Annual on ~1 QBTU  A 47% reduction in buildings energy use will take ALL cars off the road! Source: Buildings Energy Data Book & US EIA

Motivation – On Average  It can be done (1 st three examples from recent HPB)! A Grander View , Ontario Canada - 22Kft^2 office David Brower Center , Ontario Canada - 80% Energy savings as recorded in first year - 45Kft^2 office / group meetings - Most energy efficient office in CA - 42.4 % Energy savings as recorded in 11 months. The Energy Lab , Kamuela Hawaii - 5.9Kft^2 Educational - 75% Energy savings compared to CBECS - 1 st year generated 2x electricity that it used

Motivation – On Average  It will be done…  DoD is the single largest energy user in U.S. Legislation: EPA2005:Section 109. Federal Building Performance Standards amended the Energy Conservation and Production Act11 by adopting the 2004 International Energy Conservation Code, and requiring revised energy efficiency standards and a 30% reduction in energy consumption of new federal buildings over the previous standards. EISA2007: Section 431. Energy Reduction Goals for Federal Buildings amends the National Energy Conservation Policy Act (NECPA)13 by mandating a 30% energy reduction in federal buildings by 2015 relative to a 2005 baseline. EISA2007: Section 433. Federal Building Energy Efficiency Performance Standards requires 55% reduced fossil energy use in new federal buildings and major renovations by 2010 relative to a 2003 baseline, and 100% by 2030. Net Zero will require ~70% reduction in energy use

Motivation – On Variance Southern CA Edison (2010) 500  Some aspects of the 450 400 design of the power grid are 350 based on long tail demand 300 Data: CA OASIS Hours concerns. 250 200 150 100 50 0 Only used 10 days a 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 Power [MW] 4 year… x 10 Top 25% of power only 2.74% of year. Load Duration Curve 4 Load Duration Curves x 10 2.4 Southern CA Edison 2.2 Pacific Gas & Elec. 2 1.8 Power [MW] 1.6 1.4 1.2 1 0.8 0.6 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 Hours

Motivation – On Variance  Similar long tail distributions are seen at the building level (no surprise) Life Sciences Building Student Resources Building 1400 1200 1200 1000 1000 800 800 Hours Hours 600 600 400 400 200 200 0 0 0 100 200 300 400 500 600 700 800 900 0 500 1000 1500 Energy [BTU] Energy [BTU] Top 25% of power only 0.41% of year. Top 25% of power only 0.99% of year. Data: Cooling energy for two buildings @ UCSB

Motivation  Pitfalls  “….these strategies must be applied together and properly integrated in the design and operation to realize energy Modeling savings. There is no single efficiency measure or checklist of measures to achieve low-energy buildings. “  “… dramatic improvement in Monitoring performance with monitoring and correcting some problem areas identified by the metering “  “There was often a lack of control Control software or appropriate control logic to allow the technologies to work well together “ [Lessons Learned from Case Studies of Six High-Performance Buildings, P. Torcellini, S. Pless, M. Deru, B. Griffith, N. Long, [Frankel 2008] R. Judkoff, 2006, NREL Technical Report.]

Summary Energy Visualization Energy/Comfort Failure Mode Effect Analysis Optimization Seasonal Consumption - Cooling Seasonal Consumption - Heating 400 400 Frequency 300 Frequency 300 200 200 100 100 0 0 0.5 1 1.5 0.5 1 1.5 2 Total Power 6 Total Power 6 x 10 x 10 Seasonal Consumption - Cooling PMV Avg. - Heating 300 300 Frequency Frequency 200 200 100 100 0 0 -1.5 -1 -0.5 0 -2 -1.5 -1 -0.5 0 PMV Avg. PMV Avg. Data analysis toolkits Uncertainty Analysis Uncertain Inputs Uncertain Outputs Building Model 120 100 700 80 600 Occurrences 500 60 ? Frequency 400 40 300 20 200 100 0 1600 1650 1700 1750 1800 1850 1900 1950 2000 0 VAV 3 Availability Manager 0.5 1 1.5 2 2.5 3 3.5 Heating [J] - Yearly Sum x 10 11 250 300 250 200 Occurrences 200 150 Frequency 150 100 100 50 50 0 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 0 Pumps [J] - Yearly Sum x 10 11 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 BLDG L IGHT S CH Advanced Energy Modeling 120 180 100 160 140 80 120 Occurrences Frequency 100 60 80 40 60 40 20 20 0 0.9 0.95 1 1.05 1.1 1.15 1.2 1.25 Interior Lighting [J] - Yearly Peak 8 0 x 10 1600 1650 1700 1750 1800 1850 1900 1950 2000 VAV 3 Availability Manager Sensitivity Decomposition methods

Modelling / Analysis

Energy Modeling  Energy models capture both the architectural components of the building as well as its thermal physics  Typical software contains front-end for drawing purposes, with mathematical engine for computation Models are built with highschool / undergrad help Ryan Casey Erika Building Equations / design Physics / etc.

Energy Modeling – Uses Reasons for modeling (entire building)  Compliance  Leadership in Energy and Environmental Design (LEED)  ASHRAE  Rebates for efficient design  Design trades  Usually very few performed in design firm  Academic Studies  Prediction of un-sensed data  Uncertainty / Sensitivity Analysis  Optimization (design / operation)  ….  Very little control design is performed with these models at the building level (some work at the component level).  Whole-building energy models not connected to grid.

Energy Modeling & Uncertainty  Decades spent on developing energy models  Most are validated on a component basis  At the systems level, the most advanced energy models, are still do not predict consumption accurately during the design stage Actual Prediction * Stanford Y2E2 Building

Energy Modeling & Uncertainty  Discrepancy is often introduced because of uncertainty  Commissioning / Operation  Material selection  Usage  … Other unknowns  Sensitivity / Uncertainty Analysis helps manage these concerns Sensitivity / Uncertainty Analysis Uncertain Inputs Uncertain Outputs Building Model 120 100 700 80 600 Occurrences 500 60 400 Frequency 40 300 200 20 ? 100 0 1600 1650 1700 1750 1800 1850 1900 1950 2000 VAV 3 Availability Manager 0 0.5 1 1.5 2 2.5 3 3.5 Heating [J] - Yearly Sum x 10 11 250 300 200 250 Occurrences 200 150 Frequency 150 100 100 50 50 0 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 Pumps [J] - Yearly Sum 11 0 x 10 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 BLDG L IGHT S CH 120 180 100 160 140 80 120 Occurrences Frequency 100 60 80 60 40 40 20 20 0 0.9 0.95 1 1.05 1.1 1.15 1.2 1.25 Interior Lighting [J] - Yearly Peak x 10 8 0 1600 1650 1700 1750 1800 1850 1900 1950 2000 VAV 3 Availability Manager O (10) O (1000)

Energy Modeling & Uncertainty Sampling Uncertainty Analysis • O.A.T. • STD(), VAR() • Monte Carlo • COV • Latin Hypercube • Amplification • Quasi-Monte factors Carlo (deterministic) Sensitivity Analysis • Elementary Effects / screening & local methods • Morris Method • ANOVA • Derivative-based • Propagation analysis through decomposition Red: In this talk

UA / SA – Historically (Building Sys.) Author(s) # Param. Technique Notes Rahni [1997] 390->23 Pre-screening Brohus [2009] 57->10 Pre-screening / ANOVA Spitler [1989] 5 OAT / local Residential housing Struck [2009] 10 Lomas [1992] 72 Local methods Lam [2008] 10 OAT 10 different building types Firth [2010] 27 Local Household models de Wit [2009] 89 Morris Room air distribution model Corrado [2009] 129->10 LHS / Morris Heiselberg [2009] 21 Morris Elementary effects of a building model Mara [2008] 35 ANOVA Identify important parameters for calibration also. Capozzoli [2009] 6 Architectural parameters Eisenhower [2011] 1009 (up Deterministic sampling, ‘All’ available parameters in building to 2000) global derivative sensitivity