Analysis Methods for Post Occupancy Evaluation

of Energy-Use in High Performance Buildings

Using Short-Term Monitoring

by

Vipul Singh

A Thesis Presented in Partial Fulfillment

of the Requirements for the Degree
Master of Science

Approved May 2011 by the

Graduate Supervisory Committee:

T Agami Reddy, Chair

Harvey Bryan
Marlin Addison

ARIZONA STATE UNIVERSITY

May 2011
 ABSTRACT

The green building movement has been an effective catalyst in reducing energy

demands of buildings and a large number of ‘green’ certified buildings have been

in operation for several years. Whether these buildings are actually performing as

intended, and if not, identifying specific causes for this discrepancy falls into the

general realm of post-occupancy evaluation (POE). POE involves evaluating

building performance in terms of energy-use, indoor environmental quality,

acoustics and water-use; the first aspect i.e. energy-use is addressed in this

thesis.

Normally, a full year or more of energy-use and weather data is required to

determine the actual post-occupancy energy-use of buildings. In many cases,

either measured building performance data is not available or the time and cost

implications may not make it feasible to invest in monitoring the building for a

whole year. Knowledge about the minimum amount of measured data needed to

accurately capture the behavior of the building over the entire year can be

immensely beneficial.

This research identifies simple modeling techniques to determine best time of the

year to begin in-situ monitoring of building energy-use, and the least amount of

data required for generating acceptable long-term predictions. Four analysis

procedures are studied. The short-term monitoring for long-term prediction

(SMLP) approach and dry-bulb temperature analysis (DBTA) approach allow

determining the best time and duration of the year for in-situ monitoring to be

performed based only on the ambient temperature data of the location.

Multivariate change-point (MCP) modeling uses simulated/monitored data to

determine best monitoring period of the year. This is also used to validate the

i
SMLP and DBTA approaches. The hybrid inverse modeling method-1 predicts

energy-use by combining a short dataset of monitored internal loads with a year

of utility-bills, and hybrid inverse method-2 predicts long term building

performance using utility-bills only.

The results obtained show that often less than three to four months of monitored

data is adequate for estimating the annual building energy use, provided that the

monitoring is initiated at the right time, and the seasonal as well as daily

variations are adequately captured by the short dataset. The predictive accuracy

of the short data-sets is found to be strongly influenced by the closeness of the

dataset’s mean temperature to the annual average temperature. The analysis

methods studied would be very useful for energy professionals involved in POE.

ii
 DEDICATION

Thank You, Ma…

iii
 ACKNOWLEDGMENTS

This thesis concludes extensive work realized through the assistance and

encouragement of many individuals. The journey of this research was possible

because of a multitude of support systems, some of whom deserve special

recognition.

I would like to thank my committee for their guidance throughout my studies at

ASU. Particularly, Professor T. Agami Reddy, my thesis Chair, for all his support

and encouragement during the course of this research. Having him as my mentor

enriched my learning experience. Under his able guidance I was able to

complete my thesis and my graduate education with ease and minimal effort. His

critiques and inputs played a very important part in the completion of this

research. My sincere thanks to Professor Harvey Bryan and Professor Marlin

Addison for their encouragement and direction that helped me shape my thesis.

This work has been made possible through the generous funding provided by

ASHRAE. The project is a joint venture between Arizona State University (ASU)

and Milwaukee School of Engineering (MSoE). I express my sincere gratitude to

Dr. Bass Abushakra, the P.I. for this project, and the ASHRAE project monitoring

sub-committee for their supervision and direction.

Finally, the unconditional love, support and faith of my family and friends made

this journey so much easier for me. I am grateful to them for being there for me

always. A special thank you to my uncle, Mr. C.K. Bhagat, without whose

support, I would have not been able to achieve this dream.

iv
 TABLE OF CONTENTS

Page

LIST OF TABLES..................................................................................................... viii

LIST OF FIGURES .................................................................................................... x

ACRONYMS ............................................................................................................ xv

CHAPTER

1 INTRODUCTION .................................................................................... 1

1.1 Overview .................................................................................. 1

1.2 Research outline ...................................................................... 3

1.3 Potential contribution ............................................................... 5

1.4 Organization ............................................................................ 5

1.5 Scope and limitations .............................................................. 6

2 ENERGY RATING SYSTEMS IN RELATION TO POE ........................ 7

2.1 Post Occupancy Evaluation (POE) ........................................ 7

2.2 Energy rating systems & POE................................................ 7

2.3 LEED rating system ................................................................ 9

2.4 BREEAM IN-USE rating system .......................................... 11

2.5 ENERGY STAR rating system ............................................. 13

2.6 TERI GRIHA rating system .................................................. 14

3 POST OCCUPANCY EVALUATION – ENERGY USE ....................... 16

3.1 Historical overview of POE of building energy use .............. 16

3.2 Energy use measurement protocols .................................... 22

4 BUILDING ENERGY USE MODELING ............................................... 26

4.1 Data driven analysis ............................................................. 27

4.2 Detailed calibrated simulations ............................................ 28

v
CHAPTER Page

4.3 Artificial intelligence .............................................................. 28

4.4 Short term to annual extrapolation ....................................... 29

4.5 Conclusions from background literature ............................... 31

5 RESEARCH METHODS ...................................................................... 33

5.1 Climate diversity test ............................................................ 34

5.2 Short Term Monitoring for Long Term Prediction (SMLP) ... 35

5.3 Outdoor Dry-Bulb Temperature Analysis (DBTA)................ 36

5.4 Multivariate Change Point (MCP) Modeling......................... 37

5.4.1 Regression based models ...................................... 37

5.4.2 Comparison of EE software with IMT ..................... 40

5.4.3 MCP modeling technique adopted ......................... 42

5.5 Hybrid Inverse Modeling ...................................................... 49

5.5.1 Method 1: utility bills + monitored data ................... 49

5.5.2 Method 2: utility bills only ........................................ 50

6 TEMPERATURE DATA ANALYSIS RESULTS................................... 52

6.1 Datasets ................................................................................ 52

6.2 Climate diversity analysis ..................................................... 53

6.3 SMLP Results ....................................................................... 56

6.4 DBTA Results ....................................................................... 59

6.5 Comparison of results of SMLP and DBTA ......................... 61

7 MCP MODELING RESULTS................................................................. 63

7.1 Multivariate change point modeling appraoch ..................... 63

7.2 Results for large hotel - Chicago, Illinois .............................. 64

7.3 Results for office building - Albuquerque, New Mexico ....... 74

vi
CHAPTER Page

7.4 Results for hotel –Washington D.C. area.............................. 79

7.5 Validation of DBTA approach ............................................... 81

7.6 Conclusions ............................................................................ 84

8 HYBRID INVERSE MODELING RESULTS.......................................... 85

8.1 Hybrid inverse modeling-method 1 ....................................... 85

8.1.1 Results for large hotel - Chicago, IL ........................ 86

8.1.2 Results for office building - Albuquerque, NM ......... 89

8.1.3 Conclusions ............................................................. 90

8.2 Hybrid inverse modeling-method 2 ....................................... 92

8.2.1 Results for large hotel - Chicago, IL ........................ 92

8.2.2 Results for office building – Albuquerque, NM ........ 94

8.2.3 Results for hotel, Washington DC area ................... 96

8.2.4 Conclusions ............................................................. 98

9 SUMMARY, CONCLUSIONS AND FUTURE WORK .......................... 99

9.1 Summary .............................................................................. 99

9.1 Conclusions .........................................................................100

9.1 Future Work ........................................................................106

REFERENCES/ BIBLIOGRAPHY ....................................................................... 107

APPENDICES

A DESCRIPTION OF BUILDINGS ANALYZED ................................... 112

B POE: IEQ, ACOUSTICS, WATER ..................................................... 121

C ANALYSIS DATA ............................................................................... 133

vii
 LIST OF TABLES

Table Page

2.1 Credits related to POE as allocated by LEED-NC v3 2009 .............. 10

2.2 Credits related to POE as allocated by TERI-GRIHA ....................... 15

5.1 Model equations for different base periods identified using

EE software and IMT ......................................................................... 42

5.2 Prediction results from EE and IMT programs .................................. 43

6.1 Summary of buildings chosen for analysis ....................................... 52

6.2 Variation in outdoor dry-bulb temperature for Chicago, Albuquerque

& Washington D.C. ........................................................................... 55

6.3 Ranking obtained from SMLP algorithm for each month of the year for

Chicago, Albuquerque & Washington DC ........................................ 58

6.4 Summary of ranking based on DBTA for each month of the year for

Chicago, Albuquerque and Washington D.C. ................................... 61

7.1 Ranking based on the MCP modeling approach for best to worst

months of the year for in-situ monitoring of WBE energy use for Large

Hotel in Chicago, IL ........................................................................... 67

7.2 Cooling energy-use prediction results from models derived using data

from the months of January-July, April-July, June-December and

August-December for Large Hotel, Chicago, IL. .............................. 69

7.3 Monitoring periods with prediction results closest to the results

obtained when monitoring is done for the whole year. Energy

Channel: CHW; Building: Large Hotel, Chicago, IL .......................... 70

7.4 Heating energy-use prediction results for June-December and

October-December for Large Hotel, Chicago, IL .............................. 72

viii
Table Page

7.5 Monitoring periods for which identified models predict energy use

closest to those for which model is identified using whole year data.

Energy Channel: HW; Building: Large Hotel, Chicago, IL. ............... 73

7.6 Monitoring periods for which identified models predict energy use

closest to those for which model is identified using whole year data.

Energy Channel: WBE; Building: Large Office, Albuquerque, NM. .. 76

7.7 Monitoring periods for which identified models predict energy use

closest to those for which model is identified using whole year data.

Energy Channel: HW; Building: Large Office, Albuquerque, NM. .... 78

7.8 Monitoring periods for which identified models predict energy use

closest to those for which model is identified using whole year data.

Energy Channel: WBE; Building: hotel, Washington DC. .................. 80

7.9 Comparison of results for DBTA and MCP modeling approach for

large hotel building in Chicago, IL. .................................................... 81

7.10 Comparison of results for DBTA and MCP modeling approach for

office building in Albuquerque, NM.................................................... 82

7.11 Comparison of Results for DBTA and MCP modeling approach for

hotel in Washington DC. .................................................................... 83

9.1 Application areas identified and proposed research methods .........105

ix
 LIST OF FIGURES

Figure Page

5.1 Different sub-categories considered suitable for this research......... 34

5.2 Typical Change Point Models (a) three-parameter (3P) heating

energy use, (b) three-parameter (3P) cooling energy use, (c) four-

parameter (4P) heating energy use model, and (d) four-parameter

(4P) cooling energy use model.......................................................... 39

5.3 Prediction accuracy – CV (%) of models identified from IMT and

Energy Explorer software. ................................................................. 44

5.4 Prediction accuracy – NMBE (%) of models identified from IMT and

Energy Explorer software. ................................................................. 44

5.5 Flowchart for multivariate change point modeling approach ............ 48

5.6 Flowchart for hybrid inverse modeling approach - method 1........... 51

6.1 Annual variation of outdoor dry bulb temperature for Chicago, IL.

Average monthly temperature, monthly temperature range and

average annual temperature (49.44°F) with ±10% variation bands

shown ................................................................................................ 53

6.2 Annual variation of outdoor dry-bulb temperature for Albuquerque, IL.

Average monthly temperature, monthly temperature range and

average annual temperature (55.75°F) with ±10% variation bands

shown ................................................................................................ 54

6.3 Annual variation of outdoor dry-bulb temperature for Washington,

D.C. Average monthly temperature, monthly temperature range and

average annual temperature (56.91°F) with ±10% variation bands

shown ................................................................................................ 54
x
Figure Page

6.4 SMLP analysis for Chicago, IL .......................................................... 56

6.5 SMLP analysis for Albuquerque, NM ................................................ 57

6.6 SMLP analysis for Washington DC Region ...................................... 57

6.7 Outdoor dry-bulb temperature analysis: Chicago, IL ........................ 59

6.8 Outdoor dry-bulb temperature analysis: Albuquerque, NM ............. 60

6.9 Outdoor dry-bulb temperature analysis: Washington DC Region .... 60

6.10 Result comparison for SMLP and DBTA for Chicago, IL. ................ 62

6.11 Result comparison of SMLP and DBTA for Albuquerque, NM. ....... 62

6.12 Result comparison of SMLP and DBTA for Washington DC ........... 62

7.1 CV (%) for energy channels WBE, CHW & HW when different lengths

of monitoring are used for predicting annual energy-use for large

hotel at Chicago, IL ............................................................................ 64

7.2 NMBE (%) for energy channels WBE, CHW & HW when different

lengths of monitoring are used for predicting annual energy-use for

large hotel at Chicago, IL................................................................... 65

7.3 Variation in CV (%) and NMBE (%) for WBE. Graphs show monitoring

periods for which predicted CV (%) is within 10% and predicted

NMBE (%) within ± 20% of actual energy-use. Building: large hotel at

Chicago, IL. ........................................................................................ 66

7.4 Variation in CV(%) and NMBE(%) for CHW Channel. Graphs show

monitoring periods for which predicted CV (%) is within 50% and

predicted NMBE (%) within ± 20% of actual energy use. Building:

large hotel at Chicago, IL................................................................... 68

xi
Figure Page

7.5 Variation in CV(%) and NMBE (%) for HW Channel. Graphs show

monitoring periods for which predicted CV (%) is within 50% and

predicted NMBE (%) within ± 20% of actual energy use. Building:

large hotel at Chicago, IL................................................................... 71

7.6 CV (%) for energy channels WBE & HW when different lengths of

monitoring are used for predicting annual energy use...................... 74

7.7 NMBE (%) for energy channels WBE & HW when different lengths of

monitoring are used for predicting annual energy use...................... 74

7.8 Variation in CV (%) and NMBE (%) for WBE. Building: office at

Albuquerque, NM .............................................................................. 75

7.9 Variation in CV (%) and NMBE (%) for HW for office at Albuquerque,

NM...................................................................................................... 77

7.10 CV (%) for WBE energy-use channel when different lengths of

monitoring are used for predicting annual energy ............................ 79

7.11 NMBE (%) for WBE energy-use channel when different lengths of

monitoring are used for predicting annual energy use for hotel,

Washington D.C area ........................................................................ 79

7.12 Variation in CV (%) and NMBE (%) for WBE. Graphs show periods

with predicted CV (%) within 50% and predicted NMBE (%) within ±

20% of actual energy use for hotel, Washington D.C Area .............. 80

7.13 Comparison of results for Simplified DBTA and MCP modeling

approach for large hotel building in Chicago, IL................................ 81

xii
Figure Page

7.14 Comparison of results for Simplified DBTA and MCP modeling

approach for office building in Albuquerque, NM .............................. 82

7.15 Comparison of results for DBTA and MCP modeling approaches for

hotel, Washington DC Region ........................................................... 83

8.1 Variation in CV (%) and NMBE (%) for different monitoring periods for

large hotel at Chicago, IL. Analysis procedure: Hybrid Inverse

Modeling (Method 1). Energy Channel: WBE .................................. 86

8.2 Variation in CV (%) and NMBE (%) for different monitoring periods for

large hotel at Chicago, IL. Analysis procedure: Hybrid Inverse

Modeling (Method 1). Energy channel: CHW ................................... 87

8.3 Variation in CV (%) and NMBE (%) for different monitoring periods for

large hotel at Chicago, IL. Analysis procedure: Hybrid Inverse

Modeling (Method 1). Energy channel: HW ..................................... 88

8.4 Variation in CV (%) and NMBE (%) for different monitoring periods for

office building at Albuquerque, NM. Analysis Procedure: Hybrid

Inverse Modeling (Method 1). Energy channel: WBE ....................... 89

8.5 Variation in CV (%) and NMBE (%) for different monitoring periods for

office building at Albuquerque, NM. Analysis Procedure: Hybrid

Inverse Modeling (Method 1). Energy channel: HW ......................... 90

8.6 WBE energy use profiles for the entire year obtained from the

simulated data & the prediction results of the hybrid inverse modeling

methods 1 & 2 for large hotel, Chicago, IL ........................................ 93

xiii
Figure Page

8.7 Comparison of NMBE(%) and CV(%) for WBE for Large Hotel,

Chicago, IL at monthly and annual timescales for Hybrid Inverse

Modeling analysis methods 1 & 2...................................................... 93

8.8 WBE profiles for the entire year obtained using the simulated data &

the prediction results of the Hybrid Inverse Modeling Methods 1 & 2

for office building at Albuquerque, NM .............................................. 95

8.9 Comparison of NMBE(%) and CV(%) for WBE for office at

Albuquerque, NM at monthly and annual timescales for Hybrid

Inverse Modeling Analysis Methods 1 & 2 ........................................ 95

8.10 WBE Consumption for full service hotel, Washington DC Region.

Actual monitored data and predicted energy use based on utility bills

are shown .......................................................................................... 97

8.11 NMBE(%) and CV(%) for WBE Predictions using utility history only.

The variations in prediction and average errors at monthly as well as

annual timescales have been indicated for hotel, Washington D.C.

area. Analysis procedure: Hybrid Inverse Modeling (Method 2). ..... 97

xiv
 ACRONYMS

ANSI American National Standard Institute

ASHRAE American Society of Heating, Refrigerating and Air-Conditioning

Engineers

BRE Building Research Establishment

BREEAM Building Research Establishment Environmental Assessment

Method

CASBEE Comprehensive Assessment System for Building Environmental

Efficiency

CBECS Commercial Building Energy Consumption Survey

CEC California Energy Commission

CHW Cooling Energy Use

CIBSE Chartered Institute of Building Services Engineers

DBT Dry Bulb Temperature

DHW Domestic Hot Water

ECM Energy Conservation Measure

EIA Energy Information Administration

EPA Environmental Protection Agency

EUI Energy Use Intensity

FEMP Federal Energy Management Program

GBCI Green Building Certification Institute

HK-BEAM Hong Kong Building Environmental Assessment Method

HVAC Heating Ventilation and Air Conditioning

HW Heating Energy Use

IEQ Indoor Environmental Quality

xv
IMT Inverse Modeling Toolkit

IPMVP International Performance Measurement & Verification Protocols

LBNL Lawrence Berkeley National Laboratory

LEED Leadership in Energy and Environmental Design

LTEQ Light and Equipment

M&V Measurement and Verification

NEMVP North American Measurement and Verification Protocol

NREL National Renewable Energy Laboratory

OA Outside Air

PMP Performance Measurement Protocols

POE Post Occupancy Evaluation

RP Research Project

TERI GRIHA The Energy Research Institute-Green Rating for Integrated Habitat

Assessment

xvi
 Chapter 1

INTRODUCTION

1.1. Overview

Buildings consume substantial amounts of energy, water and raw materials and

have a sizeable environmental footprint. They are big end users of energy and

account for 20-40% of the energy demands in developed nations (Birt 2009).

Over the past few years, the green building movement has been an effective

catalyst in reducing the energy demand of buildings. A large number of early

generation ‘green’ certified commercial buildings have been built and are in

operation for several years. Although these buildings seem to consume less

energy compared to conventional buildings, there are still areas of improvement.

It has become necessary to investigate if these so called ‘green’ buildings are

actually achieving their intended performance targets, and what can be done to

improve their performance further.

Energy performance of individual projects is highly variable and it is important

that actual building performance data be gathered and analyzed to compare the

design-performance with design-intent. Post Occupancy Evaluation (POE) is the

process of evaluating buildings in a systematic and rigorous manner after they

have been built and occupied for some time. It focuses on the requirements of

building occupants; and includes issues such as health, safety, security,

functionality and efficiency, psychological comfort and aesthetic quality. “It

constitutes any activity that originates out of an interest in learning how a building

performs once it is built and how satisfied building users are with the environment

that has been created.” (Preiser et al., 2009).

1
POE can thus be seen as a practice aimed at understanding design criteria,

predicting effectiveness of emerging designs and for linking user response to the

performance of the building. It therefore helps determine how well design

concepts work once implemented.

An important component of POE relates to energy use. Normally a full year or

more of energy use and weather data is needed to determine the actual energy

use of the building. However, in many cases, either a full year of measured

building performance data is not available or the time and cost implications may

not make it feasible for the building owner to invest in monitoring the building for

the entire year. Knowledge about what minimum amount of measured data will

be sufficient to accurately capture the behavior of the building over the entire

year can prove to be immensely beneficial in such cases.

The main focus of this research is to develop and assess methods by which

short-term in-situ monitoring of building can be used as a workable alternative to

yearlong monitoring and verification (M&V). The intent obviously would be to

gather and analyze data for as short a period as possible. Thus the monitored

data should not only capture the diurnal variations of the weekdays and

weekends but also allow acceptable predictions to be made for seasonal

variations over the whole year.

2
1.2. Research outline

With the aim of promoting sustainable buildings, numerous countries have

developed their own green building programs. At the crux of each program is the

use of an integrated design approach and a point scheme that allots credits for

building design. The energy points allotted are typically based on the size of the

predicted reductions versus a locally specified baseline. The first buildings that

received accreditation under these green building rating schemes have now been

occupied for a reasonable amount of time, and it is germane to ask whether

these buildings are living up to expectations?

This research addresses the following three closely related aspects associated

with building in-use assessment or post-occupancy evaluation (POE):

1. What is the role of POE in the current green building rating systems?

Most rating systems are quite elaborate and exhaustive in their content with

regards to the design of buildings; however, is there enough emphasis being

placed on proper operation of the building once it is occupied? These rating

systems have been thoroughly explored and compared previously (Cole

1998; Crawley et al., 1999; Todd et al., 2001 Bosch et al., 2003; Fenner et

al., 2008; Lee et al., 2008). In this research four green rating systems,

namely, Leadership in Energy and Environmental Design (LEED), Building

Research Establishment Environmental Assessment Method (BREEAM),

Energy Star, and The Energy Research Institute-Green Rating for Integrated

Habitat Assessment (TERI-GRIHA) have been reviewed and the categories

that relate to post occupancy evaluation have been identified and outlined.

The aim is to understand whether these rating systems place adequate

emphasis on compliance, i.e., assuring that the buildings are actually

3
 achieving the ‘green’ goals that were intended during its design and

construction.

2. How is the post-occupancy performance of a building evaluated?

In order to ascertain whether the design concepts have actually been

transformed to reality, it is important that occupied buildings be verified for

their actual performance. Over the years, a number of studies have

attempted to find the difference between the predicted and the actual energy

use in high performance buildings. However, these studies were never based

on any consistent set of guidelines and methods for measuring, expressing

and comparing energy use making it difficult to compare the results from one

study to another (Haberl et al., 2006). Recently, the United States Green

Building Council (USGBC) in association with the American Society of

Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) & the

Chartered Institute of Building Services Engineers (CIBSE) has been

instrumental in the development of the “Performance Measurement Protocols

for Commercial Buildings (2010)” for assessing the performance of

commercial buildings. In Chapter 3, the major studies that have been

undertaken in the field of post occupancy evaluation of high performance

buildings are reviewed and the procedure for measuring energy use in

commercial buildings is outlined.

3. Can the performance of a building be assessed based on short-interval data?

Almost all POEs adopt the common approach to monitor the actual building

performance for the whole year. While sound, this is not only time consuming,

but also extremely expensive. The third part of this research, evaluates the

effectiveness of using short-term monitoring for predicting energy use in

4
 buildings. The aim is to develop analysis methods by which the time period

for field monitoring of energy use in buildings can be reduced to less than a

whole year (preferably less than three months) while satisfying preset

accuracy levels of annual energy performance verification. Such a method

will prove to be a cost effective alternative to year-long monitoring and

verification of buildings.

Developing a methodology for determining the best period of the year in which to

conduct in-situ monitoring so as to verify long-term building performance is the

primary focus of this research.

1.3. Potential contribution

At present there is no systematic guidance available to determine either the best

time of the year, or the minimum duration of the year, for in-situ monitoring that

can most accurately predict the energy use over the whole year. Many

engineering, consulting and services companies would like to see the

development of M&V alternatives which provide the sought-after verification with

shorter periods of data monitoring. This research aims to develop analysis

methods that can be used for making informed decisions in regards to the best

short-term M&V period for building performance assessment.

1.4. Organization

This thesis has been structured as follows: Chapter 1 gives an overview of the

research topic and outlines different sub-categories within the research

undertaken. The scope & limitations of the methods adopted have also been

stated. Chapters 2, 3 and 4 are the culmination of the literature review. Chapter 2

reviews issues related to the emphasis placed by the popular energy rating

systems on post occupancy evaluation of building energy performance. Four

5
energy rating systems have been studied, namely, LEED, BREEAM, Energy Star

and TERI-GRIHA. Chapter 3 reviews the major studies that have been

undertaken in the past for evaluating high performance buildings. The methods

that can be adopted for measuring, expressing and comparing energy use in

buildings are discussed. Chapter 4 reviews the existing literature on building

energy use modeling and the significant work that has already been done in the

field of short-term monitoring of building energy use for long term prediction of

building performance. The research methods adopted for analysis are discussed

in detail in Chapter 5. Chapters 6, 7 and 8 demonstrate how the research

methods were applied to data obtained from three buildings. The results of the

data analysis have also been presented and analyzed. Chapter 9 concludes this

thesis with a summary of findings and it also outlines potential future research.

1.5. Scope and limitations

This report provides a theoretical insight into post occupancy evaluation of

buildings in terms of energy use only. A brief description of the other aspects, i.e.

Indoor environmental quality (IEQ), acoustics and water use has been included in

Appendix B. The analysis methods have been evaluated and refined using year-

long data from two synthetic data-sets, a large hotel building of 619,200 ft2 in

Chicago, IL and a 17,430 ft2 office building in Albuquerque, NM, and one actual

measured data-set from a full service hotel building1 of 212,000ft2 located in the

Washington D.C region. As future scope for the project, the same analysis

methodology can be extended to a variety of geographic locations and building

types.

1
Due to confidentiality agreement, the name, location and description of the hotel
has been with held.
6
 Chapter 2

ENERGY RATING SYSTEMS IN RELATION TO POE

2.1. Post occupancy evaluation (POE)

Post Occupancy Evaluation (POE) is the process of evaluating a building’s

performance once it has been occupied. It focuses on the requirements of

building occupants, including health, safety, security, functionality and efficiency,

psychological comfort and aesthetic quality. POE can be seen as a practice

aimed at:

i. Understanding design criteria,

ii. Predicting effectiveness of emerging designs, and

iii. Linking user response to the performance of the building.

It can be used to measure functionality and appropriateness of design and

establish conformance with performance requirements as intended in the

functional program. It therefore helps determine how well concepts work once

applied.

2.2. Energy rating systems & post occupancy evaluation

With the increasing awareness of sustainable development in the construction

industry, implementation of an energy rating procedure to assess actual

performance of buildings is becoming extremely important. Some of the most

representative building schemes are LEED, BREEAM and Environmental

Protection Agency (EPA) Energy Performance Rating System. TERI-GRIHA has

been recently developed in India for assessing both non-air-conditioned and

partially conditioned buildings. Some other successful international rating

programs are: Comprehensive Assessment System for Building Environmental

Efficiency (CASBEE- Japan), Hong Kong Building Environmental Assessment

7
Method (HK-BEAM), Green Globes (Canada) and Green Star (Australia).

Roderick et al., (2009) compared the energy performance assessment between

LEED, BREEAM and ENERGY STAR.

An extremely important document that can have a considerable impact on the

green rating criteria in the future is ASHRAE 189.1: Standard for the design of

high-performance green buildings which was approved by American National

Standard Institute (ANSI) in January 2010. ASHRAE 189.1 has been designed to

provide a “green” foundation for buildings, and is set to complement LEED

requirements. Unlike LEED, ASHRAE 189.1 has been created with the eye

towards being incorporated into building codes, which would require mandatory

compliance. It provides the minimum requirements for the siting, design,

construction, and plans for operation of high performance green buildings. Where

a requirement is contained in this standard, it supersedes the requirements in all

the other standards.

Sections 2.3 to 2.7 that follow are limited to a discussion of the credits/points

related to post-occupancy evaluation within the LEED, BREEAM, EPA energy

performance (Energy Star) rating and the TERI-GRIHA schemes.

8
2.3. LEED rating system

Leadership in Energy and Environmental Design (LEED) is the most recognized

building environment scheme. The current version for new construction is LEED-

NC v3, which like the earlier versions, is based on a set of prerequisites and

credits. The credits are distributed amongst the following main categories:

sustainable sites (SS), water efficiency (WE), energy and atmosphere (EA),

materials and resources (MR), indoor environmental quality (IEQ) and innovation

& design (ID). There are up to 110 points that can be achieved under various

categories and based on the awarded points. There are four levels of

performance rating that the building can qualify for i.e. Certified (40+ points),

Silver (50+ points), Gold (60+ points) and Platinum (80+ points).

LEED-NC specifies two approaches to assess building energy performance.

These are included under the EA Credit 1-Optimize Energy performance. The

first is the Prescriptive Compliance Path, which allows projects to achieve certain

points when they meet the prescriptive measures of the ASHRAE Advanced

Energy Design Guides. The other approach is the Whole Building Energy

Simulation, which allows up to 19 points when the building demonstrates

improvement on energy cost compared to a normalized building. For both

approaches, the assessed building needs to meet a minimum performance level,

which is 2 points. This is equivalent to 10% improvement in the proposed building

performance rating for new buildings, or a 5% improvement in the proposed

building performance rating for major renovations to existing buildings (Roderick

et al., 2009).

EA Credit3: Measurement and Verification allows 3 credit points for new

construction. The intent is to provide for the ongoing accountability of building

9
energy consumption over time. The credit requires development and

implementation of a M&V plan consistent with either Option D: Calibrated

Simulation (Savings Estimation Method 2) or Option B: Energy Conservation

Measure Isolation as specified in the International Performance Measurement &

Verification Protocols (IPMVP) Volume III: Concepts and Options for Determining

Energy Savings in New Construction, April 2003. The M&V period must cover

atleast one year of post construction occupancy.

In addition, the new LEED 2009 rating system requires that all certified projects

commit to sharing with USGBC and/or Green Building Certification Institute

(GBCI) all available actual whole-project energy and water usage data for a

period of at least 5 years. USGBC is convinced that ongoing monitoring and

reporting of data is “the single best way to drive higher building performance

because it will bring to light external issues such as occupant behavior or

unanticipated building usage patterns” (Horst 2009). At present there is no option

to revoke an already granted certification based on monitored building

performance. USGBC recommends that projects provide performance data and

in case a project refuses to comply with this “Minimum Program Requirement”

criteria, then it not be certified at the onset. No concrete rules regarding de-

certification has yet been framed in case the building is found to under-achieve.

Table 2.1 summarizes the points related to building commissioning and post

occupancy measurement and verification under the LEED-NC v3.

Table 2.1
Credits related to POE as allocated by LEED-NC v3 2009.
LEED V3 2009: Credits Related to POE Points Remarks
EA Prereq 1 Fundamental Commissioning of Building Energy Systems Mandatory
EA Prereq 2 Minimum Energy Performance Mandatory
EA Credit 1 Optimize Energy Performacne 1 to 19
EA Credit 2 Enhanced Commissioning 2
EA Credit 3 Measurement and Verification 3
10
2.4. BREEAM-IN-USE rating system

Building Research Establishment’s Environmental Assessment Method

(BREEAM) was developed in the United Kingdom in 1990 and is one of the

earliest environmental assessment methods. It covers a range of building types

including offices, homes, industrial units, retail units and schools. The latest

version for BREEAM was released in 2008. Similar to the credit rating system in

LEED, BREEAM 2008 defines categories of credits according to the building

impact on the environment including management, health and well being, energy,

transport, water materials, waste, land use & ecology and pollution. The total

score is calculated based on credits available, number of credits achieved for

each category and a weighing factor. The overall performance is awarded a

‘Pass’, ‘Good’, ‘Very Good’ or ‘Excellent’ rating based on the score achieved.

The BREEAM-in-Use standard, released in October 2009, has been produced to

enable provision of information about the environmental performance of

buildings, building operation and how clients are managing their activities within

the building. The standard is arranged in three parts and it covers major

environmental issues that affect buildings throughout their operational life. The

three broad categories of the BREEAM-in-Use are:

i. Asset performance: inherent performance characteristics based on built

form, construction and services.

ii. Building management performance: the management policies,

procedures and practices related to the operation of the building, the

consumption of key resources such as energy, water and other consumables and

environmental impacts such as carbon and waste generation.

11
 iii. Organizational effectiveness: the understanding and implementation of

management policies, procedures and practices; staff engagement; and delivery

of key outputs.

The BREEAM In-Use certification scheme enables organizations to assess the

performance of individual buildings or a portfolio of buildings, and the

effectiveness of their designs. The scheme has a simple to use online system

and uses licensed BREEAM In-Use auditors to conduct audits, provide an

independent perspective and help organizations maximize the benefits of the

scheme. The online system enables building managers to conduct a pre-

assessment to quantify the impact of their building, existing systems and

initiatives, as well as the potential impact of proposed upgrades.

Following a successful audit, licensed Auditors are able to issue reports and

certificates on Building Research Establishment (BRE) Global’s behalf which

highlight how well a building and organization are currently performing and ways

for future improvement.

The BREEAM In-Use criteria are currently part of the United Kingdom building

regulation codes of practice, climatic conditions and energy methodology.

BREEAM has made it mandatory for buildings to obtain a BREEAM In-Use

certification within the first three years of the buildings operation and use. A

building should be occupied for at least a year before getting this certification to

ensure that enough data is collected for performance verification.

12
2.5. ENERGY STAR rating system

The EPA (Environmental Protection Agency) Rating is administered through the

ENERGY STAR program (EPA 2010b) and the rating is delivered by two online

tools:

• Portfolio Manager (for existing buildings), and

• Target Finder (for design projects)

Portfolio Manager is an interactive energy management tool that allows the

tracking and assessing of energy and water consumption through a secure online

environment. It requires one year of monthly energy consumption data to

benchmark existing building energy performance. The rating is done on a scale

of 1–100 relative to similar buildings nationwide. Higher the score, better the

building performance.

The building being assessed is not compared to other buildings; instead,

statistically representative models are used to compare the building against

similar buildings from a national survey conducted by the Energy Information

Administration (EIA) of the Department of Energy.

This national survey, known as the Commercial Building Energy Consumption

Survey (CBECS), is conducted every four years, and gathers data on building

characteristics and energy use from thousands of buildings across the United

States. The building’s peer group of comparison is those buildings in the CBECS

survey that have similar building and operating characteristics. A rating of 50

indicates that the building, from an energy consumption standpoint, performs

better than 50% of all similar buildings nationwide, while a rating of 75 indicates

that the building performs better than 75% of all similar buildings nationwide.

13
Thus, Portfolio Manager provides a platform to track energy and water use trends

as compared with the costs of these resources. This is a valuable tool for

understanding the relative costs associated with a given level of performance.

The Target Finder (EPA 2010c) sets energy performance rating targets and

compares the estimated annual energy use of a building design to the measured

energy consumption. It is a no-cost online tool that enables architects and

building owners to set energy standards and receive an EPA energy

performance score for projects during the design process. The projects that earn

a score of 75 or higher are eligible for the ‘Designed to Earn the Energy Star’

certification.

2.6. TERI GRIHA rating system

TERI-Green Rating for Integrated Habitat Assessment (GRIHA) is a five star

rating system in India. The rating emphasizes passive solar techniques for

optimizing indoor thermal and visual comfort. GRIHA encourages the

optimization of building design to reduce conventional energy demand. A building

is assessed on its predicted performance over its entire life cycle from inception

through operation. GRIHA was developed as an indigenous building rating

system, particularly to address and assess non-air conditioned or partially air-

conditioned buildings. It integrates all relevant Indian codes and standards for

buildings and acts as a tool to facilitate their implementation.

Up to 104 points over 34 different categories can be earned under the GRIHA

rating system; there are five levels of star-rating: one star (50-60 points), two

stars (61-70 points), three stars (71-80 points), four stars (81-90 points) and five

stars (91-100 points). Out of the 34 categories, two are specifically associated

with building performance evaluation. These are listed in Table 2.2 below.

14
Table 2.2

Credits related to POE as allocated by TERI GRIHA.

TERI GRIHA: Credits Related to POE Points Remarks

Credit 32 Energy Audit and Validation Mandatory
Operations and Maintenance Protocol for Electrical and
2
Credit 33 Mechanical Equipment Mandatory

• TERI-GRIHA Criterion 32: Energy Audit and Validation

Under this criterion, all buildings certified under the GRIHA rating have to get an

audit report prepared by approved auditors of the Bureau of Energy Efficiency

(BEE), Government of India. This is done to validate the predicted energy

consumption, thermal comfort and visual comfort criteria. In addition to

preparation of the audit report, TERI also performs thermal performance

monitoring and visual comfort monitoring for typically representative days to

verify the data provided in various documents, and for which points have already

been awarded. This is a mandatory provision and a building that does not adhere

to this category is not awarded the GRIHA Rating.

• TERI-GRIHA Criterion 33: Operation and Maintenance

In order to validate and maintain ‘green’ performance levels and propagate green

practices and concepts, TERI has made it mandatory that all Electrical and

mechanical systems be maintained by the owner, supplier or contractor and this

have to be ensured by means of an official contract.

15
 Chapter 3

POST OCCUPANCY EVALUATION OF ENERGY USE

A number of studies and extensive literature is available that describes various

procedures relevant to performance measurement of occupied buildings. This

database is often used by energy engineers and commissioning agents for the

diagnosis of operational problems and commissioning errors in buildings.

Knowledge of these techniques can help architects, designers and engineers

evaluate how design concepts actually work once applied, and can help them to

make informed design decisions.

Building In-Use assessment involves the measurement of its performance under

the following sub-categories:

i. Energy Use
ii. Indoor environmental quality (IEQ)
a. Thermal comfort
b. Indoor air quality
c. Lighting/daylighting
iii. Acoustics.
iv. Water Use

This chapter reviews the published work on post-occupancy performance of

green buildings in terms of energy use only. The other aspects, namely, IEQ,

Acoustics and Water Use have been briefly outlined in Appendix B.

3.1. Historical overview of POE of building energy use

The history of M&V of building energy use parallels the development and use of

computerized energy calculations in the 1960’s, with a much accelerated

awareness in 1973, resulting from the Middle East oil crisis. Prior to that energy

was cheap and abundant and M&V of energy use in a building was limited to
16
simple, unadjusted comparisons of monthly utility bills. Some of the earliest

efforts to develop standardized methods for the evaluation of building energy use

began with efforts to normalize residential heating energy use in single-family

and multi-family buildings (Socolow 1978). Procedures and methodologies to

baseline energy use in commercial buildings began to appear only in the 1980’s

and the early 1990’s (for example Haberl et al., 1988; Claridge et al., 1991).

Modeling toolkits, software, and measurement procedures were developed to aid

performance evaluation of buildings and HVAC system components. The

prominent ones amongst these include RP-1050 for calculating linear inverse

building energy analysis models (Kissock et al., 2001), and RP-1093 for

calculating diversity factors for various loads (Abushakra et al., 2001).

Efforts in several states in the United States for measuring the energy and

demand savings from retrofits in existing buildings culminated in the development

of the U.S. Department of Energy’s (USDOE) 1996 North American M&V

Protocol (NEMVP), which was accompanied by USDOE’s 1996 Federal Energy

Management Program (FEMP) Guidelines. In 1997, the NEMVP was updated

and republished as the International Performance Measurement and Verification

Protocols (IPMVP). The IPMVP was then expanded in 2001 into two volumes:

Volume I covering Energy and Water savings, and Volume II covering Indoor

Environmental Quality. In 2003 Volume III of the IPMVP was published that

covered protocols for New Construction. Finally in 2002, ASHRAE released

Guideline 14-2002: Measurement of Energy and Demand Savings intended to

serve as the technical document for the IPMVP.

Over the last three decades, significant work has been done with regards to the

development of methods for measuring, reporting and validating building energy

17
use. Much of the foundation for the ASHRAE and IPMVP publications mentioned

above was developed through the Texas A&M University’s Loan STAR (Loan to

Save Taxes and Resources) project, which developed energy monitoring

guidelines for use in Texas (Haberl et al., 1992, 1996). The Loan STAR program

was established in 1988 by the Governor’s Energy Office of Texas and aimed at

using a revolving loan financing mechanism to fund energy-conserving retrofits in

state, public schools and local government buildings. A unique feature of the

Loan-STAR was the ‘Monitoring and Analysis Program’ (MAP) (1992) that was

established to measure and report the energy savings from the retrofits. The

program resulted in the formulation of a complete workbook that included

descriptions and installation instructions for sensors needed for measuring,

procedures for retrieving data from remote buildings, and an overview of analysis

methods for building energy monitoring.

Torcellini et al., (2004) monitored and evaluated the energy performance of six

high-performance buildings around the United States. Evaluations began with

extensive one-year minimum building energy use monitoring; the data were then

used to calibrate energy simulation models. The energy performance was

compared with other buildings and to code-compliant, base case buildings.

These case studies proved instrumental in highlighting what could be repeated

and/or avoided in future buildings energy-wise. All the buildings were found to

perform better than typical buildings; however, none of them performed as well

as initially predicted. Torcellini and Deru (2005) also conducted extensive

research to identify performance metrics which have the greatest value for

determining energy performance. Their research served as a starting point to

develop standard definitions and methods for measuring and reporting building

18
performance. Their findings were published in the National Renewable Energy

Laboratory’s (NREL) ‘Performance Metrics Research Project’ (October 2005).

Brook et al., (2005) from California Energy Commission (CEC) and Haves et al.,

(2005) from the Lawrence Berkeley National Laboratory (LBNL) worked on

“Development of a Model Specification for Performance Monitoring Systems for

Commercial Buildings” and focused on four key aspects of performance

monitoring: (i) performance metrics, (ii) measurement system requirements, (iii)

data acquisition and archiving and, (iv) data visualization and reporting. The

study reviewed the potential benefits of performance monitoring and briefly

discussed the ways in which it could be implemented. It also gave insights into

reliable and effective database management and data visualization tools for

proper management, archiving and visualization of the data obtained from

building monitoring.

Haberl et al., (2006) reviewed significant work that had been done in the field of

energy use assessment. The ‘Scoping Study: Protocols for Measuring and

Reporting the On-site Performance of Buildings except Low-Rise Residential’

helped determine shortcomings in existing measurement protocols and make

recommendations to ASHRAE on how to proceed. This study formed the basis

for the development of the ASHRAE Performance Measurement Protocol for

Commercial Buildings (2010).

A large number of ASHRAE research projects have been instrumental in

providing the basis for several ASHRAE guidelines and standards related to

performance measurement. In ASHRAE RP-1286, “Evaluation of Building Energy

Performance Protocols”, Glazer (2006), developed guidance regarding base-

lining of building energy use. The project also provided an evaluation of common

19
building performance rating tools such as LEED, CAL-ARCH (California building

energy reference tool), BREEAM, etc.

More recently, a number of researchers have conducted studies to evaluate the

performance of many existing ‘green’ certified buildings. Turner et.al., (2007),

assessed eleven LEED certified buildings in the Cascadia Region, USA, by

comparing actual utility usage to three different metrics; namely, design energy

use, energy use compared to a code-compliant baseline and average energy use

of the commercial building block. Turner and Frankel (2008) undertook the study

of 121 North American LEED New Construction (NC) buildings. The study

analyzed whole-building energy usage with three different metrics, namely, (i)

Energy Use Intensity (EUI) comparison of LEED and national building stock

(CBECS national average for commercial building stock), (ii) Energy Star ratings

of LEED buildings, and (iii) Measured results compared to initial design and

baseline modeling. The authors concluded that on an average, LEED buildings

were delivering anticipated results and all the three views of building

performance consistently showed that average LEED building energy use was

25-30% better than the national average. They reported that although energy

modeling is a good indicator of program wide performance, individual project

modeling predictions varies widely from actual project performance outcomes.

They suggested improvements in LEED program quality control and follow-ups to

help maintain savings. To this end, awarding of LEED credits for advanced

commissioning and M&V were recommended.

Fowler and Rauch (2008) reported on a comprehensive study of 12 General

Services Administration (GSA) buildings. The buildings chosen were

predominantly federal buildings designed with either a green intent or LEED

20
certified. The study investigated the performance of these buildings by reviewing

water and energy use, waste management, occupancy satisfaction,

transportation, operation and maintenance, etc. Energy use was determined from

atleast 12 months of utility bills. The actual building use was compared with the

CBECS national and regional averages.

In each of the studies mentioned above there was a noticeable difference

between the predicted and the measured energy use of the building. The reason

for this variation could be attributed to a number of factors such as occupancy

patterns, difference between as-built and the initial design intent, variation in the

equipment schedules and improper commissioning of buildings. Most of these

factors are unavoidable under practical circumstances and can result in buildings

to perform quite differently than designed or predicted. It, therefore, becomes

extremely important to verify the performance of the building through actual

monitoring once the building is in operation.

A research project called “Probe” (Post-Occupancy Review of Buildings and their

Engineering) was carried out by Bordass et al., 1995-2002, under the Partners in

Innovation scheme (jointly funded by the UK Government and The Builder

Group, publishers of Building Services Journal). The project involved reviewing

high performance buildings through site documentation, technical surveys (walk-

through and spot checks), energy survey’s (CIBSE TM22 analysis), envelope

pressure tests, occupant questionnaire surveys, management interviews and

designer’s responses. The project provided tremendous information regarding

the actual performance of a number of buildings in the United Kingdom. Bordass

proposed that placing a post occupancy or as-built certification label on the

21
building can drive the need for greater emphasis on build quality/process and

commissioning of the building.

USGBC in association with ASHRAE & CIBSE has been instrumental in

developing the ‘Performance Measurement Protocols for Commercial Buildings’

(2010) for assessing the performance of commercial buildings. The document

provides a consistent set of guidelines and methods for measuring, expressing

and comparing energy use of occupied buildings.

3.2. Energy use measurement protocols

The commonly referenced standards for energy-use measurement in buildings

include:

• ASHRAE Guideline 14, Measurement of Energy and Demand Savings

(ASHRAE 2002),

• ANSI/ASHRAE Standard 105, Standard Methods for Measuring,

Expressing and Comparing Building Energy Performance (ASHRAE

2007a),

• LEED for Existing Buildings Operation and Maintenance, International

Performance Measurement and Verification Protocols (IPMVP) DOE

(1997), etc.

The ASHRAE Performance Measurement Protocols for Commercial Buildings

(2010) identifies the following commonly followed techniques for measuring the

energy-use in a building:

i. Actual physical measurement

ii. Forward and inverse modeling

iii. Whole building calibrated simulation

22
This section limits itself to the measurement protocols for actual physical

measurement of building energy use. Forward/Inverse Modeling and Whole

Building Calibrated Simulation are briefly discussed in Chapter 4.

Based on the detail, rigor, and accuracy of the measurements required to meet

the performance objectives, three levels of performance measurement protocols

have been identified, namely; Basic, Intermediate and Advanced.

The basic performance measurement method is intended to provide the

foundation for characterizing and understanding building energy performance in

a way that management can understand and incorporate into planning. The

estimated expense for basic level performance verification can range from

$1,000-2,000 per project approximately. The measured/required data are of three

types:

i. Basic building characteristics (project address, location, gross floor area,

operation, occupancy patterns, building function etc.).

ii. Whole-building annual energy use and costs of all Electricity and fuels used,

including the highest annual peak demand for each fuel. Actual site energy

use and energy cost for all forms of energy over a period of 12 consecutive

months (365 days, annual total) must be reported.

iii. Energy cost indices i.e. energy cost per unit area of the building.

Total Energy Use2 Index (EUI) = Total Annual Energy Use / Gross Floor Area, kBtu/ ft2 ...........(Eq. 3.1)

2
Total Energy Use: The total energy is the sum of all energy used in the
building (excluding source energy supplied by heat pumps). It is equal to the
energy imported (purchased) to the facility plus on-site generated energy minus
energy exported (sold) from the facility.

23
Net Energy Use Index3 (Net EUI) = Net Annual Energy Use / Gross Floor Area, $US/ ft2 .........(Eq. 3.2)

Energy Cost Index (ECI) = Net Annual Energy Cost4/ Gross Floor Area, $US/ ft2 .....................(Eq. 3.3)

The intermediate performance assessment method provides measurement and

evaluation to support an enhanced level of understanding of buildings and end

use performance so as to identify possible areas of performance improvement.

This level begins by incorporating the system-level (energy end-use) effects and

adds initial systems and end-use assessment to the basic performance

determination. The intermediate methods build on the basic annual energy use

data and associated EUI’s and ECI’s by going deeper into monthly and weekly

data for the whole building and for major energy end uses. In addition to energy

data, the Electrical demand for each of the months in the 12-month period is also

reported, and the Electrical Load Factor for each month is calculated as:

Electrical Load Factor, ELF5 = Electric Use (kWh) for the month ______________
(Monthly) Electric Demand for the month (kW) X No. of days in the month X 24
…………….. (Eq. 3.4)
3
Net Energy Use: The net energy is the sum of the imported (purchased)
energy minus exported (sold) energy.
4
Annual Energy Cost: The total cost for each energy form used. Monetary
compensation for energy exported (sold) from the facility is recorded as a
negative number. The cost of stored purchased energy used in the building is
determined using the cost of the oldest fuel in storage and not its replacement
cost.
5
Electric use profiles and ELF values should be compared against expected
patterns. As occupancy factors reduce, ELF values should also go lower.
Buildings with efficient energy use often have ELF values of 30% or less.
Buildings that have 100% occupancy factors will often have ELF values of 60 %
or more.
24
The Electric load profiles and ELF values are compared against expected

patterns. Intermediate instrumentation and analysis to determine this level of

performance may cost in the range of $2,000 to $10,000 per project

approximately.

Advanced performance assessment method is used to track daily/hourly whole

building energy performance through sub-metered data. This can help identify

additional potential improvements to increase building performance and to

develop advanced models of building performance.

This level requires that all basic results be updated and reported on a daily basis,

i.e., updated every day for the 365-day period. Second, major end uses of energy

must be either modeled or measured on an annual basis. Measurement can be

by sub- meters installed for HVAC total electric, HVAC fan electric, non-electric

heating, indoor lighting, miscellaneous electric etc.

The modeling of major end-uses can be through a combination of analysis and

short-term or diagnostic metering. The end uses that should be determined are:

heating energy, cooling energy, fan energy for heating/ cooling/ ventilation and/or

exhaust, indoor lighting, major equipment centers and all other electric loads.

Due to the extensive instrumentation and sophisticated analytical techniques

required, advanced performance evaluation may cost anywhere from $10,000-

100,000 per building. This detailed level of analysis is likely to be justified only for

case studies or for buildings that are the subject of a research or detailed case-

study or demonstration project.

25
 Chapter 4

BUILDING ENERGY USE MODELING

A model is a pattern, plan, representation, or description designed to show the

structure or the workings of an object, system or concept. It can also be a

simplified representation (usually mathematical) used to explain the workings of a

real-world system or event. (Source: Wikipedia)

Procedures and methodologies to baseline energy use in commercial buildings

began to appear in the 1980’s and early 1990’s. Since then, a number of

modeling toolkits and software have been developed that are useful in

developing performance metrics for buildings, as well as HVAC system

components. Some of these are: the Princeton Scorekeeping Software (PRISM)

(Fels et al., 1995), ASHRAE’s HVAC01 software for modeling primary HVAC

systems such as boilers and chillers (Lebrun et al., 1999), ASHRAE’s HVAC02

software for modeling secondary HVAC systems such as air handlers, blowers,

cooling coils and terminal boxes (Brandemuehl et al., 1993). ASHRAE research

project 1050-RP (Kissock et al., 2001) dealt with creating toolkits for building

energy analysis, e.g., a toolkit for calculating linear, change-point linear and

multiple-linear Inverse building energy analysis models. This chapter first outlines

some of the commonly used methods for building energy use modeling.

The most common approaches and techniques for modeling building energy use

can be classified into:

(i) Data driven analysis approach

(ii) Detailed calibrated simulations approach, and

(iii) Approaches based on artificial intelligence concepts.

26
These have been discussed in Sections 4.1-4.3. Over the past fifteen years or

so, a few researchers have attempted to use short-term monitoring for the

purpose of baselining building energy use. An overview of these studies is

provided in Section 4.4.

4.1. Data driven analysis: inverse, forward and hybrid modeling

a. Inverse modeling has been the mainstay of energy analysts and

researchers throughout the world for many years. The procedure involves using

daily or hourly whole-building energy use to develop a number of feasible

regression models for the building, comparing the results, and then selecting the

best model using appropriate statistical or model performance factors. The most

appropriate statistical indices for model selection factors are usually the model

goodness of fit (R2) and co-efficient of variance of the root mean square error

(CV -RMSE).

The analysis is conducted on the energy response of the building as it relates to

one or more driving forces or parameters. To develop an inverse model, one

must assume a physical configuration of the building or system, and then identify

the parameter of interest using statistical analysis. Two types of inverse models

have been reported in the literature: steady state and dynamic inverse models.

The simplest steady-state inverse models use linear regression to estimate

average behavior, such as average monthly billed utility energy use as a function

of average billing period temperatures. More robust methods include multiple

linear regression, change point linear regression, and variable-base degree day

regressions. Simple steady state inverse models may prove to be insensitive to

dynamic effects (i.e., thermal mass) and other variables (i.e., humidity and solar

27
gain), and are difficult to apply to certain building types; for example, buildings

that have strong on/off schedule dependent loads, or buildings that display

multiple change-points.

b. In Forward modeling, a thermodynamic model of the building is created

using fundamental engineering principles to predict the energy use for 8760

hours of the year, given the location and weather conditions. This requires a

complete description of the building, system, or component of interest, as well as

the physical description of the building geometry, geographical location, system

type, wall insulation, etc.

c. Hybrid modeling combines forward and inverse as used in calibrated

simulations. For example, when a traditional fixed-schematic simulation program

such as DOE-2 is used to simulate energy use of an existing building then one

has a forward analysis method that is being used in an inverse application

(Energy Management Handbook, 2009).

4.2. Detailed calibrated simulations

The process involves using a building simulation program (such as eQuest,

EnergyPlus, etc.) to tune or adjust inputs until the predictions match the

measured data within criteria deemed acceptable. Calibrated simulations are a

powerful tool for savings estimation and for measurement and verification

purposes (Reddy et al., 2007). Calibrated simulations are not within the scope of

this research project.

4.3. Artificial intelligence

Artificial intelligence algorithms have the ability to solve complex non-linear

problems, the ones that were solved in the past using only human intuition and

28
experience. For example, neural networks are artificial intelligence based

methods that attempt to model the working of the human brain (Kreider et al.,

1995). A number of researchers have attempted to use artificial intelligence as a

tool for building energy use prediction.

4.4. Short term to annual extrapolation

There are no absolute rules for determining the minimum acceptable length of

monitored data for the regression model to accurately predict long-term

performance. A full year of energy consumption data is likely to encompass the

entire range of variation of both climate conditions and the different operating

modes of the building HVAC systems. However, in many cases a full year of data

is not available and one is constrained to develop models using less than a full

year of data. The literature shows that only a few studies have attempted to use

short-term monitoring for the purpose of base-lining building energy use.

The first attempts in this field started emerging around 15 years ago. Kissock et

al., (1993) examined the accuracy with which single-variarte standard inverse

temperature-dependent regression models, identified from short data sets, could

be used to predict annual energy use of buildings. Katipamula et al., (1995)

examined the same problem using multi-variate linear regression models. All of

these studies concluded that regression modeling could be accurate and reliable

only when several months (more than six months) of daily data are used to

develop the model. They also noted that excluding the effect of seasonal

variation of the outdoor dry-bulb and dew point temperature in the models

developed from short data sets can cause significant prediction errors.

In an attempt to find how much data is required for hourly regression models for

accurate long term prediction of building energy use, Abushakra (1996) studied

29
the effect of the length of monitoring periods on long-term prediction of energy

consumption of an office building in Montreal., The study used a total of 28

different combinations of regressors to develop 28 different stepwise multiple

linear regression models. The data was also divided into two seasons: heating

and cooling. For each season, each of these 28 models were developed with

one-week, two-week, one-month, two-month, three-month, four-month, five-

month and six-month periods of monitoring. He concluded that NMBE did not

change substantially as one goes beyond a two-week period of monitored data.

A few more recent studies have suggested analysis methods involving a few

weeks of hourly data which provide insight into internal loads and the manner in

which the building is operated, in addition to utility bills, which would capture the

widest range and the annual average weather variables such as dry-bulb

temperature and humidity ratio (Abushakra et al., 2000). Abushakra also

developed a procedure for selecting the two week period of the year that has the

widest range of dry bulb temperature and humidity ratio while capturing the

yearly mean of these two variables.

Tests with synthetic data found that these observations are applicable with 4-P

models as well (Reddy et al., 1998). The best predictors of both cooling and

heating annual energy use are models from datasets with mean temperatures

close to the annual mean temperature and with the range of variation of daily

temperature values in the dataset encompassing as much annual variation as

possible. Thus, one month dataset in spring and fall would frequently be a better

predictor of annual energy than five month data sets from a portion of winter and

summer.

30
4.5. Conclusions from the background literature

Energy rating systems place a lot of emphasis on assessing the actual

performance of buildings. Making building-in-use evaluation mandatory for

obtaining green certification is a significant step in the right direction. Ongoing

monitoring and reporting of buildings is possibly the best way to ensure higher

building performance because it will put to test the strategies and concepts that

are being adopted for the design and planning of high performance buildings.

To allow for the comparison of the performance of one building with another, it is

important that all the buildings be assessed based on a standard set of

guidelines and methods. Comparing the energy use of different high performing

buildings can provide the much needed insight into what concepts really work

when put to use. Extensive research undertaken over the past three decades has

been instrumental in the development of the standardized protocols for the

measurement of building performance under various categories (energy, IEQ,

acoustics, water use). This can be the starting point for a standardized procedure

to be adopted for performance assessment in buildings.

Generally, a full year or more of energy use and weather data is used to

construct empirical models for assessing building energy use. However, in many

cases, either a full year of measured building performance data is not available

or the time & cost implications may not make it viable for the building owner to

invest in monitoring the building for the entire year. In such cases, short-term

monitoring may be considered an alternative.

For short-term in-situ monitoring, the time-interval over which the measurements

are taken is extremely critical., The intent obviously is to gather and analyze data

for as short a period as possible. As a minimum, one would monitor all the
31
necessary variables for atleast one week to capture the diurnal variations of the

weekdays and the weekends. However, one week or so of hourly/daily data

might not be enough for generating acceptable predictions for the whole season

or the whole year.

While many previous studies have attempted to use short-term monitoring for

long-term prediction, it was felt that additional research is needed to investigate

about factors such as: (i) optimum length of monitoring period, (ii) optimum time

or season for monitoring, (iii) necessary variables to monitor, and (iv) effective

and simple modeling techniques that can be easily adopted by practitioners. The

primary research objective of this thesis is to provide insights and

recommendations into some of these factors.

32
 Chapter 5

RESEARCH METHODS

This research attempts to develop simple modeling techniques which allow

identifying the best periods to start in-situ monitoring of the energy use of a

building based on climatic variability, and also to determine the least amount of

building energy use data that would be enough to generate acceptable long term

predictions. Statistical methods have been adopted for analyzing this issue. As

stated earlier, calibrated simulations are not within the scope of this project.

Flowchart in Figure 5.1 summarizes the different sub-categories that have been

identified and considered suitable for this research. Three application areas are

identified. These include:

(i) When detailed audit for investment grade energy conservation measures

(ECM’s) are required,

(ii) When claims made by the newly constructed green or high performance

buildings are to be ascertained, and

(iii) When savings from already installed ECM’s are to be verified against pre-

retrofit claims using pre-post monitored data. The type of analysis method to be

adopted is governed by the type of building performance data available for

analysis.

The analysis procedure that can be adopted for each of these application areas

is governed by the type of data that is available for analysis. Four broad

categories of analysis procedures are proposed and studied as part of this thesis

research: (i) using ambient temperature data only, (ii) using simulated/ monitored

data, (iii) using utility bills and a short monitored dataset, and (iv) using utility bills

only.

33
 Application Areas (Based on Available Data)

Detailed Audit Green Building Verification Pre-Post M&V

No Simulated Utility Bills Pre-Retrofit Pre Retrofit

Monitored Data (1Yr) Monitored Utility Bills
Data Data (1Yr)

Simulated Utility Bills

Data (1Yr)

Figure 5.1 Different sub-categories considered suitable for this research.

5.1. Climate diversity test/ analysis

Past studies have demonstrated that when a model is identified from short-term

data that does not span the entire range of variation of the driving variables (say

outdoor dry bulb temperature), erroneous/misleading predictions can result if

used outside the range (Reddy et al., 1988). Thus, even before one attempts to

develop a meaningful model from the monitored data, the range of associated

climatic data should meet certain criteria.

Abushakra (2000) developed an algorithm that checks for the closeness of the

outdoor dry-bulb temperature and humidity ratio of any consecutive two-week

period of the year to the annual averages, while checking at the same time the

amplitude of its dry-bulb temperature and humidity ratio ranges against the

annual averages. The algorithm allows the ranking of all possible consecutive

two-week periods of the year from best to worst. The method was based on the

findings from the past studies (Kissock et al., 1993), in which it was established

that building load prediction accuracy will be best when models are identified
34
from data periods during which outdoor dry bulb temperature (which is usually

the single most influential driver of building energy use) is closest to the annual

mean and has a large day-to-day variability.

The ‘procedure to select the “best” two-week period’ developed by Abushakra

(2000), termed as the Short Term Monitoring for Long Term Prediction (SMLP)

method forms the starting point for this research. The SMLP method has been

explained in detail in Section 5.2.

5.2. Short - term Monitoring for Long-term Prediction (SMLP)

Abushakra (1999) developed a procedure for selecting the two week period of

the year that has the widest range of dry bulb temperature and humidity ratio

while capturing the yearly mean of these two variables. The same algorithm has

been applied for monthly time intervals to determine the best month of the year

when in-situ monitoring is likely to yield a regression model that is most accurate

in its long term predictions. It is based on using the following screening indices:

1. The yearly average (YA) and yearly range (YR) of outdoor dry bulb
temperature from hourly/ daily values.

2. Monthly average (MiAi) and monthly range (MiRi) of outdoor dry bulb
temperature from hourly/daily values, for all months of the year.

3. For each of the monthly periods obtained above, the outdoor temperature
Average Error , accounts for how close the two week average is to the yearly
average:
AEi = I MiAi – YA I …………….. (Eq. 5.1)
YA

4. The Relative Average Error , accounts for the relative performance of each
period in its closeness to the yearly average:
RAEi = I AEi – AE minI …………….. (Eq. 5.2)
AE min

35
5. The Normalized Error 1 , is meant to avoid differences in the order of
magnitude while adding different errors:
NE1i = RAEi …………….. (Eq. 5.3)
RAEmax

6. Range Error , accounts for the differences between the two-week outdoor dry
bulb temperature range and temperature range:
REi = I MiRi – YR I …………….. (Eq. 5.4)
YR

7. Relative Range Error , accounts for the relative performance of each period in
its coverage to the yearly range:
RREi = I REi – RE minI …………….. (Eq. 5.5)
RE min

8. Normalized Error 2 , is meant to avoid differences in the order of magnitude

while adding different errors:
NE2i = RREi___ …………….. (Eq. 5.6)
RREmax

9. Normalized Average Error, resulting from averaging NE1 and NE2:

NAEi =NE1i + NE2i …………….. (Eq. 5.7)
2
10. Assign ranks. The “best” one-month period for monitoring is that with the
minimum NAE.

The SMLP method has been tested by Abushakra (1999) to show a good

consistency on the degradation of the predictions as one uses models from best

to worst.

5.3. Outdoor Dry-Bulb Temperature Analysis (DBTA)

The SMLP method explained above, although quite accurate, uses a lengthy

procedure to reach the desired output. Since the aim of this research is to

develop easily usable methods for analysis, a simpler and quicker method called

‘Dry-Bulb Temperature Analysis’ (DBTA) for deriving the best periods for in-situ

monitoring has been devised. The analysis is again based on findings by Kissock

et al., (1993) proposed and evaluated building load prediction accuracy being

36
best when models are identified from data periods during which outdoor dry bulb

temperature is closest to the annual mean and has a large day-to-day variability.

The DBTA method uses a sliding window technique to compare the average

outdoor temperatures of the different periods for in-situ monitoring with the

annual average. For example, for the month of January, the average

temperatures for the periods January, January-February, January-March and so

on till January-December are computed and compared to annual average

temperature to determine closeness. The analysis is done for each month of the

year taken as the starting period for monitoring.

Starting with each month of the year, the results obtained are analyzed to

determine how many months of temperature data are needed for the average for

the period to reach the annual average. The values obtained are visually

translated into a graphical format for easy comprehension. The DBTA method

provides a way of ranking the time periods (based on temperature data only) in

terms of the expected predictive accuracy of the regression models.

5.4. Multivariate Change Point (MCP) Modeling

Prior to describing the modeling technique adopted, an insight into regression

based modeling and an overview of the software that has been used for analysis

is appropriate for better comprehension.

5.4.1. Regression based models

Regression analysis is a statistical technique used to relate variables. The basic

aim is to build a mathematical model to relate dependent variables to

independent variables. In general, a regression model will be defined as a single

algebraic equation of the form (Draper and Smith, 1981)

37
 Z = f(X1,X2,……..Xk) + u …………….. (Eq. 5.8)

where, Z is a variable whose movements and values may be described or

explained by the variables X1,X2,……..Xk . The letters are known as regressors

and assumed to have a causal relationship to the dependent variable Z. The

additional term u is a random variable, which is included to account for the fact

that movements in Z are not completely explained by the variables.

In building energy study, the building energy consumption is considered to be a

dependent variable, and the other parameters such as weather and non- weather

data are taken as independent variables. Three kinds of regression models have

been developed, namely;

i. Variable-based degree-day model,

ii. Linear regression model (single-variate and multi-variate), and

iii. Change-point models. (Kissock et al., 2003)

When only one independent variable (mean temperature) is used in the

development of the model, the model is called single-variate linear regression

model. When there are more than one independent variables is used in the

development of the model, the regression model is called a multiple or a

multivariate linear regression model.

Generally, there exists a non-linear relationship between heating and cooling

energy use and the ambient temperature caused by system effects. The Change-

point models (Figure 5.2) are able to successfully capture this non-linear

relationship. Four basic types of typical change-point models can be identified:

i. Three-parameter (3P) heating energy use model,

ii. Three-parameter (3P) cooling energy use model,

iii. Four-parameter (4P) heating energy use model, and

38
 iv. Four-parameter (4P) cooling energy use model.

Figure 5.2 Typical Change Point Models (a) three-parameter (3P) heating energy use, (b)

three-parameter (3P) cooling energy use, (c) four-parameter (4P) heating energy use

model, and (d) four-parameter (4P) cooling energy use model. (Source: Energy

Management Handbook, 2009)

a. Three Parameter (3P) Model: When energy use appears to be linearly

correlated with an independent variable over part of the range of variation of the

independent variable and has another type of variation over the other part, 3P

multivariate change point regression models are appropriate. They are of the

form:

Y = b1 + b2 ( X1 - b3 )+ + b4 X2 + b5 X3 …………….. (Eq. 5.8)

where, b1 is the y-coordinate of the change point, b2 is the slope term, b3 is the x-

coordinate of the change-point, and b4 - b5 are regression coefficients of the

optional independent variables X2 through X4. The ( )+ and ( )- notations indicate

39
that the values of the parenthetic term shall be set to zero when they are

negative and positive respectively.

b. Four Parameter (4P) Model: The 4P model consists of two linear

segments joined at a change-point. When energy use appears to be non-linearly

correlated with an independent variable, a 4P model may provide a good fit to the

data. Typically, 4P Multivariate Regression models can have up to three

independent variables, and are of the form:

Y = b1 + b2 (X1 - b4)- + b3 ( X1 - b4 )+ + b5 X2 + b6 X3 …………….. (Eq. 5.9)

where, b1 is the y-coordinate of the change point, b2 is the left slope, b3 is the

right slope, b4 is the x-coordinate of the change point, and b5 - b6 are regression

coefficients of the optional independent variables X2 and X3. The ( )+ and ( )-

notations indicate that the values of the parenthetic term shall be set to zero

when they are negative and positive respectively.

5.4.2. Comparison of Energy Explorer software with Inverse Modeling

Toolkit

In 1994, ASHRAE began developing the Guideline GPC-14P, a guideline for

measuring retrofit savings. In support of Guideline-14P, ASHRAE initiated RP-

1050 (Kissock et al., 2001) to develop a toolkit for calculating linear, change-point

linear, and multiple linear inverse energy models. The toolkit was named the

‘Inverse Modeling Toolkit’ (IMT). The IMT was developed as a FORTRAN90

application for developing regression models of building energy use. The toolkit

can be used to identify single and multi-variable least-squares regression

models, variable-base degree-day and single and multi-variable change-point

models.

40
Kissock (2008) recently developed the ‘Energy Explorer’ tool for analyzing

building and facility energy-use data. It integrates the laborious tasks of data

processing, graphing and statistical modeling through a user-friendly, graphical

interface. It allows the user to determine baseline energy use, understand factors

that influence energy use, calculate retrofit savings and identify operational and

maintenance problems.

Energy Explorer (EE) includes a full package of statistical models specifically

designed for analyzing building and facility energy use. Models include mean,

median, simple and multiple-linear regression. In addition, specially-developed

2, 3, 4 and 5-parameter change-point models allow the user to precisely and

easily quantify relationships between building energy use, weather and other

energy drivers. Change-point models accurately model the non-linear energy

use patterns characteristic of whole-building Electric, steam, heat-pump, and

cooling energy use data. Modeling results are displayed numerically and

graphically to facilitate a quick and complete understanding of the model and it’s

fit to the data. In addition, retrofit savings and energy breakdowns can be

calculated from the regression models.

Since IMT is the standard software recognized by ASHRAE for modeling change

point behavior of energy use in buildings, a partial validation of the results

obtained from the Energy Explorer Tool against the output from the IMT was

done to provide credibility to the analysis results (Endurthy, 2010). The objective

of this exercise was to:

i. Compare differences in linear & change point model parameters when

both IMT and EE programs are applied to the same dataset, and

41
ii. Compare prediction accuracies of models identified from IMT and EE

programs when applied to the same dataset.

The year-long synthetic dataset of daily energy use values for the large hotel

located in Chicago with three available energy use channels; namely, whole

building electric (WBE), cooling energy use (CHW) and heating energy use (HW)

were used for this analysis. The two regressor variables used are outdoor dry-

bulb temperature (DBT) and internal electric loads (LTEQ).

Table 5.1

Model equations for different base periods identified using EE software and IMT

Base Period
used to
Response Model EE Model IMT Model
identify
model
Model Equation Model Equation

63.35+5.77(DBT- 62.56+5.78(DBT-
WBE 3P +
43.46) +1.03LTEQ 43.22)++1.02LTEQ

October 149.69+108.23(DBT- 141.67+109.98(DBT-

CHW 3P
Only 50.83)++0.65LTEQ 50.98)++0.67LTEQ

2825.30+79.69(66.02-DBT)+- 2793.85+79.88(66.49-DBT)+-
HW 3P
2.11LTEQ 2.12LTEQ
45.35+0.37(40- 44.64+0.38(40.01-
WBE 4P DBT)++6.16(DBT- +
DBT) +6.21(DBT-
40)++1.01LTEQ 40.01)++1.01LTEQ
288.32+1.43(45.04- 308.7+0.63(45.79-
DBT)++90.72(DBT- +
Oct- Jan CHW 4P DBT) +95.04(DBT-
45.04)++0.12LTEQ 45.79)++0.11LTEQ
3381.85+201.85(56.99- 3555.42+202.63(55.89-
HW 4P DBT)+-72.69(DBT-56.99)+- DBT)+-86.22(DBT-55.89)+-
1.88LTEQ 1.85LTEQ
54.57-0.12(43.25- 56.58-0.22(43.96-
WBE 4P DBT)++7.12(DBT- +
DBT) +7.32(DBT-
43.25)++1.01LTEQ 43.96)++1.01LTEQ
413.86-3.94(49.59- 440.72-4.89(50.06-
DBT)++128.22(DBT- +
Oct- May CHW 4P DBT) +131.31(DBT-
49.59)++0.10 LTEQ 50.06)++0.10LTEQ
+
2818.98+205.34(59.58- 3444.19+207(56.15-DBT) -
DBT)+-67.66(DBT-59.58)+- +
HW 4P 107.07(DBT-56.15) -
1.60LTEQ 1.56LTEQ
Table 5.1 assembles the models identified using IMT and EE software for all

three energy use channels. Four different base periods have been used to

42
identify the models. The ‘October only’ period implies that daily data for the

month of October only was used to identify the model, while ‘October-January’

indicates that data from all four months were used for model identification. There

are small differences in model parameters between IMT and EE software, the

differences are small. This partially validates the use of EE software.The better

test is to gauge differences in CV and NMBE between both models when used

for prediction. The linear & change point model outputs from EE and IMT

programs are summarized in Table 5.2. Comparison is based on the variation on

CV-RMSE (%) and NMBE (%) for predictions for each of the channels of various

base periods.

Table 5.2

Prediction results from EE and IMT software.

Response
EE Model IMT Model
Base Period used for model variable
identification
NMBE CV NMBE
CV (%)
(%) (%) (%)
WBE 6.46 -1.62 6.65 -2.43
October CHW 70.00 -30.32 70.31 -29.67
HW 68.59 -38.24 68.48 -38.49
WBE 6.30 -1.60 6.30 -3.17
Oct- Jan CHW 74.80 -38.70 73.42 -37.26
HW 24.20 -3.20 24.32 -1.80
WBE 5.50 -2.60 5.41 -2.52
Oct - May CHW 61.80 -30.90 61.13 -30.82
HW 23.30 -2.00 24.05 0.79
WBE 3.10 0.00 3.44 0.88
Oct- Sep CHW 24.10 -0.30 24.13 -0.22
HW 22.80 0.00 22.76 -0.04
Figure 5.3 and Figure 5.4 shows comparison of prediction accuracies of models

identified from IMT and EE software when applied to the same dataset. The

results for both were found to be consistent and generally very close. The

information from Table 5.2 is plotted in these figures for easier comprehension.
43
 10.00 80.00
Oct-Jan
Oct-Jan
9.00 October
70.00
8.00
60.00 Oct-May
7.00
October
October Oct-Jan 50.00
6.00

CV(%)
CV(%)
Oct-May
5.00 40.00

4.00 30.00
Oct-Sep
3.00 Oct-Sep Oct-Sep
20.00
2.00 WBE CHW
10.00
1.00

0.00 0.00

0 1 2 3 4 5 0 1 2 3 4 5
Base Period Base Period

100.00

90.00

80.00

70.00 Oct ober

60.00
CV(%)

50.00

40.00

30.00
Oct-Jan Oct-M ay Oct -Sep
20.00
HW EE
10.00
IMT
0.00
0 1 2 3 4 5
B ase Per i o d

Figure 5.3 Prediction accuracy- CV (%) of models identified from IMT and EE software.
2.00 10.00

1.00 0.00 Oct-Sep

Oct-Sep
0 1 2 3 4 5

0.00 -10.00
0 1 2 3 4 Oct-Sep 5

-1.00 -20.00

October Oct -Jan

-2.00 -30.00 October
October Oct-M ay
October Oct-M ay
ay
Oct-M Oct-Jan
Oct-Jan
-3.00 -40.00
Oct -Jan
WBE CHW
-4.00 -50.00
B ase Per i o d B ase Per io d
5.00

0.00 Oct-May Oct-Sep

Oct-Jan Oct-May
0 1 2 Oct-Jan 3 4 5
-5.00

-10.00

-15.00

-20.00

-25.00

-30.00
EE
-35.00

-40.00
October
HW IMT
-45.00
Base Period

Figure 5.4 Prediction accuracy-NMBE(%) of models identified from IMT and EE software.

44
5.4.3. MCP modeling technique adopted

The analysis approach is to predict whole building electric (WBE), cooling energy

use (CHW), and heating energy use (HW) using a multivariate change point

(MCP) model derived from the short data set. The MCP models for the three

energy-use channels are generated using daily averages for outdoor dry bulb

temperature (DBT) and lighting & equipment load (LTEQ) as the regressors

using Energy Explorer software. 2-P, 3-P & 4-P change-point models were

evaluated of which 4-P turned out to be the best, and so this was used in all

subsequent analysis. The MCP model generated has the form:

Ei = A + B (X1 -DBT) + + C(DBT - X1) + D(LTEQ) ………….. (Eq. 5.10)

Where, Ei is the energy use, X1 is the x-coordinate of the change point for

outdoor dry bulb temperature (DBT) and internal loads (LTEQ). A, B, C & D are

regression coefficients. ( )+ notation indicates that the value of the parenthetic

term shall be set to zero when it is negative.

The accuracy of the model largely depends on the starting date and the end date

of data collection. The start and the end date are influential in that the data within

this period should capture the yearly fluctuations in temperature and humidity. In

order to determine the best starting month and duration for in-situ monitoring, a

systematic approach has been adopted. The models are generated for each

starting month of the year (January to December), with each selection

subsequently expanded in increments of one month to mimic different durations

of monitoring. For example, for the starting month of January, the first model is

generated using the data for January only, which is then subsequently increased

in increments of one month i.e. January-February, January-March, January-April

and so on until the whole year of data is used for generating the model. The

45
same process is repeated for each month of the year taken as the start period of

in-situ measurement.

To determine the accuracy of the models derived from the short data-sets, the

values of annual energy use predicted by models obtained from short data-sets

are compared to the actual energy use in the original dataset. The predictive

accuracy of the models is evaluated based on two statistical indices: Coefficient

of Variation of the Root Mean Square Error2or CVRMSE (%) (Written as CV (%) in

this report), and the Normalized Mean Bias Error or NMBE (%) defined as:

………….. (Eq. 5.11)

………….. (Eq. 5.12)

where,
yActual,i the actual value of the dependent variable corresponding to a
particular set of values of the independent variables
yPredicted,I the predicted dependent variable value for the same set of
independent variables
y Actual the mean value of the dependent variable of the actual data set
n the number of data points in the actual data set
p the total number of regression parameters in the model.

Since the CV (%) is calculated as the ratio of the root mean squared error

(RMSE) to the mean of the dependent variable, it describes the model fit in terms

of the relative sizes of the squared residuals and mean outcome values. Lower

CV (%) implies smaller residuals relative to the predicted value. NMBE (%),

often simply stated as ‘bias error’ refers to how far the average statistic lies from
46
the parameter it is estimating, i.e., the biased error which arises when estimating

a quantity. Thus, low CV (%) and NMBE (%) values are indicative of a good

model fit.

When analyzing the results of the regression models for WBE, CHW and HW, it

is necessary to understand the difference between the importance of CV (%) and

NMBE (%) for the three cases. For WBE, the demand is more important, while

for CHW and HW, the consumption is more important. The CV (%) delivers more

information about the demand accuracy than does NMBE (%). A model that has

high CV (%) value cannot be used to estimate a demand value. Utility companies

often charge larger energy consumers by demand use (i.e., the maximum hourly

use during the given month) along with the consumption. CHW and HW are

usually analyzed by consumption. A model that has a low NMBE (%) can

accurately predict the total consumption even if the CV (%) is high. Thus for

WBE, a model with lower CV (%) is deemed to be better, and for CHW and HW,

a model with lower NMBE (%) is the better choice.

As already mentioned, Energy Explorer software is used to generate the model

equations, a MATLAB code (Endurthy, 2011) has been written to calculate

external CV & NMBEs for various model equations generated. The Flowchart of

the modeling technique adopted is shown in Figure 5.5.

47
 Hourly Data Analyze daily average DBT
data in terms of the SMLP
Method and the Simplified
Dry- Bulb Temperature (DBT),
DBTA
Energy Whole Building (WBE),
Light and Equipment (LTEQ),
Choose data for the starting
Cooling Energy Use (CHW), month and length of period
Heating Energy Use (HW) to be used for identifying the
model
Humidity Potential (w)

Identify Multivariate Change

Compute Daily Averages Point Regression models for
the three energy-use
channels (WBE, CHW &
Dry- Bulb Temperature (DBT), HW) using outdoor dry bulb
Pass into
Energy Whole Building (WBE), temperature (DBT) and
Energy
lighting & equipment load
Light and Equipment (LTEQ), Explorer
(LTEQ) as the regressors.
Software
Cooling Energy Use (CHW),
Heating Energy Use (HW)
Humidity Potential (w)
Predict WBE/CHW/HW for
each day of the year based
on models derived in Energy
Explorer software

Pass into MATLAB to

calculate NMBE(%) & CV(%)
for models.

Repeat for all Base Periods

Figure 5.5 Flowchart for Multivariate Change Point (MCP) modeling approach.

48
5.5. Hybrid Inverse Modeling

5.5.1. Method1: utility bills + monitored data

This approach for predicting energy use combines the monitored daily energy

use and internal loads with atleast one year of recent utility bills (representing the

long-term data) to predict the building energy performance for the whole year

(Abushakra et al.,1999). Thus, in this modeling technique, information from

monthly utility bills is applied to the model. The utility bills are accurate and easily

acceptable sources of building energy information. In order to minimize the

confounding effects of co-linearity between regression variables, the regression

is completed in two stages (Abushakra, 2000).The first stage of the regression

model has the form: B () + + C()

Ek=a + b (X1 -DBTk ) + c(DBTk - X1) + d(wk-0.009)+ where, k=1 to 12

(indicating the 12 months) ………….. (Eq. 5.13)

where, E is the monthly energy use, DBTk is average monthly temperature, and

(wk-0.009)+ is the average specific humidity potential.,

The second stage uses the daily equipment and lighting loads (LTEQ), which

also take into account the occupancy patterns for the building, so as to create a

hybrid model. The coefficients d and e are found by regressing following:

Ek - b (X1 -DBTk ) - c(DBTk - X1) - d(wk-0.009)+ =

e+ f(LTEQ) ………….. (Eq.5.14)

Finally, the model equations obtained in stage one and two are combined to

derive the final model for predicting energy use. This equation has the form:

Ei=d + b(X1 -DBTk ) + c(DBTk - X1) + d(wk-0.009)+ e (LTEQ) ………….. (Eq. 5.15)

Abushakra developed the hybrid models by combining monthly utility bills along

with monitored hourly data. He analyzed all consecutive two-week periods of the

year and ranked them from “best” to “worst” in terms of their prediction accuracy.
49
In this research, the analysis has been limited to daily timescales instead of

hourly. Since the research aims at finding the shortest period suitable for in-situ

energy use monitoring, the length of the period for this analysis is limited to a

maximum of three consecutive months. Thus, models are generated for each

starting month of the year (January to December), with each selection

subsequently expanded in increments of one month to a maximum of three

consecutive months to mimic different durations of monitoring. For example, for

the starting month of January, the first model is generated using the data for

January only and then subsequently increased in increments of one month i.e.

January-February and January- February-March. The same is repeated for each

starting month of the year.

The NMBE (%) and CV (%) (Section 5.4) indices estimate the predictive

accuracy of the results from the regression models generated. The time plots of

CV (%) and NMBE (%) for each type of energy use and for different lengths of

monitoring periods allow easier understanding of the prediction patterns.

5.5.2. Method 2: utility bills only

Method 2 looks at predicting building performance at daily timescales using utility

bills only. The models generated have the form:

Ek=a + b (X1 -DBTk ) + c(DBTk - X1) + d(wk-0.009)+ where, k=1 to 12 (indicating

the 12 months) …….. (Eq. 5.16)

This equation is used for predicting energy use at daily time scales. The NMBE

(%) and CV (%) (Refer to Section 5.4) statistical parameters estimate the

predictive accuracy of the results from the regression models generated. The

results are shown at monthly as well as annual time scales. This helps evaluate

50
the effectiveness of using utility bills in predicting the energy use for each month

of the year as well as annually.

The Flowchart of the hybrid inverse modeling technique (method 1) is shown in

Figure 5.6 below:

Monthly Utility Bill Data

One Year of Utility Data
(WBE, CHW, HW)

Monthly Average Outdoor Choose Length of Regression

Dry-Bulb Temperature Period
1Month – 12 Months
Daily Measured Weather Data

Pass into Energy Explorer Stage 2 - Regression

Software
Regress Daily Residuals against
Stage 1 - 3P MCP Model
daily monitored LTEQ
Ek - bDBTk - c(wk-0.009)+ = d + e(LTEQ)
Ek=a + Ek=a + b (X1 -DBTk ) +
c(DBTk - X1) + d(wk-0.009)+

k=1 to 12 (indicating the 12

months) Use Equation
Ei=d + bDBTk + c(wk-0.009)+ e (LTEQ)

To predict entire year

LTEQ
(Weekend / Weekday)

Figure5.6 Flowchart for Hybrid Inverse Modeling approach (Method 1).

51
 Chapter 6

TEMPERATURE DATA ANALYSIS RESULTS

6.1. Datasets

The methodology described in Chapter 5 has been applied to three buildings

(two synthetic and one actual) for which a full year of data was available for

analysis. Table 6.1 summarizes the key features of buildings chosen for analysis.

Detailed descriptions of these buildings are provided in Appendix A.

Table 6.1

Summary of buildings chosen for analysis.

Building Summary
Area Actual(A) /
No Building Description Data Channels
(Sqft) Synthetic(S)
Response
Regressor
(Energy)
Large Hotel, Chicago IL WBE, DBT,
1 619,200 S
(06/06-05/07 Data) CHW, HW LTEQ

Office Building,
DBT,
2 Albuquerque, NM 17,430 S WBE, HW
LTEQ
(2004 Data)
Full Service Hotel
3 Washington DC Region 212,000 A WBE DBT
(2009 Data)

Three energy use channels are considered for analysis, namely, whole building

electric (WBE), cooling energy use (CHW), and heating energy use (HW). The

analysis has been done at daily timescales. The variation in NMBE (%) & CV (%)

for predictive accuracy of these three energy channels when different in-situ

monitoring periods are selected for model building forms the basis of evaluation.

Following sections summarize the analysis results obtained for each of the three

buildings analyzed.

52
6.2. Climate diversity analysis

Outdoor dry-bulb temperature is usually the single most influential driver of

building energy use. Figures 6.1 to 6.3 show the variation of the outdoor dry-bulb

temperature for the three different locations selected. The dots indicate the

monthly mean temperatures and the vertical whiskers represent the monthly

temperature range i.e., range of daily temperature values. The annual average

temperature and its bands of ±10% variation (represented by the dashed

rectangular area) for each location has also been plotted. The range can be said

to represent the swing season for the location. In most cases it is within this

temperature range that the “change point” is expected to occur due to a change

in the season. This range is therefore also referred to as the ‘change-point’

range.

Chicago, Illinois
120
Annual Avg Temp: 49.44 °F
100
Temperature (Deg F)

80

60

40

20

0
1 2 3 4 5 6 7 8 9 10 11 12
Month of Year

Figure 6.1 Annual variation of outdoor dry bulb temperature for Chicago, IL. Average

monthly temperature, monthly temperature range and average annual temperature

(49.44°F) with bands of ±10% variation shown.

53
 Albuquerque, New Mexico
120
Annual Avg Temp: 55.75 °F
100

Temperature (Deg F)
80

60

40

20

0
1 2 3 4 5 6 7 8 9 10 11 12
Month of Year

Figure 6.2 Annual variation of outdoor dry-bulb temperature for Albuquerque, IL. Average

monthly temperature, monthly temperature range and average annual temperature

(55.75°F) with bands of ±10% variation shown.

Washington DC Region
120
Annual Avg Temp: 56.91 °F
110
100
90
Temperature (Deg F)

80
70
60
50
40
30
20
10
0
1 2 3 4 5 6 7 8 9 10 11 12
Month of Year

Figure 6.3 Annual variation of outdoor dry-bulb temperature for Washington, D.C.

Average monthly temperature, monthly temperature range and average annual

temperature (56.91°F) with bands of ±10% variation shown.

54
The circles on the x-axis highlight the months that are expected to be the best for

in-situ monitoring (due to their monthly mean temperatures being closest to the

annual average and large day-to-day variability). This investigation is just a visual

assessment of the temperature data and needs to be further verified by analysis.

The results are summarized in Table 6.2.

Table 6.2

Variation in outdoor dry-bulb temperature for Chicago, Albuquerque & Washington D.C.

Annual Average Expected Best

Annual Temperature
Location Temperature Months for In-Situ
Range Max-Min (°F)
(°F) Monitoring

Chicago, Illinois 49.44 83.67 – (-3.29) = 86.96 April, October

Albuquerque,
55.75 83.15 - 25.09 = 58.07 April, October
New Mexico

Washington DC March, April,

56.91 77.75 - 31.19 = 46.56
Region October

55
6.3. Results of the SMLP method

The climate data is analyzed to determine the predicted best month of the year

for data monitoring based on the SMLP procedure described in Section 5.2.

a. Results for Chicago, Illinois

The criterion for the SMLP method has been discussed earlier in Section 5.2.

The method gives an insight into time of the year when in-situ monitoring is likely

to yield a regression model that is most accurate in its long term predictions.

Figure 6.4 shows the results of the SMLP analysis for Chicago, IL. The ranking

obtained for all months of the year is represented graphically. The three best

months for in-situ monitoring (ranks 1-3) are indicated in Figure 6.4. The ranking

(1 to 12), obtained from the algorithm, for all the months of the year is assembled

in Table 6.3.

Selection of Best Month for Building Monitoring

from Temperature Data for Chicago, IL
1.2
Normalized Average Error

1.0

0.8

0.6

0.4
3 2
0.2 1

0.0
November

December
January

August

October
September
February

July
May
March

April

June

Figure 6.4 SMLP analysis for Chicago, IL

56
 b. Results for Albuquerque, New Mexico

Figure 6.5 shows the results of the SMLP analysis when applied to Albuquerque,

NM. The best months for in-situ monitoring (ranks 1-3) are indicated. The ranking

(1 to 12) obtained from the algorithm, for all the months of the year, is provided in

Table 6.3.

Selection of Best Month for Building Monitoring

from Temperature Data for Albuquerque, NM
1.2
Normalized Average Error

1.0

0.8

0.6

0.4
3
0.2 2
1
0.0
January

August
June
February

July
April

November

December
September

October
May
March

Figure 6.5 SMLP Analysis for Albuquerque, NM

c. Results for Washington DC Area

Figure 6.6 shows the results of the SMLP analysis when applied to the

Washington DC Region. The best months for in-situ monitoring (ranks 1-3) are

indicated. The ranking (1 to 12) obtained from the algorithm, for all the months of

the year, is provided in Table 6.3.

Selection of Best Month for Building Monitoring
from Temperature Data for WashingtonDC
1.0
0.9
Normalized Average Error

0.8
0.7
0.6
0.5
0.4
0.3
1
0.2 3
2
0.1
0.0
January

August
June
February

July
April

November

December
September

October
May
March

Figure 6.6 SMLP Analysis for Washington DC Region.

57
Table 6.3

Ranking obtained from the SMLP Algorithm for each month of the year for Chicago,

Albuquerque and Washington D.C.

Ranking Based on SMLP Analysis

Rank Chicago , IL Albuquerque, NM Washington D.C. Area
1 October October April
2 November April March
3 April November October
4 December March November
5 March May December
6 May February February
7 January June May
8 February January September
9 September August June
10 June December January
11 August September August
12 July July July

58
6.4. Results of the DBTA method

The temperature data has been analyzed to determine time periods expected to

be the best for in-situ monitoring based on the procedure described in Section

5.3,.

The numbers 1-12 on the x-axis represent increase in the monitoring periods in

successive increments of a month, i.e., sliding window lengths for each month of

the year taken as the starting period for monitoring. Each month is ranked based

on the length of period required to reach the yearly average temperature value.

The results for all the three locations are shown graphically in Figure 6.7 to

Figure 6.9 and in a tabular format in Table 6.4.

Temperature Analysis : Chicago,IL January

80 February
March
70
Temperature (deg F)

April
60
May
50
June
40
July
30
August
20 September
10 October
0 November
1 2 3 4 5 6 7 8 9 10 11 12 December
Number of Consecutive Months in Dataset Annual Average

Figure 6.7 Outdoor Dry-Bulb Temperature Analysis: Chicago, IL. Annual

average temperature: 49.44 °F

59
 Temperature Analysis: Albuquerque, NM January
90 February
80 March
Temperature (deg F)

70 April
60 May
50 June
40 July
30 August
20 September
10 October
0 November
1 2 3 4 5 6 7 8 9 10 11 12
December
Number of Consecutive Months in Dataset Annual Average

Figure 6.8 Outdoor Dry-Bulb Temperature Analysis: Albuquerque, NM. Annual average

temperature: 55.75 °F

Temperature Analysis: Washington DC January

90 February
80 March
Temperature (deg F)

70 April
60 May
50 June
40 July
30 August
20 September
10 October
0 November
1 2 3 4 5 6 7 8 9 10 11 12
December
Number of Consecutive Months in Dataset Annual Average

Figure 6.9 Outdoor Dry-Bulb Temperature Analysis: Washington DC. Annual average

temperature: 56.91 °F

60
Table 6.4

Summary of ranking based on DBTA method for each month of the Year for Chicago,

Albuquerque and Washington D.C.

Ranking is based on the number of months required to come close to yearly average temperature
Chicago, Illinois Albuquerque, New Mexico Washington DC Region

Start Month No. Of No. Of

No. Of Months
Months of Months of
End Month of Data Ranking End Month Ranking End Month Ranking
Data Data
Required
Required Required
January July 7 7 July 7 7 July-August 7.5 7
February June-July 6 6 June 5 5 June 5 5
March May-June 3 3 May 3 3 May-June 3.5 3
April April-May 2 1 April 1 1 April 1 1
May March 11 11 February 10 11 Feb-March 10.5 11
June February 9 9 Jan-Feb 8.5 9 Jan-Feb 8.5 9
July January 7 7 January 7 7 Jan-Feb 7.5 7
August December 5 5 December 5 5 Dec-Jan 5.5 5
September Nov-Dec 3 3 November 3 3 Nov-Dec 3.5 3
October November 2 1 October 1 1 Oct-Nov 1.5 1
November September 11 11 September 11 12 September 11 12
December Aug-Sept 9 10 August 9 10 August 9 10

6.5. Comparison of results of SMLP and DBTA methods

This section investigates into the consistency between the rankings obtained

from the SMLP Algorithm and the DBTA methods. The SMLP rankings are based

on the algorithm developed by Abushakra (1999) which requires several indices

to rank the different months of year based on the closeness of the monthly mean

temperatures to the annual average. The DBTA, on the other hand, ranks the

months based on the length of data required to reach the annual average

temperatures for each starting month. While the rankings obtained are not

exactly the same, they are somewhat related, and consistency between them is

clearly evident from the Figure 6.10 to Figure 6.12. The analysis methods clearly

indicate March, April and October to be the best periods with the other months

with extremes in weather conditions ranked poorly in both analysis methods. In

Chapter 7 the rankings obtained from the DBTA are analyzed further against

actual modeling of energy-use data using the MCP modeling explained in

Chapter 5.

61
 Result Comparison : SMLP & DBTA for Chicago, IL
14

12
Ranking ( Best to Worst) 10

August
July
June
January

November

December
September
February

May

October
April
March

Chicago SMLP Ranking Chicago DBTA Ranking

Figure 6.10 Result comparison for SMLP and DBTA for Chicago, IL.

Result Comparison : SMLP & DBTA for Albuquerque,NM

14

12
Ranking ( Best to Worst)

10

0
August
July
June
January

November

December
September
February

May

October
April
March

Albuquerque SMLP Ranking Albuquerque DBTA Ranking

Figure 6.11 Result comparison of SMLP and DBTA for Albuquerque, NM.

Result Comparison : SMLP & DBTA for Washington DC

14

12
Ranking ( Best to Worst)

10

0
August
July
June
January

November

December
September
February

May

October
April
March

East Coast SMLP Ranking East Coast DBTA Ranking

Figure 6.12 Result comparison of SMLP and DBTA for Washington DC.

62
 Chapter 7

MCP MODELING RESULTS

7.1. Multivariate change point modeling approach

The results obtained from the methods in Chapter 6 are based only on the

temperature data. It is important to verify these rankings with actual modeling of

the energy use data.

This section assembles the results of the MCP modeling approach explained in

Section 5.4. The MCP models for the three energy-use channels, namely, whole

building electric (WBE), cooling energy use (CHW), and heating energy use

(HW) are generated using daily averages for outdoor dry bulb temperature

(DBT) and lighting & equipment load (LTEQ) as the regressors.

The models are generated for each starting month of the year (January to

December), with each selection subsequently expanded in increments of one

month to mimic different durations of monitoring.

To determine the accuracy of the models derived from the short data-sets, the

values of annual energy use predicted by models from short data-sets are

compared to the actual energy use. The predictive accuracy of the models is

evaluated based on two statistical indices: Coefficient of Variation of the Root

Mean Square Error or CV (%), and the Normalized Mean Bias Error or NMBE

(%). The results are presented graphically below.

(Refer to Appendix C for all regression model equations and corresponding CV

(%) and NMBE (%) results for the three buildings).

63
7.2. Results for the large hotel - Chicago, Illinois

The time series plots of CV (%) and NMBE (%) of each type of energy use

channels (WBE, CHW & HW) and for different lengths of monitoring are

displayed in Figure 7.1 and Figure 7.2 respectively. For the sake of clarity, the

graphs for each energy channel have been split into two. The graphs on the left

assemble results when starting periods for monitoring are from January to June.

The ones on the right have monitoring periods starting from July to December.

Numbers 1-12 on the x-axis denote the length in months as the monitoring period

is increased in successive increments of a month.

CV(%):Whole Building Energy Use CV(%):Whole Building Energy Use
(Starting Months January to June) (Starting Months July to December)
100 100

80 80

60 60

40 40

20 20

0 0
1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12
Number of Consecutive Months in Dataset Number of Consecutive Months in Dataset
January February March April May June July August September October November December

CV(%) : Cooling Thermal Use CV(%) : Cooling Thermal Use

(Starting Months January to June) (Starting Months July to December)
100 100

80 80

60 60

40 40

20 20

0 0
1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12
Number of Consecutive Months in Dataset Number of Consecutive Months in Dataset
January February March April May June July August September October November December

CV(%) : Heating Thermal Use CV(%) : Heating Thermal Use

(Starting Months January to June) (Starting Months July to December)
100 100

80 80

60 60

40 40

20 20

0 0
1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12
Number of Consecutive Months in Dataset Number of Consecutive Months in Dataset
January February March April May June July August September October November December

Figure 7.1 CV (%) for energy channels WBE, CHW & HW when different lengths of

monitoring are used for predicting annual energy-use for the Large Hotel at Chicago, IL.

64
 NMBE(%):Whole Building Energy Use NMBE(%):Whole Building Energy Use
(Starting Months January to June) (Starting Months July to December)
100 100
80 80
60 60
40 40
20 20
0 0
-20 1 2 3 4 5 6 7 8 9 10 11 12 -20 1 2 3 4 5 6 7 8 9 10 11 12
-40 -40
-60 -60
-80 -80
-100 -100
Number of Consecutive Months in Dataset Number of Consecutive Months in Dataset
January February March April May June July August September October November December

NMBE(%) : Cooling Thermal Use NMBE(%) : Cooling Thermal Use

(Starting Months January to June) (Starting Months July to December)
100 100
80 80
60 60
40 40
20 20
0 0
-20 1 2 3 4 5 6 7 8 9 10 11 12 -20 1 2 3 4 5 6 7 8 9 10 11 12
-40 -40
-60 -60
-80 -80
-100 -100
Number of Consecutive Months in Dataset Number of Consecutive Months in Dataset

January February March April May June July August September October November December

NMBE(%) : HeatingThermal Use NMBE(%) : Heating Thermal Use

(Starting Months January to June) (Starting Months July to December)
100 100

80 80

60 60

40 40

20 20

0 0

-20 1 2 3 4 5 6 7 8 9 10 11 12 ‐20 1 2 3 4 5 6 7 8 9 10 11 12

-40 ‐40

-60 ‐60

-80 ‐80

-100 ‐100
Number of Consecutive Months in Dataset Number of Consecutive Months in Dataset

January February March April May June July August September October November December

Figure 7.2 NMBE (%) for energy channels WBE, CHW & HW when different lengths of

monitoring are used for predicting annual energy-use for the Large Hotel at Chicago, IL

7.2.1. Results for Whole Building Electric for Large Hotel, Chicago, IL

Figure 7.3 below shows results for WBE from Figures 7.1-7.2 in more detail. All

the twelve starting months have been combined into one figure. The graph for

CV(%) shows all the monitoring periods that predicted WBE within 10% of the

actual energy use while the graph for NMBE (%) shows monitoring periods within

a bias of ± 20%. Clearly, November, March and October are the best months to

start in-situ monitoring with energy use prediction varying from annual energy

use by only 5.2%, 5.3% and 6.4% respectively. The average interval prediction
65
errors of models identified from one month of data only are 2.65%, 3.01% and

3.72%, respectively, for these three months.

CV(%):Whole Building Electric (WBE)

January
10
February
March
8
April
May
6
CV (%)

June
July
4
August
2 September
October
0 November
1 2 3 4 5 6 7 8 9 10 11 12 December
Number of Consecutive Months in Dataset

NMBE(%):Whole Building Electric (WBE)

20 January
February
March
10
April
May
NMBE(%)

0 June
1 2 3 4 5 6 7 8 9 10 11 12 July
August
-10
September
October
-20 November
Number of Consecutive Months in Dataset December

Figure 7.3 Variation in CV (%) and NMBE (%) for WBE. Graphs show monitoring periods

for which predicted CV (%) is within 10% and predicted NMBE (%) within ± 20% of actual

energy-use. Building: Large Hotel at Chicago, IL.

Intuitively, one would expect prediction accuracy of the models to improve with

the length of the data-set. This improvement is not that significant in the case of

WBE. For example, for the starting month of November, the prediction improves

66
only by a small margin of 0.2% when the monitoring period is increased from one

month to six-months. The difference in the predictive accuracy of short data set

of the month of November compared to the prediction using the model obtained

from whole year of data is merely 2.11%. Thus, in this case, for WBE only one

month of data is adequate for predicting the building performance for the whole

year within acceptable limits.

The worst monitoring periods for monitoring WBE are July, February and January

with CV(%) of 53.58%, 33.88% and 18.14% respectively and NMBE(%) of

-41.5%, 22.91% and 12.16% respectively.

The ranking for all the one-month periods for in-situ monitoring of WBE, as

predicted by the MCP modeling approach, are given in Table 7.1.

Table 7.1

Ranking (based on the MCP Approach) of best to worst months of the year for in-situ

monitoring of WBE energy use for Large Hotel in Chicago, IL.

Ranking: WBE Prediction

Rank Month Rank Month
1 November 7 September
2 March 8 August
3 October 9 December
4 June 10 January
5 April 11 February
6 May 12 July

7.2.2. Results for Cooling Energy Use (CHW) for Large Hotel, Chicago, IL

Figure 7.4 shows results for CHW from Figures 7.1-7.2 in more detail. All the

twelve starting months have been combined into one figure. The graph for CV

(%) shows all the monitoring periods that predicted CHW within 50% of the actual

energy use while the graph for NMBE (%) shows monitoring periods with bias of

± 20%.

67
 On an average, the CV (%) for cooling energy use models never really

decreases below 24%, irrespective of the length of the dataset used for

modeling. The average errors range from -22.57% to 118.88% for different base

periods (outliers such as predictions using July data are ignored due to marked

inconsistency with other results). Since for cooling energy use (CHW) no single

month data is enough for extrapolating the performance of the building over the

whole year, longer periods of monitoring are therefore required.

CV(%) : Cooling Thermal Use (CHW)

50 January
February
40 March
April
30 May
CV (%)

June
20
July
August
10
September
October
0
November
1 2 3 4 5 6 7 8 9 10 11 12
Number of Consecutive Months in Dataset December

NMBE(%) : Cooling Thermal Use (CHW)

20 January
February
March
10
April
NMBE (%)

May
0 June
1 2 3 4 5 6 7 8 9 10 11 12 July
August
-10
Sepember
October
-20 November
Number of Consecutive Months in Dataset December

Figure 7.4 Variation in CV(%) and NMBE(%) for CHW Channel. Graphs show

monitoring periods with predicted CV (%) within 50% and predicted NMBE (%) within ±

20% of actual energy use. Building: Large Hotel at Chicago, IL

68
The general trend observed is that models identified from low or high

temperature regions tend to seriously err in the prediction of annual energy use.

For example, when in-situ monitoring is started from the month of January or

June (Table 7.2) at least seven months of data is required for the prediction of

energy use to reach acceptable levels of accuracy. On the other hand, if

monitoring is started in the months from the swing season, say April or August,

then only four months of data is enough to achieve similar results. In all the

cases, irrespective of the start time for monitoring, as we include data from the

months during which the outdoor dry bulb temperature is close to the annual

mean (March-April and October-November in this case), the prediction results of

the identified model gradually becomes more accurate.

Table 7.2

Cooling energy-use prediction results of models derived using data from the

months of January-July, April-July, June-December and August-December for

the Large Hotel, Chicago, IL.

Monitoring Period Duration of Prediction Accuracy -

Monitoring
Start Month End Month (Months) CV (%) NMBE (%)

January July 7 25 5
1
April July 4 27 8
June December 7 25 2.57
2
August December 5 26 5.6

Conversely, models become poorer in the annual predictive ability as more

monitored data is added which include months during which the outdoor dry bulb

temperature is not close to the annual mean. Thus, considering the case of

October as the starting month for monitoring, predictions worsen as data for

almost 6 additional months is added for identifying the energy use model. It

69
starts stabilizing again when more data from months of March-April is used.

Table 7.3 gives five options of short term monitoring periods for CHW energy

channel which give prediction results similar to the results obtained when the

model is generated using whole year data.

Table 7.3

Monitoring periods with prediction results closest to the results obtained when

monitoring is done for the whole year. Energy Channel: CHW; Building: Large

Hotel, Chicago, IL.

Monitoring Period Duration of Prediction Accuracy (CHW)

S.No. Monitoring
Start Month End Month (Months) CV (%) NMBE (%)

January December 12 24.15 0.28

1 February July 6 25.02 4.96
2 July December 6 25.38 ‐4.87
3 April August 5 26.03 3.99
4 September December 4 26.06 4.45
5 August December 5 26.33 ‐5.6

7.2.3. Results for Heating Energy Use (HW) for Large Hotel, Chicago, IL

Figure 7.5 shows results for HW from Figures 7.1-7.2 in more detail. All the

twelve starting months have been combined into one figure. The graph for CV

(%) shows all the monitoring periods that predicted HW within 50% of the actual

energy use while the graph for NMBE (%) shows monitoring periods with bias of

± 20%.

70
 CV(%) : Heating Thermal Use (HW)
50 January
February
40 March
April
30 May
CV (%)

June
20 July
August
10 September
October
0 November
1 2 3 4 5 6 7 8 9 10 11 12
December
Number of Consecutive Months in Dataset

NMBE(%) : HeatingThermal Use (HW)

20 January
February
March
10
April
May
NMBE (%)

0 June
1 2 3 4 5 6 7 8 9 10 11 12 July
August
-10 September
October
November
-20
Number of Consecutive Months in Dataset December

Figure 7.5 Variation in CV(%) and NMBE (%) for HW Channel. Graphs show monitoring

periods with predicted CV (%) within 50% and predicted NMBE (%) within ± 20% of

actual energy use. Building: Large Hotel at Chicago, IL

71
Similar to CHW, the CV (%) for heating thermal use energy models never really

decrease below 23%, irrespective of the length of the dataset used for modeling.

The average errors range from -45.68% to 81.74% for different base periods

(outliers such as predictions using October data are ignored due to marked

inconsistency with other results). Since for heating energy use prediction no

single month data is enough for accurately extrapolating the performance of the

building over the whole year, longer periods of monitoring are therefore required.

Similar to the predictions of cooling thermal use, for heating also the general

trend observed is that models constructed in low or high temperature regions

tend to seriously err in the prediction of annual energy use. For example, when

in-situ monitoring is started from the month of June (Table 7.4) at least seven

months of data are required to identify a model whose prediction of energy use is

within acceptable accuracy levels. On the other hand, if monitoring is started in

the months of the swing season, say October, then only three months of data is

enough to achieve the same results. In all the cases, irrespective of the start time

for monitoring, as we add data from the months during which the outdoor dry

bulb temperature is close to the annual mean (October-November in this case),

the prediction results start becoming more accurate.

Table 7.4

Heating energy-use prediction results for June-December and October-

December for Large Hotel, Chicago, IL.

Monitoring Period Duration of Prediction Accuracy - HW
S.No. Monitoring
Start Month End Month (Months) CV (%) NMBE (%)

June December 7 24.1 3.3

Case 1
October December 3 24.38 -0.28

72
Table 7.5

Monitoring periods for which identified models predict energy use closest to

those for which model is identified using whole year data. Energy Channel: HW;

Building: Large Hotel, Chicago, IL.

Monitoring Period Duration of Prediction Accuracy (HW)

S.No. Monitoring
Start Month End Month (Months) CV (%) NMBE (%)

January December 12 22.76 -0.16

1 March June 4 23.14 ‐3.21
2 September January 5 23.70 5.08
3 April September 6 24.04 6.16
4 October December 3 24.38 ‐0.28
5 February June 5 24.45 ‐6.66

Table 7.5 gives five options of short-term monitoring periods for heating energy

use that give prediction results very close to the ones derived from models

generated using whole year data.

73
7.3. Results for Office Building - Albuquerque, New Mexico

The time series plots of CV (%) and NMBE (%) for each type of energy use are

displayed in Figure 7.6 and Figure 7.7 respectively.

CV(%):Whole Building Energy Use CV(%):Whole Building Energy Use
(Starting Months January to June) (Starting Months July to December)
100 100

80 80

60 60

40 40

20 20

0 0
1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12
Number of Consecutive Months in Dataset Number of Consecutive Months in Dataset
January February March April May June July August September October November December

CV(%) : Heating Thermal Use CV(%) : Heating Thermal Use

(Starting Months January to June) (Starting Months July to December)
100 100

80 80

60 60

40 40

20 20

0 0
1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12
Number of Consecutive Months in Dataset Number of Consecutive Months in Dataset
January February March April May June July August September October November December

Figure 7.6 CV (%) for energy channels WBE & HW when different lengths of monitoring

are used for predicting annual energy use. Building: Office at Albuquerque, NM.
NMBE(%):Whole Building Energy Use NMBE(%):Whole Building Energy Use
(Starting Months January to June) (Starting Months July to December)
100 100
80 80
60 60
40 40
20 20
0 0
-20 1 2 3 4 5 6 7 8 9 10 11 12 -20 1 2 3 4 5 6 7 8 9 10 11 12
-40 -40
-60 -60
-80 -80
-100 -100
Number of Consecutive Months in Dataset Number of Consecutive Months in Dataset
January February March April May June July August September October November December

NMBE(%) : HeatingThermal Use NMBE(%) : Heating Thermal Use

(Starting Months January to June) (Starting Months July to December)
100 100
80 80
60 60
40 40
20 20
0 0
-20 1 2 3 4 5 6 7 8 9 10 11 12 -20 1 2 3 4 5 6 7 8 9 10 11 12
-40 -40
-60 -60
-80 -80
-100 -100
Number of Consecutive Months in Dataset Number of Consecutive Months in Dataset

January February March April May June July August September October November December

Figure 7.7 NMBE (%) for energy channels WBE & HW when different lengths of

monitoring are used for predicting annual energy use for office at Albuquerque, NM.

74
Albuquerque does not have a predominant cooling season; thus only two

channels, WBE and HW have been analyzed for this case.

7.3.1. Results for WBE energy use for office at Albuquerque, NM

Figure 7.8 shows results for whole building Electric from Figures 7.6 - 7.7 in more

detail. All the twelve starting months have been combined into one figure. The

graph for CV(%) shows all monitoring periods that predicted WBE within 50% of

the actual energy use while the graph for NMBE(%) shows monitoring periods

which resulted in predictions with bias of ± 20%

. CV(%):Whole Building Electric (WBE)

January
50
February
March
40
April
May
30
CV (%)

June
July
20
August
10 September
October
0 November
1 2 3 4 5 6 7 8 9 10 11 12 December
Number of Consecutive Months in Dataset

NMBE(%):Whole Building Electric (WBE)

20 January
February
March
10
April
May
NMBE(%)

0 June
1 2 3 4 5 6 7 8 9 10 11 12 July
August
-10
September
October
-20 November
Number of Consecutive Months in Dataset December

Figure 7.8 Variation in CV (%) and NMBE (%) for Whole Building Electric.

Building: Office at Albuquerque, NM

75
On an average, the predicted CV (%) for whole building electric never decrease

below 20%, irrespective of the length of the dataset used for modeling. Clearly,

for Albuquerque, no single month data is adequate for extrapolating the

performance of the building over the whole year; longer periods of monitoring

therefore are required. Minimum of four months of in-situ monitoring is required

to yield predictions within acceptable accuracy.

Table 7.6 summarizes the five best options of short-term monitoring periods for

the WBE channel. The prediction results of the models obtained from the data

from these periods is very close to the results derived from models generated

using whole year data.

Table 7.6

Monitoring periods for which identified models predict energy use closest to

those for which model is identified using whole year data. Energy Channel: WBE;

Building: Large Office, Albuquerque, NM.

Prediction Accuracy
Monitoring Period Duration of
Whole Building Electric
S.No. Monitoring
Start Month End Month (Months) CV (%) NMBE (%)

January December 12 20.13 0.04

1 February June 5 20.29 ‐0.34
2 March June 4 20.57 ‐0.68
3 August November 4 20.48 2.06
4 November April 6 21.83 0.32
5 December April 5 21.42 ‐0.62
Note that, each of the above cases includes periods with outdoor dry bulb

temperatures close to the annual mean.

76
7.3.2. Results for Heating Energy Use for Office at Albuquerque, NM

Figure 7.9 shows results for HW from Figures 7.6 – 7.7 in more detail. All the

twelve starting months have been combined into one figure. The graph for CV

(%) shows all the monitoring periods that predicted HW within 50% of the actual

energy use while the graph for NMBE (%) shows monitoring periods with bias of

±20%.

NMBE(%) : Heating Energy Use (HW)

20 January
February
March
10
April
May
NMBE (%)

0 June
1 2 3 4 5 6 7 8 9 10 11 12 July
August
-10 September
October
November
-20
Number of Consecutive Months in Dataset December

CV(%) : Heating Energy Use (HW)

January
50
February
40 March
April
30 May
CV (%)

June
20 July
August
10 September
October
0
November
1 2 3 4 5 6 7 8 9 10 11 12
December
Number of Consecutive Months in Dataset

Figure 7.9 Variation in CV (%) and NMBE (%) for HW for office at Albuquerque, NM.

77
The predicted CV (%) for heating energy use in this case never decreases below

11%, irrespective of the length of the dataset used for modeling. Table 7.7

summarizes the five best options of short-term monitoring periods for the HW

channel. The prediction results of the models obtained from the data from these

periods is very close to the results derived from models generated using whole

year data.

Table 7.7

Monitoring periods for which identified models predict energy use closest to

those for which model is identified using whole year data. Energy Channel: HW;

Building: Large Office, Albuquerque, NM.

Prediction Accuracy
Monitoring Period Duration of
Heating Energy-Use
S.No. Monitoring
Start Month End Month (Months) CV (%) NMBE (%)

January December 12 11.26 -0.02

1 January May 5 11.38 0.17
2 February May 4 11.33 ‐0.11
3 August December 5 11.36 ‐0.39
4 September January 5 12.3 0.81
5 December May 6 11.9 ‐0.31

78
7.4. Results for Full Service Hotel – Washington DC Region

This section presents analysis results of the actual measured data for whole

building electric energy-use channel. The time series plots of CV (%) and NMBE

(%) are displayed in Figure 7.10 and Figure 7.11.

CV(%):Whole Building Energy Use CV(%):Whole Building Energy Use

(Starting Months January to June) (Starting Months July to December)
100 100

80 80

60 60

40 40

20 20

0 0
1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12
Number of Consecutive Months in Dataset Number of Consecutive Months in Dataset
January February March April May June July August September October November December

Figure 7.10 CV (%) for WBE energy-use channel when different lengths of monitoring are

used for predicting annual energy for hotel in Washington D.C. area.

NMBE(%):Whole Building Energy Use NMBE(%):Whole Building Energy Use

(Starting Months January to June) (Starting Months July to December)
100 100
80 80
60 60
40 40
20 20
0 0
-20 1 2 3 4 5 6 7 8 9 10 11 12 -20 1 2 3 4 5 6 7 8 9 10 11 12
-40 -40
-60 -60
-80 -80
-100 -100
Number of Consecutive Months in Dataset Number of Consecutive Months in Dataset
January February March April May June July August September October November December

Figure 7.11 NMBE (%) for WBE energy-use channel when different lengths of monitoring

are used for predicting annual energy use for hotel in Washington D.C area.

Figure 7.12 below shows results for WBE from Figures 7.10 – 7.11 in more

detail. All the twelve starting months have been combined into one figure. The

graph for CV(%) shows all monitoring periods that predicted WBE value within

50% of the actual energy use while the graph for NMBE(%) shows monitoring

periods with bias of ± 20%.

79
 CV(%):Whole Building Electric (WBE)
January
50
February
March
40
April
May
30

CV (%)
June
July
20
August
September
10
October
November
0
1 2 3 4 5 6 7 8 9 10 11 12 December
Number of Consecutive Months in Dataset

NMBE(%):Whole Building Electric (WBE)

20 January
February
March
10
April
May
NMBE(%)

0 June
1 2 3 4 5 6 7 8 9 10 11 12 July
August
-10
September
October
-20 November
Number of Consecutive Months in Dataset December

Figure 7.12 Variation in CV (%) and NMBE (%) for WBE. Graphs show periods with

predicted CV (%) within 50% and predicted NMBE (%) within ± 20% of actual energy use

for full service hotel located in Washington D.C area.

Table 7.8

Monitoring periods for which identified models predict energy use closest to

those for which model is identified using whole year data. Energy Channel: WBE;

Building: Full service hotel, Washington DC.

Monitoring Period Duration of Prediction Accuracy (WBE)
S.No. Monitoring
Start Month End Month (Months) CV (%) NMBE (%)

January December 12 9.13 0.00

1 March May 3 10.2 3.5
2 August November 4 9.59 0.79
3 September November 3 9.62 0.76
4 October December 3 9.31 0.59
5 February June 5 9.28 1.16
80
7.5. Validation of DBTA approach

In this section, the rankings obtained from the DBTA and MCP Modeling

Approach have been compared. The results for the Large Hotel, Chicago, IL are

shown in Table 7.9 and Figure 7.13 below.

Table 7.9

Comparison of results for DBTA and MCP Modeling approaches for the large

hotel building in Chicago, IL.

Monitoring Period for Best Building Performance Prediction

Ranking
End Month
Start Month Temperature MCP Method Temperature MCP Method
Analysis WBE CHW HW Analysis WBE CHW HW
January July July Aug-Sept June 7 7 8 7
February June-July July Aug-Sept June 6 5 7 5
March May-June June August June 3 3 3 3
April April-May May May May-June 1 1 1 1
May March Nov-Dec Dec-Jan December 11 9 8 10
June February January Nov-Dec December 9 10 6 8
July January January Oct-Nov January 7 7 5 8
August December January Oct-Nov December 5 5 3 5
September Nov-Dec December Nov-Dec December 3 3 2 3
October November November August Nov-Dec 2 1 12 1
November September July August July 11 11 11 12
December Aug-Sept August Aug-Sept July 10 11 10 10

Result Comparison: DBTA and MCP Approach

14
Ranking (Best to Worst)

12
10
8
6
4
2
0
1 2 3 4 5 6 7 8 9 10 11 12
Start Month for In-Situ Monitoring
Temperature Analysis WBE CHW HW
Figure 7.13 Comparison of results for Simplified DBTA and MCP modeling approaches

for the large hotel building in Chicago, IL.

81
The results for the Office Building, Albuquerque, NM are shown in Table 7.10

and Figure 7.14.

Table 7.10

Comparison of results for simplified DBTA and MCP Modeling approaches for

the office building in Albuquerque, NM.

Monitoring Period for Best Building Performance

Ranking for Each Starting
Prediction
Month
End Month
Start Month Temperature MCP Method Temperature MCPMethod
Analysis WBE HW Analysis WBE HW
January July June July-Aug 7 7 8
February June June June 5 6 5
March May April June-July 3 3 4
April April June May 1 4 1
May February January January 11 11 11
June Jan-Feb December Dec-Jan 9 9 10
July January Dec-Jan December 7 8 7
August December November December 5 5 5
September November Oct-Nov December 3 2 3
October October October Nov-Dec 1 1 2
November September July June-July 12 11 12
December August June June 10 9 9

Result Comparison: DBTA and MCP Approach

14
12
10
Ranking

8
6
4
2
0
1 2 3 4 5 6 7 8 9 10 11 12
Start Month for In-Situ Monitoring
Temperature Analysis WBE HW

Figure 7.14 Comparison of results for Simplified DBTA and MCP modeling approaches

for the large hotel building in Albuquerque, NM.

82
The results for the full service hotel, Washington DC Region are shown in Table

7.11 and Figure 7.15 below.

Table 7.11

Comparison of results for simplified DBTA and MCP Modeling approaches for the

hotel, Washington DC area.

Monitoring Period for Best Building
Ranking for Each Starting
Performance Prediction
Month
End Month
Start Month Temperature MCP Method Temperature MCP Method
Analysis WBE Analysis WBE
January July-August June 7 7
February June June 5 5
March May-June June 3 4
April April April 1 1
May Feb-March December 11 10
June Jan-Feb Dec-Jan 9 9
July Jan-Feb Dec-Jan 7 8
August Dec-Jan December 6 5
September Nov-Dec Nov-Dec 3 3
October Oct-Nov October 2 1
November September July 12 11
December August August 10 11
Result Comparison: DBTA and MCP Approach
14
Ranking (Best to Worst)

12
10
8
6
4
2
0
1 2 3 4 5 6 7 8 9 10 11 12
Start Month for In-Situ Monitoring
Temperature Analysis WBE

Figure 7.15 Comparison of results for DBTA and MCP modeling approaches for the hotel,

Washington DC area.

83
7.6. Conclusion

The above ranking is based on the duration of building monitoring needed such

that models identified from the data provide predictions which are closest to

those when a whole year worth of data is used. The objective is to identify the

most suitable month to install data acquisition equipment to in the building and

the length of monitoring needed to make accurate annual predictions.

Outdoor temperature is the most important factor influencing the energy use in a

building. In the above analysis, a distinct pattern linking the outdoor dry-bulb

temperatures to actual building energy use emerged.

Clearly, from Figures 7.13, 7.14 & 7.15 it is evident that the results obtained from

the simplified DBTA and the MCP approach are fairly consistent for all the three

buildings analyzed. As seen here, the months of April and October prove to be

the best to begin in-situ monitoring of the building energy performance.

Beginning in these months, only two to three months of data is enough to allow

models to be identified which would predict the long-term energy performance of

the buildings within acceptable accuracy levels.

Considering a scenario where neither any simulated data nor any utility history

regarding the building performance is available, and, if the building operator has

the option of choosing when to install the data acquisition equipment, the

simplified DBTA can be used as a tool to make recommendations for the best

time of the year to start energy-use measurement and the minimum duration of

monitoring required for predicting building energy use.

84
 Chapter 8

HYBRID INVERSE MODELING RESULTS

This chapter assembles the results of the two hybrid inverse modeling

approaches explained in Sections 5.5.1 and 5.5.2. The hybrid inverse modeling

(method 1) predicts energy use by combining a short data-set of monitored daily

energy use and internal loads with atleast one year of recent utility bills

(representative of the long-term building energy use behavior data). The hybrid

inverse modeling (method 2), on the other hand, uses information from only the

utility bills for estimating building energy-use.

8.1. Hybrid Inverse Modeling-Method 1

In this method, the model identification is done in two stages explained in Section

5.5. First stage involves regressing the monthly energy use with outdoor dry bulb

temperature and average specific humidity potential as the regressors. The

second stage then uses the model residuals of the first stage and finds another

regression model using the daily equipment and lighting loads as the regressor.

Finally, the model equations obtained from stages 1 and 2 are combined to

derive the final model for predicting energy use.

Since the research aims at finding the shortest period suitable for in-situ energy

use monitoring, the length of the period for this analysis is limited to a maximum

of three consecutive months. Thus, models are generated for each starting

month of the year (January to December), with each selection subsequently

expanded in increments of one month to a maximum of three consecutive

months to mimic different durations of monitoring.

85
8.1.1. Results for large hotel - Chicago, Illinois

The time series plots of CV (%) and NMBE (%) and for the three energy use

channels namely, whole building Electric, cooling energy use and heating energy

use (WBE, CHW & HW) are displayed in Figures 8.1 to 8.3, respectively, for

different lengths of monitoring. The results for the MCP approach adopted

previously have also been shown on the graphs to allow for a comparison. There

is a clear advantage of using hybrid inverse modeling approach compared to the

MCP approach since the CV(%) and NMBE(%) values are lower.
CV (%):Whole Building Electric NMBE(%):Whole Building Electric
100% 100%

HYBRID INVERSE MODELING MCP APPROACH HYBRID INVERSE MODELING MCP APPROACH
80% 80%

60% 60%
NMBE (%)

NMBE (%)

40% 40%

20% 20%

0% 0%
January

November

January

November
July

July
Sep

Sep
May

May
March

March
-20% -20%

CV (%):Whole Building Electric NMBE(%):Whole Building Electric

100% 100%
HYBRID INVERSE MODELING MCP APPROACH HYBRID INVERSE MODELING MCP APPROACH
80% 80%

60% 60%
NMBE (%)

NMBE (%)

40% 40%

20% 20%

0% 0%
Sep Oct

Sep Oct
Jul Aug

Jul Aug
Jan Feb

Mar Apr

Jan Feb

Mar Apr
May Jun

May Jun
Nov Dec

Nov Dec

-20% -20%

CV (%):Whole Building Electric NMBE(%):Whole Building Electric

100% 100%
HYBRID INVERSE MODELING MCP APPROACH HYBRID INVERSE MODELING MCP APPROACH
80% 80%

60% 60%
NMBE (%)

NMBE (%)

40% 40%

20% 20%

0% 0%
Sep Oct Nov

Sep Oct Nov

Jul Aug Sep

Jul Aug Sep

Jan Feb Mar

Jan Feb Mar

Mar Apr May

May Jun Jul

Mar Apr May

May Jun Jul

Nov Dec Jan

Nov Dec Jan

-20% -20%

Figure 8.1 Variation in CV (%) and NMBE (%) for different monitoring periods for the

large hotel at Chicago, IL. Analysis procedure: Hybrid Inverse Modeling (Method 1).

Energy Channel: WBE. 86

 CV(%):Cooling Energy Use NMBE(%):Cooling Energy Use
100% 100%

HYBRID INVERSE MODELING MCP APPROACH HYBRID INVERSE MODELING MCP APPROACH
80% 80%

60% 60%
NMBE (%)

NMBE (%)
40% 40%

20% 20%

0% 0%

November

November
January

January
July

July
Sep

Sep
May

May
March

March
-20% -20%

CV(%):Cooling Energy Use NMBE(%):Cooling Energy Use

100% 100%
HYBRID INVERSE MODELING MCP APPROACH HYBRID INVERSE MODELING MCP APPROACH
80% 80%

60% 60%
NMBE (%)

NMBE (%)
40% 40%

20% 20%

0% 0%
Sep Oct

Sep Oct
Jul Aug

Jul Aug
Mar Apr

Mar Apr
Jan Feb

May Jun

Jan Feb

May Jun
Nov Dec

Nov Dec
-20% -20%

CV(%):Cooling Energy Use NMBE(%):Cooling Energy Use

100% 100%
HYBRID INVERSE MODELING MCP APPROACH HYBRID INVERSE MODELING MCP APPROACH
80% 80%

60% 60%
NMBE (%)

NMBE (%)

40% 40%

20% 20%

0% 0%
Jul Aug Sep

Sep Oct Nov

Jul Aug Sep

Sep Oct Nov

Jan Feb Mar

Mar Apr May

Jan Feb Mar

Mar Apr May

May Jun Jul

May Jun Jul

Nov Dec Jan

Nov Dec Jan

-20% -20%

Figure 8.2 Variation in CV (%) and NMBE (%) for different monitoring periods for the

large hotel at Chicago, IL. Analysis procedure: Hybrid Inverse Modeling (Method 1).

Energy channel: CHW.

87
 CV(%):Heating Energy Use NMBE(%):Heating Energy Use
100% 100%

HYBRID INVERSE MODELING MCP APPROACH HYBRID INVERSE MODELING MCP APPROACH
80% 80%

60% 60%
NMBE (%)

NMBE (%)
40% 40%

20% 20%

0% 0%

November

November
January

January
July

July
Sep

Sep
May

May
March

March
-20% -20%

CV(%):Heating Energy Use NMBE(%):Heating Energy Use

100% 100%
HYBRID INVERSE MODELING MCP APPROACH HYBRID INVERSE MODELING MCP APPROACH
80% 80%

60% 60%
NMBE (%)

NMBE (%)
40% 40%

20% 20%

0% 0%
Sep Oct

Sep Oct
Jul Aug

Jul Aug
Jan Feb

Mar Apr

Jan Feb

Mar Apr
May Jun

May Jun
Nov Dec

Nov Dec
-20% -20%

CV(%):Heating Energy Use NMBE(%):Heating Energy Use

100% 100%
HYBRID INVERSE MODELING MCP APPROACH HYBRID INVERSE MODELING MCP APPROACH
80% 80%

60% 60%
NMBE (%)

NMBE (%)

40% 40%

20% 20%

0% 0%
Sep Oct Nov

Sep Oct Nov

Jul Aug Sep

Jul Aug Sep

Jan Feb Mar

Jan Feb Mar

Mar Apr May

May Jun Jul

Mar Apr May

May Jun Jul

Nov Dec Jan

Nov Dec Jan

-20% -20%

Figure 8.3 Variation in CV (%) and NMBE (%) for different monitoring periods for the

large hotel at Chicago, IL. Analysis procedure: Hybrid Inverse Modeling (Method 1).

Energy channel: HW.

88
8.1.2. Results for office building - Albuquerque, New Mexico

The time series plots of CV (%) and NMBE (%)for both energy use channels,

namely, whole building electric and heating energy use (WBE & HW) are

displayed in Figure 8.4 and Figure 8.5, respectively, for different lengths of

monitoring . The results for the MCP approach adopted previously have also

been shown on the graphs to allow for a comparison. Note that the CV (%) and

NMBE (%) values for the hybrid model vary little with the monitoring length

chosen or the month in which monitoring is initiated.

CV(%):Whole Building Electric NMBE(%):Whole Building Electric
100% 100%

80%
HYBRID INVERSE MODELING MCP APPROACH 80%
HYBRID INVERSE MODELING MCP APPROACH

60% 60%
NMBE (%)
CV (%)

40% 40%

20% 20%

0% 0%
November

November
January

January
October

October
February

February
August

August
September

December

September

December
July

July
June

June
May

May
April

April
March

-20% -20% March

CV(%):Whole Building Electric NMBE(%):Whole Building Electric

100% 100%

80%
HYBRID INVERSE MODELING MCP APPROACH 80%
HYBRID INVERSE MODELING MCP APPROACH

60% 60%
NMBE (%)
CV (%)

40% 40%

20% 20%

0% 0%
April May

April May
Mar Apr

Mar Apr
June July

June July
Jan Feb

Jan Feb
Aug Sept

Aug Sept
Feb March

Sep Oct

Feb March

Sep Oct
Oct Nov

Oct Nov
Jul Aug

Dec Jan

Jul Aug

Dec Jan
Nov Dec

Nov Dec
May Jun

May Jun

-20% -20%

CV(%):Whole Building Electric NMBE(%):Whole Building Electric

100% 100%

80%
HYBRID INVERSE MODELING MCP APPROACH 80%
HYBRID INVERSE MODELING MCP APPROACH

60% 60%
NMBE (%)
CV (%)

40% 40%

20% 20%

0% 0%
Sep Oct Nov

Sep Oct Nov

Jan Feb Mar

May Jun Jul

Jan Feb Mar

May Jun Jul

Mar Apr May

Mar Apr May

Nov Dec Jan

Dec Jan Feb

Nov Dec Jan

Dec Jan Feb

Feb Mar Apr

Feb Mar Apr

Apr May June

Apr May June

June July Aug

Aug Sept Oct

June July Aug

Aug Sept Oct

Oct Nov Dec

Oct Nov Dec

Jul Aug Sep

Jul Aug Sep

-20% -20%

Figure 8.4 Variation in CV (%) and NMBE (%) for different monitoring periods for office

building at Albuquerque, NM. Analysis Procedure: Hybrid Inverse Modeling (Method 1).

Energy channel: WBE.

89
 CV(%):Heating Energy Use NMBE(%):Heating Energy Use
100% 100%

80%
HYBRID INVERSE MODELING MCP APPROACH 80%
HYBRID INVERSE MODELING MCP APPROACH

60% 60%

NMBE (%)
CV (%)

40% 40%

20% 20%

0% 0%

November

November
January

January
October

October
February

February
August

August
September

December

September

December
July

July
June

June
May

May
April

April
March

March
-20% -20%

CV(%):Heating Energy Use NMBE(%):Heating Energy Use

100% 100%
HYBRID INVERSE MODELING MCP APPROACH HYBRID INVERSE MODELING MCP APPROACH
80% 80%

60% 60%

NMBE (%)
CV (%)

40% 40%

20% 20%

0% 0%

June July
April May

Aug Sept
Mar Apr

June July
Jan Feb

Aug Sept

Sep Oct
Feb March

Sep Oct

Feb March

Dec Jan
Oct Nov
Oct Nov
Jul Aug

Dec Jan

Jul Aug
April May
Mar Apr
Jan Feb

May Jun
Nov Dec

Nov Dec
May Jun

-20% -20%

CV(%):Heating Energy Use NMBE(%):Heating Energy Use

100% 100%

80%
HYBRID INVERSE MODELING MCP APPROACH 80%
HYBRID INVERSE MODELING MCP APPROACH

60% 60%
NMBE (%)
CV (%)

40% 40%

20% 20%

0% 0%
Sep Oct Nov

Sep Oct Nov

Jan Feb Mar

May Jun Jul

Jan Feb Mar

May Jun Jul

Mar Apr May

Mar Apr May

Nov Dec Jan

Dec Jan Feb

Nov Dec Jan

Dec Jan Feb

Feb Mar Apr

Feb Mar Apr

Apr May June

Apr May June

June July Aug

Aug Sept Oct

June July Aug

Aug Sept Oct

Oct Nov Dec

Oct Nov Dec

Jul Aug Sep

-20% -20% Jul Aug Sep

Figure 8.5 Variation in CV (%) and NMBE (%) for different monitoring periods for office

building at Albuquerque, NM. Analysis Procedure: Hybrid Inverse Modeling (Method 1).

Energy channel: HW.

8.1.3. Conclusions

The results in Sections 8.1.1 and 8.1.2 show a clear advantage of using the

hybrid inverse modeling approach in predicting energy use in a building

compared with the MCP modeling approach. Since the method uses utility history

to represent the long-term data, a much shorter period of in-situ monitoring can

be used for estimating energy use. The long-term predictions for the energy

90
channels are found to be almost the same irrespective of the time of the year

chosen for monitoring. Not much improvement in the long term predictions is

observed when the period from which the regression models are generated is

increased from one to three months.

Looking closely at the results, one finds that the models regressed with data from

the swing seasons tend to have marginally better predictions compared to the

data obtained from the peak summer or winter periods. In the case of the building

in Chicago, if the data acquisition system could be applied for only a period of

one month, April or October would be the best choices. In the case of

Albuquerque, March and October are found to be the best for measuring the

whole building Electric and any month from April to September is good for

predicting heating energy use of the building.

If the building owner has the option to keep the data acquisition equipment for a

longer period, it would be better not just to obtain data from the swing season

(spring or fall) but also capture a little more variability by extending the

measurement into the cooling or heating season depending upon which energy

channel is being modeled.

Combination of information from the utility bills with a short period of monitored

daily energy use data is therefore a good source of information for predicting long

term building performance. In most cases, only one month of monitored data is

sufficient in predicting long term energy use of the building.

91
8.2. Hybrid Inverse Modeling-Method 2

The hybrid inverse modeling (method 2) aims at predicting long-term building

performance at daily timescales using utility bills only with no monitoring at all.

The analysis method has been tested only for the WBE energy channel for the

three datasets. The intent is to determine if utility bills alone can be used to

predict the performance of buildings and by how much does the predictive

accuracy decrease when no information about daily energy use and internal

loads is added to the regression model.

In this case, the statistical parameters, NMBE(%) and CV(%), are analyzed at

monthly as well as annual time scales. This would help provide an insight into

how well the model generated from utility history alone is able to predict the

energy use for each month of the year as well as for the whole year.

8.2.1 Results for the large hotel – Chicago, IL

In Figure 8.6, the whole building electricity consumption for the entire year

obtained from the simulated data and the prediction results from the hybrid

inverse modeling (method 1) and hybrid inverse modeling (method 2) are shown.

Monitored data for the two months (March-April) is used for identifying the

regression model for hybrid inverse modeling (method 1). The inaccuracy in

prediction of energy-use for the entire year using the model equation obtained

from utility data only is clearly evident. Since the utility data does not capture the

daily trends of internal loads in the building, the model over-predicts the energy

use over the weekends for a major portion of the year and under-predicts for

some portions. On the other hand, the energy use prediction profile obtained

from the regression model of the hybrid inverse modeling (method 2) is very

close to the profile derived from the simulated data.

92
 Whole Building Electric Consumption for Whole Year
1800

1600

1400
WBE (kWh)

1200

1000

800

600
1
8
15
22
29
36
43
50
57
64
71
78
85
92
99
106
113
120
127
134
141
148
155
162
169
176
183
190
197
204
211
218
225
232
239
246
253
260
267
274
281
288
295
302
309
316
323
330
337
344
351
358
365
Day of Year
Simulated Data Prediction Using Utility Bills + Short Data-Set (March-April) Prediction Using Utility Bills Only

Figure 8.6 WBE use profiles for the entire year obtained from the simulated data & the

prediction results of the hybrid inverse modeling methods 1 &2 for the large hotel,

Chicago, IL.

NMBE(%) for Whole Building Electric

5%
Monthly Prediction
4%
(Utility Only)
3%
2% Annual Prediction
1% (Utility Only)
0%
‐1% 1 Monthly Prediction
2 3 4 5 6 7 8 9 10 11 12
MONTHS (Utility Bills +Monitored
‐2% Data)
‐3% Annual Prediction
‐4% (Utility Bills +Monitored
Data)
‐5%

CV(%) for Whole Building Electric

20%
18% Monthly Prediction
(Utility Only)
16%
14% Annual Prediction
12% (Utility Only)
10%
8% Monthly Prediction
6% MONTHS (Utility Bills +Monitored
Data)
4%
Annual Prediction
2%
(Utility Bills +Monitored
0% Data)
1 2 3 4 5 6 7 8 9 10 11 12

Figure 8.7 Comparison of NMBE(%) and CV(%) for WBE for the Large Hotel, Chicago,

IL at monthly and annual timescales for Hybrid Inverse Modeling analysis methods 1 & 2.

The time series plots of CV (%) and NMBE (%) in Figure 8.7 compare the

predictive accuracy of the models obtained using utility bills and a short dataset

93
and utility bills alone, i.e. hybrid inverse modeling methods 1 and 2 respectively.

The average errors do not vary more than ±5% for either case. The errors from

method 1, however, are slightly lower compared to the results from method 2.

Since the CV (%) is a better indicator of the predictive accuracy when analysis is

done at monthly timescales, clearly, the hybrid inverse analysis (method 1) yields

more accurate predictions on both monthly as well as annual timescales. The

monthly CV(%), when only utility data (hybrid inverse modeling method 2) is

used, varies from 12% to 16%. It is further reduced by approximately 10% when

information from monitored data from just two months (March-April in this case)

is used in addition to utility history for generating the regression model. The

CV(%) values of monthly predictions for hybrid inverse modeling (method 1)

ranges from 4.3%-2.4%. The annual CV (%) for hybrid inverse method 1 and

method 2 is 3.38% and 14.26%, respectively.

8.2.2. Results for office building – Albuquerque, NM

Figure 8.8 shows the whole building Electricity consumption for the entire year

obtained from the simulated data and the prediction results from the hybrid

inverse modeling (method 1) and hybrid inverse modeling (method 2). Monitored

data for the two months (March-April) is used for obtaining the regression model

for hybrid inverse approach (method 1).

Analysis of the office building at Albuquerque also reveals (as was the case for

the large hotel in Chicago) that the utility data fails to capture the daily trends of

internal loads in the building. Not surprisingly, the model generated using utility

data only over-predicts the energy use for the weekends when internal loads are

lower. The energy use prediction profile obtained from the regression model of

94
the hybrid inverse modeling (method 2) is very close to the profile derived from

the simulated or as-designed dataset.

Whole Building Electric Consumption for Whole Year

60

50

40
WBE (kWh)

30

20

10

0
1
8
15
22
29
36
43
50
57
64
71
78
85
92
99
106
113
120
127
134
141
148
155
162
169
176
183
190
197
204
211
218
225
232
239
246
253
260
267
274
281
288
295
302
309
316
323
330
337
344
351
358
365
Day of Year
Simulated Data Prediction Using Utility Bills Only Utility Bills + Monitored Data (March-April)

Figure 8.8 WBE profiles for the entire year obtained using the simulated data & the

prediction results of the Hybrid Inverse Modeling Methods 1 &2 for the office building at

Albuquerque, NM.

NMBE(%) for Whole Building Electric

20%
Monthly Prediction
(Utility Only)
10%
Annual Prediction
0% (Utility Only)
MONTHS
1 2 3 4 5 6 7 8 9 10 11 12
Monthly Prediction
‐10%
(Utility Bills +Monitored
Data)
‐20% Annual Prediction
(Utility Bills +Monitored
Data)
‐30%

60%
CV(%) for Whole Building Electric
Monthly Prediction
50% (Utility Only)

40% Annual Prediction

(Utility Only)
30%
Monthly Prediction
20% (Utility Bills +Monitored
Data)
10%
Annual Prediction
MONTHS (Utility Bills +Monitored
0% Data)
1 2 3 4 5 6 7 8 9 10 11 12

Figure 8.9 Comparison of NMBE(%) and CV(%) for Whole Building electric for the Office

at Albuquerque, NM at Monthly and Annual Timescales for Hybrid Inverse Modeling

Analysis Methods 1 & 2.

95
The time plots of CV (%) and NMBE (%) in Figure 8.9 allow comparison of the

predictive accuracy of the models obtained using utility bills and a short dataset

(two months) and utility bills alone, i.e. hybrid inverse modeling methods 1 and 2

respectively. The average errors for method 1 range from -15% to 13%, and for

method 2 they vary from -9% to -26.8%. Since the CV (%) is a better indicator of

the predictive accuracy when analysis is done at monthly timescales, clearly, the

hybrid inverse analysis (method 1) yields more accurate predictions monthly as

well as annually. The monthly CV(%) value, when only utility data (hybrid inverse

modeling method 2) is used, varies from 40% to 55%, Further, it reduces to

approximately half when information from monitored data from two months

(March-April in this case) is used in addition to utility history for generating the

regression model. The CV(%) values of monthly predictions for hybrid inverse

modeling (method 1) ranges from 12%-28%. The annual CV (%) for hybrid

inverse method 1 and method 2 is 22.9% and 53.5% respectively.

8.2.3. Results for hotel – Washington DC area

The data available for the full service hotel located in the Washington DC region

includes only monitored data for WBE and DBT for one complete year. Since

there is no information available regarding the LTEQ loads of the building, only

utility bills have been used for predicting the performance of the building. The

humidity potential (w) was also not measured, therefore, only DBT is used as the

regressor. The analysis procedure adopted is explained in Section 5.5.2. The

energy use profiles for actual monitored data and the energy-use predicted using

the model obtained from utility data only are shown in Figure 8.10.

96
 Whole Building Electric Consumption for Whole Year
1800
Monitored Data Prediction Using Utility Bills
1600
1400
WBE (kWh) 1200
1000
800
600
400
200
0
1
8
15
22
29
36
43
50
57
64
71
78
85
92
99
106
113
120
127
134
141
148
155
162
169
176
183
190
197
204
211
218
225
232
239
246
253
260
267
274
281
288
295
302
309
316
323
330
337
344
351
358
365
Day of Year

Figure 8.10 WBE Consumption for full service hotel, Washington DC Region. Actual

monitored data and predicted energy use based on utility bills are shown.

NMBE(%) for WBE CV(%): WBE

20% 20%
Monthly Prediction Monthly Prediction
15%
Annual Prediction Annual Prediction
10% 15%

5%
10%
0%
‐5% 1 2 3 4 5 6 7 8 9 10 11 12
5%
‐10%
‐15% 0%
‐20% 1 2 3 4 5 6 7 8 9 10 11 12
MONTHS MONTHS
Figure 8.11 NMBE (%) and CV %) for WBE Predictions using utility history only. The

variations in prediction and average errors at monthly as well as annual timescales have

been indicated for hotel, Washington D.C. Region. Analysis procedure: Hybrid Inverse

Modeling (Method 2).

The time series plots of NMBE (%) and CV (%) in Figure 8.11 show the

predictive accuracy of the model obtained using utility bills only. The average

errors for predictions for different months of the year vary from -11.3% to 2.9%.

The CV(%) also shows large variations over the year and ranges between 7% to

16.9%. The annual CV (%) for this case is 11%. In this case, the utility bills prove

to be a good source for accurately predicting the building energy use. However,

additional knowledge regarding internal loads would have probably made the

prediction results even more accurate.

97
8.2.4. Conclusions

Clearly, there is an advantage in using some amount of monitored data for

predicting building performance. As seen above, adding only two-months of

monitored data to the utility information greatly improves the prediction results.

However, in case no monitored data is available, utility bills can be used to

provide a fairly accurate estimate of the long term energy use patterns of the

building energy use at both monthly as well as annual timescales.

98
 Chapter 9

SUMMARY, CONCLUSIONS & FUTURE WORK

9.1. Summary

The main focus throughout this research was to investigate if short-term in-situ

monitoring of energy use in buildings can be adopted as a workable alternative to

year-long monitoring and verification. Simple modeling techniques to determine

best time of the year to begin in-situ monitoring of building energy-use, and the

least amount of data required for generating acceptable long-term predictions

were identified. To determine the amount of short-term building energy-use data

that would be enough to generate acceptable annual predictions, four different

analysis methods were tested on year-long datasets from three buildings (two

synthetic and one actual) from diverse geographical locations within the United

States.

Three application areas for a research of this nature were identified; namely,

detailed audits, green building performance verification and verification of post-

retrofit claims using pre-post monitored data. The analysis methods proposed

and studied were categorized based on the type of data that could be available

for analysis. These categories include use of: (i) ambient temperature data only,

(ii) simulated/ monitored year-long data, (iii) year-long utility bills and a short

monitored dataset, and (iv) year-long utility bills only.

The short-term monitoring for long-term prediction (SMLP) and dry-bulb

temperature analysis (DBTA) approaches used only temperature data to

determine the best time and duration of the year for in-situ monitoring.

Multivariate change-point (MCP) modeling technique was used to determine best

monitoring period of the year based on simulated/monitored data. The method

99
was also used to verify the findings of the temperature analysis using DBTA

method. The hybrid inverse modeling method-1 predicted energy-use by

combining a short dataset of monitored internal loads with a year of utility-bills,

and hybrid inverse method-2 predicted long term building performance using

utility-bills only. The results of each of these analysis methods are summarized

in Section 9.2.

9.2. Conclusions

a. DBTA and SMLP Approaches

The DBTA method, devised as a simple and easy-to-use alternative to the SMLP

method (Abushakra, 2000), was used for analyzing the temperature data for the

three different locations. The rankings obtained from the SMLP method and the

DBTA showed similar patterns with months from the swing seasons ranked best

in both analysis procedures. March, April and October were found to be the best

periods for starting in-situ monitoring, while, the other months with extremes in

weather conditions ranked poorly in both analysis methods. If monitoring was to

be initiated from these months, only two to three months of data was found to be

adequate to identify models which would predict the long-term energy

performance of the buildings within acceptable accuracy levels. The DBTA

method was also found to be quite accurate in capturing the information

regarding the duration of monitoring required for making acceptable energy-use

predictions.

The rankings obtained from the DBTA were further verified by means of actual

modeling of energy-use data using the MCP modeling method. The results

obtained from the DBTA and the MCP approaches were found to be fairly

consistent for all the three buildings analyzed. A distinct pattern linking the

100
outdoor dry-bulb temperatures to actual building energy-use emerged. It was

established that the annual predictive ability of short data-sets is strongly

influenced by the closeness of the dataset’s mean temperature to the annual

mean temperature.

Considering a scenario where neither any simulated data nor any utility history

regarding the building performance is available, and, if the building operator has

the option of choosing the time of the year when to install the data acquisition

equipment, the simplified DBTA can be used as a tool to make recommendations

for best time of the year to start energy-use measurement and the minimum

duration of monitoring required for predicting building energy use.

b. MCP Modeling

This analysis approach was used to predict whole building electric (WBE),

cooling energy use (CHW), and heating energy use (HW) using a multivariate

change point (MCP) models derived from short datasets. The outdoor dry bulb

temperature (DBT) and lighting and equipment loads (LTEQ) were used as

regressors. To determine the accuracy of the models derived from the short data-

sets, the values of annual energy use predicted by models obtained from short

data-sets were compared to the actual energy use in the original dataset. The

predictive accuracy of the models was evaluated based on two statistical indices;

CVRMSE (%) and the NMBE (%).

The accuracy of the models was found to largely depend on the starting date and

the end date of data collection. The months from the swing season, namely,

March, April, October and November were found to be the best months to begin

in-situ monitoring of building energy use. The general trend observed was that

101
models identified from low or high temperature regions tend to seriously err in the

prediction of annual energy use.

The results from the MCP modeling method gave insight into the importance of

the length of duration of monitoring the energy-use for predicting performance.

Intuitively, one would expect the prediction accuracy of the annual energy to

improve as the length of the data-set increases. In many cases, the average

annual prediction was seen to decline as the length of the data-set was

increased. Thus, one may infer that for accurate prediction results, not only is it

important to know the best starting period to begin in-situ measurements, but

correctly determining the length of data-set is of equal significance. For the three

buildings analyzed, three to four months of monitored data was found to be

adequate for making long term predictions of building performance within

acceptable accuracy levels.

c. Hybrid Inverse Modeling

The Hybrid Inverse Modeling (method 1) aimed at predicting building energy-use

by combining a short data-set of monitored daily energy-use and internal loads

with atleast one year of recent utility bills (representing the long-term data). The

results showed a clear advantage of using hybrid inverse modeling approach in

predicting energy use in a building when compared with the MCP approach.

Since the method uses utility history to represent the long-term data, a much

shorter period was found to be sufficient for estimating long-term energy-use.

The long-term predictions for the energy channels were found to be almost the

same irrespective of the time of the year chosen for monitoring. Not much

improvement in the long term predictions was observed when the period from

which the regression models were generated was increased from one to three

102
months. In almost all cases, only one month of monitored data was sufficient in

predicting long term energy use of the building within acceptable accuracy levels.

Models regressed with data from the swing seasons were found to have

marginally better predictions compared to the data obtained from the peak

summer or winter periods. In the case of the building in Chicago, if the data

acquisition system could be applied for only a period of one month, April or

October would be the best choices. In the case of Albuquerque, March and

October were found to be the best for measuring the whole building electric and

any month from April to September were good for predicting heating energy use

of the building. Combination of information from the utility bills with a short period

of monitored daily energy use data is therefore a good way of acquiring data from

which models can be identified which would predict long term building

performance.

In the hybrid inverse modeling (method 2) the information from only the utility bills

was used to identify models for estimating building energy-use. The advantage of

using some amount of monitored data, as in the case of method 1, for predicting

building performance was clearly demonstrated. Adding only two-months of

monitored data to the utility information was able to improve the prediction results

by almost 10-15%. However, in case no monitored data is available, utility bills

were found to provide a fairly accurate estimate of the long term energy use

patterns of the building energy use, at both monthly as well as annual timescales,

provided there are no marked differences in the internal lights and equipment

loads between weekdays and weekends.

103
To conclude, these findings and proposed strategies might be unacceptable in

cases when one does not have the luxury of waiting until the climatic conditions

are favorable to perform the in-situ tests. In case such an option is available, the

analysis procedures can be used for finding the optimum start time and optimum

length to monitor the energy-use in the building for predicting annual

performance within acceptable accuracy levels. This can result in both time and

cost benefits for the building owner.

In Chapter 5, three application areas for a research of this nature were identified.

Table 9.1 summarizes the different application areas identified and the proposed

research method that can be adopted to determine the best period for carrying

out building energy-use monitoring. In cases where neither simulated nor any

monitored data is available for analysis, the SMLP and DBTA approaches can be

used to estimate the best time of the year and minimum duration for carrying out

in-situ monitoring of building energy use. These methods utilize information from

the weather variables for predicting best time of monitoring. In cases where

simulated building performance data is available, the MCP modeling technique

can be adopted. This method can be used for performance verification of the

newly constructed green buildings for which simulated data is usually available.

In cases where a building has been in operation for some time, and, there is

atleast a year worth of utility data available for analysis, the hybrid inverse

modeling methods can provide information regarding best periods of the year to

begin in-situ monitoring of the building energy-use.

104
Table 9.1

Application Areas Identified and the Proposed Research Methods.

Summary of Cases and Analysis Methods

Type of Baseline Energy
Recommended Analysis
Application Areas Use Data Available for
Strategy
Analysis
Climate Diversity Test +
Weather Variables Only
DBTA / SMLP Approach
Climate Diversity Test +
Weather Variables + DBTA / SMLP Approach
Simulated Data
MCP Modeling
Detailed Audit Climate Diversity Test +
Weather Variables +
DBTA / SMLP Approach
Short Term Monitored
Data
MCP Modeling
Climate Diversity Test +
Weather Variables + DBTA / SMLP Approach
Utility Bills (One Year)+
Green Building Short Monitored Dataset Hybrid Inverse Modeling –
Performance Verification Method 1
Climate Diversity Test +
Weather Variables + DBTA / SMLP Analysis
Utility Bills (One Year)+
Short Monitored Dataset Hybrid Inverse Modeling –
Method 2
Climate Diversity Test +
Pre-Post M&V
DBTA / SMLP Approach
Weather Variables+
Pre-Retrofit Monthly
Hybrid Inverse Modeling –
Data (1 year)+
Method 1/
Post-Retrofit Monitored
MCP Modeling
Data
(depending on availability of
Utility Bills)

105
9.3. Future Work

Broader sample of buildings from diverse geographical locations should be used

to generalize the results obtained. The research methods discussed in this report

can be further refined by analyzing more datasets from actual buildings and of

different building types. The actual building energy-use data analyzed in this

research consisted of the whole building electric use data only. The pre-post

retrofit cases were not investigated at in this research and could form a potential

study for the future. In this study, a MATLAB code was written to obtain the CV

(%) and NMBE (%) indices for the predictions made by the regression models

generated in the Energy Explorer software. The procedure adopted is a bit more

time consuming since the data has to be manually extracted from one software

and fed into another. Automating the entire process would result energy

engineers as well as designers and building operators adopt these approaches in

order to determine best periods of the year to begin in-situ monitoring of their

buildings and the associated lengths of monitoring.

106
 REFERENCES/ BIBLIOGRAPHY

Abushakra, B. (2000). “Short-term Monitoring Long-term Prediction of Energy

Use in Large Commercial and Institutional Buildings”, Ph.D. Dissertation
defended August 2000, Department of Mechanical Engineering, Texas A&M
University.

Abushakra, B., & Mitchell, P. (2010). “RP-1404 measurement, modeling, analysis

and reporting protocols for short-term M&V of whole building energy
perofrmance. phase 1- literature review, monitoring protocol and test matrix”.
Unpublished manuscript, January.

Abushakra, B., Claridge, D. E., & Reddy, T. A. (1999). “Investigation on the Use
of Short-term Monitored Data for Long-term Prediction of Building Energy
Use”. Proceedings of the Renewable and Advanced Energy Systems for the
21st Century, ASME Solar Energy Conference, Hawaii.

ANSI/ASHRAE Standard 105-2007. “Standard Methods for Measuring,

Expressing and Comparing Building Energy Performance”.

ASHRAE Guideline 14-2002. "Measurement of Energy and Demand Savings".

ASHRAE Handbook 2005. "Fundamentals"

ASHRAE Standard 189.1-2009. “Standard for the Design of High Performance,

Green Buildings; Except Low Rise Residential Building”.

ASHRAE Standard 55-2004."Thermal Environmental Conditions for Occupancy".

ASHRAE Standard 62.1-2010. "Ventilation for Acceptable Air Quality".

ASHRAE Standard 90.1-2007. “Energy Standard for Buildings except Low Rise
Residential Buildings”.

ASHRAE/USGBC/CIBSE-2010. “Performance Measurement Protocols for

Commercial Buildings”.

Birt, B., & Newsham, G. R. (2009). “Post-occupancy Evaluation of Energy and

Indoor Environment Quality in Green Buildings: A Review”. 3rd International
Conference on Smart and Sustainable Environments, Delft, Netherlands.

Claridge, D. E. (1998). “A Perspective on Methods for Analysis of Measured

Energy Data from Commercial Buildings”. Journal of Solar Energy
Engineering. 150-155.

Claridge, D. E., Haberl, J. S., Sparks, R. J., Lopez, R. E., & Kissock, K. (1991).
“Monitored Commercial Building Energy Data: Reporting the Results.
Improving Energy Conservation Retrofits with Measured Savings”.

107
Claridge, D. E., Haberl, J. S., Sparks, R. J., Lopez, R.E., Kissock, K., (1991).
“Monitored Commercial Building Energy Data: Reporting the Results”.
ASHARAE Transaction.

Claridge, D. E., Haberl, J. S., Turner, W. C., (1991). “Improving Energy

Conservation retrofits with Measured Savings”. ASHARAE Journal.

Claridge, D. E., Haberl, J. S., Turner, W. D., O Neal, D., Heffington, W., Tombari,
C., & Roberts, M. (1991). “Energy conservation retrofits with measured
savings”.

Dhar, A. et al. 1998. “Modeling Hourly Energy Use in Commercial Buildings with
Fourier Series Functional Forms”, ASME Journal of Solar Energy
Engineering, vol. 120, pp.217-223, August.

Dhar, A., T. A. Reddy and D. E. Claridge (1999). "A Fourier Series Model to
Predict Hourly Heating and Cooling Energy Use in Commercial Buildings
With Outdoor Temperature as the Only Weather Variable." Journal of Solar
Energy Engineering, 47-53.

Federal Facilities Council. (2001). “Learning from our buildings: A state-of-the-

Practice Summary of Post-Occupancy Evaluation”. Washington, D.C.:
National Academy Press.

FEMP, 1996. “Measurement and Verification (M&V) Guidelines for Federal

Energy Projects”, U.S. Department of Energy, March.

Fowler, K. M., & Rauch, E. M. (2008). “Assessing Green Building Performance: A

Post Occupancy Evaluation of 12 GSA Buildings”. Richland, Washington
99352: Pacific Northwest National Laboratory.

Haberl, J. S. (1998). “Baseline Calculations for Measurement and Verification of

Energy and Demand Savings in a Revolving Loan Program in Texas”.
ASHRAE Transaction, 841-858.

Haberl, J. S. (2006). “Scoping study: Protocols for measuring and reporting the
on-site performance of buildings except low-rise residential buildings”.
ASHRAE.

Haberl, J.S., Claridge, D.E., Harrje D.T. The Design of Field Experiments and
Demonstrations, Energy Systems Laboratory, Mechanical Engineering
Department, Texas A&M University, College, Texas and Center for Energy
and Environmental Studies”, Princeton University, Princeton, New Jersey.

Haberl, J.S., Liu, Z.B., Carlos, J., Lynn, B., Underwood, D. “Development of a
Monitoring and Verification (M&V) Plan and Base Line for the Fort Hood”
ESPC Project.

IESNA Lighting Handbook.

108
IPMVP 2001. “International Performance Measurement and Verification
Protocols (IPMVP): Volume I- Concepts and Practices for Determining
Energy and Water Savings”, IPMVP Inc.

IPMVP 2002. “International Performance Measurement and Verification

Protocols (IPMVP): Volume II- Concepts and Practices for Improved Indoor
Environmental Quality”, IPMVP Inc.

IPMVP 2003. “International Performance Measurement and Verification

Protocols (IPMVP): Volume III- Concepts and Practices for Determining
Energy Savings in New Constructions”, IPMVP Inc.

Katipamula, S., & Claridge, D. E. (1998). “Use of Simplified System Models to

Measure Retrofit Energy Savings.” Journal of Energy Engineering, 57-68.

Katipamula, S., Reddy, T. A., & Claridge, D. E. (1995). “Bias in Predicting Annual
Energy Use in Commercial Buildings with Regression Models Developed
from Short Data Sets”. ASME/JSME/JSES International Solar Energy
Conference, 99-110.

Katipamula, S., Reddy, T. A., & Claridge, D. E. (1995). “Effect of time resolution
on statistical modeling of cooling energy use in large commercial buildings”.
ASHRAE Transactions, 101(2), 3894.

Kissock, J. K., Haberl, J. S., Claridge, D. E., (2003). “Inverse Model Toolkit”(RP
1050): Numerical Algorithms for Best-Fit Variable-Base Degree-Day and
Change-Point Models”, ASHRAE Transactions-Research, Vol. 109, Part 2.

Kissock, J. K., Reddy, T. A., Fletcher, D., & Claridge, D. E. (1993). “The Effect of
Short Data Periods on the Annual Prediction Accuracy of Temperature-
Dependent Regression Models of Commercial Building Energy Use”.
Proceedings of the ASME International Solar Energy Conference,
Washington D.C. 455-463.

Kissock, K., Mc Bride, J. (1998). “The effects of Varying Indoor air temperatures
and Heat Gain on measurement of Retrofit savings”.

Kreider, J. F., Curtiss, P., & Rabl, A. (2002). “Heating and Cooling of Buildings:
Design for Efficiency” (2nd ed.). New York: McGraw-Hill.

Liu, Z.B., Haberl, J.S., Carlos, J., Cho, S., Lynn, B., Cook, M. “Methodology for
Analyzing Energy and Demand Savings from Energy Services Performance
Contract Using Short Term Data”.

Nayak, J. K., Prajapati, J.A., (May 2006). Handbook on Energy Conscious

Buildings. (Pilot ed.). Prepared under the interactive R&D project between
IIT, Mumbai and Solar Energy Center, Ministry of Non-Conventional Energy
Sources, Government of India.

109
NIBS Guideline 3-2006. “Exterior Enclosure Technical Requirements for the
Commissioning Process”.

O’ Neal, D., Bryant J., “Building Energy Instrumentation for Determining Retrofit
Savings: Lessons Learned”. Preprint November 1991. Published in the 1992
proceedings of the ASME/JSES/KSES International Solar Energy
Conference.

O’ Neal, D., Bryant,J., Turner,W.D., & Glass, M. J. (1990). “Metering and

Calibration in Loan STAR Buildings”. Seventh Annual Symposium on
Improving Building Systems in Hot and Humid Climates, Fort Worth, Texas.

Reddy, T. A., & Claridge, D. E. (2000). “Uncertainty of Measured Energy Savings

from Statistical Baseline Models”. International Journal of Heating,
Ventilating, Air-Conditioning and Refrigerating Research.

Reddy, T. A., Haberl, J. S., & Elleson, J. S. (1999). “Engineering Uncertainty

Analysis in the Evaluation of Energy and Cost Savings of Cooling System
Alternatives Based on Field Monitored Data”.

Reddy, T. A., Haberl, J. S., & Elleson, J. S., “Methodology Development for
Determining Long-Term Performance of Cool Storage Systems from Short-
Term Tests”. ASHRAE Transactions: Symposia.

Reddy, T. A., Kissock, J. K. & Ruch, D. K., (August 1998). “Uncertainty in

Baseline Regression Modeling and in Determination of Retrofit Savings”.
Journal of Solar Energy Engineering, 185-192.

Reddy, T. A., Kissock, J. K., & Claridge, D. E. (1992). “Uncertainty Analysis in

Estimating Building Energy Retrofit Savings in the Loan STAR Program.
ACEEE.

Reddy, T. A., Kissock, J. K., & Ruch, D. K. (1998). “Uncertainty in Baseline

Regression Modeling and in Determination of Retrofit Savings”.

Rodriguez, J. B. (2003). “Energy and Environmental Assessments for Buildings

in Developing Countries”.

Romm, J.J., Browning, D.W., (1998). “Greening the Building and the Bottom
Line: Increasing Productivity through Energy-Efficient Design”.

Ruch D.K., Claridge, D.E.,”A development and comparison of NAC Estimates for
linear and change point energy models for commercial buildings”.

Ruch, D., Claridge, D. E. (1992). “A Four-Parameter Change Point Model for

Predicting Energy Consumption in Commercial Buildings” Journal of Energy
Engineering, 72-77.

110
Scofield, J. H., (2009). “A Re-Examination of the NBI LEED Building Energy
Consumption Study”. Proceedings of Energy Program Evolution Conference,
Portland.

Sylvester, K.E., Song, S., Haberl, J. S., Turner, W. C.,(2002). “Sustainability

Assessment of the Robert E. Johnson State Office Building”. Prepared for
the Texas State Energy Conservation Office and The General Service
Commission of Texas.

Torcellini, P. A., Pless, S., (August 2005). “Evaluation of the Energy Performance
of Six High-Performance Buildings”. Proceedings of the ISES Solar World
Congress, Orlando, Florida.

Turner, W. C. (2001). Energy management handbook (4th ed.). Lilburn, GA:

Fairmont Press.

USGBC Green Building Council, “Leadership in Energy and Environmental

Design” (LEED).

Wang, J., Reddy, T.A., Claridge, D.E., Haberl, J.S. and Pollard, W. “Data Quality
Requirements for determining Energy Savings in the Weatherization
Assistance Program (WAP).”

Web References:
http://www.assetsinuse.org
http://www.usgbc.org/
http://www.grihaindia.org/
http://www.energystar.gov/
http://www.newbuildings.org/
http://www.usablebuildings.co.uk
http://www.ashrae.org
http://www.gsa.gov

111
 APPENDIX A
DESCRIPTION OF BUILDINGS ANALYZED

A1. Large Hotel Chicago, Illinois

Description extracted from final report of ASHRAE RP 1340 (Maor, Reddy

et al. 2009).

Type of Building: Synthetic

Energy Use Channels Analyzed: Whole building Electric, cooling energy use,

heating energy use.

A synthetic DOE 2.1 E building energy simulation model is developed for the

large, 650-guest room hotel. The location selected for the hotel is Chicago, IL,

which corresponds to area 5A in the geographical locations established for this

research project. The building is a forty three story 619,200 ft2, rectangular

shaped building with 8 thermal zones. Floor 1 is lobby, shops and restaurants.

Floors 2, 3, and 4 accommodate conference rooms, banquet and offices. Floors

5 thru 42 are guest rooms in a perimeter – core layout where the core includes

corridors, shafts and service rooms. Floor 43 accommodates mechanical rooms

and service areas. Building envelope properties, systems efficiencies, etc. are

based on typical design practices for the late 1980. Sections of ASHRAE 90.1-

2004 minimum requirements used as well.

Operating schedules (lighting, occupancy, etc.) are based mainly on data from

ASHRAE 90.1-1989. Although a variety of secondary air systems (such as VAV,

Fan Coils etc.) can be found in large hotels, for simplicity, Variable Air Volume

(VAV) with Hot Water (HW) reheat is used in the lobby, conference rooms, and

other administrative areas. Four Pipe Fan Coils (FPFC) units (chilled water and

hot water) used for the guest rooms. The guest rooms floors (5-42) core areas

112
served by a 100% OA, Dedicated Outdoor Air System (DOAS) which comprises

a Reheat Fan System (RHFS). The intent of this approach is to allow introduction

of pretreated OA to each guest room directly, provisions in the model applied to

decouple the core cooling and heating load and the need of constant supply of

OA to the guest rooms. The chiller plant has a cooling capacity of 1,400 Ton,

which comprise of two (2) 700 Ton, Water Cooled, Electric, Centrifugal Chillers).

The Hot Water heating plant is 20,086 MBtu /h, which comprises two (2) 300

BHP or 10,043 MBtu/h, gas fired boilers.

Table A1.1

Building Description: Large Hotel, Chicago, Illinois

General Description
Location Chicago, IL
Floor Area (Sqft) 619,200
Above Grade Floors 43
Below Grade Floors 0
% Conditioned and Lit 100
Geometry
Footprint Shape Square (120’ x 120’)
Zoning (1st thru 4th floor) Each floor is single zone
Zoning (5th thru 42nd floor) 4 Perimeter /1 Interior
Zoning (43rd floor) Single zone
Perimeter Depth (Feet) 20
Floor to Floor Height (Feet) 13
Floor to Ceiling Height (Feet) 9
Envelope
Roof Massive, R-27
Walls Glass Curtain Wall, U=0.11(Btu/h-Ft^2-Deg F)
Foundation Slab, U=0.025 (Btu/h-Ft^2-Deg F)
Double Glazing Reflective=0.55 (Btu/h-Ft^2-DegF),
Windows
SC=0.41
Windows to Wall Ratio (%) 36.0
Exterior and Interior Shades None
Schedules
Operation schedule 24/7
Secondary Systems
Lobby, Conf. Rooms, offices VAV with HW reheats.
Guest Rooms. Four Pipe Fans Coils
Guest Rooms – DOAS/ Vent. Reheat Fan System
Mechanical Room 43rd floor SZ

113
The building loads of interest are the whole-building energy, thermal cooling

loads, the thermal heating loads and lights & equipment load (non-cooling

electrical loads). The hourly values of these four loads in conjunction with the

climatic data (such as ambient air dry-bulb temperature) are available from the

building energy simulation program results. The electric loads vary from about

550 – 1,400 kW from daytime to nighttime and from weekday to weekend with no

seasonal trend. On the other hand the thermal loads vary widely from season to

season.

The graphs in Figure A1 are scatter plots of the daily energy-use over a year for

the four energy channels mentioned above.

Large Hotel, Chicago- IL : Whole Building Energy-Use Large Hotel, Chicago-IL : Energy Consumption (LTEQ)
1600 1600

1400 1400
Electricity Consumption (kW)

Electricity Consumption (kW)

1200 1200

1000 1000

800 800
600 600
400 400
200 200
0 0
-10 0 10 20 30 40 50 60 70 80 90 100 -10 0 10 20 30 40 50 60 70 80 90 100
Outdoor Dry-Bulb Temp (deg F) Outdoor Dry-Bulb Temp (deg F)

Large Hotel, Chicago- IL: Cooling Thermal Use Large Hotel, Chicago- IL: Heating Thermal Use
16000 16000
14000 14000
12000 12000
Chilled Water (MBTU/hr)

10000
Hot Water (BTU/hr)

10000
8000 8000
6000 6000
4000 4000
2000 2000
0 0
-10 0 10 20 30 40 50 60 70 80 90 100 -10 0 10 20 30 40 50 60 70 80 90 100
Outdoor Dry-Bulb Temp (deg F) Outdoor Dry-Bulb Temp (deg F)

Figure A1.1 Scatter plots of the daily energy-use over a year for the energy use

channels for Hotel at Chicago, IL.

114
A2. Office Building, Albuquerque, NM

Type of Building: Synthetic

Energy Use Channels Analyzed: Whole building electric, heating energy use.

The Office Building, located in Albuquerque, New Mexico is a two storied building

with a net conditioned area amounting to approximately 17,430 sqft. Full year

simulated data for whole building electric and heating energy use channels are

available for analysis. The basic details of the size and orientation are

summarized in Table A2.1 and the images of the simulation model are provided

in Figure A2.1.

Table A2.1

Building Description for Office at Albuquerque, NM.

Building Description
Location Albuquerque, NM
Total Conditioned Area 17,430 sqft
Aspect Ratio 2.1 to 1
Orientation Long axis faces east/west
Floor to Floor Height/ Floor to Ceiling Height. 12ft
Window Area 1497 sqft

FigureA2.1: Simulation Model Images for the Office at Albuquerque.

115
Table A2.2

Detailed Description of the Office Building at Albuquerque, NM.

S.No Description As Designed
1 Building Description
Location Alburqueque, New Mexico
Total conditioned area 17,430 sqsft
Aspect ratio 2.1 to 1 (~132ft x 64ft)
Orientation long axis faces east/west
Floor‐to‐floor / Flr to Ceiling Ht. 12 ft/9ft
Window Area 1497 sqft
2 Exterior Wall Description
opaque wall U‐value U=0.061
wall finish color 0.5 (medium light)
3 Roof Construction
roof U‐value (R‐value) 0.031 (32.3)
roof exterior finish color 0.5 (medium light)
4 Window‐ Door Type
Glass Type Dbl Pilkington CTG 6851
U‐value (incl. frame effects) 0.34, ctr glass (0.55 w bad frm)
solar heat gain coefficient 0.38 (SC=0.44)
window wall ratio 15%
5 Window Shading
exterior shading 3 ft overhang at south & SW
interior shading None
6 Interior Wall Type
2x4in Metal Stud, 24in o.c
cavity fill batt nominal 4 in fiberglass batt (R‐
7 Floor Construction
First floor 6 inch slab on grade with carpet
Second floor 4 in concrete with carpet
Perimeter insulation 2 ft vertical ext. R‐5 EPS
8 Cooling System Type
Package Rooftop VAV, HW
area served all but electrical rooms
capacity 67 tons total, 62 tons sensible
supply CFM 24000 CFM
VAV terminal minimum flow 0.4
fan Control VSD
cooling / heating setpoint 74F / 70F
Unoccupied Set Point 82F / 64F
outside air (OA) supply 15 cfm/person (~7.5% of design
supply air temperature 55 F minimum
SAT reset resest by zone demand to 68F
fan static, supply/return 3.75/1.5 in WG
economizer high limit 65F
cooling EER 11.2
116
 Simulation 1
S.No Description As Designed
9 Heating System Type
boiler efficiency 0.85
HW loop design operation 40F DT, 180F setpt, reset to 140F
HW pipe UA 153 Btu/hr‐F
HW preheat coil 45F mixed air T
terminal reheat delta T 35F
10 DHW System Type
Electric Tank (50 gal)elec 2 ‐ 6 kW elements
DHW setpoint 126F
DHW peak demand 6.5 gal/hr (measured)
DHW recirculation 2 gpm recirc loop pump
11 Schedules
Occupancy Schedule 7am ‐ 5pm M‐Th, 1/2 on Fri
Lighting & Equipment Schedule 7am ‐ 5pm M‐Th, 1/2 on Fri

Fan Schedule 6am ‐ 6pm WD

Seasonal Occupancy Closed: Dec 24‐ Dec 31

Lighting Power Density (W/sf) ATE598 Example Bldg Ltg Loads
Receptacle Load ATE598 Example Bldg Plug
Loads

Number of Occupants actual: 90 (~200); design: 122

Occupant Sens/Lat (Btu/per) 250/200

The building loads of interest are the whole-building energy, thermal heating

loads and lights & equipment load (non-cooling Electrical loads). The hourly

values of these four loads in conjunction with the climatic data (such as ambient

air dry-bulb temperature) are available from the building energy simulation

program results. The graphs below show the annual energy-use profiles for the

three energy channels mentioned above. The data shown is at daily timescale.

117
 WBE Data : Office Building Alburquerque, New Mexico LTEQ Data : Office Building Alburquerque, New Mexico
60 30

Electricity Consumption (kWh)

Electricity Consumption (kWh)
50 25

40 20

30 15

20 10

10 5

0 0
0 20 40 60 80 100 0 20 40 60 80 100
DBT (deg F) DBT (deg F)

HW Data : Office Building Alburquerque, New Mexico

70000

60000

50000
Hot Water (Btu/hr)

40000

30000

20000

10000

0
0 20 40 60 80 100
DBT (deg F)

Figure A2.2 Scatter plots of the daily energy-use over a year for the energy use

channels for Office at Albuquerque, NM.

118
A3. Full Service Hotel, Washington DC Region

Type of Building: Actual (Whole Building Electric Utility Data for 2009 available)

Energy Use Channels Analyzed: Whole building Electric.

The data used for analysis is of a full-service, large hotel. Intensive metering and

monitoring effort aimed at understanding energy end use patterns in the hotel

was undertaken in order to come up with retrofit strategies to reduce its energy

consumption by atleast 30%. About 10% of the guest rooms (32), as well as

circuits for most of the end uses in public spaces (lighting, LTEQvators, air

handlers and other HVAC system components, and various equipment), were

equipped with meters.

Located at the Washington DC Region, USA, the building is a twelve-story, full-

service hotel with over 300 guest rooms and gross floor area of 212,000 ft². The

building energy systems are somewhat outdated. The hotel has an energy

management system (EMS) that is not connected to all building energy systems

(e.g., the newly-installed chiller, exhaust fans) and is not being used in a fully-

functional way. Also, single-pane windows and un-insulated exterior masonry

walls result in a thermally-inefficient building envelope. Most of the packaged

terminal heat pumps (PTHP) serving the guest rooms are 7-years old or more.

The baseline energy use of the project was measured and a detailed study of the

end-use energy consumption patterns was conducted. The metering provided a

detailed and accurate picture of energy consumption at this particular property.

The end use consumption data from this hotel can be considered typical of the

large hotel sector in general since the guest room equipment and various other

features of the hotel are typical of large hotels in the United States.

119
The analysis of the 2009 Electricity interval data from the utility shows a relatively

flat daily load curve in all the seasons across all days of the week. The lack of

variation between weekdays and weekends is not surprising, given the hotel has

a very high occupancy rate and serves both tourists (including weekend stays)

and business guests (typically during the week).

The graphs in Figure A3.1 below is a scatter plot of daily whole building electric

use versus outdoor DBT for the whole year of 2009. One notes that the data is

well behaved with a clear 4P change point behavior.

Full Service Hotel, Whole Building Electric Data

1600
1400 WBE
1200
WBE (kWh)

1000
800
600
400
200
0
0 10 20 30 40 50 60 70 80 90
Outdoor Dry Bulb Temperature

Figure A3.1 Scatter plot of the daily energy-use over a year for the WBE energy

channel for a Hotel at the Washington DC area, USA.

120
 APPENDIX B

POST OCCUPANCY EVALUATION: IEQ, ACOUSTICS, WATER

Post Occupancy Evaluation involves the measurement of the performance of a

building in actual use in terms of the following aspects:

i. Energy Use,

ii. Indoor environmental quality (IEQ): Thermal comfort, Indoor air quality,

Lighting/daylighting,

iii. Acoustics, and

iv. Water Use

Published work on post-occupancy performance of green buildings in terms of

energy use has been reviewed in Chapter 4. The other aspects, namely, IEQ,

acoustics and water are briefly discussed below.

B1. Indoor Environmental Quality (IEQ) Measurement

B1.1. Thermal Comfort

a. Background

Comfort conditioning is the largest component of energy consumed in buildings

and it accounts for nearly 15-20% of the total national energy use in the United

States. Thermal discomfort can be a major source of dissatisfaction for the users

of a building (Huizenga et al., 2006a). The physiological and psychological bases

of thermal comfort have been studied for many years now; however, few

benchmarks or standard procedures for evaluating a building’s thermal

performance can be found.

Thermal comfort has largely been studied in controlled laboratory experiments

which often do not represent realistic building environments. Due to this reason,

121
over the past couple of decades, many research projects involved taking

measurements of physical environment in occupied buildings and analyzing such

data.

Most high performance buildings aim to improve occupant comfort while reducing

the energy use associated with indoor environmental conditioning.

ASHRAE Standard 55: Thermal Environmental conditions for Human Occupancy

(ASHRAE 2004), relates to thermal comfort. The standard specifies the warm

and cool temperature limits for the thermal comfort as affected by four

environmental parameters, namely, air temperature, radiant temperature,

humidity and air movement and two personal variables, namely; clothing

insulation and metabolic activity level (Figure B1.1).

Figure B1.1 ASHRAE Standard 55 Comfort Zone (Source: ASHRAE 2004)

Recently, there have been a number of field studies in office buildings and

schools in which the occupants’ surveyed comfort responses were matched with

simultaneous laboratory-grade measurements of their nearby thermal

environment (Leaman and Bordass, 2006). Other occupant surveys use

questionnaires alone to directly measure occupant satisfaction with the quality of

the indoor environment, and to address the specific causes of discomfort in real

building environments.
122
b. Thermal Comfort Measurement Protocols: Basic, Intermediate & Advanced

The measurement protocol adopted for evaluating the performance of any indoor

environment is based on the complexity of the environment. Conventional and

uniform environments and occupancies can be adequately evaluated by basic

level techniques, while non-uniform environments and unusual occupancies may

need evaluation at the advanced level.

In the basic level, physical measurements of thermal conditions are optional and

are often employed only to discover the causes of problems that may have been

identified by occupant surveys. Under this level, the instruments used are

generally hand-held readout instruments, e.g., portable thermometer

(temperature), Electric or sling psychrometer (humidity), globe thermometer

(MRT), IR gun (surface temperature), shortwave-spectrum pyranometer (solar

gain), handheld anemometer (air speed), smoke tube/ chalk dust puffer (airflow),

etc.

The intermediate level protocol involves monitoring the physical environment and

concurrent ‘right-now’ occupant perceptions of the thermal environment. This

type of monitoring is used for comfort prediction, commissioning and controlling

environmental control systems and for identifying and correcting faults more

precisely than the basic level. The methods and tools adopted under this

category are: background data, ‘right-now’ subjective surveys and actual physical

data collection ranging from something as simple as reading the thermostat to

detailed workstation sensing.

The advanced level covers detailed quantification of complex comfort

environments and environmental control systems affecting the occupants.

Numerous measurement and analysis approaches can be adopted under this

123
performance measurement category. Such detailed measurements may impose

upon the occupants or disrupt normal work. The method can be used for the

analysis of asymmetrical environments (temperature stratification, radiant heating

and cooling systems, solar radiation impacts on occupants through fenestration

etc.) and transient environments (occupant experience in a series of different

environments when moving, as in transit stations, etc.).

Usually an indoor climate monitoring station is installed on the occupant desk. It

measures temperature, mean radiant temperature, relative humidity, nearly omni-

directional air speed, plane radiant temperature, etc. Solar pathfinder, fish eye

camera are used for predicting annual sunpaths at a given point indoors. Heated

object sensor can emulate the human body’s heat transfer via radiation and

convection. Vertical temperature profiles are measured using sensor string and

moving sensor arm and surface temperatures can be measured with an IR spot

meter and imaging. Air speed (omni-directional, directional, turbulence, wave

form fluctuating flow) can be measured using various types of anemometers.

Source: Performance Measurement Protocols for Commercial Buildings

(ASHRAE 2010)

B1.2. Indoor Air Quality (IAQ)

a. Background

Good Indoor Air Quality (IAQ) requires the control of contaminants in the air

supply. These contaminants may come from outside or from sources within the

building, e.g. equipment, activities, people and materials. The commonly followed

strategies to control the contamination of inside air include: general ventilation,

local exhaust ventilation, filtration, isolation or capture. In the context of

measurement of IAQ performance, two aspects need to be considered: health

124
impacts of IAQ and the comfort or perceived IAQ (Indoor Air Quality Guide: Best

Practices for Design, Construction, Commissioning (ASHRAE 2009b)). The basic

and intermediate levels of performance measurement of IAQ focus on

“perceived” IAQ. Three fundamental characteristics are assessed while

measuring the IAQ performance. These are: building system characteristics

(HVAC characteristics), environmental factors (thermal conditions, presence of

contaminants, acoustics and illumination) and occupant perception (ASHRAE

Standard 62.1, Ventilation for Acceptable Indoor Air Quality).

b. IAQ Measurement Protocols: Basic, Intermediate & Advanced

Basic measurements of IAQ are used to obtain information about the building, its

environment, and the occupant responses to the environment. The

measurements determine the quality of air at the project site methods such as

testing the quality of air on site, conducting facility pre-evaluations and surveys-

review of complaint logs and similar unstructured reports. Previous occupant

surveys can be used as the starting point for IAQ performance evaluation,

reviewing operational documentation of the facility and site assessment to

evaluate the condition of the building systems (HVAC, mechanical equipment,

OA damper operation, HVAC distribution system, coil cleanliness, filter

specifications etc.). Occupant surveys are often used to determine occupant

satisfaction with IAQ. Surveys include questions regarding the perceptions of

fresh air, stuffiness, presence of odors, temperature/humidity adequacy etc. In

order to ensure compliance with ASHRAE Standard 62.1, ventilation rates should

be measured at the OA intake to each HVAC fan system.

125
If combustion sources are present in or near the building, Carbon dioxide

concentrations should be measured in occupied spaces and compared with the

EPA ambient air levels (EPA, 2008b).

Intermediate measurements are conducted to present a level of IAQ evaluations

that give a higher level of confidence regarding the quality of indoor air. In

addition to the tools specified for the basic level, the intermediate level uses

additional analysis to obtain information about the building, its indoor

environment and the occupant responses to that environment. The building site

and the immediate surroundings are surveyed for local contaminants that may be

of concern if allowed to enter the building. Smoke tests and differential pressure

measurements are carried out to evaluate airflow patterns and duct leakages that

might have occurred. Potential contaminant sources are analyzed and evaluated.

This level also aims at measuring the OA flow rates and building pressure

differential to confirm that the exhaust ducts are at negative pressure to avoid

cross contamination.

The advanced measurements are for the purpose of pollutant identification and

source control. This is helpful in establishing the baseline and long-term

measurements for pollutants and also helps identify the events that might need

investigation and/or corrective action. In addition to the intermediate level

measurements, the advanced measurements report baseline and continuously

measured carbon dioxide concentrations, fine particulates, and TVOC’s as well

as look at any corrective action that may be required to be taken.

Source: Performance Measurement Protocols for Commercial Buildings

(ASHRAE 2010)

126
B1.3. Lighting/ Daylighting

a. Background

Lighting/ Daylighting has a very crucial role to play in the quality of the built

environment. The right amount of lighting can improve worker productivity,

enhance aesthetic appeal of the space, improve tenant retention, increase retail

sales, facilitate education and create the desired mood for visitors. Lighting

quality is defined as the degree to which luminous environment supports the

visual performance, social interaction & communication, health & safety and

aesthetic judgment of the occupants of a space (IESNA Lighting Handbook).

b. Lighting/ Daylighting Measurement Protocols: Basic, Intermediate &

Advanced

Basic measurements of the Lighting/Daylighting in the building are used to

determine the building occupant satisfaction with the lighting, comparing it

against benchmarks and identifying problems. Spot measurements of the

important photometric parameters are taken to assess the light quality in space.

The methods adopted generally include: occupant surveys (Center for the Built

Environment, CBE, has developed the Occupant Surveys that cover various

issues related to the Indoor Environmental Quality). At this level one often takes

spot measurements of the illuminance levels in the space and compares them

with recommended levels for the particular space. Spot measurements can help

the building owner discover areas within the building where the light levels are

not appropriate for the tasks being performed.

The goal of the intermediate methods is quite similar to the basic measurements

but a greater resolution is sought. The problems identified at the basic level are

diagnosed and solutions are developed. The parameters measured and

127
calculated include: Illuminance, luminance and luminance ratios and discomfort

glare. Unlike the spot measurements taken at the basic level, the intermediate

level requires the use of a regular grid of measurement points is used to

completely cover the surface under consideration. Typically, the spacing between

measurement points is set to one-fourth the spacing between luminaires. The

height of the points depends on where the primary task is performed. In addition

to actual measurements, occupant surveys can be used to evaluate the day-to-

day satisfaction of the building’s regular users with the lighting systems. The

survey generally deals with both lighting as well as daylighting.

The advanced level protocol is intended to provide high-resolution data gathering

for lighting-critical situations. The metrics mentioned in the intermediate level are

dealt with in far more detail. The conventional methods, if used, are generally

extremely time consuming and laborious. A relatively new technique, HDR

photography is now being used as a quick, high-resolution approach to evaluate

this issue.

Source: Performance Measurement Protocols for Commercial Buildings

(ASHRAE 2010)

B2. Acoustics

a. Background

Of all the factors studied for performance evaluation of a built environment,

acoustic performance can be the cause of greatest dissatisfaction for the space

users. Acoustic performance requirements are embodied in the types of activities

that the occupants engage in during their time spent in these spaces. The

building’s acoustical environment is made up of internal sound contributions from

128
occupant conversations, HVAC equipment, electrical equipment and plumbing

systems as well as from external sound penetrating the building envelope.

Numerous acoustic standards are available for the measurement of building

elements and mechanical systems. However, there are few standards being

mandated for the actual performance of architectural spaces as it relates to

human perception and performance. A majority of buildings built in the past few

decades have been shown to be acoustically challenged, and often green

decisions can lead to bad acoustic results (Jensen and Arens, 2005; Salter and

Waldeck, 2006; Abbaszadeh et al., 2006). Many green design guidelines and

rating systems have now begun address acoustic design needs for high

performance buildings.

b. Acoustics Measurement Protocols: Basic, Intermediate & Advanced

Basic measurements help assess the acoustic annoyance that may affect the

study and work performance in an environment. The basic methods of evaluating

the acoustic performance of commercial buildings include occupant surveys,

survey of the A-weighted sound pressure level that can help determine the

background noise in the room.

The acoustic measurements should be conducted with the room vacated by its

normal occupants. All the non-HVAC related equipment (computers/radios etc.)

should be turned off for the duration of the measurements. If possible the

measurement should be conducted with the system operating at full capacity, i.e.

maximum cooling for VAV system. In cases when there is intruding noise from

outside sources, testing should be scheduled during time periods when these

sounds are at maximum. If windows are designed to be opened for ventilation,

measurements are to be taken with and without windows open.

129
The intermediate level looks at accurate assessment of the background noise

and reverberation time in a room. The objective is to assess acoustic annoyance

that would affect study and work performance as well as speech and telephone

communication, listening conditions and privacy. These measurements should be

conducted by an acoustical consultant or personnel with training in acoustic

testing. The procedure requires measurement of background noise in octave

bands. Acoustic performance criteria vary from room to room and are based on

the type of activities that occupants engage in during their time spent in these

spaces. The overall result is that any commercial building will typically have

some spaces that are acceptable and others that are not acceptable. If more

than 90% of the room background noise measurements are found to be

acceptable, then the building may be considered acceptable.

The buildings that claim to meet high levels of acoustic performance require a

much higher level of proof of achieving those goals. This level of evaluating

acoustic performance has to be performed by experienced acoustical

consultants. Advanced measurements provide accurate assessment of speech

privacy, speech communication and isolation from intruding noise. The

instrumentation of the Intermediate level fulfill most of the requirements of the

advanced level.

Sound isolation or acoustic privacy can be differentiated into two categories:

freedom from intrusive noise, such as outside traffic, telephones ringing,

footsteps on the floor above, and speech privacy, or freedom from being

overheard and from over hearing others. The acoustical performance of many

building elements such as walls, windows, and doors are initially evaluated by

manufacturers in acoustical laboratories.

130
B3. Water Use

a. Background

The rising shortage of quality potable water supplies all over the globe has

resulted in an increased interest in water conservation strategies. Building

owners and operators are beginning to actively monitor and control the water use

in their buildings to conserve water reserves in their regions. Potable water

requires a great deal of energy use for its treatment, pumping and transport.

Similarly, waste water also requires energy for transport and treatment.

Conserving water can thus benefit in two ways, save water as well as save

energy.

Water-Use Measurement Protocols: Basic, Intermediate & Advanced

The objective of ‘Basic’ measurements is to determine the periodic and annual

total water use and cost delivered from the utility and used on the building site.

The lowest cost and easiest method of measuring total water usage is to use

utility water meter data shown on the building’s water utility bill. The basic level

simply measures the volume of water metered at the site and its cost on monthly

and annual basis.

In addition to tracking monthly and annual site water use, the intermediate

measurement level segregates the use and cost into the portion that enters the

wastewater system and that which does not. In addition to all measurements in

the basic level, the intermediate measurement provides improved feedback since

water usage from each flow stream is determined. This level uses the utility water

meter and a separate landscape water meter.

Based in the utility bills and landscape meter readings, water saving strategies

for each flow can be determined and then evaluated in the next billing cycle.

131
The objective of advanced level measurements is to measure as many separate

water-using components as possible to better determine usage patterns of the

total site water use. The following water meters can be used: Main utility water

meter, landscape water use, HVAC equipment water use (cooling tower, boilers),

water use for swimming pools & water features, kitchen water use, hot water use,

water use for process investment, recycled gray water use, etc. The measured

values are compared and evaluated over time to determine if the water saving

strategies adopted are effective or not.

Source: Performance Measurement Protocols for Commercial Buildings

(ASHRAE 2010)

132
 APPENDIX C

ANALYSIS DATA

C1. Regression Models for the MCP Approach

(Refer to Section 6.3 and Sections 7.6.1, 7.6.2 and 7.6.3)

The analysis technique employed in the MCP Approach is to predict whole

building Electric (WBE), cooling energy use (CHW), and heating energy use

(HW) using multiple linear regression (MCP) models derived from the short data

sets. The MCP models for the three energy-use channels are generated using

daily averages for outdoor dry bulb temperature (DBT) and lighting & equipment

load (LTEQ) as the regressors using Energy Explorer software. The models and

plots obtained for the three buildings used for analysis are given in the following

sections.

C1.1. Large Hotel at Chicago, Illinois

a. Time Series and Scatter Plots

Figures C1.1.1 to C1.1.6 are examples of the time series plots for the three

energy channels. The blue portion on the graphs represents the short data-set

used to estimate the energy use over the whole year represented by the red

portion on the graphs. The time-series graphs were plotted for each of all the

base periods shown in Table C1.1.1 Typical examples for the three energy

channels; WBE, HW and CHW are shown below.

133
FigureC1.1.1 Time Series Plots for WBE from Month of October Used as the Short Data-

Set for the Large Hotel, Chicago, IL

FigureC1.1.2 Time Series Plots for CHW for Data from Month of October Used as the

Short Data-Set for the Large Hotel, Chicago, IL

134
FigureC1.1.3 Time Series Plots for HW from Month of October Used as the Short Data-

Set for the Large Hotel, Chicago, IL

Examples of the scatter plots of the MCP models for the three energy channels

for Large Hotel, Chicago, IL are shown in Figure C1.1.4 to C1.1.6.

FigureC1.1.4 MCP Scatter Plot for WBE versus DBT & LTEQ from Month of October

Used as the Short Data-Set. Building: Large Hotel, Chicago, IL

135
FigureC1.1.5 MCP Scatter Plot for CHW versus DBT & LTEQ from Month of October

Used as the Short Data-Set. Building: Large Hotel, Chicago, IL

FigureC1.1.6 MCP Scatter Plot for HW versus DBT & LTEQ from Month of October Used

as the Short Data-Set. Building: Large Hotel, Chicago, IL

The blue points on the graphs represent the baseline data used to identify the

model and the red points indicate the whole year data. The scatter graphs shown

above were plotted for each of all the base periods shown in Table C1.1.

136
 b. Multivariate Change Point Models

Table C1.1.1

MCP Models for all Base Periods and Energy Channels for Large Hotel at Chicago, IL.

S.No. Base Period Channel Large Hotel, Chicago, IL: Model Equations (4P) R^2 RMSE CV-RMSE

1 January WBE WBE=62.32 - 16.64(0.88-DBT) + + -0.49 (DBT-0.88)+ + 1.01LTEQ 0.95 30.46 3.21%

CHW CHW=405.01 - 113.21(0.83 - DBT) + + -3.27(DBT-0.83) + + 0.09LTEQ 0.04 200.89 50.06%

HW HW = 9737.36 - -159.06(28-DBT)+ + -97.76(DBT - 28) + + -2.65LTEQ 0.73 1147.36 13.94%

2 Jan-Feb WBE WBE = 63.67 - 0.40 (38.01-DBT)+ + -3.56 (DBT - 38.01)+ +1.00 LTEQ 0.95 29.28 3.10%

CHW CHW = 416.93 - 2.60(38.01 - DBT)+ + -23.15(DBT - 38.01)+ + 0.03LTEQ 0.03 193.17 48.69%

HW HW = 13295.80 - -36.12(12-DBT)+ + -148.66(DBT - 12.00) + + -2.96 LTEQ 0.68 1076.05 12.57%

3 Jan-March WBE WBE = 53.83 - 0.18(46.01 - DBT)+ +8.74(DBT - 46.01)+ +1.01LTEQ 0.95 27.92 2.95%

CHW CHW = 358.29 - 1.18(48.02 - DBT)+ +87.61(DBT - 48.02)+ + 0.06LTEQ 0.31 184.43 45.33%

HW HW = 13055.24 - -42.07(11.99 - DBT)+ + -150.09(DBT-11.99)+ + -2.74LTEQ 0.76 900.78 11.18%

4 Jan-April WBE WBE = 60.98 - 0.31(44.68 - DBT)+ + 7.73(DBT -44.68)+ + 1.01 LTEQ 0.96 26.8 2.83%

CHW CHW = 408.18 - 2.29(45.88 -DBT)+ +71.21(DBT - 45.88)+ + 0.04 LTEQ 0.55 178.41 37.27%

HW HW = 12600.07 - -17.61(13.01 - DBT)+ + -186.53(DBT - 13.01)+ + -1.92LTEQ 0.85 945.46 13.44%

5 Jan-May WBE WBE = 55.69 - 0.31(44.07 - DBT)+ + 7.40(DBT-44.07)+ +1.01LTEQ 0.97 25.46 2.62%

CHW CHW = 606.37 - 9.73(53.40 -DBT)+ +164.63(DBT - 53.40)+ +0.07LTEQ 0.91 233.52 30.82%

HW HW = 12423.68 - -16.97(13.55 - DBT)+ + -197.34(DBT-13.55)+ + -1.71 LTEQ 0.91 975.86 16.60%

6 Jan-June WBE WBE = 48.04 - 0.32 (46 - DBT)+ + 9.50(DBT-46)+ +1.02 LTEQ 0.97 28.05 2.78%

CHW CHW = 512.93 - 11.79(55.53 - DBT)+ +243.90(DBT - 55.53)+ +0.27LTEQ 0.95 402.62 28.91%

HW HW = 1733.56 - -197.87(64.65 -DBT)+ + -15.03(DBT-64.65)+ + -1.21 LTEQ 0.93 941.6 18.87%

7 Jan-July WBE WBE = 43.26 - 0.50(48.37 - DBT)+ +11.47(DBT - 48.37)+ +1.03LTEQ 0.97 32.2 3.08%

CHW CHW = 372.60 - 12.74(56.41 - DBT) + +289.69(DBT - 56.41)+ +0.47LTEQ 0.96 529.03 24.66%

HW HW = 1497.93- -197.81(64.99 - DBT)+ + -13.50(DBT - 64.99)+ + -1.01 LTEQ 0.94 870.73 20.13%

8 Jan-August WBE WBE = 36.17 - 0.49(48.67 -DBT)+ +12.07(DBT - 48.67)+ +1.04LTEQ 0.97 33.32 3.12%

CHW CHW = 131.31 - 11.62(55.70 - DBT)+ +294.97(DBT -55.70)+ +0.70LTEQ 0.95 575.32 21.88%

HW HW = 1390.06- -198.54(64.98 - DBT)+ + -12.53(DBT - 64.98)+ + -0.92LTEQ 0.95 814.58 21.24%

9 Jan-Sept WBE WBE = 37.25 - 0.50(48.50 - DBT)+ +11.97(DBT-48.50)+ +1.04LTEQ 0.97 31.94 2.97%

CHW CHW = 151.97 - 12.34(55.24- DBT)+ + 287.17(DBT -55.24)+ + 0.69LTEQ 0.95 565.96 20.67%

HW HW = 1293.82 - -205.25(63.57 - DBT)+ + -14.24(DBT - 63.57)+ + -0.77LTEQ 0.95 801.18 23.01%

10 Jan-October WBE WBE = 47.04 - 0.75(49.17 - DBT) + +11.91(DBT - 49.17)+ +1.04LTEQ 0.97 32.96 3.08%

CHW CHW = 378.42 - 17.12(56.55 - DBT)+ + 298.21(DBT-56.55)+ + 0.62 LTEQ 0.95 576.21 22.29%

HW HW = 1422.18 - -209.42(61.80 - DBT)+ + -21.34(DBT - 61.80)+ + -0.78 LTEQ 0.94 860.6 25.83%

11 Jan-Nov WBE WBE = 52.38 - 0.81(49.20 - DBT)+ + 11.89(DBT - 49.20)+ + 1.03LTEQ 0.97 32.86 3.10%

CHW CHW = 415.18 - 17.08(56.53 - DBT)+ + 298.84 (DBT - 56.53)+ +0.55LTEQ 0.95 559.22 23.28%

HW HW = 1455.89 - -205.88(62.22 - DBT)+ + -17.57(DBT - 62.22)+ + -0.88LTEQ 0.93 868.49 24.90%

12 Jan-Dec WBE WBE = 45.68 - 0.45(48.49 - DBT)+ +11.82(DBT - 48.49)+ +1.03LTEQ 0.97 32.73 3.12%

137
S.No. Base Period Channel Large Hotel, Chicago, IL: Model Equations (4P) R^2 RMSE CV-RMSE

CHW CHW = 398.55 - 14.96(56.39 - DBT)+ +298.98(DBT-56.39)+ + 0.52LTEQ 0.95 541.41 24.25%

HW HW = 1580.05 - -209.93(62.39 - DBT)+ + -17.34(DBT - 62.39)++-1.02 LTEQ 0.94 899.5 22.81%

13 February WBE WBE= 112.52 - 2.10(38 - DBT)+ + -9.52(DBT - 38)+ + 0.97LTEQ 0.97 26.47 280.00%

CHW CHW = 741.62 - 13.91(38 - DBT)+ + -62.32(DBT - 38)+ + -0.19LTEQ 0.25 174.37 44.51%

HW HW = 12109.92 - -65.36(24 - DBT)+ + -186.31(DBT - 24)+ + -2.51LTEQ 0.94 375.19 4.21%

14 Feb-March WBE WBE= 64.09 - 0.78(46 - DBT)+ + 8.04(DBT - 46)+ + 1.01LTEQ 0.96 26.59 2.81%

CHW CHW= 441.43 - 5.24(48.74 - DBT)+ + 87.56(DBT - 48.74)+ + 0.04LTEQ 0.45 175.68 42.86%

HW HW= 9017.01 - -177.59(35.99 - DBT)+ + -137.62(DBT - 35.99)+ + -2.16LTEQ 0.88 580.99 7.30%

15 Feb-April WBE WBE= 69.12 - 0.93(45.99-DBT)+ + 7.99(DBT - 45.99)+ + 1.01LTEQ 0.97 25.43 2.67%

CHW CHW= 448.88 - 5.72(46.04 - DBT)+ + 67.49(DBT-46.04)+ + 0.03 LTEQ 0.65 169.39 33.49%

HW HW= 4579.66 - -213(52.17 - DBT)+ + 5.08(DBT - 52.17)+ + -1.47LTEQ 0.9 764.17 11.55%

16 Feb-May WBE WBE= 62.68 - 0.94(45.10 -DBT)+ + 7.39(DBT - 45.10)+ + 1.01LTEQ 0.97 23.93 2.44%

CHW CHW= 682.40 - 15.57 (54.32 - DBT)+ + 168.45(DBT-54.32)+ +0.11LTEQ 0.93 228.6 26.90%

HW HW= 3373.74 - -228(54.99 - DBT)+ + - 115.04(DBT - 54.99)++ -1.23LTEQ 0.93 814.97 15.46%

17 Feb-June WBE WBE= 56.08 - 1.02(47.31 - DBT)+ + 9.59(DBT - 47.31)+ + 1.02LTEQ 0.97 27.51 2.70%

CHW CHW= 611.06 - 18.53(56.33 - DBT)+ + 247.93(DBT - 56.33)+ + 0.31LTEQ 0.95 424.8 26.60%

HW HW= 1605.63 - -225.96(62.12 - DBT)+ + -16.78(DBT - 62.12)+ + -1LTEQ 0.95 749.72 17.35%

18 Feb-July WBE WBE= 48.44 - 1.07(49.09 - DBT)+ + 11.46(DBT - 49.09)+ + 1.04LTEQ 0.97 32.34 3.05%

CHW CHW= 403.33 - 18.41(56.74 - DBT)+ + 289.48(DBT - 56.74)+ + 0.55LTEQ 0.95 560.77 22.95%

HW HW= 1393.27 - -225.90(62.36 - DBT)++-16.12(DBT-62.36)++-0.82LTEQ 0.96 683.31 18.68%

19 Feb-August WBE WBE= 72.79 - 2.10(53.47 - DBT)+ + 13.12(DBT - 53.47)+ + 1.04LTEQ 0.97 33.52 3.08%

CHW CHW= 186.73 - 18.34(56.28)+ +297.42(DBT - 56.28)+ + 0.79LTEQ 0.95 606.49 20.53%

HW HW= 1282.39 - -226.05(62.47-DBT)+ +-14.99(DBT-62.47)++-0.72LTEQ -0.96 631.55 19.79%

20 Feb-Sept WBE WBE= 42.58 - 1.06(49.20 - DBT)+ + 11.96(DBT-49.20)+ + 1.04LTEQ 0.97 32 2.93%

CHW CHW= 170.21 - 18.20(55.57)+ + 287.03(DBT -55.57)+ + 0.79LTEQ 0.95 592.37 19.51%

HW HW= 1186.79 - -232.54(61.61-DBT)+ + -13.93(DBT-61.61)+ + -0.62LTEQ 0.96 619.55 21.57%

21 Feb-October WBE WBE= 81.70 - 2.31(53.41-DBT)+ + 12.73(DBT-53.41)+ +1.04LTEQ 0.97 32.68 3.02%

CHW CHW= 408.48 - 23.23(56.86 -DBT)++298.13(DBT-56.86)+ +0.69LTEQ 0.95 597.76 21.10%

HW HW= 1360.62--236.48(59.68 - DBT)+ +-24.49(DBT-59.68)+ +-0.61LTEQ 0.94 730.75 26.33%

22 Feb-October WBE WBE= 83.86 - 2.35(53.33-DBT)+ +12.74(DBT-53.33)++1.03LTEQ 0.97 32.68 3.05%

CHW CHW= 455.19 - 23.28(56.84-DBT)++298.74(DBT-56.84)++0.61LTEQ 0.95 577.21 22.14%

HW HW= 1473.95- -228.33(59.92 - DBT)+ + -24.36(DBT - 59.92)+ +-0.75LTEQ 0.93 775.21 25.82%

23 Feb-Nov WBE WBE= 47.21 - 0.67(48.78 - DBT)+ + 11.81(DBT - 48.78)+ + 1.03LTEQ 0.97 32.94 3.11%

CHW CHW= 408.99 - 18.31(56.58 - DBT)+ + 298.87(DBT-56.58)++0.57LTEQ 0.95 558.88 23.26%

HW HW= 1604.22 - -228.35(60.44 - DBT)+ + -23.21(DBT-60.44)+ + -0.93LTEQ 0.94 807.68 22.78%

24 Feb-Dec WBE WBE = 45.68 - 0.45(48.49 - DBT)+ +11.82(DBT - 48.49)+ +1.03LTEQ 0.97 32.73 3.12%

CHW CHW = 398.55 - 14.96(56.39 - DBT)+ +298.98(DBT-56.39)+ + 0.52LTEQ 0.95 541.41 24.25%

HW HW = 1580.05 - -209.93(62.39 - DBT)+ + -17.34(DBT - 62.39)++-1.02 LTEQ 0.94 899.5 22.81%

25 March WBE WBE=69.92 - 1.22(46-DBT)+ + 7.88(DBT-46)++1.0LTEQ 0.97 25.29 2.67%

138
S.No. Base Period Channel Large Hotel, Chicago, IL: Model Equations (4P) R^2 RMSE CV-RMSE

CHW CHW=490.52-8(48.96-DBT)++87.52(DBT-48.96)++0.01LTEQ 0.64 167.95 39.41%

HW HW=10934.76 - -6718.97(23.08-DBT)++-137.54(DBT-23.08)++-2.53LTEQ 0.9 429.12 6.05%

26 March-April WBE WBE=72.27-1.28(46-DBT)++7.76(DBT-46)++1LTEQ 0.97 24.3317 2.54%

CHW CHW=470.59-7.81(46.23-DBT)++67.51(DBT-46.23)++0.02LTEQ 0.72 163.47 29.29%

HW HW=4509.16- -207.23(52-DBT)+ +37.46(DBT-52)++-1.55LTEQ 0.89 674.01 12.12%

27 March-May WBE WBE=64.92-1.40(45.59-DBT)++7.39(DBT-45.59)++1.01LTEQ 0.98 22.65 2.29%

CHW CHW=731.44-21.46(54.77-DBT)++168.42(DBT-54.77)++0.14LTEQ 0.93 231.93 23.45%

HW HW=3641.24--226.55(52.96-DBT)++-117.69(DBT-52.96)++-1.23LTEQ 0.92 751.91 18.06%

28 March-June WBE WBE=57.76-1.52(47.84-DBT)++9.59(DBT-47.84)++1.03LTEQ 0.98 27.45 2.65%

CHW CHW=642.24-25.05(56.66-DBT)++247.71(DBT-56.66)++0.37LTEQ 0.94 457.64 24.42%

HW HW=1552.50--218.32(61.96-DBT)++-17.07(DBT-61.96)++-0.93LTEQ 0.94 684.35 20.95%

29 March-July WBE WBE=104.98-3.36(56.38-DBT)++13.14(DBT-56.38)++1.04LTEQ 0.97 32.49 3.00%

CHW CHW=447.03-25.79(57.22-DBT)++291.11(DBT-57.22)++0.63LTEQ 0.95 601.44 21.33%

HW HW=1324.07--218.23(62.21-DBT)++-16.35(DBT-62.21)++-0.73LTEQ 0.95 611.81 22.71%

30 March-Aug WBE WBE=88.80-3.2(55.49-DBT)++13.46(DBT-55.49)++1.05LTEQ 0.97 33.71 3.04%

CHW CHW=168.17-24.66(56.54-DBT)++297.07(DBT-56.54)++0.90LTEQ 0.95 643.89 19.25%

HW HW=1212.95--218.48(62.32-DBT)++-15.30(DBT-62.32)++-0.64LTEQ 0.95 557.95 24.04%

31 March-Sept WBE WBE=79.70-2.92(53.56-DBT)++12.71(DBT-53.56)++1.04LTEQ 0.97 32.24 2.90%

CHW CHW=266.77-28.34(56.40-DBT)++291.16(DBT-56.40)++0.87LTEQ 0.94 623.01 18.42%

HW HW=1124.96--226.68(61.36-DBT)++-14.30(DBT-61.36)++-0.54LTEQ 0.95 548.81 26.36%

32 March-Oct WBE WBE=110.51-3.75(56.76-DBT)++13.48(DBT-56.76)++1.04LTEQ 0.97 32.48 2.95%

CHW CHW=438.45-30.56(57.18-DBT)++298.60(DBT-57.18)++0.76LTEQ 0.94 623.74 20.05%

HW HW=1320.76--224.48(59.72-DBT)++-24.65(DBT-59.72)++-0.56LTEQ 0.92 669.44 32.29%

33 March-Nov WBE WBE=111.43-3.71(56.64-DBT)++13.51(DBT-56.64)++1.04LTEQ 0.97 32.49 3.00%

CHW CHW=498.22-30.23(57.20-DBT)++299.54(DBT-57.20)++-0.67LTEQ 0.94 598.88 21.14%

HW HW=1427.75--211.81(60.44-DBT)++-23.87(DBT-60.44)++-0.71LTEQ 0.92 706.97 29.45%

34 March-Dec WBE WBE=44.90-0.59(48.75-DBT)++11.82(DBT-48.75)++1.03LTEQ 0.97 33.4 3.13%

CHW CHW=383.92-19.73(56.65-DBT)++298.62(DBT-56.65)++0.63LTEQ 0.95 580.04 22.43%

HW HW=1587.84--223.65(60.30-DBT)++-23.19(DBT-60.30)++-0.90LTEQ 0.94 770.87 25.25%

35 March-Jan WBE WBE=43.64-0.37(48.43-DBT)++11.82(DBT-48.43)++1.03LTEQ 0.97 33.11 3.13%

CHW CHW=741.62-13.91(38-DBT)++-62.32(DBT-38)++-0.19LTEQ 0.25 174.37 44,51%

HW HW=1587.84--223.65(60.30-DBT)++-23.19(DBT-60.30)++-0.90LTEQ 0.94 770.87 25.25%

36 March-Feb WBE WBE = 45.68 - 0.45(48.49 - DBT)+ +11.82(DBT - 48.49)+ +1.03LTEQ 0.97 32.73 3.12%

CHW CHW = 398.55 - 14.96(56.39 - DBT)+ +298.98(DBT-56.39)+ + 0.52LTEQ 0.95 541.41 24.25%

HW HW = 1580.05 - -209.93(62.39 - DBT)+ + -17.34(DBT - 62.39)++-1.02 LTEQ 0.94 899.5 22.81%

37 April WBE WBE=57.26- -5.62(43-DBT)+ + 7.52(DBT-43)++1.0LTEQ 0.98 21.68 2.24%

CHW CHW=356.48 - -38.40(43.48-DBT)++64.20(DBT-43.48)++-0.02LTEQ 0.73 149.62 21.55%

HW HW=6270.93- -393.77(38.74-DBT)++-108.13(DBT-38.74)++-1.68LTEQ 0.82 442.62 11.14%

38 April-May WBE WBE=44.98- -5.07(43-DBT)++7.33(DBT-43)++1.02LTEQ 0.98 20.29 2.00%

139
S.No. Base Period Channel Large Hotel, Chicago, IL: Model Equations (4P) R^2 RMSE CV-RMSE

CHW CHW=761.70-28.29(55.42-DBT)++170.75(DBT-55.42)++0.20LTEQ 0.93 254.68 19.97%

HW HW=2112.62- -179.15(62.08-DBT)++-28.39(DBT-62.08)++-1.32LTEQ 0.84 625.72 23.42%

39 April-June WBE WBE=32.40--2.72(45.67-DBT)++9.46(DBT-45.67)++1.04LTEQ 0.98 27.92 2.61%

CHW CHW=686.79-35.23(57.20-DBT)++249.04(DBT-57.20)++0.45LTEQ 0.93 519.15 21.93%

HW HW=1480.64- -189.46(62.36-DBT)+ +-17.44(DBT-62.36)+ + -0.86LTEQ 0.89 550.03 28.04%

40 April-July WBE WBE=113.24-4.29(57.45-DBT)+ +13.20(DBT-57.45)++1.04LTEQ 0.97 33.6 3.01%

CHW CHW=418.95-34.08(57-DBT)++290.65(DBT-57.48)++0.76LTEQ 0.94 666.82 19.46%

HW HW=1230.75--189.33(62.65-DBT)++-16.64(DBT-62.64)++-0.63LTEQ 0.91 476.15 30.21%

41 April-August WBE WBE=98.27-4.18(56.81-DBT)++13.61(DBT-56.81)++1.05LTEQ 0.97 34.8 3.05%

CHW CHW=92.80-31.39(56.72-DBT)++296.11(DBT-56.72)++1.07LTEQ 0.93 700.04 17.79%

HW HW=1123.26- -190.41(62.72-DBT)++-15.51(DBT-62.72)++-0.54LTEQ 0.91 425.63 31.46%

42 April-Sept WBE WBE=108.82-4.73(56.84-DBT)++13.22(DBT-56.84)++1.05LTEQ 0.97 33.01 2.90%

CHW CHW=251.37-39.76(56.69)++290.23(DBT-56.69)++1LTEQ 0.93 667.01 17.17%

HW HW=1058.19--205.06(61.32-DBT)++-14.54(DBT-61.32)++-0.46LTEQ 0.9 428.05 34.74%

43 April-Oct WBE WBE= 116.11-4.57(57.47-DBT)++13.48(DBT-57.47)++1.05LTEQ 0.97 32.64 2.91%

CHW CHW=406.29-38.93(57.35-DBT)++297.77(DBT-57.35)++0.88LTEQ 0.93 650.88 18.58%

HW HW=1215.12- -164.79(62.48-DBT)++-17.64(DBT-62.48)++-0.61LTEQ 0.84 530.89 39.37%

44 April-Nov WBE WBE=114.60-4.13(57.16-DBT)++13.53(DBT-57.16)++1.04LTEQ 0.97 32.97 3.00%

CHW CHW=475.51-34.34(57.33-DBT)++299.13(DBT-57.33)++0.75LTEQ 0.94 631.08 20.11%

HW HW=1362.88--183.42(61.37-DBT)++-22.06(DBT-61.37)++-0.69LTEQ 0.89 590.78 32.75%

45 April-Dec WBE WBE=45.83-0.58(49.23-DBT)++11.88(DBT-49.23)++1.04LTEQ 0.97 34.06 3.15%

CHW CHW=332.89-18.86(56.66-DBT)++298.30(DBT-56.66)++0.70LTEQ 0.94 607.66 21.47%

HW HW=1593.14--224.41(59.50-DBT)++-24.04(DBT-59.50)++-0.87LTEQ 0.94 734.76 28.29%

46 April-Jan WBE WBE=42.82-0.34(48.56-DBT)++11.80(DBT-48.56)++1.03LTEQ 0.97 33.7 3.15%

CHW CHW=441.43-5.24(48.74-DBT)++87.56(DBT-48.74)++0.04LTEQ 0.45 175.68 42.68%

HW HW=9017.01--177.59(35.99-DBT)++-137.62(DBT-35.99)++-2.16LTEQ 0.88 580.99 7.30%

47 April-Feb WBE WBE=69.92 - 1.22(46-DBT)+ + 7.88(DBT-46)++1.0LTEQ 0.97 25.29 2.67%

CHW CHW=490.52-8(48.96-DBT)++87.52(DBT-48.96)++0.01LTEQ 0.64 167.95 39.41%

HW HW=10934.76 - -6718.97(23-DBT)++-137.54(DBT-23.08)++-2.53LTEQ 0.9 429.12 6.05%

48 April-March WBE WBE = 45.68 - 0.45(48.49 - DBT)+ +11.82(DBT - 48.49)+ +1.03LTEQ 0.97 32.73 3.12%

CHW CHW = 398.55 - 14.96(56.39 - DBT)+ +298.98(DBT-56.39)+ + 0.52LTEQ 0.95 541.41 24.25%

HW HW = 1580.05 - -209.93(62.39 - DBT)+ + -17.34(DBT - 62.39)++-1.02 LTEQ 0.94 899.5 22.81%

49 May WBE WBE=60.70--1522.37(47.03-DBT)++7.35(DBT-47.03)++1.03LTEQ 0.99 17.76 1.68%

CHW CHW=897.25-39.79(56.69)++175.83(DBT-56.69)++0.23LTEQ 0.92 306.18 16.67%

HW HW=2039.29--87.13(65-DBT)++-27.55(DBT-65)++-1.34LTEQ 0.8 297.91 12.12%

50 May-June WBE WBE=246.89-8.40(70-DBT)++13.15(DBT-70)++1.04LTEQ 0.97 29.01 2.60%

CHW CHW=1415.01-108.14(61.51-DBT)++264.79(DBT-61.51)++0.64LTEQ 0.9 621.06 19.47%

HW HW=1253.88--105.70(63.65-DBT)++-19.28(DBT-63.65)++-0.61LTEQ 0.8 297.05 30.59%

51 May-July WBE WBE=159.46-7.01(62.10-DBT)++13.70(DBT-62.10)++1.05LTEQ 0.96 34.87 2.99%

140
S.No. Base Period Channel Large Hotel, Chicago, IL: Model Equations (4P) R^2 RMSE CV-RMSE

CHW CHW=492.33-61.26(58.42-DBT)++291.65(DBT-58.42)++0.96LTEQ 0.9 759.8 17.59%

HW HW=971.37--102.39(64.89-DBT)++-15.56(DBT-64.89)++-0.39LTEQ 0.83 244.99 30.86%

52 May-Aug WBE WBE=142.57-7.26(61-DBT)++13.90(DBT-61)++1.06LTEQ 0.96 36.1 3.05%

CHW CHW=17.11-47.40(57.19-DBT)++295.60(DBT-57.19)++1.32LTEQ 0.89 773.03 16.36%

HW HW=908.61--106.26(64.66-DBT)++-14.15(DBT-64.66)++-0.35LTEQ 0.83 215.58 30.22%

53 May-Sept WBE WBE=198.43-9.09(64.66-DBT)++14.03(DBT-64.66)++1.06LTEQ 0.96 33.28 2.84%

CHW CHW=325.61-71.10(57.49-DBT)++289.34(DBT-57.49)++1.20LTEQ 0.9 720.9 15.99%

HW HW=938--119.29(62-DBT)++-15.14(DBT-62)++-0.33LTEQ 0.77 229.07 32.98%

54 May-Oct WBE WBE=125.79-4.89(58.55-DBT)++13.55(DBT-58.55)++1.05LTEQ 0.97 33.05 2.88%

CHW CHW=360.41-40.55(57.59-DBT)++297.36(DBT-57.59)++1.02LTEQ 0.92 695.23 17.54%

HW HW=1070.44--104.50(64.52-DBT)++-14.28(DBT-64.52)++0.53LTEQ 0.83 310.34 33.81%

55 May-Nov WBE WBE=119.86-4.33(57.76-DBT)++13.52(DBT-57.76)++1.04LTEQ 0.97 33.46 2.99%

CHW CHW=452.32-36.17(57.53-DBT)++298.80(DBT-57.53)++0.85LTEQ 0.93 668.27 19.20%

HW HW=1386.51--191.54(59.10-DBT)++-24.76(DBT-59.10)++-0.62LTEQ 0.9 530.11 35.36%

56 May-Dec WBE WBE=52.62-1.17(50.52-DBT)++12.19(DBT-50.52)++1.04LTEQ 0.97 34.6 3.15%

CHW CHW=375.42-22.68(56.99-DBT)++298.97(DBT-56.99)++0.75LTEQ 0.94 637.31 20.55%

HW HW=1704.25--240.55(57.09-DBT)++-30.59(DBT-57.09)++-0.82LTEQ 0.94 714.86 30.02%

57 May-Jan WBE WBE=37.87-0.26(48.68-DBT)++11.99(DBT-48.68)++1.04LTEQ 0.97 33.7 3.15%

CHW CHW=322.01-15.28(56.58-DBT)++299.12(DBT-56.58)++0.67LTEQ 0.94 610.33 21.88%

HW HW=1662.25-211.03(59.92-DBT)++-23.69(DBT-59.92)++-0.97LTEQ 0.94 837.68 27.19%

58 May-Feb WBE WBE=40.64-0.36(48.73-DBT)++11.99(DBT-48.73)++1.03LTEQ 0.97 33.88 3.17%

CHW CHW=360.79-15.07(56.53-DBT)++299.39(DBT-56.53)++0.6LTEQ 0.95 585.3 22.79%

HW HW=1657.26--216.47(60.48-DBT)++-23.04(DBT-60.48)++-0.99LTEQ 0.94 908.04 25.10%

59 May-March WBE WBE=41.46-0.36(48.63-DBT)++12.01(DBT-48.63)++1.03LTEQ 0.97 33.22 3.14%

CHW CHW=381.86-15.15(56.46-DBT)++299.59(DBT-56.46)++0.55LTEQ 0.95 561.82 23.70%

HW HW=1642.54--213.22(61.64-DBT)++-20.94(DBT-61.64)++-1.03LTEQ 0.94 917.9 23.30%

60 May-April WBE WBE = 45.68 - 0.45(48.49 - DBT)+ +11.82(DBT - 48.49)+ +1.03LTEQ 0.97 32.73 3.12%

CHW CHW = 398.55 - 14.96(56.39 - DBT)+ +298.98(DBT-56.39)+ + 0.52LTEQ 0.95 541.41 24.25%

HW HW = 1580.05 - -209.93(62.39 - DBT)+ + -17.34(DBT - 62.39)++-1.02 LTEQ 0.94 899.5 22.81%

61 June WBE WBE=187.42-6.84(65-DBT)++9.89(DBT-65)++1.08LTEQ 0.97 28.1 2.38%

CHW CHW=1558.08-164.85(64-DBT)++228.11(DBT-64)++1.83LTEQ 0.87 629.97 13.73%

HW HW=625.57- -15.43(68.70-DBT)++-9.21(DBT-68.70)++-0.12LTEQ 0.88 32.28 6.25%

62 June-July WBE WBE=350.31-10.68(79.84-DBT)++25.55(DBT-79.84)++1.09LTEQ 0.97 32.89 2.70%

CHW CHW=5678.70-245.93(80-DBT)++424.17(DBT-80)++2.03LTEQ 0.85 720.52 12.91%

HW HW=582.51--16.88(69.72)++-8.53(DBT-69.72)++-0.11LTEQ 0.9 27.34 5.69%

63 June-Aug WBE WBE=343.19-11.30(78-DBT)++19.11(DBT-78)++1.09LTEQ 0.96 33.28 2.71%

CHW CHW=1781.17-211.98(64-DBT)++272.04(DBT-64)++2.01LTEQ 0.84 698.01 12.25%

HW HW=590.85--16.16(69.59-DBT)++-8.70(DBT-69.59)++-0.12LTEQ 0.9 24.05 5.03%

64 June-Sept WBE WBE=354.35-11.42(77.80-DBT)++19.01(DBT-77.80)++1.08LTEQ 0.97 30.23 2.51%

141
S.No. Base Period Channel Large Hotel, Chicago, IL: Model Equations (4P) R^2 RMSE CV-RMSE

CHW CHW=-151.67--196.69(56.17-DBT)++273.72(DBT-56.17)++1.74LTEQ 0.89 642.36 12.38%

HW HW=821.97--1291.25(54.32-DBT)++-13.96(DBT-54.32)++-0.11LTEQ 0.87 43.09 8.41%

65 June-Oct WBE WBE=112.24-4.87(57.49-DBT)++13.10(DBT-57.49)++1.06LTEQ 0.97 31.41 2.70%

CHW CHW=152.84-44.41(57-DBT)++288.39(DBT-57)++1.34LTEQ 0.93 646.56 14.70%

HW HW=936--108.70(63.86-DBT)++-11.34(DBT-63.86)++0.41LTEQ 0.84 297.15 36.35%

66 June-Nov WBE WBE=110.02-4.37(56.86-DBT)++13.20(DBT-56.86)++1.05LTEQ 0.97 32.39 2.87%

CHW CHW=287.44-39.04(56.92-DBT)++291.06(DBT-56.92)++1.08LTEQ 0.94 627.85 16.70%

HW HW=1204.15--196.67(59.02-DBT)++-21.47(DBT-59.02)++-0.47LTEQ 0.91 543.04 35.86%

67 June-Dec WBE WBE=38.89-0.52(49.39-DBT)++12.30(DBT-49.39)++1.04LTEQ 0.97 33.26 3.02%

CHW CHW=55.50-16.17(55.43-DBT)++286.98(DBT-55.43)++0.93LTEQ 0.95 598.58 18.29%

HW HW=1464.45--234.75(58.47-DBT)++-20.15(DBT-58.47)++-0.78LTEQ 0.94 745.27 28.93%

68 June-Jan WBE WBE=33.92-0.16(48.31-DBT)++12.12(DBT-48.31)++1.04LTEQ 0.97 32.85 3.03%

CHW CHW=21.90-9.83(54.72-DBT)++283.45(DBT-54.72)++0.80LTEQ 0.96 567.58 19.51%

HW HW=1516.14--209.07(60.77-DBT)++-18.62(DBT-60.77)++-0.91LTEQ 0.94 870.92 26.45%

69 June-Feb WBE WBE=37.36-0.28(48.40-DBT)++12.13(DBT-48.40)++1.04LTEQ 0.97 32.4 3.03%

CHW CHW=74.78-9.60(54.65-DBT)++284.01(DBT-54.65)++0.71LTEQ 0.96 541.68 20.43%

HW HW=1530.93--213.41(61.45-DBT)++-18.63(DBT-61.45)++-0.95LTEQ 0.94 939.3 24.28%

70 June-March WBE WBE=38.87-0.28(48.35-DBT)++12.16(DBT-48.35)++1.03LTEQ 0.97 31.79 3.01%

CHW CHW=110.88-9.27(54.59-DBT)++284.80(DBT-54.59)++0.64LTEQ 0.93 254.68 19.97%

HW HW=1457.87--207.92(63.35-DBT)++-10.79(DBT-63.35)++-1.00LTEQ 0.94 940.77 22.41%

71 June-April WBE WBE=43.76-0.39(48.19-DBT)++11.95(DBT-48.19)++1.03LTEQ 0.97 31.36 2.99%

CHW CHW=198.42-10.55(54.77-DBT)++285.43(DBT-54.77)++0.58LTEQ 0.92 306.18 16.67%

HW HW=1446.29--206.18(63.57-DBT)++-10.98(DBT-63.57)++-0.99LTEQ 0.94 917.39 21.96%

72 June-May WBE WBE = 45.68 - 0.45(48.49 - DBT)+ +11.82(DBT - 48.49)+ +1.03LTEQ 0.97 32.73 3.12%

CHW CHW = 398.55 - 14.96(56.39 - DBT)+ +298.98(DBT-56.39)+ + 0.52LTEQ 0.95 541.41 24.25%

HW HW = 1580.05 - -209.93(62.39 - DBT)+ + -17.34(DBT - 62.39)++-1.02 LTEQ 0.94 899.5 22.81%

73 July WBE WBE = 163.78 - -15.03(67.12 - DBT)+ +13.30(DBT - 67.12)+ +1.16LTEQ 0.99 20.44 1.62%

CHW CHW = 1290.92 -- 9144.49(65.10 - DBT)+ +260.56(DBT-65.10)+ + 3.26LTEQ 0.93 417.14 6.38%

HW HW = 564.92 - 19.10(69.69 - DBT)+ + -7.80(DBT - 69.69)++-0.10 LTEQ 0.91 17.63 3.96%

74 July-August WBE WBE = 249.07 - 9.73(71.03 - DBT)+ +15.11(DBT - 71.03)+ +1.11LTEQ 0.98 22.67 1.81%

CHW CHW = 6493.50 - 288.34(81.00 - DBT)+ +-4.71(DBT-81.00)+ + 2.47LTEQ 0.93 420.88 6.75%

HW HW = 579.34 - -12.90(70.30 - DBT)+ + -8.79(DBT - 70.30)++-0.12 LTEQ 0.93 15.48 3.37%

75 July-Sept WBE WBE = 283.92 - 11.59(72 - DBT)+ +15.81(DBT - 72)+ +1.09LTEQ 0.98 22.52 1.86%

CHW CHW = -401.06 -- 243.04(56.20 - DBT)+ +290.23(DBT-56.20)+ + 1.99LTEQ 0.94 469.08 8.71%

HW HW = 789.02 - -161.75(56.87 - DBT)+ + -13.08(DBT - 56.87)++-0.13 LTEQ 0.9 39.79 7.78%

76 July-October WBE WBE = 122.62 - 5.29(59.37 - DBT)+ +14.58(DBT - 59.37)+ +1.07LTEQ 0.98 27.59 2.38%

CHW CHW = 70.50 - 41.39(57.36 - DBT)+ +309.79(DBT-57.36)+ + 1.40LTEQ 0.95 568.59 13.10%

HW HW = 1004.10 - -106.03(64.47 - DBT)+ + -12.26(DBT - 64.47)++-0.49 LTEQ 0.84 317.96 35.71%

77 July-Nov WBE WBE = 105.96 - 4.26(57.46 - DBT)+ +14.45(DBT - 57.46)+ +1.06LTEQ 0.98 29.13 2.60%

142
S.No. Base Period Channel Large Hotel, Chicago, IL: Model Equations (4P) R^2 RMSE CV-RMSE

CHW CHW = 218.92 - 37.37(57.23 - DBT)+ +311.74(DBT-57.23)+ + 1.14LTEQ 0.96 539.83 15.01%

HW HW = 1317.93 - -194.44(59 - DBT)+ + -25.21(DBT - 59)++-0.55 LTEQ 0.91 589.04 34.45%

78 July-Dec WBE WBE = 40.42 - 0.65(50.38 - DBT)+ +13.13(DBT - 50.38)+ +1.05LTEQ 0.98 31.26 2.87%

CHW CHW = 111.77 -18.48(56.39 - DBT)+ +312.07(DBT-56.39)+ + 0.95LTEQ 0.96 517.2 16.91%

HW HW = 1611.61 - -233.70(58.35 - DBT)+ + -24.21(DBT - 58.35)++-0.88 LTEQ 0.94 800.62 27.49%

79 July-Jan WBE WBE = 38.01 - 0.32(49.53 - DBT)+ +12.99(DBT - 49.53)+ +1.04LTEQ 0.98 31.14 2.91%

CHW CHW = 47.94 -10.39(55.31 - DBT)+ +304.40(DBT-55.31)+ + 0.79LTEQ 0.97 489.09 18.28%

HW HW = 1649.29 - -207.48(60.88 - DBT)+ + -21.89(DBT - 60.88)++-1.03 LTEQ 0.94 925.24 25.15%

80 July-Feb WBE WBE = 41.35 - 0.41(49.57 - DBT)+ +13.00(DBT - 49.57)+ +1.04LTEQ 0.98 30.82 2.92%

CHW CHW = 108.75 -10.15(55.24 - DBT)+ +304.39(DBT-55.24)+ + 0.70LTEQ 0.97 465.44 19.29%

HW HW = 1662.82 - -211.59(61.61 - DBT)+ + -22.03(DBT - 61.61)++-1.06 LTEQ 0.93 990.5 23.13%

81 July-March WBE WBE = 42.62 - 0.41(49.46 - DBT)+ +13.02(DBT - 49.46)+ +1.03LTEQ 0.98 30.3 2.91%

CHW CHW = 146.65 -9.84(55.17 - DBT)+ +304.84(DBT-55.17)+ + 0.62LTEQ 0.97 444.07 20.30%

HW HW = 1570.58 - -206.18(63.59 - DBT)+ + -13.09(DBT - 63.59)++-1.11 LTEQ 0.93 983.88 21.38%

82 July-April WBE WBE = 44.74 - 0.39(48.56 - DBT)+ +12.51(DBT - 48.56)+ +1.03LTEQ 0.97 30.06 2.90%

CHW CHW = 229.29 -10.94(55.29 - DBT)+ +305.08(DBT-55.29)+ + 0.57LTEQ 0.97 426.78 20.92%

HW HW = 1544.47 - -204.60(63.80 - DBT)+ + -12.97(DBT - 63.80)++-1.09 LTEQ 0.93 954.61 21.03%

83 July-May WBE WBE = 46.81 - 0.44(48.80 - DBT)+ +12.26(DBT - 48.80)+ +1.03LTEQ 0.97 32.18 3.10%

CHW CHW = 407.88 -14.53(56.79 - DBT)+ +317.75(DBT-56.79)+ + 0.50LTEQ 0.96 495.81 24.53%

HW HW = 1698.79 - -208.76(62.41 - DBT)+ + -20.93(DBT - 62.41)++-1.12 LTEQ 0.93 933.41 21.96%

84 July-June WBE WBE = 45.68 - 0.45(48.49 - DBT)+ +11.82(DBT - 48.49)+ +1.03LTEQ 0.97 32.73 3.12%

CHW CHW = 398.55 - 14.96(56.39 - DBT)+ +298.98(DBT-56.39)+ + 0.52LTEQ 0.95 541.41 24.25%

HW HW = 1580.05 - -209.93(62.39 - DBT)+ + -17.34(DBT - 62.39)++-1.02 LTEQ 0.94 899.5 22.81%

85 August WBE WBE =293.83 - 9.76(72.82 - DBT)+ +24.50(DBT - 72.82)+ +1.08LTEQ 0.99 11.91 0.96%

CHW CHW = 4300.48 - 237.40(72.28 - DBT)+ +421.14(DBT-72.28)+ + 1.87LTEQ 0.97 259.1 4.36%

HW HW = 587.85 - -11.55(70 - DBT)+ + -10.30(DBT - 70)++-0.12 LTEQ 0.95 11.98 2.53%

86 Aug-Sept WBE WBE =315.15 - 11.34(73 - DBT)+ +25.41(DBT - 73)+ +1.06LTEQ 0.99 16.51 1.39%

CHW CHW = 4857.63 - 268.89(72.62 - DBT)+ +449.19(DBT-72.62)+ + 1.23LTEQ 0.95 411.42 8.58%

HW HW = 829.53 - -159.53(56.71- DBT)+ + -15.11(DBT - 56.71)++-0.14 LTEQ 0.9 43.33 7.98%

87 Aug-October WBE WBE =186.60 - 6.43(63.33 - DBT)+ +17.42(DBT -63.33)+ +1.04LTEQ 0.98 27.66 2.45%

CHW CHW = 794.22 - 55.59(59.44 - DBT)+ +348.91(DBT-59.44)+ + 0.88LTEQ 0.94 560.82 15.59%

HW HW = 1169.69 - -105.96(64.12- DBT)+ + -15.59(DBT - 64.12)++-0.64 LTEQ 0.82 367.38 35.31%

88 Aug-Nov WBE WBE =139.99 - 4.60(59.66 - DBT)+ +15.95(DBT -59.6)+ +1.04LTEQ 0.97 29.9 2.76%

CHW CHW = 698.31 - 39.88(58.63 - DBT)+ +344.59(DBT-58.63)+ + 0.70LTEQ 0.95 522.82 18.35%

HW HW = 1519.58 - -194.07(68.62- DBT)+ + -31.90(DBT - 58.62)++-0.68 LTEQ 0.9 658.7 32.43%

89 Aug-Dec WBE WBE =55.21 - 0.63(50.34 - DBT)+ +13.09(DBT -50.34)+ +1.03LTEQ 0.97 32.53 3.08%

CHW CHW = 452.76 - 18.52(57.40 - DBT)+ +338.79(DBT-57.40)+ + 0.59LTEQ 0.96 497.82 21.15%

HW HW = 1855.27 - -233.67(58.00- DBT)+ + -30.63(DBT - 58.00)++-1.06 LTEQ 0.93 876.99 25.70%

90 Aug-Jan WBE WBE =51.87 - 0.32(49.45 - DBT)+ +12.90(DBT -49.45)+ +1.02LTEQ 0.97 32.13 3.10%

143
S.No. Base Period Channel Large Hotel, Chicago, IL: Model Equations (4P) R^2 RMSE CV-RMSE

CHW CHW = 464.88 - 13.85(57.13 - DBT)+ +339.19(DBT-57.13)+ + 0.46LTEQ 0.96 468.25 23.13%

HW HW = 1880.76 - -207.46(60.50- DBT)+ + -28.92(DBT - 60.50)++-1.20 LTEQ 0.93 999.35 23.66%

91 Aug-Feb WBE WBE =53.86 - 0.41(49.48 - DBT)+ +12.90(DBT -49.48)+ +1.02LTEQ 0.97 31.61 3.08%

CHW CHW = 500.01 - 13.54(57.06 - DBT)+ +339.25(DBT-57.06)+ + 0.40LTEQ 0.96 442.52 24.46%

HW HW = 1879.74 - -211.58(61.23- DBT)+ + -29.84(DBT - 61.23)++-1.21 LTEQ 0.92 1059.69 21.88%

92 Aug-March WBE WBE =54.23 - 0.41(49.40 - DBT)+ +12.93(DBT -49.40)+ +1.02LTEQ 0.97 30.92 3.05%

CHW CHW = 501.01 - 13.23(56.86 - DBT)+ +336.88(DBT-56.86)+ + 0.35LTEQ 0.96 420.38 25.75%

HW HW = 1749.48 - -206.20(63.36- DBT)+ + -17.28(DBT - 63.36)++-1.26 LTEQ 0.92 1044.29 20.35%

93 Aug-April WBE WBE =55.21 - 0.39(48.39 - DBT)+ +12.29(DBT -48.39)+ +1.02LTEQ 0.97 30.52 3.02%

CHW CHW = 517.52 - 13.40(56.72 - DBT)+ +333.58(DBT-56.72)+ + 0.34LTEQ 0.96 402.82 26.34%

HW HW = 1704.24 - -204.63(63.56- DBT)+ + -17.28(DBT - 63.56)++-1.22 LTEQ 0.92 1006.85 20.12%

94 Aug-May WBE WBE =55.57 - 0.44(48.39 - DBT)+ +11.72(DBT -48.39)+ +1.02LTEQ 0.96 32.43 3.20%

CHW CHW = 641.10 - 15.31(57.48 - DBT)+ +331.38(DBT-57.48)+ + 0.27LTEQ 0.94 479.18 30.70%

HW HW = 1928.34 - -209.26(61.93- DBT)+ + -31.89(DBT - 61.73)++-1.24 LTEQ 0.92 978.79 21.10%

95 Aug-June WBE WBE =51.16 - 0.39(47.72 - DBT)+ +11.14(DBT -47.72)+ +1.02LTEQ 0.97 32.49 3.16%

CHW CHW = 555.70 - 14.98(56.40 - DBT)+ +291.66(DBT-56.40)+ + 0.34LTEQ 0.94 526.79 28.75%

HW HW = 1697.66 - -209.94(62.25- DBT)+ + -18.87(DBT - 62.25)++-1.13 LTEQ 0.93 939.66 22.02%

96 Aug-July WBE WBE = 45.68 - 0.45(48.49 - DBT)+ +11.82(DBT - 48.49)+ +1.03LTEQ 0.97 32.73 3.12%

CHW CHW = 398.55 - 14.96(56.39 - DBT)+ +298.98(DBT-56.39)+ + 0.52LTEQ 0.95 541.41 24.25%

HW HW = 1580.05 - -209.93(62.39 - DBT)+ + -17.34(DBT - 62.39)++-1.02 LTEQ 0.94 899.5 22.81%

97 September WBE WBE =212.89 - 9.03(63.00 - DBT)+ +12.24(DBT - 63.00)+ +1.03LTEQ 0.99 16.72 1.49%

CHW CHW = 579.60 - -53.04(56.52 - DBT)+ +268.25(DBT-56.52)+ + 0.95LTEQ 0.92 435.51 12.04%

HW HW = 880.35 - -156.04(56.51 - DBT)+ + -17.38(DBT - 56.51)++-0.16 LTEQ 0.89 56.9 9.22%

98 Sept-Oct WBE WBE =141.19 - 4.83(57.15 - DBT)+ +11.82(DBT - 57.15)+ +1.02LTEQ 0.97 27.34 2.56%

CHW CHW = 1122.45 - -50.84(58.50 - DBT)+ +291.99(DBT-58.50)+ + 0.36LTEQ 0.89 556.34 23.10%

HW HW = 1704.63 - -116.49(59.00 - DBT)+ + -45.13(DBT - 59.00)++-0.76 LTEQ 0.78 444.76 33.49%

99 Sept-Nov WBE WBE =107.73 - 3.41(52.57 - DBT)+ +10.77(DBT - 52.57)+ +1.02LTEQ 0.97 29.26 2.84%

CHW CHW = 959.88 - -37.00(57.40 - DBT)+ +279.53(DBT-57.40)+ + 0.29LTEQ 0.91 490.97 27.33%

HW HW = 1782.90 - -193.99(58.00 - DBT)+ + -40.20(DBT - 58.00)++-0.85LTEQ 0.87 762.52 29.77%

100 Sept-Dec WBE WBE =55.54 - 0.01(44.70 - DBT)+ +9.50(DBT - 44.70)+ +1.01LTEQ 0.97 29.81 2.96%

CHW CHW = 689.27 - 17.41(55.27 - DBT)+ +258.89(DBT-55.27)+ + 0.20LTEQ 0.92 454.87 31.41%

HW HW = 2109.59 - -237.68(57.31 - DBT)+ + -35.94(DBT - 57.31)++-1.24LTEQ 0.91 979.94 24.03%

101 Sept-Jan WBE WBE =50.59 - -0.21(44.32 - DBT)+ +9.50(DBT - 44.32)+ +1.01LTEQ 0.96 29.84 3.00%

CHW CHW = 498.27 - 6.47(53.38 - DBT)+ +241.19(DBT-53.38)+ + 0.13LTEQ 0.92 422.31 34.29%

HW HW = 2239.96 - -207.98(59.42 - DBT)+ + -46.11(DBT - 59.42)++-1.37LTEQ 0.91 1095.17 21.97%

102 Sept-Feb WBE WBE =52.17 - -0.02(44.56 - DBT)+ +9.50(DBT - 44.56)+ +1.01LTEQ 0.96 29.64 3.00%

CHW CHW = 504.33 - 6.65(53.35 - DBT)+ +241.28(DBT-53.35)+ + 0.12LTEQ 0.92 394.53 35.81%

HW HW = 2127.21 - -211.37(60.74 - DBT)+ + -39.76(DBT - 60.74)++-1.37LTEQ 0.9 1146.23 20.50%

103 Sept-March WBE WBE =53.76 - -0.04(44.83 - DBT)+ +9.61(DBT - 44.83)+ +1.01LTEQ 0.96 29.11 2.97%

144
S.No. Base Period Channel Large Hotel, Chicago, IL: Model Equations (4P) R^2 RMSE CV-RMSE

CHW CHW = 505.21 - 6.41(53.36 - DBT)+ +242.92(DBT-53.36)+ + 0.09LTEQ 0.92 371.82 37.07%

HW HW = 2027.90 - -206.32(62.65 - DBT)+ + -30.14(DBT - 62.65)++-1.42LTEQ 0.9 1117.42 19.22%

104 Sept-April WBE WBE =57.55 - 0.11(45.20 - DBT)+ +9.65(DBT - 45.20)+ +1.01LTEQ 0.96 28.54 2.91%

CHW CHW = 624.92 - 10.33(54.58 - DBT)+ +257.58(DBT-54.58)+ + 0.09LTEQ 0.91 356.42 36.95%

HW HW = 1876.60 - -204.38(63.31 - DBT)+ + -19.66(DBT - 63.31)++-1.34LTEQ 0.9 1068.71 19.14%

105 Sept-May WBE WBE =54.59 - -0.10(44.65 - DBT)+ +8.88(DBT - 44.65)+ +1.01LTEQ 0.96 28.67 2.90%

CHW CHW = 637.15 - 11.38(54.55 - DBT)+ +230.08(DBT-53.36)+ + 0.11LTEQ 0.89 398.89 37.50%

HW HW = 2133.29 - -209.26(61.20 - DBT)+ + -39.44(DBT - 61.20)++-1.35LTEQ 0.91 1031.53 20.18%

106 Sept-June WBE WBE =49.63 - -0.11(45.23 - DBT)+ +9.45(DBT - 45.23)+ +1.02LTEQ 0.97 28.94 2.87%

CHW CHW = 518.36 - 11.42(54.57 - DBT)+ +238.40(DBT-54.57)+ + 0.24LTEQ 0.93 432.88 30.64%

HW HW = 1873.25 - -210.56(61.70 - DBT)+ + -24.02(DBT - 61.70)++-1.22LTEQ 0.92 985.77 21.17%

107 Sept-July WBE WBE = 49.21 - 0.39(47.86 - DBT)+ +11.17(DBT - 47.86)+ +1.02LTEQ 0.97 31.66 3.07%

CHW CHW = 521.70 - 14.96(56.47 - DBT)+ +287.10(DBT-56.47)+ + 0.38LTEQ 0.95 504.55 26.72%

HW HW = 1671.19 - -209.86(62.28 - DBT)+ + -17.98(DBT - 62.28)++-1.10 LTEQ 0.93 939.86 22.04%

108 Sept-Aug WBE WBE = 45.68 - 0.45(48.49 - DBT)+ +11.82(DBT - 48.49)+ +1.03LTEQ 0.97 32.73 3.12%

CHW CHW = 398.55 - 14.96(56.39 - DBT)+ +298.98(DBT-56.39)+ + 0.52LTEQ 0.95 541.41 24.25%

HW HW = 1580.05 - -209.93(62.39 - DBT)+ + -17.34(DBT - 62.39)++-1.02 LTEQ 0.94 899.5 22.81%

109 October WBE WBE = 128.02 - 4.31(57.38 - DBT)+ +7.43(DBT - 57.38)+ +1.03LTEQ 0.97 25.18 2.48%

CHW CHW = 843.46 - 38.02(58.61 - DBT)+ +159.93(DBT-58.61)+ + 0.49LTEQ 0.9 231.46 18.71%

HW HW = 4917.98 - 19566.40(39.03 - DBT)+ + -81.06(DBT - 39.03)++-1.99 LTEQ 0.76 437.89 21.72%

110 Oct-Nov WBE WBE = 67.89 - 1.60(43.65 - DBT)+ +6.34(DBT - 43.65)+ +1.01LTEQ 0.96 27.78 2.82%

CHW CHW = 879.46 - 32.86(57.99 - DBT)+ +165.11(DBT-57.99)+ + 0.31LTEQ 0.89 233.21 25.89%

HW HW = 3573.14 --194.52(52 - DBT)+ + -80.85(DBT - 52)++-1.58 LTEQ 0.8 882.84 25.10%

111 Oct-Dec WBE WBE = 43.88 - -0.70(39.48 - DBT)+ +6.08(DBT - 39.48)+ +1.01LTEQ 0.96 28.57 2.95%

CHW CHW = 280.81 -- 2.45(45.01 - DBT)+ +90.72(DBT-45.01)+ + 0.12LTEQ 0.85 243.36 33.19%

HW HW = 3595.40 --232.82(53.81 - DBT)+ + -68.36(DBT - 53.81)++-1.91 LTEQ 0.88 1101.91 20.74%

112 Oct-Jan WBE WBE = 45.35 - -0.37(40.00 - DBT)+ +6.16(DBT - 40.00)+ +1.01LTEQ 0.96 28.89 3.00%

CHW CHW = 288.32 -- 1.43(45.04 - DBT)+ +90.72(DBT-45.04)+ + 0.12LTEQ 0.83 231.26 35.60%

HW HW = 3381.42 --201.85(56.99 - DBT)+ + -72.69(DBT - 56.99)++-1.88 LTEQ 0.86 1191.79 19.70%

113 Oct-Feb WBE WBE = 52.13 - -0.00(42.14 - DBT)+ +6.68(DBT - 42.14)+ +1.01LTEQ 0.96 28.88 3.01%

CHW CHW = 307.18 -- 0.18(45.21 - DBT)+ +90.80(DBT-45.21)+ + 0.11LTEQ 0.82 223.76 37.18%

HW HW = 10722.87 --133.53(23.11 - DBT)+ + -214.35(DBT - 23.11)++-1.79 LTEQ 0.86 1204.27 18.30%

114 Oct-March WBE WBE = 54.53 - -0.08(42.79 - DBT)+ +6.92(DBT - 42.79)+ +1.01LTEQ 0.96 28.39 2.96%

CHW CHW = 382.54 - 2.24(47.89 - DBT)+ +105.30(DBT-47.89)+ + 0.08LTEQ 0.8 216.62 37.88%

HW HW = 10699.01 --133.45(23.28 - DBT)+ + -209.12(DBT - 23.28)++-1.78 LTEQ 0.85 1146.47 17.19%

115 Oct-April WBE WBE = 57.35 - 0.12(43.27 - DBT)+ +7.00(DBT - 43.27)+ +1.01LTEQ 0.96 27.72 2.89%

CHW CHW = 394.41 - 2.46(47.56 - DBT)+ +100.71(DBT-47.89)+ + 0.08LTEQ 0.8 210.87 35.79%

HW HW = 10554.93 --135.87(23.01 - DBT)+ + -207.66(DBT - 23.01)++-1.62 LTEQ 0.86 1091.16 17.35%

116 Oct-May WBE WBE = 54.57 - 0.12(43.25 - DBT)+ +7.12(DBT - 43.25)+ +1.01LTEQ 0.96 26.74 2.75%

145
S.No. Base Period Channel Large Hotel, Chicago, IL: Model Equations (4P) R^2 RMSE CV-RMSE

CHW CHW = 413.86 - 3.94(49.59 - DBT)+ +128.22(DBT-49.59)+ + 0.10LTEQ 0.88 241.35 32.25%

HW HW = 2818.98 --205.34(59.58 - DBT)+ + -67.66(DBT - 59.58)++-1.60 LTEQ 0.89 1069.26 18.87%

117 Oct-June WBE WBE = 49.88 - 0.11(45.10 - DBT)+ +9.05(DBT - 45.10)+ +1.02LTEQ 0.97 29.06 2.92%

CHW CHW = 634.47 - 13.60(56.64 - DBT)+ +251.89(DBT-56.64)+ + 0.21LTEQ 0.93 382.96 32.72%

HW HW = 2028.05 --205.81(62.49 - DBT)+ + -21.62(DBT - 62.49)++-1.40 LTEQ 0.91 1019.85 20.00%

118 Oct-July WBE WBE = 51.02 - 0.45(48.28 - DBT)+ +11.25(DBT - 48.28)+ +1.02LTEQ 0.97 32.71 3.20%

CHW CHW = 560.42 - 14.21(57.43 - DBT)+ +300.72(DBT-57.43)+ + 0.33LTEQ 0.95 490.93 28.58%

HW HW = 1814.85 --205.77(62.82 - DBT)+ + -20.28(DBT - 62.82)++-1.23 LTEQ 0.92 967.92 20.93%

119 Oct-Aug WBE WBE = 45.58 - 0.45(48.61 - DBT)+ +11.88(DBT - 48.61)+ +1.03LTEQ 0.97 33.79 3.24%

CHW CHW = 377.03 - 13.57(56.77 - DBT)+ +306.66(DBT-56.77)+ + 0.50LTEQ 0.95 541.1072 25.67%

HW HW = 1654.01 --205.48(63.26 - DBT)+ + -15.83(DBT - 63.26)++-1.14 LTEQ 0.93 922.78 21.76%

120 Oct-Sept WBE WBE = 45.68 - 0.45(48.49 - DBT)+ +11.82(DBT - 48.49)+ +1.03LTEQ 0.97 32.73 3.12%

CHW CHW = 398.55 - 14.96(56.39 - DBT)+ +298.98(DBT-56.39)+ + 0.52LTEQ 0.95 541.41 24.25%

HW HW = 1580.05 - -209.93(62.39 - DBT)+ + -17.34(DBT - 62.39)++-1.02 LTEQ 0.94 899.5 22.81%

121 November WBE WBE = 73.11 - 0.60(44.39 - DBT)+ +7.66(DBT - 44.39)+ +0.99LTEQ 0.95 30.23 3.17%

CHW CHW = 722.60 - 12.00(50.36 - DBT)+ +129.27(DBT-50.36)+ + -0.14LTEQ 0.83 204.4 36.96%

HW HW = 5447.28 --122.44(46.01 - DBT)+ + -171.85(DBT - 46.01)++-1.23 LTEQ 0.71 802.63 15.83%

122 Nov-Dec WBE WBE = 47.75 -- 0.71(43.00 - DBT)+ +7.70(DBT - 43.00)+ +1.01LTEQ 0.96 29.31 3.10%

CHW CHW = 416.60 -- 2.44(47.85 - DBT)+ +121.80(DBT-47.85)+ + -0.05LTEQ 0.71 201.78 42.27%

HW HW = 8169.40 --218.73(35.99 - DBT)+ + -171.19(DBT - 35.99)++-2.14 LTEQ 0.83 1051.18 15.04%

123 Nov-Jan WBE WBE = 47.15 -- 0.40(43.41 - DBT)+ +7.76(DBT - 43.41)+ +1.01LTEQ 0.95 29.34 3.10%

CHW CHW = 383.14 -- 1.44(47.89 - DBT)+ +121.49(DBT-47.89)+ + -0.00LTEQ 0.62 198.63 43.97%

HW HW = 12777.64 --107.14(12.01 - DBT)+ + -182.33(DBT - 12.01)++-2.29 LTEQ 0.79 1129.88 15.25%

124 Nov-Feb WBE WBE = 49.66 -- 0.15(43.83 - DBT)+ +7.75(DBT - 43.83)+ +1.01LTEQ 0.95 29.23 3.09%

CHW CHW = 384.47 -- 0.15(48.04 - DBT)+ +121.32(DBT-48.04)+ + -0.02LTEQ 0.57 196.24 44.83%

HW HW = 10953.60 --134.61(23.84 - DBT)+ + -194.96(DBT - 23.84)++-2.18 LTEQ 0.77 1114.89 14.37%

125 Nov-March WBE WBE = 51.58 -- 0.07(44.36 - DBT)+ +7.95(DBT - 44.36)+ +1.01LTEQ 0.95 28.45 3.01%

CHW CHW = 389.28 -0.16(48.32 - DBT)+ +115.01(DBT-48.32)+ + 0.01LTEQ 0.56 192.24 44.16%

HW HW = 12698.06 --95.35(12.99 - DBT)+ + -176.84(DBT - 12.99)++-2.16 LTEQ 0.78 1023.9 13.43%

126 Nov-April WBE WBE = 55.62 -0.01(44.08 - DBT)+ +7.71(DBT - 44.08)+ +1.01LTEQ 0.96 27.7 2.92%

CHW CHW = 387.62 -0.27(46.52 - DBT)+ +91.80(DBT-46.52)+ + 0.02LTEQ 0.61 190.07 39.74%

HW HW = 10742.14 --135.49(23.00 - DBT)+ + -195.34(DBT - 23.00)++-1.82 LTEQ 0.83 997.03 14.20%

127 Nov-May WBE WBE = 52.10 -0.01(43.68 - DBT)+ +7.48(DBT - 43.68)+ +1.01LTEQ 0.96 26.61 2.76%

CHW CHW = 415.50 -2.32(49.52 - DBT)+ +135.42(DBT-49.52)+ + 0.05LTEQ 0.89 225.34 33.29%

HW HW = 10634.89 --135.61(23.01 - DBT)+ + -201.04(DBT - 23.01)++-1.70 LTEQ 0.89 1006.09 16.23%

128 Nov-June WBE WBE = 47.47 -0.03(45.38 - DBT)+ +9.42(DBT - 45.38)+ +1.02LTEQ 0.97 28.47 2.87%

CHW CHW = 506.37 -9.74(55.36 - DBT)+ +244.13(DBT-55.36)+ + 0.21LTEQ 0.94 370.58 31.90%

HW HW = 1969.50 --199.53(64.49 - DBT)+ + -15.96(DBT - 64.49)++-1.45 LTEQ 0.92 967.35 17.61%

129 Nov-July WBE WBE = 43.32 -0.19(47.56 - DBT)+ +11.31(DBT - 47.56)+ +1.03LTEQ 0.97 31.89 3.12%

146
S.No. Base Period Channel Large Hotel, Chicago, IL: Model Equations (4P) R^2 RMSE CV-RMSE

CHW CHW = 402.64 -10.69(56.32 - DBT)+ +291.04(DBT-56.32)+ + 0.37LTEQ 0.96 484.11 27.31%

HW HW = 1733.41 --199.46(64.88 - DBT)+ + -14.14(DBT - 64.88)++-1.27LTEQ 0.94 912.25 18.54%

130 Nov-Aug WBE WBE = 39.37 -0.27(48.24 - DBT)+ +11.99(DBT - 48.24)+ +1.03LTEQ 0.97 32.92 3.15%

CHW CHW = 202.33 -9.70(55.60 - DBT)+ +296.05(DBT-55.60)+ + 0.56LTEQ 0.96 530.96 24.17%

HW HW = 1616.86 --200.05(64.91 - DBT)+ + -12.90(DBT - 64.91)++-1.17LTEQ 0.94 865.98 19.38%

131 Nov-Sept WBE WBE = 37.37 -0.20(47.77 - DBT)+ +11.82(DBT - 47.77)+ +1.03LTEQ 0.97 31.81 3.02%

CHW CHW = 164.54 -9.00(54.77 - DBT)+ +284.61(DBT-54.77)+ + 0.57LTEQ 0.96 527.52 22.69%

HW HW = 1518.12 --206.05(63.69 - DBT)+ + -14.08(DBT - 63.69)++-1.02LTEQ 0.95 852.58 20.68%

132 Nov-Oct WBE WBE = 45.68 - 0.45(48.49 - DBT)+ +11.82(DBT - 48.49)+ +1.03LTEQ 0.97 32.73 3.12%

CHW CHW = 398.55 - 14.96(56.39 - DBT)+ +298.98(DBT-56.39)+ + 0.52LTEQ 0.95 541.41 24.25%

HW HW = 1580.05 - -209.93(62.39 - DBT)+ + -17.34(DBT - 62.39)++-1.02 LTEQ 0.94 899.5 22.81%

133 December WBE WBE = 50.44 - -3.81(20.00 - DBT)+ +0.46(DBT - 20.00)+ +1.00LTEQ 0.97 25.3 2.69%

CHW CHW = 332.77 -- 25.21(20.00 - DBT)+ +3.01(DBT-20.00)+ + 0.01LTEQ 0.33 166.41 41.18%

HW HW = 12140.93 - -193.36(20.00 - DBT)+ + -143.90(DBT - 20.00)++-2.63 LTEQ 0.91 573.23 6.48%

134 Dec-Jan WBE WBE = 55.01 - -1446.23(-2.97 - DBT)+ +-0.34(DBT -- 2.97)+ +1.02LTEQ 0.96 28.72 3.04%

CHW CHW = 361.43 -- 28589.32(-2.99 - DBT)+ +-2.24(DBT--2.99)+ + 0.11LTEQ 0.08 189.26 47.00%

HW HW = 14588.56 - -859.93(-0.37 - DBT)+ + -132.82(DBT -- 0.37)++-3.01 LTEQ 0.72 1047.1 12.26%

135 Dec-Feb WBE WBE = 48.46 - -734.62(-2.93 - DBT)+ +-0.05(DBT -- 2.93)+ +1.01LTEQ 0.95 28.69 3.04%

CHW CHW = 317.83 -- 654.12(-2.45 - DBT)+ +-0.34(DBT--2.45)+ + 0.10LTEQ 0.05 189.1 47.36%

HW HW = 14699.95 - -857.97(-0.81 - DBT)+ + -135.02(DBT -- 0.81)++-2.73 LTEQ 0.72 968.65 11.19%

136 Dec-March WBE WBE = 43.12 - -0.16(48.68 - DBT)+ +12.13(DBT - 48.68)+ +1.01LTEQ 0.96 28.15 2.98%

CHW CHW = 284.51 -- 1.04(49.06 - DBT)+ +103.83(DBT-49.06)+ + 0.10LTEQ 0.25 185.73 45.73%

HW HW = 14741.77 - -860.60(-0.97 - DBT)+ + -142.06(DBT -- 0.97)++-2.62 LTEQ 0.78 871.24 10.55%

137 Dec-April WBE WBE = 52.39 - -0.33(44.20 - DBT)+ +7.76(DBT - 44.20)+ +1.01LTEQ 0.96 27.55 2.90%

CHW CHW = 353.54 - 0.17(45.69 - DBT)+ +72.69(DBT-45.69)+ + 0.06LTEQ 0.48 183 39.49%

HW HW = 10676.91 - -128.63(25.00 - DBT)+ + -197.84(DBT -25.00)++-1.94LTEQ 0.85 936.06 12.64%

138 Dec-May WBE WBE = 49.16 - -0.02(43.68 - DBT)+ +7.50(DBT - 43.68)+ +1.01LTEQ 0.97 26.3 2.72%

CHW CHW = 351.78 - 1.49(49.38 - DBT)+ +135.79(DBT-49.38)+ + 0.10LTEQ 0.9 228.89 32.82%

HW HW = 10524.84 - -128.96(25.00 - DBT)+ + -206.96(DBT -25.00)++-1.77LTEQ 0.91 970.14 15.91%

139 Dec-June WBE WBE = 43.39 -0.03(45.69 - DBT)+ +9.58(DBT - 45.69)+ +1.02LTEQ 0.97 28.33 2.84%

CHW CHW = 400.75 - 8.10(54.64 - DBT)+ +236.94(DBT-54.64)+ + 0.28LTEQ 0.95 383.41 30.72%

HW HW = 1888.42 - -200.14(65.16 - DBT)+ + -12.19(DBT -65.16)++-1.42LTEQ 0.93 950.5 17.11%

140 Dec-July WBE WBE = 36.01 -0.11(47.51 - DBT)+ +11.38(DBT - 47.51)+ +1.03LTEQ 0.97 31.93 3.10%

CHW CHW = 348.49 - 10.59(56.23 - DBT)+ +289.79(DBT-56.23)+ + 0.44LTEQ 0.96 502.15 26.11%

HW HW = 1675.49 - -200.46(65.28 - DBT)+ + -13.84(DBT -65.28)++-1.21LTEQ 0.94 889.25 18.14%

141 Dec-Aug WBE WBE = 32.43 -0.21(48.30 - DBT)+ +12.10(DBT - 48.30)+ +1.04LTEQ 0.97 32.98 3.13%

CHW CHW = 133.61 - 9.52(55.53 - DBT)+ +295.17(DBT-55.53)+ + 0.64LTEQ 0.96 548.14 23.06%

HW HW = 1553.09 - -201.10(65.31 - DBT)+ + -12.19(DBT -65.31)++-1.11LTEQ 0.95 839.27 19.07%

142 Dec-Sept WBE WBE = 30.35 -0.10(47.69 - DBT)+ +11.88(DBT - 47.69)+ +1.04LTEQ 0.97 31.76 2.99%

147
S.No. Base Period Channel Large Hotel, Chicago, IL: Model Equations (4P) R^2 RMSE CV-RMSE

CHW CHW = 99.60 - 8.75(54.72 - DBT)+ +284.18(DBT-54.72)+ + 0.64LTEQ 0.96 542.09 21.69%

HW HW = 1427.84 - -207.17(64.11 - DBT)+ + -11.94(DBT -64.11)++-0.96LTEQ 0.95 829.91 20.60%

143 Dec-Oct WBE WBE = 41.07 -0.41(48.56 - DBT)+ +11.88(DBT - 48.56)+ +1.03LTEQ 0.97 32.81 3.10%

CHW CHW = 364.75 - 15.05(56.39 - DBT)+ +298.27(DBT-56.39)+ + 0.58LTEQ 0.95 555.88 23.33%

HW HW = 1512.00 - -211.79(62.49 - DBT)+ + -16.97(DBT -62.49)++-0.96LTEQ 0.94 889.58 23.15%

148
 c. Monthly Prediction Results:

Figures C1.1.7 to C1.1.10 are typical examples of plots for monthly prediction

results obtained from the MCP models for the three energy channels. The plots

show how well the models generated from the short dataset are able to predict

the energy use for each month of the year. Examples for datasets derived from

one month data from the months of January, April, July and October have been

included in this appendix. The graphs summarize the NMBE (%) and CV (%)

values obtained for the three channels under analysis, namely, Whole building

Electric (WBE), cooling energy use (CHW), and heating energy use (HW) for

each month of the year.

149
 NMBE %: EWB (STARTING MONTH JANUARY) CV %: EWB (STARTING MONTH JANUARY)
100% 100%
January Only January to April January to August January Only January to April January to August

80% 80%

60% 60%

40% 40%

20% 20%

0% 0%
0 1 2 3 4 5 6 7 8 9 10 11 12 13 0 1 2 3 4 5 6 7 8 9 10 11 12 13
Month Month
‐20% ‐20%

NMBE % : CHW (STARTING MONTH JANUARY) CV % : CHW (STARTING MONTH JANUARY)

100% 100%
January Only January to April January to August January Only January to April January to August

80% 80%

60% 60%

40% 40%

20% 20%

0% 0%
0 1 2 3 4 5 6 7 8 9 10 11 12 13 0 1 2 3 4 5 6 7 8 9 10 11 12 13
Month Month
‐20% ‐20%

NMBE % : HW (STARTING MONTH JANUARY) CV %: HW (STARTING MONTH JANUARY)

100% 100%
January Only January to April January to August January Only
January to April
80% 80%
January to August
60% 60%

40% 40%

20% 20%

0% 0%
0 1 2 3 4 5 6 7 8 9 10 11 12 13 0 1 2 3 4 5 6 7 8 9 10 11 12 13
‐20%
Month Month
‐20%

Figure C1.1.7 Monthly Prediction Results obtained from the MCP models.
Dataset: January. Building: Large Hotel, Chicago, IL.
NMBE % : EWB (STARTING MONTH APRIL) CV % : EWB (STARTING MONTH APRIL)
100% 100%
April Only April to July April to November April Only April to July April to November

80% 80%

60% 60%

40% 40%

20% 20%

0% 0%
0 1 2 3 4 5 6 7 8 9 10 11 12 13 0 1 2 3 4 5 6 7 8 9 10 11 12 13
Month Month
‐20% ‐20%

NMBE % : CHW (STARTING MONTH APRIL) CV % : EWB (STARTING MONTH APRIL)

100% 100%
April Only April to July April to November April Only April to July April to November

80% 80%

60% 60%

40% 40%

20% 20%

0% 0%
0 1 2 3 4 5 6 7 8 9 10 11 12 13 0 1 2 3 4 5 6 7 8 9 10 11 12 13
Month Month
‐20% ‐20%

NMBE % : HW (STARTING MONTH APRIL) CV % : CHW (STARTING MONTH APRIL)

100% 100%
April Only April to July April to November April Only April to July April to November
80% 80%

60% 60%

40% 40%

20% 20%

0% 0%
0 1 2 3 4 5 6 7 8 9 10 11 12 13 0 1 2 3 4 5 6 7 8 9 10 11 12 13
Month Month
‐20% ‐20%

Figure C1.1.8 Monthly Prediction Results obtained from the MCP models.
Dataset: April. Building: Large Hotel, Chicago, IL.

150
 NMBE % : EWB (STARTING MONTH JULY) CV % : EWB (STARTING MONTH JULY)
100% 100%
July Only July to October July to February July Only July to October July to February
80% 80%

60% 60%

40% 40%

20% 20%

0% 0%
0 1 2 3 4 5 6 7 8 9 10 11 12 13 0 1 2 3 4 5 6 7 8 9 10 11 12 13
Month Month
‐20% ‐20%

NMBE % : CHW (STARTING MONTH JULY) CV %: CHW (STARTING MONTH JULY)

100% 100%
July Only July to October July to February July Only July to October July to February

80% 80%

60% 60%

40% 40%

20% 20%

0% 0%
0 1 2 3 4 5 6 7 8 9 10 11 12 13 0 1 2 3 4 5 6 7 8 9 10 11 12 13
Month Month
‐20% ‐20%

NMBE % : HW (STARTING MONTH JULY) CV % : HW (STARTING MONTH JULY)

100% 100%
July Only July to October July to February July Only July to October July to February

80% 80%

60% 60%

40% 40%

20% 20%

0% 0%
0 1 2 3 4 5 6 7 8 9 10 11 12 13 0 1 2 3 4 5 6 7 8 9 10 11 12 13
Month Month
‐20% ‐20%

Figure C1.1.9 Monthly Prediction Results obtained from the MCP models.
Dataset: July. Building: Large Hotel, Chicago, IL.
NMBE : EWB (STARTING MONTH OCTOBER) CV : EWB (STARTING MONTH OCTOBER)
100% 100%
October Only October to January October to May October Only October to January October to May

80% 80%

60% 60%

40% 40%

20% 20%

0% 0%
0 1 2 3 4 5 6 7 8 9 10 11 12 13 0 1 2 3 4 5 6 7 8 9 10 11 12 13
‐20% Month ‐20% Month

NMBE : CHW (STARTING MONTH OCTOBER) CV : CHW (STARTING MONTH OCTOBER)

100% 100%
October Only October to January October to May October Only October to January October to May

80% 80%

60% 60%

40% 40%

20% 20%

0% 0%
0 1 2 3 4 5 6 7 8 9 10 11 12 13 0 1 2 3 4 5 6 7 8 9 10 11 12 13
‐20%
Month ‐20%
Month

NMBE : HW (STARTING MONTH OCTOBER) CV % : HW (STARTING MONTH OCTOBER)

100% 100%
October Only October to January October to May October Only October to January October to May

80% 80%

60% 60%

40% 40%

20% 20%

0% 0%
0 1 2 3 4 5 6 7 8 9 10 11 12 13 0 1 2 3 4 5 6 7 8 9 10 11 12 13
Month Month
‐20% ‐20%

Figure C1.1.10 Monthly Prediction Results obtained from the MCP models.
Dataset: July. Building: Large Hotel, Chicago, IL

151
C1.2. Office Building at Albuquerque, New Mexico

a. Time Series and Scatter Plots

Typical examples of the time series plots for the two energy channels; namely,

WBE and HW, are shown in Fig C1.2.1 to Fig. C1.2.2. The blue portion on the

graphs represents the short data-set used to estimate the energy use over the

whole year represented by the red portion on the graphs. The time-series graphs

are plotted for each of all the base periods shown in Table C2.1.

FigureC1.2.1 Time Series Plots for WBE for October used as the Short Data-Set for the

Office, Albuquerque, NM.

Figure C1.2.2 Time Series Plots for HW for October used as the Short Data-Set for the

Office, Albuquerque, NM

152
Figures C1.2.3 to C1.2.4 are examples of the scatter plots of the MCP models for

the two energy channels, WBE and HW.

FigureC1.2.3 MCP Scatter Plot for WBE versus DBT & LTEQ for October Used as the

Short Data-Set for the Office, Albuquerque, NM

Figure C1.2.4 MCP Scatter Plot for HW versus DBT & LTEQ for October Used as the

Short Data-Set for the Office, Albuquerque, NM

153
 b. Multivariate Change Point Models

Table C1.2.1

MCP Models for all Base Periods and Energy Channels for Office Building at

Albuquerque, NM.

S.No. Base Period Channel Equation R^2 RMSE CV-RMSE

1 January WBE WBE=0.37 - 0.20(32.00-DBT) + + -0.12 (DBT-32.00)+ + 1.18LTEQ 0.96 2 9.24%

HW HW =26308.16 - -49.02(41-DBT)+ + -1152.86(DBT - 41) + + 645.92LTEQ 0.7 3415 8.99%

2 Jan-Feb WBE WBE = 0.42 -- 0.01 (49.93-DBT)+ + -18.81 (DBT - 49.93)+ +1.14 LTEQ 0.95 2 9.49%

HW HW = 24859.23 - -179.25(50.03-DBT)+ + -9055.33(DBT - 50.03) + + 575.35LTEQ 0.58 4035 10.56%

3 Jan-March WBE WBE = 1.12 - 0.06(41.00 - DBT)+ +-0.20(DBT - 41.00)+ +1.14LTEQ 0.9 3 12.67%

HW HW = 29948.87 - -1626.60(28.00 - DBT)+ + -196.33(DBT-28.00)+ + 485.72LTEQ 0.47 4333 11.58%

4 Jan-April WBE WBE = -1.29 - 0.00(51.00 - DBT)+ + 0.34(DBT -51.00)+ + 1.24LTEQ 0.9 3 13.10%

HW HW = 31918.15 - -1601.50(28.00 - DBT)+ + -240.38(DBT - 28.00)+ + 394.49LTEQ 0.46 4130 11.43%

5 Jan-May WBE WBE = 1.87 - 0.15(71.00 - DBT)+ + -2.00(DBT-71.00)+ +1.33LTEQ 0.85 4 17.37%

HW HW = 26532.67 - -289.07(54.00 - DBT)+ + -95.50(DBT-54.00)+ + 317.88 LTEQ 0.48 3872 11.01%

6 Jan-June WBE WBE = -5.65 - -0.02(47.01 - DBT)+ + 0.29(DBT-47.01)+ +1.45 LTEQ 0.86 5 17.59%

HW HW = 27468.93 - -290.19(54.16 -DBT)+ + -90.90(DBT-54.16)+ + 267.60 LTEQ 0.49 3636 10.53%

7 Jan-July WBE WBE = -7.23 - -0.03(47.99 - DBT)+ +0.33(DBT - 47.99)+ +1.52LTEQ 0.83 5 19.75%

HW HW = 27582.76- -283.66(57.00 - DBT)+ + -49.02(DBT - 57.00)+ + 227.02LTEQ 0.5 3431 10.08%

8 Jan-August WBE WBE = -8.70 --0.04(47.01 -DBT)+ +0.32(DBT - 47.01)+ +1.59LTEQ 0.85 5 18.92%

HW HW = 27905.95- -283.02(57.98 - DBT)+ + -35.01(DBT - 57.98)+ + 197.72LTEQ 0.5 3251 9.65%

9 Jan-Sept WBE WBE = -8.90 - -0.04(47.01 - DBT)+ +0.31(DBT-47.01)+ +1.60LTEQ 0.84 5 19.30%

HW HW = 28232.01 - -283.68(58.14 - DBT)+ + -33.32(DBT - 58.14)+ + 178.48LTEQ 0.51 3093 9.25%

10 Jan-October WBE WBE = -7.83 - 0.01(51 - DBT) + +0.33(DBT - 51)+ +1.58LTEQ 0.83 6 19.64%

HW HW = 28782.84 - -305.99(55.00 - DBT)+ + -43.93(DBT - 55.00)+ + 167.82 LTEQ 0.49 3006 9.03%

11 Jan-Nov WBE WBE = -8.57 - -0.11(44 - DBT)+ + 0.30(DBT - 44.00)+ + 1.54LTEQ 0.82 6 20.43

HW HW = 28259.91 - -259.77(57.47 - DBT)+ + -45.22(DBT - 57.47)+ + 189.33LTEQ 0.45 3271 9.76%

12 Jan-Dec WBE WBE =-7.77 - 0.12(44.00 - DBT)+ +0.30(DBT - 44.00)+ +1.50LTEQ 0.82 5 20.28%

HW HW = 27047.32 - -269.50(58.00 - DBT)+ + -49.03(DBT - 58.00)++251.04 LTEQ 0.44 3855 11.32%

13 February WBE WBE= 1.26 - 0.00(49.90 - DBT)+ + -18.35(DBT - 49.90)+ + 1.11LTEQ 0.94 2 10.09%

HW HW = 26205.60 - -230.18(50 - DBT)+ + -8123.05(DBT - 50)+ + 498.13LTEQ 0.51 4730 12.30%

14 Feb-March WBE WBE= 2.04 - 0.15(41 - DBT)+ +-0.26(DBT - 41)+ + 1.12LTEQ 0.87 3 13.99%

HW HW= 29058.09 - -577.55(36.00 - DBT)+ + -161.23(DBT - 36.00)+ + 424.1716LTEQ 0.38 4726 12.74%

15 Feb-April WBE WBE= -1.79 - 0.01(51.00-DBT)+ + 0.32(DBT - 51.00)+ + 1.28LTEQ 0.88 3 13.91%

HW HW= 27419.61 - -321.80(51.00 - DBT)+ + -122.81(DBT - 51.00)+ + 304.24LTEQ 0.4 4197 11.83%

16 Feb-May WBE WBE= 1.14 - 0.18(70.99 -DBT)+ +- 2.20(DBT - 70.99)+ + 1.39LTEQ 0.83 5 18.00%

HW HW= 28055.60 - -327.81(53.84 - DBT)+ + - 71.32(DBT - 53.84)++226.29LTEQ 0.44 3731 10.83%

154
S.No. Base Period Channel Equation R^2 RMSE CV-RMSE

17 Feb-June WBE WBE= -5.97 - 0.08(51.00 - DBT)+ + 0.29(DBT - 51.00)+ + 1.52LTEQ 0.85 5 17.91%

HW HW= 28869.48 - -333.06(53.99 - DBT)+ + -68.83(DBT - 53.99)+ + 181.67LTEQ 0.46 3393 10.04%

18 Feb-July WBE WBE= 0.97 - 0.24(82.01 - DBT)+ + 10.35(DBT - 82.01)+ + 1.61LTEQ 0.82 6 19.89%

HW HW= 29294.79 - -339.15(54.51 - DBT)++-49.33(DBT-54.51)++149.01LTEQ 0.47 3126 9.37%

19 Feb-August WBE WBE= -0.21 - 0.25(82.01 - DBT)+ + 10.62(DBT - 82.01)+ + 1.68LTEQ 0.85 5 18.73%

HW HW= 29588.25 - -340.79(55.00-DBT)+ +-41.68(DBT-55.00)++126.10LTEQ 0.48 2912 8.81%

20 Feb-Sept WBE WBE= -0.44 - 0.24(82.01 - DBT)+ + 10.87(DBT-82.01)+ + 1.67LTEQ 0.84 6 19.14%

HW HW= 29755.87 - -341.49(55.31-DBT)+ + -38.25(DBT-55.31)+ + 112.59LTEQ 0.49 2737 8.33%

21 Feb-October WBE WBE= -0.24 - 0.24(82.01-DBT)+ + 11.00(DBT-82.01)+ +1.65LTEQ 0.83 6 19.58%

HW HW= 30001.95--361.43(53.25 - DBT)+ +-41.71(DBT-53.25)+ +109.13LTEQ 0.47 2654 8.10%

22 Feb-Nov WBE WBE= -10.68 - -0.33(39.00 - DBT)+ + 0.29(DBT - 39.00)+ + 1.58LTEQ 0.81 6 20.58%

HW HW= 29224.84 - -263.27(57.03 - DBT)+ + -43.23(DBT-57.03)+ + 140.99LTEQ 0.4 3051 9.23%

23 Feb-Dec WBE WBE = -8.56 --0.09(44.00 - DBT)+ +0.30(DBT - 44.00)+ +1.54LTEQ 0.82 6 20.37%

HW HW = 27758.72 - -274.60(57.68 - DBT)+ + -46.36(DBT - 57.68)++214.55 LTEQ 0.41 3789 11.26%

24 Feb-Jan WBE WBE =-7.77 - 0.12(44.00 - DBT)+ +0.30(DBT - 44.00)+ +1.50LTEQ 0.82 5 20.28%

HW HW = 27047.32 - -269.50(58.00 - DBT)+ + -49.03(DBT - 58.00)++251.04 LTEQ 0.44 3855 11.32%

25 March WBE WBE=0.08 - 0.45(36-DBT)+ + -0.09(DBT-36)++1.22LTEQ 0.96 2 7.88%

HW HW=30616.69 - -1038.56(35-DBT)++-162.56(DBT-35)++343.15LTEQ 0.9 429 6.05%

26 March-April WBE WBE=-3.93-0.00(47-DBT)++0.23(DBT-47)++1.39LTEQ 0.97 24 2.54%

HW HW=32364.08-972.62(35-DBT)+ +-190.30(DBT-35)++229.28LTEQ 0.89 674 12.12%

27 March-May WBE WBE=-0.39-0.20(70.93-DBT)++-2.27(DBT-70.93)++1.49LTEQ 0.87 4 16.31%

HW HW=33316.84-884.73(35-DBT)++-158.59(DBT-35.00)++157.67LTEQ 0.32 3045 9.16%

28 March-June WBE WBE=-8.71-0.01(46.99-DBT)++0.27(DBT-46.99)++1.62LTEQ 0.88 5 16.28%

HW HW=33728.23-823.15(35.00-DBT)++-139.15(DBT-35.00)++122.26LTEQ 0.33 2701 8.25%

29 March-July WBE WBE=-0.79-0.25(82.00-DBT)++10.78(DBT-82.00)++1.70LTEQ 0.85 6 18.69%

HW HW=29329.03--160.44(66.01-DBT)++-5.47(DBT-66.01)++98.62LTEQ 0.34 2445 7.54%

30 March-Aug WBE WBE=-1.89-0.26(82.00-DBT)++11.05(DBT-82.00)++1.76LTEQ 0.87 5 17.40%

HW HW=29585.87--162.42(66.25-DBT)++3.06(DBT-66.25)++82.53LTEQ 0.35 2240 6.95%

31 March-Sept WBE WBE=-1.88-0.25(82.00-DBT)++11.23(DBT-82.00)++1.75LTEQ 0.85 5 18.08%

HW HW=29940.44--170.41(64.40-DBT)++-10.44(DBT-64.40)++72.54LTEQ 0.35 2084 6.49%

32 March-Oct WBE WBE=-1.41-0.24(82.02-DBT)++11.50(DBT-82.02)++1.71LTEQ 0.84 6 18.76%

HW HW=30571.00--184.34(57.48-DBT)++-40.23(DBT-57.23)++71.08LTEQ 0.31 2030 6.32%

33 March-Nov WBE WBE=0.22-0.26(82.00-DBT)++10.67(DBT-82.00)++1.64LTEQ 0.82 6 20.05%

HW HW=28849.23--138.33(70.00-DBT)++28.60(DBT-70.00)++110.95LTEQ 0.31 2610 8.03%

34 March-Dec WBE WBE=-10.21--0.32(40.00-DBT)++0.29(DBT-40.00)++1.58LTEQ 0.83 6 20.10%

HW HW=28266.14--241.55(58.42-DBT)++-47.60(DBT-58.42)++188.72LTEQ 0.37 3625 10.91%

35 March-Jan WBE WBE=-8.58--0.15(44.00-DBT)++0.30(DBT-44.00)++1.54LTEQ 0.83 5 19.94%

HW HW=26731.70--227.55(63.37-DBT)++-7.18(DBT-63.37)++230.58LTEQ 0.41 3732 11.08%

155
S.No. Base Period Channel Equation R^2 RMSE CV-RMSE

36 March-Feb WBE WBE =-7.77 - 0.12(44.00 - DBT)+ +0.30(DBT - 44.00)+ +1.50LTEQ 0.82 5 20.28%

HW HW = 27047.32 - -269.50(58.00 - DBT)+ + -49.03(DBT - 58.00)++251.04 LTEQ 0.44 3855 11.32%

37 April WBE WBE=-3.40- 0.06(64.21-DBT)+ + 0.72(DBT-64.21)++1.52LTEQ 0.97 2 7.41%

HW HW=30667.31- 47.01(61.00-DBT)++-285.03(DBT-61.00)++104.12LTEQ 0.3 1302 4.03%

38 April-May WBE WBE=-2.80-0.15(70.77-DBT)++-2.13(DBT-70.77)++1.61LTEQ 0.87 5 17.18%

HW HW=31025.24- 14.93(61.00-DBT)++-98.00(DBT-61.00)++61.66LTEQ 0.25 1032 3.24%

39 April-June WBE WBE=-3.72-0.17(78.00-DBT)++0.92(DBT-78.00)++1.74LTEQ 0.88 5 16.23%

HW HW=31225.89- -4.16(61.00-DBT)+ +-65.41(DBT-61.00)+ + 41.50LTEQ 0.24 879 2.78%

40 April-July WBE WBE=-3.21-0.23(82.01-DBT)+ +11.74(DBT-82.01)++1.82LTEQ 0.85 6 18.65%

HW HW=30495.77--40.39(77.87-DBT)++52.56(DBT-77.87)++31.14LTEQ 0.24 775 2.45%

41 April-August WBE WBE=-4.10-0.23(82.01-DBT)++11.96(DBT-82.01)++1.87LTEQ 0.87 5 16.97%

HW HW=30693.91- -40.85(75.68-DBT)++20.33(DBT-75.68)++24.74LTEQ 0.24 697 2.21%

42 April-Sept WBE WBE=-3.71-0.23(82.01-DBT)++12.02(DBT-82.01)++1.83LTEQ 0.86 6 17.89%

HW HW=30757.74--40.96(74.19-DBT)++8.57(DBT-74.19)++23.52LTEQ 0.23 646 2.05%

43 April-Oct WBE WBE= -2.80-0.23(82.00-DBT)++11.74(DBT-82.00)++1.78LTEQ 0.85 6 18.78%

HW HW=31320.72- -12.86(61.00-DBT)++-42.29(DBT-61.00)++28.30LTEQ 0.23 702 2.22%

44 April-Nov WBE WBE=-0.60-0.27(82.02-DBT)++11.08(DBT-82.02)++1.69LTEQ 0.82 6 20.35%

HW HW=29853.16--123.38(66.43-DBT)++-10.07(DBT-66.43)++75.94LTEQ 0.26 2131 6.64%

45 April-Dec WBE WBE=-10.90--0.37(39.24-DBT)++0.29(DBT-39.24)++1.61LTEQ 0.83 6 20.58%

HW HW=28599.82--267.27(56.64-DBT)++-42.63(DBT-56.64)++172.20LTEQ 0.37 3489 10.59%

46 April-Jan WBE WBE=-8.97--0.14(44.00-DBT)++0.30(DBT-44.00)++1.56LTEQ 0.83 6 20.45%

HW HW=27643.76--268.58(57.00-DBT)++-45.42(DBT-57.00)++221.00LTEQ 0.43 3637 10.87%

47 April-Feb WBE WBE=-8.07 - -0.12(44-DBT)+ + 0.30(DBT-44)++1.52LTEQ 0.82 6 20.81%

HW HW=27161.45 - -290.23(56.99-DBT)++-43.65(DBT-56.99)++243.41LTEQ 0.46 3781 11.16%

48 April-March WBE WBE =-7.77 - 0.12(44.00 - DBT)+ +0.30(DBT - 44.00)+ +1.50LTEQ 0.82 5 20.28%

HW HW = 27047.32 - -269.50(58.00 - DBT)+ + -49.03(DBT - 58.00)++251.04 LTEQ 0.44 3855 11.32%

49 May WBE WBE=-4.24-0.09(70.49-DBT)++-2.03(DBT-70.49)++1.70LTEQ 0.82 6 22.93%

HW HW=31043.28-0.04(69-DBT)++-69.11(DBT-69)++24.38LTEQ 0.19 458 1.45%

50 May-June WBE WBE=-6.28-0.09(78-DBT)++0.92(DBT-78)++1.85LTEQ 0.87 5 17.93%

HW HW=-31377.87-86.62(53.68-DBT)++-17.47(DBT-53.68)++13.06LTEQ 0.14 339 1.08%

51 May-July WBE WBE=-8.08-0.10(73.00-DBT)++0.45(DBT-73.00)++1.92LTEQ 0.84 6 19.66%

HW HW=31032.68--16.84(77.69-DBT)++12.24(DBT-77.69)++8.97LTEQ 0.15 279 0.89%

52 May-Aug WBE WBE=-5.88-0.20(82.01-DBT)++12.72(DBT-82.01)++1.95LTEQ 0.87 6 17.65%

HW HW=31091.05--16.25(76.31-DBT)++5.19(DBT-76.31)++7.30LTEQ 0.14 243 0.78%

53 May-Sept WBE WBE=-4.89-0.22(82.00-DBT)++12.30(DBT-82.00)++1.89LTEQ 0.85 6 18.57%

HW HW=31044.12--15.25(78-DBT)++12.17(DBT-78)++9.24LTEQ 0.16 244 0.78%

54 May-Oct WBE WBE=-3.46-0.25(82.01-DBT)++11.94(DBT-82.01)++1.82LTEQ 0.84 6 19.58%

HW HW=31841.67-5685.75(41.13-DBT)++-25.75(DBT-41.13)++16.72LTEQ 0.22 473 1.50%

156
S.No. Base Period Channel Equation R^2 RMSE CV-RMSE

55 May-Nov WBE WBE=-0.94-0.28(82.00-DBT)++10.86(DBT-82.00)++1.71LTEQ 0.81 6 21.28%

HW HW=29928.19--140.57(66.08-DBT)++-10.30(DBT-66.08)++72.50LTEQ 0.29 2190 6.83%

56 May-Dec WBE WBE=-11.42--0.37(40.00-DBT)++0.30(DBT-40.00)++1.62LTEQ 0.82 6 21.49%

HW HW=28002.99--246.74(61.02-DBT)++-23.58(DBT-61.02)++179.85LTEQ 0.39 3620 20.97%

57 May-Jan WBE WBE=-9.56--0.18(44.00-DBT)++0.32(DBT-44.00)++1.57LTEQ 0.82 6 21.29%

HW HW=26733.69--237.92(63.26-DBT)++-10.45(DBT-63.26)++232.55LTEQ 0.44 3751 11,17%

58 May-Feb WBE WBE=-8.47--0.15(44.00-DBT)++0.31(DBT-44.00)++1.52LTEQ 0.81 6 21.62%

HW HW=26242.15--257.98(62.68-DBT)++-8.63(DBT-62.68)++256.09LTEQ 0.47 3886 11.42%

59 May-March WBE WBE=-8.06--0.15(44.17-DBT)++0.31(DBT-44.17)++1.50LTEQ 0.82 6 21.00%

HW HW=26068.89--242.80(63.79-DBT)++-7.92(DBT-63.79)++263.81LTEQ 0.45 3948 11.55%

60 May-April WBE WBE =-7.77 - 0.12(44.00 - DBT)+ +0.30(DBT - 44.00)+ +1.50LTEQ 0.82 5 20.28%

HW HW = 27047.32 - -269.50(58.00 - DBT)+ + -49.03(DBT - 58.00)++251.04 LTEQ 0.44 3855 11.32%

61 June WBE WBE=-8.14-0.20(73.05-DBT)++-0.16(DBT-73.05)++2.02LTEQ 0.96 3 9.45%

HW HW=31249.24 -5.28(71.00-DBT)++-3.58(DBT-71.00)++1.08LTEQ 0.19 30 0.09%

62 June-July WBE WBE=-6.96-0.18(82.18-DBT)++15.39(DBT-82.18)++2.02LTEQ 0.87 6 17.54%

HW HW=31248.25-3.47(71-DBT)++-1.77(DBT-71.00)++0.68LTEQ 0.13 23 0.07%

63 June-Aug WBE WBE=-7.31-0.17(82.00-DBT)++12.83(DBT-82.00)++2.03LTEQ 0.9 5 14.94%

HW HW=31257.76-30.20(66.00-DBT)++-3.09(DBT-66.00)++1.70LTEQ 0.15 46 0.15%

64 June-Sept WBE WBE=-5.26-0.28(82.00-DBT)++11.93(DBT-82.00)++1.95LTEQ 0.87 6 17.14%

HW HW=31127.07--11.48(78.00-DBT)++6.65(DBT-78.00)++5.55LTEQ 0.16 142 0.45%

65 June-Oct WBE WBE=-3.54-0.28(82.01-DBT)++11.64(DBT-82.01)++1.85LTEQ 0.85 6 18.71%

HW HW=31946.88-858.76(41.99-DBT)++-28.42(DBT-41.99)++14.88LTEQ 0.25 476 1.52%

66 June-Nov WBE WBE=-0.69-0.30(82.00-DBT)++10.57(DBT-82.00)++1.71LTEQ 0.82 6 20.84%

HW HW=29478.98--131.94(69.99-DBT)++18.26(DBT-69.99)++82.96LTEQ 0.3 2342 7.28%

67 June-Dec WBE WBE=-11.09--0.18(44.00-DBT)++0.34(DBT-44.00)++1.62LTEQ 0.83 6 20.86%

HW HW=27194.20--228.18(64.34-DBT)++-4.17(DBT-64.34)++205.49LTEQ 0.39 3836 11.55%

68 June-Jan WBE WBE=-9.50--0.20(44.00-DBT)++0.33(DBT-44.00)++1.55LTEQ 0.83 6 20.84%

HW HW=25471.65--212.08(70.00-DBT)++81.79(DBT-70.00)++260.80LTEQ 0.45 3921 11.59%

69 June-Feb WBE WBE=-7.65--0.06(51.00-DBT)++0.40(DBT-51.00)++1.50LTEQ 0.82 6 21.18%

HW HW=25537.78--252.57(64.00-DBT)++1.96(DBT-64.00)++284.26LTEQ 0.47 4039 11.77%

70 June-March WBE WBE=-7.08--0.06(51.00-DBT)++0.40(DBT-51.00)++1.48LTEQ 0.82 5 20.47%

HW HW=24816.17--219.09(70.00-DBT)++90.91(DBT-70.00)++290.39LTEQ 0.45 4086 11.85%

71 June-April WBE WBE=-6.49--0.02(51.00-DBT)++0.37(DBT-51.00)++1.48LTEQ 0.83 5 19.76%

HW HW=25705.85--244.69(63.80-DBT)++6.31(DBT-63.80)++273.22LTEQ 0.44 3983 11.62%

72 June-May WBE WBE =-7.77 - 0.12(44.00 - DBT)+ +0.30(DBT - 44.00)+ +1.50LTEQ 0.82 5 20.28%

HW HW = 27047.32 - -269.50(58.00 - DBT)+ + -49.03(DBT - 58.00)++251.04 LTEQ 0.44 3855 11.32%

73 July WBE WBE = -6.37 - 0.33(82.25 - DBT)+ +15.69(DBT - 82.25)+ +2.04LTEQ 0.79 8 23.31%

HW HW = 31242.52 - 2.38(75.00 - DBT)+ + -0.36(DBT - 75.00)++0.32 LTEQ 0.1 11 0.03%

157
S.No. Base Period Channel Equation R^2 RMSE CV-RMSE

74 July-August WBE WBE = -7.12 - 0.24(82.01 - DBT)+ +12.56(DBT - 82.01)+ +2.05LTEQ 0.88 6 17.17%

HW HW = 31199.45 - -7.33(77.98 - DBT)+ + 3.19(DBT - 77.98)++2.16 LTEQ 0.23 50 0.16%

75 July-Sept WBE WBE = -4.58 - 0.33(82.00 - DBT)+ +11.51(DBT - 82.00)+ +1.93LTEQ 0.84 6 19.09%

HW HW = 31093.00 - -14.62(78.00 - DBT)+ + 10.14(DBT - 78.00)++7.11 LTEQ 0.21 156 0.50%

76 July-October WBE WBE = -9.90 - 0.00(64.00 - DBT)+ +0.53(DBT - 64.00)+ +1.80LTEQ 0.82 6 20.49%

HW HW = 31884.44 - 507.83(42.62 - DBT)+ + -28.32(DBT - 42.62)++18.37 LTEQ 0.25 523 1.66%

77 July-Nov WBE WBE = 0.24 - 0.29(82.00 - DBT)+ +10.46(DBT - 82.00)+ +1.65LTEQ 0.79 7 22.53%

HW HW = 29050.41 - -128.62(71.30 - DBT)+ + 39.14(DBT - 71.30)++99.81 LTEQ 0.29 2550 7.88%

78 July-Dec WBE WBE = -9.89 -- 0.18(44.00 - DBT)+ +0.34(DBT - 44.00)+ +1.56LTEQ 0.81 6 22.21%

HW HW = 26506.02 - -229.75(64.46 - DBT)+ + 4.22(DBT - 64.46)++237.81 LTEQ 0.39 4092 12.20%

79 July-Jan WBE WBE = -8.28 - -0.19(44.00 - DBT)+ +0.33(DBT - 44.00)+ +1.49LTEQ 0.81 6 21.89%

HW HW = 24584.78 - -213.87(70.75 - DBT)+ + 124.15(DBT - 70.75)++296.23 LTEQ 0.45 4119 12.04%

80 July-Feb WBE WBE = -6.56 - -0.07(51.01 - DBT)+ +0.40(DBT - 51.01)+ +1.44LTEQ 0.8 6 22.03%

HW HW = 24840.67 - -254.69(64.00 - DBT)+ + 10.47(DBT - 64.00)++317.56 LTEQ 0.47 4215 12.15%

81 July-March WBE WBE = -6.02 -- 0.06(51.01 - DBT)+ +0.39(DBT - 51.01)+ +1.42LTEQ 0.81 6 21.09%

HW HW = 24015.63 - -219.61(70.81 - DBT)+ + 136.04(DBT - 70.81)++321.39 LTEQ 0.45 4242 12.81%

82 July-April WBE WBE = -5.51 -- 0.02(51.00 - DBT)+ +0.36(DBT - 51.00)+ +1.43LTEQ 0.82 5 20.25%

HW HW = 25153.19 - -245.98(63.86 - DBT)+ + 14.65(DBT - 63.86)++299.09 LTEQ 0.44 4126 11.93%

83 July-May WBE WBE = -6.85 - -0.12(44.00 - DBT)+ +0.29(DBT - 44.00)+ +1.46LTEQ 0.97 32 3.10%

HW HW = 26671.58 - -271.63(57.68 - DBT)+ + -48.08(DBT - 57.68)++272.62 LTEQ 0.44 3984 11.62%

84 July-June WBE WBE =-7.77 - 0.12(44.00 - DBT)+ +0.30(DBT - 44.00)+ +1.50LTEQ 0.82 5 20.28%

HW HW = 27047.32 - -269.50(58.00 - DBT)+ + -49.03(DBT - 58.00)++251.04 LTEQ 0.44 3855 11.32%

85 August WBE WBE =-8.01 - 0.82(70.00 - DBT)+ +-0.05(DBT - 70.00)+ +2.02LTEQ 0.97 3 8.00%

HW HW = 31240.88 - -108.35(66.40 - DBT)+ + -3.67(DBT - 66.40)++3.53 LTEQ 0.24 67 0.21%

86 Aug-Sept WBE WBE =-9.32 - -0.58(68.00 - DBT)+ +0.46(DBT - 68.00)+ +1.87LTEQ 0.88 5 16.57%

HW HW = 31240.05 - 140.87(63.30- DBT)+ + -11.34(DBT - 63.30)++10.04 LTEQ 0.2 187 0.60%

87 Aug-October WBE WBE =-8.44 - 0.02(64.00 - DBT)+ +0.47(DBT -64.00)+ +1.74LTEQ 0.85 6 18.77%

HW HW = 31243.48 - -9.80(64.00- DBT)+ + -45.35(DBT - 64.00)++26.49 LTEQ 0.24 597 1.89%

88 Aug-Nov WBE WBE =-11.34 - -0.51(39.00 - DBT)+ +0.31(DBT -39.00)+ +1.57LTEQ 0.8 6 21.46%

HW HW = 28349.02 - -128.37(72.44- DBT)+ + 133.44(DBT - 72.44)++127.17 LTEQ 0.29 2835 8.69%

89 Aug-Dec WBE WBE =-8.21 --0.31(39.99 - DBT)+ +0.28(DBT -39.99)+ +1.47LTEQ 0.82 6 21.03%

HW HW = 24576.67 - -210.04(71.80- DBT)+ + 287.59(DBT - 71.80)++286.00 LTEQ 0.39 4404 12.94%

90 Aug-Jan WBE WBE =-6.40 - -0.14(43.99 - DBT)+ +0.29(DBT -43.99)+ +1.42LTEQ 0.83 5 20.27%

HW HW = 23213.66 - -216.27(72.30- DBT)+ + 373.88(DBT - 72.30)++346.78 LTEQ 0.45 4350 12.54%

91 Aug-Feb WBE WBE =-4.77 -- 0.04(51.00 - DBT)+ +0.36(DBT -51.00)+ +1.37LTEQ 0.82 5 20.55%

HW HW = 22723.38 - -228.18(72.18- DBT)+ + 394.33(DBT - 72.18)++366.02 LTEQ 0.47 4411 12.53%

92 Aug-March WBE WBE =-4.41 - -0.04(51.00 - DBT)+ +0.35(DBT -51.00)+ +1.36LTEQ 0.83 5 19.46%

HW HW = 22845.89 - -220.42(72.31- DBT)+ + 390.91(DBT - 72.31)++362.88 LTEQ 0.44 4415 12.51%

158
S.No. Base Period Channel Equation R^2 RMSE CV-RMSE

93 Aug-April WBE WBE =-5.04 - -0.09(44.00 - DBT)+ +0.27(DBT -44.00)+ +1.37LTEQ 0.84 5 18.58%

HW HW = 25721.64 - -271.09(57.00- DBT)+ + -66.36(DBT - 57.00)++332.16 LTEQ 0.43 4285 12.26%

94 Aug-May WBE WBE -5.51 - -0.08(43.99 - DBT)+ +0.26(DBT -43.99)+ +1.41LTEQ 0.83 5 19.50%

HW HW = 26298.62 - -272.94(57.00- DBT)+ + -63.18(DBT - 57.00)++299.94 LTEQ 0.13 23 0.07%

95 Aug-June WBE WBE =-6.72 - -0.10(44.00 - DBT)+ +0.28(DBT -44.00)+ +1.46LTEQ 0.84 5 19.18%

HW HW = 26820.77 - -271.68(57.00- DBT)+ + -67.17(DBT - 57.00)++274.36 LTEQ 0.44 3986 11.62%

96 Aug-July WBE WBE =-7.77 - 0.12(44.00 - DBT)+ +0.30(DBT - 44.00)+ +1.50LTEQ 0.82 5 20.28%

HW HW = 27047.32 - -269.50(58.00 - DBT)+ + -49.03(DBT - 58.00)++251.04 LTEQ 0.44 3855 11.32%

97 September WBE WBE =-6.70 - -1.69(68.00 - DBT)+ +1.60(DBT - 68.00)+ +1.56LTEQ 0.82 6 20.69%

HW HW = 31117.79 - 12.96(71.00 - DBT)+ + -81.73(DBT - 71.00)++17.44 LTEQ 0.27 241 0.77%

98 Sept-Oct WBE WBE =-4.50 - -0.03(68.00 - DBT)+ +1.32(DBT - 68.00)+ +1.51LTEQ 0.8 6 21.43%

HW HW = 31076.09 - 2.18(64.00 - DBT)+ + -81.24(DBT - 64.00)++42.66 LTEQ 0.25 700 2.21%

99 Sept-Nov WBE WBE =-0.96 - 0.14(68.00 - DBT)+ +1.28(DBT - 68.00)+ +1.36LTEQ 0.76 6 23.47%

HW HW = 31843.98 - 32.33(44.00 - DBT)+ + -154.32(DBT - 44.00)++162.31LTEQ 0.28 3223 9.74%

100 Sept-Dec WBE WBE =-0.10 - 0.09(68.00 - DBT)+ +1.46(DBT - 68.00)+ +1.29LTEQ 0.79 5 21.87%

HW HW = 32565.49 - 435.01(33.40 - DBT)+ + -241.08(DBT - 33.40)++336.03LTEQ 0.4 4758 13.70%

101 Sept-Jan WBE WBE =0.30 - 0.07(68.00 - DBT)+ +1.53(DBT - 68.00)+ +1.26LTEQ 0.82 5 20.38%

HW HW = 31058.79 - 102.46(34.00 - DBT)+ + -241.06(DBT - 34.00)++403.40LTEQ 0.45 4604 13.01%

102 Sept-Feb WBE WBE =0.61 - 0.07(68.00 - DBT)+ +1.62(DBT - 68.00)+ +1.23LTEQ 0.8 5 20.38%

HW HW = 31160.25 - 120.38(33.01 - DBT)+ + -245.32(DBT - 33.01)++415.87LTEQ 0.46 4625 12.90%

103 Sept-March WBE WBE =0.42 - 0.06(68.00 - DBT)+ +1.66(DBT - 68.00)+ +1.23LTEQ 0.82 5 18.98%

HW HW = 30900.49 - 40.70(34.00 - DBT)+ + -236.68(DBT - 34.00)++403.90LTEQ 0.44 4605 12.84%

104 Sept-April WBE WBE =0.54 - 0.09(68.00 - DBT)+ +1.43(DBT - 68.00)+ +1.27LTEQ 0.83 4 18.15%

HW HW = 25395.65 - -272.76(55.00 - DBT)+ + -108.73(DBT - 55.00)++375.41LTEQ 0.43 4452 12.57%

105 Sept-May WBE WBE =-4.26 - -0.13(39.99 - DBT)+ +0.20(DBT - 39.99)+ +1.33LTEQ 0.82 5 19.56%

HW HW = 26117.07 - -275.81(55.00 - DBT)+ + -99.23(DBT - 55.00)++334.79LTEQ 0.43 4283 12.25%

106 Sept-June WBE WBE =-5.42 - -0.08(44.00 - DBT)+ +0.26(DBT - 44.00)+ +1.40LTEQ 0.83 5 19.50%

HW HW = 26701.12 - -274.82(55.14 - DBT)+ + -95.52(DBT - 55.14)++303.53LTEQ 0.44 4128 11.93%

107 Sept-July WBE WBE = -6.74 - -0.12(44.00 - DBT)+ +0.30(DBT - 44.00)+ +1.45LTEQ 0.81 5 20.81%

HW HW = 26802.67 - -271.88(57.02 - DBT)+ + -61.78(DBT - 57.02)++274.85 LTEQ 0.44 3986 11.62%

108 Sept-Aug WBE WBE =-7.77 - 0.12(44.00 - DBT)+ +0.30(DBT - 44.00)+ +1.50LTEQ 0.82 5 20.28%

HW HW = 27047.32 - -269.50(58.00 - DBT)+ + -49.03(DBT - 58.00)++251.04 LTEQ 0.44 3855 11.32%

109 October WBE WBE = -5.11 - -0.19(67.68 - DBT)+ +4.88(DBT - 67.68)+ +1.40LTEQ 0.83 5 19.85%

HW HW = 29851.91 - -88.32(50.77 - DBT)+ + 58.14(DBT - 50.77)++83.48 LTEQ 0.35 856 2.67%

110 Oct-Nov WBE WBE = 1.79 - 0.22(57.00 - DBT)+ +-0.34(DBT - 57.00)+ +1.20LTEQ 0.74 6 23.09%

HW HW = 30508.49 -30.00(44.00 - DBT)+ + -163.26(DBT - 44.00)++230.43 LTEQ 0.25 3870 11.41%

111 Oct-Dec WBE WBE = 1.23 - 0.12(58.00 - DBT)+ +-0.30(DBT - 58.00)+ +1.20LTEQ 0.82 4 19.75%

HW HW = 30966.62 -536.73(33.58 - DBT)+ + -285.12(DBT - 33.58)++437.12 LTEQ 0.4 5249 14.67%

159
S.No. Base Period Channel Equation R^2 RMSE CV-RMSE

112 Oct-Jan WBE WBE = -1.52 - -0.12(39.00 - DBT)+ +0.10(DBT - 39.00)+ +1.19LTEQ 0.84 4 18.20%

HW HW = 20666.50 --257.91(64.80 - DBT)+ + 1208.83(DBT - 64.80)++508.53 LTEQ 0.44 4892 13.45%

113 Oct-Feb WBE WBE = -0.83 - -0.09(39.00 - DBT)+ +0.07(DBT - 39.00)+ +1.17LTEQ 0.82 4 18.51%

HW HW = 20776.65 --265.21(64.79 - DBT)+ + 1193.18(DBT - 64.79)++503.41 LTEQ 0.45 4852 13.20%

114 Oct-March WBE WBE = 0.04 - 0.01(68.01 - DBT)+ +5.13(DBT - 68.01)+ +1.18LTEQ 0.84 4 16.97%

HW HW = 21562.18 --246.86(64.95 - DBT)+ + 1119.75(DBT - 64.95)++476.30 LTEQ 0.42 4808 13.14%

115 Oct-April WBE WBE = -1.88 - -0.08(39.80 - DBT)+ +0.13(DBT - 39.80)+ +1.23LTEQ 0.85 4 16.50%

HW HW = 24253.58 --282.94(55.00 - DBT)+ + -27.26(DBT - 55.00)++424.77 LTEQ 0.42 4640 12.89%

116 Oct-May WBE WBE = -3.21 - -0.10(39.99 - DBT)+ +0.17(DBT - 39.99)+ +1.29LTEQ 0.83 5 18.65%

HW HW = 25351.52 --280.23(54.86 - DBT)+ + -79.67(DBT - 54.86)++372.13 LTEQ 0.42 4459 12.59%

117 Oct-June WBE WBE = -4.84 - -0.08(44.00 - DBT)+ +0.26(DBT - 44.00)+ +1.37LTEQ 0.83 5 18.95%

HW HW = 26137.67 --276.36(54.99 - DBT)+ + -90.85(DBT - 54.99)++333.27 LTEQ 0.43 4289 12.27%

118 Oct-July WBE WBE = -5.89 - -0.06(47.01 - DBT)+ +0.33(DBT - 47.01)+ +1.43LTEQ 0.82 5 20.59%

HW HW = 26357.60 --270.92(56.99 - DBT)+ + -59.49(DBT - 56.99)++299.04 LTEQ 0.44 4132 11.95%

119 Oct-Aug WBE WBE = -7.49 - -0.12(44.00 - DBT)+ +0.31(DBT - 44.00)+ +1.49LTEQ 0.83 5 20.06%

HW HW = 26786.16 --271.48(57.36 - DBT)+ + -52.75(DBT - 57.36)++271.33 LTEQ 0.44 3988 11.64%

120 Oct-Sept WBE WBE =-7.77 - 0.12(44.00 - DBT)+ +0.30(DBT - 44.00)+ +1.50LTEQ 0.82 5 20.28%

HW HW = 27047.32 - -269.50(58.00 - DBT)+ + -49.03(DBT - 58.00)++251.04 LTEQ 0.44 3855 11.32%

121 November WBE WBE = -2.06 - -0.51(40.00 - DBT)+ +0.66(DBT - 44.00)+ +1.05LTEQ 0.72 5 23.55%

HW HW = 28562.32 -200.17(47.00 - DBT)+ + -695.37(DBT - 47.00)++439.01 LTEQ 0.38 4587 12.82%

122 Nov-Dec WBE WBE = -1.14 -- 0.15(44.00 - DBT)+ +0.35(DBT - 44.00)+ +1.11LTEQ 0.83 4 18.43%

HW HW = 28373.28 -272.49(37.00 - DBT)+ + -472.42(DBT - 37.00)++619.44 LTEQ 0.45 5588 14.84%

123 Nov-Jan WBE WBE = -0.92 -- 0.17(40.00 - DBT)+ +0.17(DBT - 40.00)+ +1.12LTEQ 0.87 3 15.92%

HW HW = 27533.81 -128.73(37.00 - DBT)+ + -445.74(DBT - 37.00)++636.49 LTEQ 0.5 4931 13.06%

124 Nov-Feb WBE WBE = -0.80 -- 0.10(51.00 - DBT)+ +0.80(DBT - 51.00)+ +1.10LTEQ 0.85 4 16.66%

HW HW = 25471.23 --123.96(47.00 - DBT)+ + -874.44(DBT - 47.00)++601.90 LTEQ 0.49 4881 12.87%

125 Nov-March WBE WBE = -0.67 -- 0.06(51.00 - DBT)+ +0.37(DBT - 51.00)+ +1.14LTEQ 0.86 3 15.31%

HW HW = 28817.65 -119.40(34.00 - DBT)+ + -284.10(DBT - 34.00)++541.48 LTEQ 0.44 4907 13.09%

126 Nov-April WBE WBE = -1.01 --0.02(50.01 - DBT)+ +0.37(DBT - 50.01)+ +1.19LTEQ 0.87 4 15.19%

HW HW = 29969.61 -109.47(34.00 - DBT)+ + -294.93(DBT - 34.00)++474.58 LTEQ 0.43 4758 12.98%

127 Nov-May WBE WBE = -2.75 --0.06(44.00 - DBT)+ +0.23(DBT - 44.00)+ +1.27LTEQ 0.84 4 18.13%

HW HW = 30861.10 -40.54(34.00 - DBT)+ + -243.43(DBT - 34.00)++406.05 LTEQ 0.43 4603 12.82%

128 Nov-June WBE WBE = -4.87 --0.10(44.00 - DBT)+ +0.29(DBT - 44.00)+ +1.36LTEQ 0.84 5 18.42%

HW HW = 25484.16 --256.10(57.02 - DBT)+ + -110.14(DBT - 57.02)++365.07 LTEQ 0.44 4423 12.52%

129 Nov-July WBE WBE = -6.39 --0.13(44.00 - DBT)+ +0.32(DBT - 44.00)+ +1.42LTEQ 0.82 5 20.31%

HW HW = 24557.45 --235.82(66.01 - DBT)+ + 14.07(DBT - 66.01)++323.69LTEQ 0.45 4248 12.19%

130 Nov-Aug WBE WBE = -7.64 --0.13(44.00 - DBT)+ +0.32(DBT - 44.00)+ +1.49LTEQ 0.84 5 19.76%

HW HW = 24995.00 --236.02(66.70 - DBT)+ + 42.23(DBT - 66.70)++292.65LTEQ 0.45 4086 11.84%

160
S.No. Base Period Channel Equation R^2 RMSE CV-RMSE

131 Nov-Sept WBE WBE = -7.96 --0.13(44.00 - DBT)+ +0.31(DBT - 44.00)+ +1.50LTEQ 0.83 5 20.04%

HW HW = 25888.32 --243.90(64.07 - DBT)+ + -2.12(DBT - 64.07)++268.37LTEQ 0.46 3941 11.52%

132 Nov-Oct WBE WBE =-7.77 - 0.12(44.00 - DBT)+ +0.30(DBT - 44.00)+ +1.50LTEQ 0.82 5 20.28%

HW HW = 27047.32 - -269.50(58.00 - DBT)+ + -49.03(DBT - 58.00)++251.04 LTEQ 0.44 3855 11.32%

133 December WBE WBE = 0.31 - 0.43(29.00 - DBT)+ +-0.08(DBT - 29.00)+ +1.16LTEQ 0.97 2 8.86%

HW HW = 28405.82 - 542.42(39.00 - DBT)+ + -2129.50(DBT - 39.00)++852.15 LTEQ 0.69 4971 12.54%

134 Dec-Jan WBE WBE = 0.23 - 0.24(29.00 - DBT)+ +-0.06(DBT - 29.00)+ +1.17LTEQ 0.96 2 8.88%

HW HW = 27417.06 - 302.07(39.00 - DBT)+ + -1457.69(DBT -39.00)++777.08 LTEQ 0.66 4330 11.19%

135 Dec-Feb WBE WBE = -0.03 - -0.01(49.95 - DBT)+ +-19.06(DBT -49.95)+ +1.15LTEQ 0.95 2 9.23%

HW HW = 24278.60 - -127.97(50.07 - DBT)+ + -11659.86(DBT -50.07)++672.60 LTEQ 0.57 4679 12.11%

136 Dec-March WBE WBE = 0.31 - 0.00(45.90 - DBT)+ +-0.28(DBT - 45.90)+ +1.15LTEQ 0.92 3 11.94%

HW HW = 28767.00 - 163.01(34.00 - DBT)+ + -314.25(DBT -34.00)++574.06 LTEQ 0.48 4844 12.77%

137 Dec-April WBE WBE = -1.06 - 0.00(50.99 - DBT)+ +0.35(DBT - 50.99)+ +1.23LTEQ 0.91 3 12.55%

HW HW = 30189.38 - 139.64(34.00 - DBT)+ + -321.55(DBT -34.00)++487.98LTEQ 0.47 4709 12.79%

138 Dec-May WBE WBE = -2.83 - -0.02(43.99 - DBT)+ +0.20(DBT - 43.99)+ +1.31LTEQ 0.86 4 16.78%

HW HW = 24994.33 - -288.23(54.01 - DBT)+ + -126.75(DBT -54.01)++415.49LTEQ 0.47 4550 12.67%

139 Dec-June WBE WBE = -4.80 --0.03(47.00 - DBT)+ +0.29(DBT - 47.00)+ +1.40LTEQ 0.87 4 17.24%

HW HW = 25390.71 - -279.08(57.00 - DBT)+ + -96.21(DBT -57.00)++359.84LTEQ 0.47 4344 12.32%

140 Dec-July WBE WBE = -6.31 --0.04(47.77 - DBT)+ +0.33(DBT - 47.77)+ +1.47LTEQ 0.84 5 19.49%

HW HW = 24877.54 - -257.82(64.00 - DBT)+ + 1.45(DBT -64.00)++314.46LTEQ 0.48 4153 11.95%

141 Dec-Aug WBE WBE = -7.65 --0.05(47.01 - DBT)+ +0.32(DBT - 47.01)+ +1.54LTEQ 0.85 5 18.89%

HW HW = 25195.29 - -250.90(66.01 - DBT)+ + 42.05(DBT -66.01)++281.27LTEQ 0.48 3975 11.57%

142 Dec-Sept WBE WBE = -7.94 --0.06(47.01 - DBT)+ +0.32(DBT - 47.01)+ +1.55LTEQ 0.84 5 19.27%

HW HW = 26022.29 - -257.48(64.00 - DBT)+ + 8.47(DBT -64.00)++255.77LTEQ 0.25 3870 11.41%

143 Dec-Oct WBE WBE = -7.02 -0.00(51.00 - DBT)+ +0.33(DBT - 51.00)+ +1.54LTEQ 0.84 5 19.59%

HW HW = 27265.25 - -295.64(56.99 - DBT)+ + -43.32(DBT -56.99)++238.00LTEQ 0.47 3722 10.99%

144 Dec-Nov WBE WBE =-7.77 - 0.12(44.00 - DBT)+ +0.30(DBT - 44.00)+ +1.50LTEQ 0.82 5 20.28%

HW HW = 27047.32 - -269.50(58.00 - DBT)+ + -49.03(DBT - 58.00)++251.04 LTEQ 0.44 3855 11.32%

161
C1.3. Hotel, Washington D.C Region

a. Time Series and Scatter Plots

Typical examples of the time series and scatter plots for the WBE energy

channel for the hotel in Washington D.C. are shown in Fig C1.3.1 and Fig.

C1.3.2, respectively. The blue portion on the graphs represents the short data-

set used to estimate the energy use over the whole year represented by the red

portion on the graphs. The graphs were plotted for each of all the base periods

shown in Table C1.3.

Figure C1.3.1 Time Series Plots for WBE for October used as the Short Data-Set for

Service Hotel, Washington D.C.

Figure C1.3.2 Scatter Plot for WBE for October used as the Short Data-Set for the Hotel,

Washington D.C.
162
 b. Multivariate Change Point Models

Table C1.3.1

MCP Models for all Base Periods and Energy Channels for Full Service Hotel,

Washington D.C.

S.No. Base Period Channel Equation R^2 RMSE CV-RMSE

1 January WBE WBE=1014.60 - -25.57 (33.20-DBT) + + -39.67 (DBT-33.20)+ + 0.0 LTEQ 0.87 73.75 7.05%

2 Jan-Feb WBE WBE = 690.42 -- 31.22 (42.36-DBT)+ + -17.18 (DBT - 42.36)+ + 0.00 LTEQ 0.92 73.73 7.88%

3 Jan-March WBE WBE = 636.47 - -33.52(43.02 - DBT)+ +-15.69(DBT - 43.02)+ +0.00LTEQ 0.93 75.01 9.02%

4 Jan-April WBE WBE = 382.32 - -29.39(53.36 - DBT)+ + 10.19(DBT -53.36)+ + 0.00LTEQ 0.93 76.35 10.21%

5 Jan-May WBE WBE = 390.62 - -29.39(53.08 - DBT)+ + 9.26(DBT-53.08)+ +0.00LTEQ 0.94 70.88 10.15%

6 Jan-June WBE WBE = 377.35 - -29.39(53.53 - DBT)+ + 11.46(DBT-53.53)+ +0.00 LTEQ 0.93 67.73 9.91%

7 Jan-July WBE WBE = 370.80 - -29.38(53.76 - DBT)+ +12.29(DBT - 53.76)+ +0.00LTEQ 0.93 65.04 9.58%

8 Jan-August WBE WBE = -373.02 --29.41(53.68 -DBT)+ +12.31(DBT - 53.68)+ +0.00LTEQ 0.92 63.2 9.34%

9 Jan-Sept WBE WBE = 369.52 - -29.41(53.80 - DBT)+ +12.38(DBT-53.80)+ +0.00LTEQ 0.92 61.35 9.16%

10 Jan-October WBE WBE = 375.37 - -28.54(54.24 - DBT) + +12.35(DBT - 54.24)+ +0.00LTEQ 0.91 61.45 9.34%

11 Jan-Nov WBE WBE =372.04 - -28.53(54.33- DBT)+ + 12.54(DBT - 54.33)+ + 0.00LTEQ 0.92 61.35 9.35

12 Jan-Dec WBE WBE =372.74 - -28.56(54.32 - DBT)+ +12.49(DBT - 54.32)+ +0.00LTEQ 0.92 61.39 9.17%

13 February WBE WBE= 718.80 - -34.51(40.40 - DBT)+ + -16.76(DBT - 40.40)+ + 0.00LTEQ 0.92 70.35 8.67%

14 Feb-March WBE WBE= 685.87 - -38.29(40.16 - DBT)+ +-16.07(DBT - 40.16)+ + 0.00LTEQ 0.93 70.27 9.78%

15 Feb-April WBE WBE= 383.48 - -28.46(53.50-DBT)+ + 10.22(DBT - 53.50)+ + 0.00LTEQ 0.9 76.49 11.88%

16 Feb-May WBE WBE= 391.76 - -28.46(53.20 -DBT)+ + 9.26(DBT - 53.20)+ + 0.00LTEQ 0.9 69.49 11.42%

17 Feb-June WBE WBE= 378.76 - -28.46(53.66 - DBT)+ + 11.46(DBT - 53.66)+ + 0.00LTEQ 0.89 65.87 10.83%

18 Feb-July WBE WBE= 372.34 - -28.45(53.89 - DBT)+ + 12.29(DBT - 53.89)+ + 0.00LTEQ 0.89 62.96 10.22%

19 Feb-August WBE WBE= 374.68 - -28.46(53.80 - DBT)+ + 12.13(DBT - 53.80)+ + 0.00LTEQ 0.88 61.1 9.81%

20 Feb-Sept WBE WBE= 371.17 - -28.45(53.93 - DBT)+ + 12.39(DBT-53.93)+ + 0.00LTEQ 0.87 59.2 9.53%

21 Feb-October WBE WBE= 378.16 - -27.33(54.45-DBT)+ + 12.34(DBT-54.45)+ +0.00LTEQ 0.86 59.43 9.68%

22 Feb-Nov WBE WBE= 373.89 - -27.23(54.70 - DBT)+ + 12.68(DBT - 54.70)+ + 0.00LTEQ 0.87 59.51 9.66%

23 Feb-Dec WBE WBE = 374.56 --27.81(54.48 - DBT)+ +12.51(DBT - 54.48)+ +0.00LTEQ 0.89 59.84 9.43%

24 Feb-Jan WBE WBE =372.74 - -28.56(54.32 - DBT)+ +12.49(DBT - 54.32)+ +0.00LTEQ 0.92 61.39 9.17%

25 March WBE WBE=700.79 - -44.01(38.02-DBT)+ + -15.62(DBT-38.02)++0.00LTEQ 0.94 62.97 9.92%

26 March-April WBE WBE=377.98--27.32(53.83-DBT)++10.78(DBT-53.83)++0.00LTEQ 0.86 77.59 13.68%

27 March-May WBE WBE=389.21--27.32(53.42-DBT)++9.55(DBT-53.42)++0.00LTEQ 0.85 68.14 12.47%

28 March-June WBE WBE=376.30 --27.32(53.89-DBT)++11.75(DBT-53.89)++0.00LTEQ 0.84 63.74 11.35%

29 March-July WBE WBE=369.51--27.12(54.26-DBT)++12.66(DBT-54.26)++0.00LTEQ 0.84 60.49 10.43%

30 March-Aug WBE WBE=372.58--27.33(54.02-DBT)++12.35(DBT-54.02)++0.00LTEQ 0.83 58.75 9.89%

31 March-Sept WBE WBE=368.95--27.13(54.27-DBT)++12.69(DBT-54.27)++0.00LTEQ 0.82 56.79 9.52%

32 March-Oct WBE WBE=377.01--24.06(55.96-DBT)++13.40(DBT-55.96)++0.00LTEQ 0.81 56.79 9.60%

33 March-Nov WBE WBE=370.71--24.99(55.89-DBT)++13.67(DBT-55.89)++0.00LTEQ 0.84 57.29 9.61%

163
34 March-Dec WBE WBE=368.93--26.17(55.63-DBT)++13.57(DBT-55.63)++0.00LTEQ 0.87 58.31 9.43%

35 March-Jan WBE WBE=370.48--28.54(54.43-DBT)++12.68(DBT-54.43)++0.00LTEQ 0.92 60.27 9.16%

36 March-Feb WBE WBE =372.74 - -28.56(54.32 - DBT)+ +12.49(DBT - 54.32)+ +0.00LTEQ 0.92 61.39 9.17%

37 April WBE WBE=377.12- -13.53(61.11-DBT)+ + 19.33(DBT-61.11)++0.00LTEQ 0.72 55.31 11.13%

38 April-May WBE WBE=405.95--12.12(60.26-DBT)++14.99(DBT-60.26)++0.00LTEQ 0.67 48.5 9.67%

39 April-June WBE WBE=403.55--11.39(61.07-DBT)++16.83(DBT-61.07)++0.00LTEQ 0.79 45.04 8.39%

40 April-July WBE WBE=400.54--11.38(61.33-DBT)+ +17.21(DBT-61.33)++0.00LTEQ 0.83 43.25 7.64%

41 April-August WBE WBE=406.01--12.12(60.25-DBT)++15.28(DBT-60.25)++0.00LTEQ 0.81 46.1 7.87%

42 April-Sept WBE WBE=406.37-12.12(60.22-DBT)++15.24(DBT-60.22)++0.00LTEQ 0.81 45.21 7.66%

43 April-Oct WBE WBE= 409.02--14.22(60.19-DBT)++15.07(DBT-60.19)++0.00LTEQ 0.8 45.37 7.75%

44 April-Nov WBE WBE=383.23--23.28(56.66-DBT)++13.58(DBT-56.66)++0.00LTEQ 0.83 50.07 8.47%

45 April-Dec WBE WBE=378.06--25.39(56.09-DBT)++13.44(DBT-56.09)++0.09LTEQ 0.88 52.35 8.49%

46 April-Jan WBE WBE=373.01--26.99(55.69-DBT)++13.43(DBT-55.69)++0.00LTEQ 0.91 63.19 7.77%

47 April-Feb WBE WBE=378.48 - -27.95(54.71-DBT)+ + 12.48(DBT-54.71)++0.00LTEQ 0.93 57.55 8.55%

48 April-March WBE WBE =372.74 - -28.56(54.32 - DBT)+ +12.49(DBT - 54.32)+ +0.00LTEQ 0.92 61.39 9.17%

49 May WBE WBE=460.48--0.55(64.50-DBT)++15.50(DBT-64.50)++0.00LTEQ 0.67 37.82 7.47%

50 May-June WBE WBE=453.10--1.50(64.35-DBT)++17.35(DBT-64.35)++0.00LTEQ 0.83 37.68 6.77%

51 May-July WBE WBE=466.18-0.51(65.95-DBT)++18.54(DBT-65.95)++0.00LTEQ 0.85 37.46 6.38%

52 May-Aug WBE WBE=432.67--6.74(62.00-DBT)++15.26(DBT-62.00)++0.00LTEQ 0.79 43.18 7.11%

53 May-Sept WBE WBE=-458.12--0.84(63.93-DBT)++15.61(DBT-63.93)++0.00LTEQ 0.79 42.64 7.01%

54 May-Oct WBE WBE=425.80--13.42(61.25-DBT)++15.00(DBT-61.25)++0.00LTEQ 0.8 41.96 7.00%

55 May-Nov WBE WBE=-380.49-23.28(57.02-DBT)++14.00(DBT-57.02)++0.00LTEQ 0.84 48.66 8.05%

56 May-Dec WBE WBE=374.32--25.54(56.27-DBT)++13.77(DBT-56.27)++0.00LTEQ 0.88 51.42 8.16%

57 May-Jan WBE WBE=368.26--27.11(55.83-DBT)++13.76(DBT-55.83)++0.00LTEQ 0.93 55.07 8.12%

58 May-Feb WBE WBE=375.29--28.30(54.59-DBT)++12.57(DBT-54.59)++0.00LTEQ 0.93 57.32 8.31%

59 May-March WBE WBE=367.32--28.82(54.33-DBT)++12.77(DBT-54.33)++0.00LTEQ 0.92 61.41 8.96%

60 May-April WBE WBE =372.74 - -28.56(54.32 - DBT)+ +12.49(DBT - 54.32)+ +0.00LTEQ 0.92 61.39 9.17%

61 June WBE WBE=662.09-16.18(76.60-DBT)++26.38(DBT-76.60)++0.00LTEQ 0.82 38.08 6.26%

62 June-July WBE WBE=624.13-15.17(75.10-DBT)++23.94(DBT-75.10)++0.00LTEQ 0.82 36.33 5.76%

63 June-Aug WBE WBE=716.00-17.50(79.60-DBT)++-2.77(DBT-79.60)++0.00LTEQ 0.69 41.8 6.52%

64 June-Sept WBE WBE=753.89-16.02(82.48-DBT)++-14.45(DBT-82.48)++0.00LTEQ 0.73 42.21 6.66%

65 June-Oct WBE WBE=436.86--12.66(61.84-DBT)++14.91(DBT-61.84)++0.00LTEQ 0.77 42.36 6.84%

66 June-Nov WBE WBE=366.44--24.54(56.76-DBT)++14.62(DBT-56.76)++0.00LTEQ 0.83 48.62 7.82%

67 June-Dec WBE WBE=362.80--26.38(56.13-DBT)++14.33(DBT-56.13)++0.00LTEQ 0.89 51.45 7.93%

68 June-Jan WBE WBE=356.67--27.71(55.78-DBT)++14.37(DBT-55.71)++0.00LTEQ 0.93 55.41 7.93%

69 June-Feb WBE WBE=-359.80--27.51(55.74-DBT)++14.21(DBT-55.74)++0.00LTEQ 0.93 57.83 8.13%

70 June-March WBE WBE=359.29--28.83(54.61-DBT)++13.43(DBT-54.61)++0.00LTEQ 0.92 62.21 8.85%

71 June-April WBE WBE=368.01--28.55(54.49-DBT)++12.93(DBT-54.49)++0.00LTEQ 0.92 62.22 9.09%

72 June-May WBE WBE =372.74 - -28.56(54.32 - DBT)+ +12.49(DBT - 54.32)+ +0.00LTEQ 0.92 61.39 9.17%

73 July WBE WBE = 546.94 - -3.42(72.18 - DBT)+ +24.69(DBT - 72.18)+ +0.00LTEQ 0.83 32.72 5.01%

164
74 July-August WBE WBE = 683.81 - 29.44(76.67 - DBT)+ +3.88(DBT - 76.67)+ +0.00LTEQ 0.58 41.55 6.32%

75 July-Sept WBE WBE = 753.57 - 15.78(82.67 - DBT)+ +-15.25(DBT - 82.67)+ +0.00LTEQ 0.69 43.7 6.80%

76 July-October WBE WBE = 439.07 --12.30(61.72 - DBT)+ +14.53(DBT - 61.72)+ +0.00LTEQ 0.76 43.62 7.03%

77 July-Nov WBE WBE = 369.28 - -24.55(56.64 - DBT)+ +14.31(DBT - 56.64)+ +0.00LTEQ 0.84 50.38 8.07%

78 July-Dec WBE WBE = 366.14 -- 26.31(56.01 - DBT)+ +14.01(DBT - 56.01)+ +0.00LTEQ 0.89 53.25 8.12%

79 July-Jan WBE WBE = 360.82 - -27.70(55.64 - DBT)+ +14.01(DBT - 55.64)+ +0.00LTEQ 0.93 57.2 8.03%

80 July-Feb WBE WBE = 372.18 - -28.29(54.71 - DBT)+ +12.85(DBT - 54.71)+ +0.00LTEQ 0.93 59.66 8.24%

81 July-March WBE WBE = 362.68 -- 28.83(54.49 - DBT)+ +13.13(DBT - 54.49)+ +0.00LTEQ 0.92 64.17 8.99%

82 July-April WBE WBE = 370.85 -- 28.55(54.39 - DBT)+ +12.68(DBT - 54.39)+ +0.00LTEQ 0.92 63.92 9.23%

83 July-May WBE WBE = 375.10 - -28.54(54.24 - DBT)+ +12.27(DBT - 54.24)+ +0.00LTEQ 0.92 62.84 9.31%

84 July-June WBE WBE =372.74 - -28.56(54.32 - DBT)+ +12.49(DBT - 54.32)+ +0.00LTEQ 0.92 61.39 9.17%

85 August WBE WBE =676.49 - 29.89(76.30 - DBT)+ +-0.61(DBT - 76.30)+ +0.00LTEQ 0.46 36.31 5.49%

7 Aug-Sept WBE WBE =665.70 -16.98(76.30 - DBT)+ +6.54(DBT - 76.30)+ +0.00LTEQ 0.7 44.28 6.95%

87 Aug-October WBE WBE =443.53 - -12.46(61.17 - DBT)+ +13.53(DBT -61.17)+ +0.00LTEQ 0.76 44.18 7.24%

88 Aug-Nov WBE WBE =375.51 - -24.54(56.39 - DBT)+ +13.67(DBT -56.39)+ +0.00LTEQ 0.85 52.34 8.49%

89 Aug-Dec WBE WBE =372.30 --26.30(55.78 - DBT)+ +13.41(DBT -55.78)+ +0.00LTEQ 0.9 55.2 8.41%

90 Aug-Jan WBE WBE =369.44 - -27.94(55.14 - DBT)+ +13.17(DBT -55.14)+ +0.00LTEQ 0.93 59.28 8.21%

91 Aug-Feb WBE WBE =377.81 -- 28.30(54.51 - DBT)+ +12.32(DBT -54.51)+ +0.00LTEQ 0.93 61.55 8.38%

92 Aug-March WBE WBE =367.65 - -28.83(54.32 - DBT)+ +12.64(DBT -54.32)+ +0.00LTEQ 0.83 43.25 7.64%

93 Aug-April WBE WBE =374.17 - -28.55(54.27 - DBT)+ +12.34(DBT -54.27)+ +0.00LTEQ 0.92 65.7 9.43%

94 Aug-May WBE WBE= 379.06 - -28.60(54.07 - DBT)+ +11.77(DBT -54.07)+ +0.00LTEQ 0.92 64.28 9.49%

95 Aug-June WBE WBE =375.74 - -28.55(54.22 - DBT)+ +12.16(DBT -54.22)+ +0.00LTEQ 0.92 62.67 9.34%

96 Aug-July WBE WBE =372.74 - -28.56(54.32 - DBT)+ +12.49(DBT - 54.32)+ +0.00LTEQ 0.82 5.46 20.28%

97 September WBE WBE =687.84 - 13.80(79.50 - DBT)+ +23.76(DBT - 79.50)+ +0.00LTEQ 0.83 41.8 6.83%

98 Sept-Oct WBE WBE =437.63 - -12.30(61.84 - DBT)+ +14.96(DBT - 61.84)+ +0.00LTEQ 0.83 39.47 6.76%

99 Sept-Nov WBE WBE =368.00 - -24.54(56.70 - DBT)+ +14.51(DBT - 56.70)+ +0.00LTEQ 0.88 52.57 8.74%

100 Sept-Dec WBE WBE =365.66 - -26.30(56.03 - DBT)+ +14.07(DBT - 56.03)+ +0.00LTEQ 0.91 56.2 8.58%

101 Sept-Jan WBE WBE =360.24 -- 27.70(55.66 - DBT)+ +14.09(DBT - 55.66)+ +0.00LTEQ 0.94 60.91 8.29%

102 Sept-Feb WBE WBE =374.16 - -28.29(54.64 - DBT)+ +12.54(DBT - 54.64)+ +0.00LTEQ 0.94 63.42 8.49%

103 Sept-March WBE WBE =364.60 - -28.83(54.42 - DBT)+ +12.80(DBT - 54.42)+ +0.00LTEQ 0.93 68.39 9.37%

104 Sept-April WBE WBE =372.96 - -28.56(54.31 - DBT)+ +12.28(DBT - 54.31)+ +0.00LTEQ 0.93 67.55 9.63%

105 Sept-May WBE WBE =-379.45 - -28.61(54.05 - DBT)+ +11.48(DBT - 54.05)+ +0.00LTEQ 0.93 65.72 9.68%

106 Sept-June WBE WBE =374.69 - -28.55(54.26 - DBT)+ +12.16(DBT - 54.26)+ +0.00LTEQ 0.93 63.88 9.50%

107 Sept-July WBE WBE = 371.05 - -28.55(54.38 - DBT)+ +12.63(DBT - 54.38)+ +0.00LTEQ 0.92 62.39 9.31%

108 Sept-Aug WBE WBE =372.74 - -28.56(54.32 - DBT)+ +12.49(DBT - 54.32)+ +0.00LTEQ 0.82 5.46 20.28%

109 October WBE WBE = 434.11 - -12.30(62.13 - DBT)+ +17.80(DBT - 62.13)+ +0.00LTEQ 0.82 37.7 6.77%

110 Oct-Nov WBE WBE = 370.22 - -24.54(56.60 - DBT)+ +14.39(DBT - 56.60)+ +0.00LTEQ 0.89 57.76 9.69%

111 Oct-Dec WBE WBE = 372.96 - -26.30(55.75 - DBT)+ +12.92(DBT - 55.75)+ +0.00LTEQ 0.92 60.43 9.03%

112 Oct-Jan WBE WBE = 383.53 - -28.20(54.46 - DBT)+ +10.49(DBT - 54.46)+ +0.00LTEQ 0.94 64.71 8.64%

113 Oct-Feb WBE WBE = 400.93 - -28.53(53.53 - DBT)+ +8.24(DBT - 53.53)+ +0.00LTEQ 0.94 66.4 8.59%

165
114 Oct-March WBE WBE =412.62 - -29.59(52.28 - DBT)+ +5.72(DBT - 52.28)+ +0.00LTEQ 0.93 71.21 9.50%

115 Oct-April WBE WBE = 387.75 - -28.76(53.66 - DBT)+ +9.67(DBT - 53.66)+ +0.00LTEQ 0.93 70.09 9.82%

116 Oct-May WBE WBE = 392.63 - -28.75(53.50 - DBT)+ +9.24(DBT - 53.50)+ +0.00LTEQ 0.93 67.34 9.80%

117 Oct-June WBE WBE = 379.53 - -28.60(54.05 - DBT)+ +11.53(DBT - 54.05)+ +0.00LTEQ 0.93 65.67 9.68%

118 Oct-July WBE WBE = 373.59 - -28.56(54.29 - DBT)+ +12.39(DBT - 54.29)+ +0.00LTEQ 0.92 63.99 9.46%

119 Oct-Aug WBE WBE = 375.07 - -28.55(54.24 - DBT)+ +12.31(DBT - 54.24)+ +0.00LTEQ 0.92 55.2 8.41%

120 Oct-Sept WBE WBE =372.74 - -28.56(54.32 - DBT)+ +12.49(DBT - 54.32)+ +0.00LTEQ 0.82 5.46 20.28%

121 November WBE WBE = 422.38 - -29.81(51.77 - DBT)+ +-3.26(DBT - 51.77)+ +0.00LTEQ 0.94 54.96 8.63%

122 Nov-Dec WBE WBE = 426.93 -- 29.07(52.13 - DBT)+ +-3.50(DBT - 52.13)+ +0.00LTEQ 0.94 58.51 8.05%

123 Nov-Jan WBE WBE = 428.34 -- 29.59(51.95 - DBT)+ +-3.60(DBT - 51.95)+ +0.00LTEQ 0.94 64.24 7.70%

124 Nov-Feb WBE WBE = 447.66 -- 29.69(51.16 - DBT)+ +-4.72(DBT - 51.16)+ +0.00LTEQ 0.94 66.08 7.97%

125 Nov-March WBE WBE = 513.61 -- 30.84(48.21 - DBT)+ +-10.47(DBT - 48.21)+ +0.00LTEQ 0.93 69.52 7.75%

126 Nov-April WBE WBE = 383.62 --29.38(53.37 - DBT)+ +9.14(DBT - 53.37)+ +0.00LTEQ 0.93 71.34 9.63%

127 Nov-May WBE WBE = 388.76 --29.37(53.20 - DBT)+ +9.13(DBT - 53.20)+ +0.00LTEQ 0.94 67.93 9.63%

128 Nov-June WBE WBE = 376.62 --29.37(53.62 - DBT)+ +11.43(DBT - 53.62)+ +0.00LTEQ 0.93 66.17 9.53%

129 Nov-July WBE WBE = 368.97 --29.18(54.01 - DBT)+ +12.47(DBT - 54.01)+ +0.00LTEQ 0.93 64.25 9.32%

130 Nov-Aug WBE WBE = 370.89 --29.18(53.94 - DBT)+ +12.35(DBT - 53.94)+ +0.00LTEQ 0.92 62.88 9.16%

131 Nov-Sept WBE WBE = 368.22 --29.19(54.03 - DBT)+ +12.55(DBT - 54.03)+ +0.00LTEQ 0.92 61.27 9.01%

132 Nov-Oct WBE WBE =372.74 - -28.56(54.32 - DBT)+ +12.49(DBT - 54.32)+ +0.00LTEQ 0.82 5.46 20.28%

133 December WBE WBE = 576.86 -- 28.52(47.30 - DBT)+ +-13.63(DBT - 47.30)+ +0.00LTEQ 0.91 63.19 7.77%

134 Dec-Jan WBE WBE = 573.70 - -29.21(47.30 - DBT)+ +-13.34(DBT - 47.30)+ +0.00LTEQ 0.91 68.9 7.41%

135 Dec-Feb WBE WBE = 511.30 - -29.54(49.13 - DBT)+ +-8.41(DBT -49.13)+ +0.00LTEQ 0.92 69.55 7.79%

136 Dec-March WBE WBE = 521.85 -- 30.72(47.99 - DBT)+ +-10.72(DBT - 47.99)+ +0.00LTEQ 0.93 72.86 8.81%

137 Dec-April WBE WBE = 383.90 - -29.31(53.42 - DBT)+ +10.06(DBT - 53.42)+ +0.00LTEQ 0.93 73.55 9.66%

138 Dec-May WBE WBE = 391.51 - -29.31(53.17 - DBT)+ +9.22(DBT - 53.17)+ +0.00LTEQ 0.77 42.36 6.84%

139 Dec-June WBE WBE = 378.62 --29.31(53.60 - DBT)+ +11.43(DBT - 53.60)+ +0.00LTEQ 0.93 66.92 9.53%

140 Dec-July WBE WBE = 372.35 --29.31(53.82 - DBT)+ +12.24(DBT - 53.82)+ +0.00LTEQ 0.93 64.68 9.29%

141 Dec-Aug WBE WBE = 374.56 --29.32(53.74 - DBT)+ +12.09(DBT - 53.74)+ +0.00LTEQ 0.92 63.09 9.11%

142 Dec-Sept WBE WBE = 371.15 --29.32(53.85 - DBT)+ +12.34(DBT - 53.85)+ +0.00LTEQ 0.92 61.44 8.98%

143 Dec-Oct WBE WBE = 376.02 --28.56(54.23 - DBT)+ 12.31(DBT - 54.23)+ +0.00LTEQ 0.92 61.48 9.14%

144 Dec-Nov WBE WBE =372.74 - -28.56(54.32 - DBT)+ +12.49(DBT - 54.32)+ +0.00LTEQ 0.82 5.46 20.28%

166
C2. Hybrid Inverse Modeling Approach-Method 1

(Refer Section 6.5 and Sections 9.1.1 and 9.1.2)

This approach for predicting energy use combines the monitored daily energy

use and internal loads with atleast one year of recent utility bills (representing the

long-term data) to provide a prediction of the building energy performance for the

whole year (Abushakra, Reddy, Claridge 1999). Thus, in this modeling technique,

information from monthly utility bills is applied to the model. The utility bills are

accurate and easily acceptable sources of building energy information. In order

to minimize the confounding effects of co-linearity between regression variables,

the regression is completed in two stages (Abushakra 2000).

C2.1. Large Hotel at Chicago, Illinois

Table C2.1.1

One Year of Monthly Utility History (Daily Average Values) for Large Hotel at Chicago, IL.

Monthly Utility History

S.No. Year Month DBTk (wk-0.009)
+ WBEk HWk CHWk
k Wh MBTU/hr MBTU/hr

1 2007 1 23.6 0.00000 948 8232 401

2 2007 2 28.0 0.00000 944 8918 392
3 2007 3 34.9 0.00000 947 7096 426
4 2007 4 47.1 0.00000 967 3975 694
5 2007 5 59.5 0.00027 1057 1411 1837
6 2006 6 70.0 0.00141 1181 517 4588
7 2006 7 74.3 0.00263 1259 445 6540
8 2006 8 71.2 0.00315 1243 474 5937
9 2006 9 64.6 0.00113 1123 617 3618
10 2006 10 53.1 0.00015 1018 2016 1237
11 2006 11 39.6 0.00007 953 5070 553
12 2006 12 27.4 0.00000 939 8847 404

167
 a. Regression Equations Obtained from Utility Bills (Stage 1):

Table C2.1.2

Regression Equations for Stage-1 of Hybrid Inverse Modeling (Method 1) for Large Hotel

at Chicago, IL

Model Model Equation R^2 RMSE CV‐RMSE

WBEk = 945.70 ‐ 0.00 (44.32 ‐ DBTk)+ + 7.07(DBTk ‐ 44.32)+ + 35757.29 (wk‐0.009)+ 1 5.49 0.50%
4P + + +
HWk = 664.71 ‐‐226.29 (61.40 ‐ DBTk) + 0.00(DBTk ‐ 61.40) +‐ 72833.71 (wk‐0.009) 0.98 532.01 13.40%
Model + + +
CHWk = 469.37 ‐ 0.00 (50.33 ‐ DBTk) + 156.92(DBTk ‐ 50.33) + 767034.10 (wk‐0.009) 0.99 192.24 8.70%

b. Regression Equations Obtained from Utility Bills and Monitored Data

(Stage 2):

Table C2.1.3

Regression Equations for Stage-2 of Hybrid Inverse Modeling (Method 1) for Large Hotel

at Chicago, IL

Base
Period Model Model Equation
2E*=-954.52- 0.00 (44.32 - DBTk)+ + 7.07(DBTk - 44.32)+ + 35757.29 (wk-0.009)+ +2.09LTEQ(WD)+2.76LTEQ(WE)
2P
January 2C*=-6314.82 --226.29 (61.40 - DBTk)+ + 0.00(DBTk - 61.40)+ +- 72833.71 (wk-0.009)++ 7.20LTEQ(WD) + 11.64 LTEQ(WE)
Linear
2H* = 9478.60- 0.00 (50.33 - DBTk)+ + 156.92(DBTk - 50.33)+ + 767034.10 (wk-0.009)+ + -10.57 LTEQ(WD)+ -
16.18LTEQ(WE)
2E*=-866.30- 0.00 (44.32 - DBTk)+ + 7.07(DBTk - 44.32)+ + 35757.29 (wk-0.009)+ +1.99LTEQ(WD)+2.61LTEQ(WE)
2P
Jan-Feb 2C*=-5646.07 --226.29 (61.40 - DBTk)+ + 0.00(DBTk - 61.40)+ +- 72833.71 (wk-0.009)++ 6.47LTEQ(WD) + 10.48 LTEQ(WE)
Linear
2H* = -4498.33- 0.00 (50.33 - DBTk)+ + 156.92(DBTk - 50.33)+ + 767034.10 (wk-0.009)+ + 5.24 LTEQ(WD)+ 9.59LTEQ(WE)

2E*=-997.35- 0.00 (44.32 - DBTk)+ + 7.07(DBTk - 44.32)+ + 35757.29 (wk-0.009)+ +2.13LTEQ(WD)+2.84LTEQ(WE)

2P
Jan-March 2C*=-5109.42 --226.29 (61.40 - DBTk)+ + 0.00(DBTk - 61.40)+ +- 72833.71 (wk-0.009)++ 5.89LTEQ(WD) + 9.56 LTEQ(WE)
Linear
2H* = 5779.94- 0.00 (50.33 - DBTk)+ + 156.92(DBTk - 50.33)+ + 767034.10 (wk-0.009)+ + -5.54 LTEQ(WD)+- 7.95LTEQ(WE)

2E*=-1574.13- 0.00 (44.32 - DBTk)+ + 7.07(DBTk - 44.32)+ + 35757.29 (wk-0.009)+ +2.75LTEQ(WD)+3.85LTEQ(WE)

2P
March 2C*=-1867.58 --226.29 (61.40 - DBTk)+ + 0.00(DBTk - 61.40)+ +- 72833.71 (wk-0.009)++ 7.20LTEQ(WD) + 11.64 LTEQ(WE)
Linear
2H* = -36271.21- 0.00 (50.33 - DBTk)+ + 156.92(DBTk - 50.33)+ + 767034.10 (wk-0.009)+ + 40.05 LTEQ(WD)+
66.42LTEQ(WE)
2E*=73.10- 0.00 (44.32 - DBTk)+ + 7.07(DBTk - 44.32)+ + 35757.29 (wk-0.009)+ +0.99LTEQ(WD)+0.98LTEQ(WE)
2P
March-April 2C*=4578.98 --226.29 (61.40 - DBTk)+ + 0.00(DBTk - 61.40)+ +- 72833.71 (wk-0.009)+- 4.45LTEQ(WD) + - 7.25 LTEQ(WE)
Linear
2H* = -52457.90 - 0.00 (50.33 - DBTk)+ + 156.92(DBTk - 50.33)+ + 767034.10 (wk-0.009)+ + 57.38 LTEQ(WD)+
94.84LTEQ(WE)
2E*=-35.91- 0.00 (44.32 - DBTk)+ + 7.07(DBTk - 44.32)+ + 35757.29 (wk-0.009)+ +1.11LTEQ(WD)+1.17LTEQ(WE)
2P 2C*=-11579.46 --226.29 (61.40 - DBTk)+ + 0.00(DBTk - 61.40)+ +- 72833.71 (wk-0.009)++ 12.87LTEQ(WD) + 21.21
March-May
Linear LTEQ(WE)
2H* = -35635.27 - 0.00 (50.33 - DBTk)+ + 156.92(DBTk - 50.33)+ + 767034.10 (wk-0.009)+ + 39.21
LTEQ(WD)+65.24LTEQ(WE)
2E*=-296.81- 0.00 (44.32 - DBTk)+ + 7.07(DBTk - 44.32)+ + 35757.29 (wk-0.009)+ +1.39LTEQ(WD)+1.64LTEQ(WE)

May 2P 2C*=-17348.51 --226.29 (61.40 - DBTk)+ + 0.00(DBTk - 61.40)+ +- 72833.71 (wk-0.009)++ 18.96LTEQ(WD) + 31.43

168
 Base
Period Model Model Equation
Linear LTEQ(WE)

2H* = -2826.49 - 0.00 (50.33 - DBTk)+ + 156.92(DBTk - 50.33)+ + 767034.10 (wk-0.009)+ + 3.50 LTEQ(WD)+ 6.82LTEQ(WE)

2E*=784.88- 0.00 (44.32 - DBTk)+ + 7.07(DBTk - 44.32)+ + 35757.29 (wk-0.009)+ +0.23LTEQ(WD)+-0.28LTEQ(WE)

2P
May-June 2C*=7784.02 --226.29 (61.40 - DBTk)+ + 0.00(DBTk - 61.40)+ +- 72833.71 (wk-0.009)++-8.13LTEQ(WD) +- 13.49 LTEQ(WE)
Linear
2H* = -582.94- 0.00 (50.33 - DBTk)+ + 156.92(DBTk - 50.33)+ + 767034.10 (wk-0.009)+ + 1.14 LTEQ(WD)+ 2.49LTEQ(WE)

2E*=630.86- 0.00 (44.32 - DBTk)+ + 7.07(DBTk - 44.32)+ + 35757.29 (wk-0.009)+ +0.40LTEQ(WD)+0.00LTEQ(WE)

2P 2C*=10073.86 --226.29 (61.40 - DBTk)+ + 0.00(DBTk - 61.40)+ +- 72833.71 (wk-0.009)+ + -10.44LTEQ(WD) + -
May-July
Linear 17.34LTEQ(WE)
2H* = 1013.30 - 0.00 (50.33 - DBTk)+ + 156.92(DBTk - 50.33)+ + 767034.10 (wk-0.009)+ + -0.53 LTEQ(WD)+- 0.45LTEQ(WE)

2E*=358.69- 0.00 (44.32 - DBTk)+ + 7.07(DBTk - 44.32)+ + 35757.29 (wk-0.009)+ +0.71LTEQ(WD)+0.51LTEQ(WE)

2P 2C*=13256.23 --226.29 (61.40 - DBTk)+ + 0.00(DBTk - 61.40)+ +- 72833.71 (wk-0.009)++ -13.53LTEQ(WD) + -22.66
July
Linear LTEQ(WE)
2H* = 3728.96 - 0.00 (50.33 - DBTk)+ + 156.92(DBTk - 50.33)+ + 767034.10 (wk-0.009)+ + -3.36 LTEQ(WD)+ -
5.46LTEQ(WE)
2E*=814.15- 0.00 (44.32 - DBTk)+ + 7.07(DBTk - 44.32)+ + 35757.29 (wk-0.009)+ +0.20LTEQ(WD)+-0.35LTEQ(WE)
2P 2C*=23941.41 --226.29 (61.40 - DBTk)+ + 0.00(DBTk - 61.40)+ +- 72833.71 (wk-0.009)++-25.36LTEQ(WD)+ -42.57
July-August
Linear LTEQ(WE)
2H* = 3577.26 - 0.00 (50.33 - DBTk)+ + 156.92(DBTk - 50.33)+ + 767034.10 (wk-0.009)+ + -3.16 LTEQ(WD)+ -5.15LTEQ(WE)

2E*=671.64- 0.00 (44.32 - DBTk)+ + 7.07(DBTk - 44.32)+ + 35757.29 (wk-0.009)+ +0.36LTEQ(WD)+-0.10LTEQ(WE)

July- 2P 2C*=18192.01 --226.29 (61.40 - DBTk)+ + 0.00(DBTk - 61.40)+ +- 72833.71 (wk-0.009)+ + -19.12LTEQ(WD) + -
September Linear 32.41LTEQ(WE)
2H* = 2820.77 - 0.00 (50.33 - DBTk)+ + 156.92(DBTk - 50.33)+ + 767034.10 (wk-0.009)+ + -2.42 LTEQ(WD)+- 3.93LTEQ(WE)

2E*=297.48- 0.00 (44.32 - DBTk)+ + 7.07(DBTk - 44.32)+ + 35757.29 (wk-0.009)+ +0.77LTEQ(WD)+0.57LTEQ(WE)

2P 2C*=6037.19 --226.29 (61.40 - DBTk)+ + 0.00(DBTk - 61.40)+ +- 72833.71 (wk-0.009)+ + -5.91LTEQ(WD) + -10.84
September
Linear LTEQ(WE)
2H* = 7363.87 - 0.00 (50.33 - DBTk)+ + 156.92(DBTk - 50.33)+ + 767034.10 (wk-0.009)+ + -7.53 LTEQ(WD)+ -
12.34LTEQ(WE)
2E*=696.41- 0.00 (44.32 - DBTk)+ + 7.07(DBTk - 44.32)+ + 35757.29 (wk-0.009)+ +0.33LTEQ(WD)+-0.13LTEQ(WE)
2P 2C*=11434.23 --226.29 (61.40 - DBTk)+ + 0.00(DBTk - 61.40)+ +- 72833.71 (wk-0.009)++-11.92LTEQ(WD)+ -
Sept-Oct
Linear 20.23LTEQ(WE)
2H* = -18452.46 - 0.00 (50.33 - DBTk)+ + 156.92(DBTk - 50.33)+ + 767034.10 (wk-0.009)+ + 20.25 LTEQ(WD)+
33.82LTEQ(WE)
2E*=496.82-0.00 (44.32 - DBTk)+ + 7.07(DBTk - 44.32)+ + 35757.29 (wk-0.009)+ +0.54LTEQ(WD)+0.23LTEQ(WE)
Sept- 2P
2C*=5589.33 --226.29 (61.40 - DBTk)+ + 0.00(DBTk - 61.40)+ +- 72833.71 (wk-0.009)+ + -5.60LTEQ(WD) + -9.50LTEQ(WE)
November Linear
2H* = -23430.92- 0.00 (50.33 - DBTk)+ + 156.92(DBTk - 50.33)+ + 767034.10 (wk-0.009)+ + 25.57
LTEQ(WD)+42.76LTEQ(WE)
2E*=-660.90- 0.00 (44.32 - DBTk)+ + 7.07(DBTk - 44.32)+ + 35757.29 (wk-0.009)+ +1.78LTEQ(WD)+2.29LTEQ(WE)
2P
November 2C*=-3357.27 --226.29 (61.40 - DBTk)+ + 0.00(DBTk - 61.40)+ +- 72833.71 (wk-0.009)+ + 4.04LTEQ(WD) + 6.89 LTEQ(WE)
Linear
2H* = -52807.28 - 0.00 (50.33 - DBTk)+ + 156.92(DBTk - 50.33)+ + 767034.10 (wk-0.009)+ + 57.22 LTEQ(WD)+
94.74LTEQ(WE)
2E*=-396.78- 0.00 (44.32 - DBTk)+ + 7.07(DBTk - 44.32)+ + 35757.29 (wk-0.009)+ +1.49LTEQ(WD)+1.80LTEQ(WE)
2P
Nov-Dec 2C*=-3789.35--226.29 (61.40 - DBTk)+ + 0.00(DBTk - 61.40)+ +- 72833.71 (wk-0.009)++4.50LTEQ(WD)+ 7.42LTEQ(WE)
Linear
2H* =-43343.35 - 0.00 (50.33 - DBTk)+ + 156.92(DBTk - 50.33)+ + 767034.10 (wk-0.009)+ + 47.16 LTEQ(WD)+
78.43LTEQ(WE)
2E*=-462.60-0.00 (44.32 - DBTk)+ + 7.07(DBTk - 44.32)+ + 35757.29 (wk-0.009)+ +1.56LTEQ(WD)+1.91LTEQ(WE)
2P
Nov-Jan 2C*=-3924.56 --226.29 (61.40 - DBTk)+ + 0.00(DBTk - 61.40)+ +- 72833.71 (wk-0.009)+ + 4.64LTEQ(WD) + 7.60LTEQ(WE)
Linear
2H* = -19761.35 - 0.00 (50.33 - DBTk)+ + 156.92(DBTk - 50.33)+ + 767034.10 (wk-0.009)+ + 21.45
LTEQ(WD)+36.33LTEQ(WE)

169
C2.2 Office Building at Albuquerque, New Mexico

Table C2.2.1

One Year of Monthly Utility History (Daily Average Values) for Office Building at

Albuquerque, NM.

Monthly Utility History

+ EWBk
(wk-0.009) HWk
S.No. Year Month DBTk kWh Mbtu/hr

1 2004 1 35.6 0.00000 22 37984

2 2004 2 39.7 0.00000 23 38448
3 2004 3 44.7 0.00000 25 35875
4 2004 4 55.1 0.00079 27 32309
5 2004 5 65.9 0.00026 28 31468
6 2004 6 72.9 0.00184 33 31258
7 2004 7 77.9 0.00405 34 31247
8 2004 8 74.2 0.00416 33 31286
9 2004 9 68.9 0.00190 30 31417
10 2004 10 56.0 0.00019 25 31998
11 2004 11 43.4 0.00000 23 35781
12 2004 12 34.7 0.00000 21 39657

a. Regression Equations Obtained from Utility Bills (Stage 1):

Table C2.2.2

Regression Equations for Stage-1 of Hybrid Inverse Modeling (Method-1) for Office at

Albuquerque, NM.

Model
Model Equation R^2 RMSE CV‐RMSE
Type

+ + +
4P WBEk = 21.57 ‐ 0.00 (34.71 ‐ DBTk) + 0.22(DBTk ‐ 34.71) + 777.72 (wk‐0.009) 0.97 0.93 3.5
Model + + +
HWk =31447.21 ‐‐337.56 (57.70 ‐ DBTk) + 0.00(DBTk ‐ 57.70) + ‐45822.31(wk‐0.009) 0.98 490.83 1.40%

170
Table C2.2.3

Regression Equations for Stage-2 of Hybrid Inverse Modeling (Method-1) for Office at

Albuquerque, NM.

Base Period Model Model Equation

WBE*=‐4 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.17LTEQ(WD)+1.29LTEQ(WE)
2P HW* =23926.30 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +
January Linear 424.80 LTEQ(WD)+ ‐19.40LTEQ(WE)
WBE*= ‐3.71 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.12LTEQ(WD)+1.19LTEQ(WE)
2P HW* =26527.53 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +
Jan-Feb Linear 319.89 LTEQ(WD)+ ‐141.97LTEQ(WE)
WBE*= ‐3.45 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.11LTEQ(WD)+1.05LTEQ(WE)
2P HW* =27469.35 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +
Jan-March Linear 260.26 LTEQ(WD)+ ‐180.65LTEQ(WE)
WBE*=‐3.93 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.10LTEQ(WD)+1.13LTEQ(WE)
2P HW* =33454.84 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +
February Linear 45.90 LTEQ(WD)+ ‐667.97LTEQ(WE)
WBE*= ‐3.25 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.09LTEQ(WD)+0.92LTEQ(WE)
2P HW* =31781.28 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +
Feb-March Linear 80.80 LTEQ(WD)+ ‐494.16LTEQ(WE)
WBE*= ‐4.24 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.16LTEQ(WD)+0.82LTEQ(WE)
2P HW* =31760.39 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +
Feb-April Linear 49.72 LTEQ(WD)+ ‐406.83LTEQ(WE)
WBE*= ‐2.43 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.08LTEQ(WD)+0.69LTEQ(WE)
2P HW* =29885.95 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +
March Linear 128.57 LTEQ(WD)+ ‐297.85LTEQ(WE)
WBE*= ‐4.38 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.19LTEQ(WD)+0.66LTEQ(WE)
2P HW* =31007.31 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +
March-April Linear 51.42 LTEQ(WD)+ ‐285.80LTEQ(WE)
WBE*= ‐4.91 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.22LTEQ(WD)+0.58LTEQ(WE)
2P HW* =30858.94 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +
March-May Linear 45.56 LTEQ(WD)+ ‐153.91LTEQ(WE)
WBE*= ‐6.57 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
April +1.32LTEQ(WD)+0.66LTEQ(WE)
2P HW* =32450.94 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ + ‐
Linear 42.00 LTEQ(WD)+ ‐306.44LTEQ(WE)
WBE*= ‐6.29 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.30LTEQ(WD)+0.55LTEQ(WE)
2P HW* =31553.16‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +‐6.52
April-May Linear LTEQ(WD)+ ‐116.49LTEQ(WE)
WBE*= ‐9.33 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐
0.009)++1.41LTEQ(WD)+0.75LTEQ(WE)
2P HW* =31495.92 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ + ‐
April-June Linear 4.81 LTEQ(WD)+ ‐81.89LTEQ(WE)
WBE*= ‐6.04‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.28LTEQ(WD)+0.47LTEQ(WE)
2P HW* =30449.91 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +
May Linear 36.72 LTEQ(WD)+ 74.52LTEQ(WE)
WBE*= ‐11.14 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.47LTEQ(WD)+0.85LTEQ(WE)

171
Base Period Model Model Equation
2P HW* =30906.52‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +18.01
May-June
Linear LTEQ(WD)+ 34.63LTEQ(WE)
WBE*= ‐11.19 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐
0.009)++1.45LTEQ(WD)+0.74LTEQ(WE)
2P HW* =31108.20 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +
May-July
Linear 11.03 LTEQ(WD)+ 20.29LTEQ(WE)
WBE*= ‐16.24‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.66LTEQ(WD)+1.20LTEQ(WE)
June 2P HW* =31359.12 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ + ‐
Linear 0.48 LTEQ(WD)+ ‐3.25LTEQ(WE)
WBE*= ‐13.41 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.52LTEQ(WD)+0.82LTEQ(WE)
2P HW* =31402.42‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +‐0.49
June0July
Linear LTEQ(WD)+ ‐2.17LTEQ(WE)
WBE*= ‐13.23 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐
0.009)++1.49LTEQ(WD)+0.70LTEQ(WE)
2P HW* =31383.94 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ + 1.66
June-August
Linear LTEQ(WD)+ 1.54LTEQ(WE)
WBE*= ‐10.67‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.37LTEQ(WD)+0.46LTEQ(WE)
2P HW* =31416.16 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ + 0.72
July Linear LTEQ(WD)+ 1.30LTEQ(WE)
WBE*= ‐11.69 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.40LTEQ(WD)+0.46LTEQ(WE)
2P HW* =31384.34‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +3.23
July-August Linear LTEQ(WD)+ 4.75LTEQ(WE)
WBE*= ‐10.29 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐
0.009)++1.33LTEQ(WD)+0.38LTEQ(WE)
July- 2P HW* =31276.09 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ + 9.01
September Linear LTEQ(WD)+ 12.58LTEQ(WE)
WBE*= ‐13.13‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.44LTEQ(WD)+0.51LTEQ(WE)
2P HW* =31364.25 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ + 5.21
August Linear LTEQ(WD)+ 7.01LTEQ(WE)
WBE*= ‐10.31 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.31LTEQ(WD)+0.36LTEQ(WE)
2P HW* =31211.22‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +12.88
August-Sept Linear LTEQ(WD)+ 17.61LTEQ(WE)
WBE*= ‐8.64 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐
0.009)++1.24LTEQ(WD)+0.48LTEQ(WE)
August- 2P HW* =31055.20 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +
October Linear 13.44 LTEQ(WD)+ ‐27.85LTEQ(WE)
WBE*= ‐7.483‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.18LTEQ(WD)+0.21LTEQ(WE)
2P HW* =31058.20 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +
September Linear 20.55 LTEQ(WD)+ 27.59LTEQ(WE)
WBE*=‐6.52 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.15LTEQ(WD)+0.48LTEQ(WE)
2P HW* =30932.41‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +16.25
Sept-Oct Linear LTEQ(WD)+ ‐48.50LTEQ(WE)
WBE*=‐2.02 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐
0.009)++0.96LTEQ(WD)+0.25LTEQ(WE)
Sept- 2P HW* =27327.78 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +
November Linear 180.83 LTEQ(WD)+ 186.73LTEQ(WE)
WBE*= ‐5.86‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.13LTEQ(WD)+0.71LTEQ(WE)
2P HW* =30977.53 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +
October Linear 4.35 LTEQ(WD)+ ‐124.16LTEQ(WE)
WBE*=0.45 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+0.86LTEQ(WD)+0.26LTEQ(WE)
October- 2P HW* =25588.87‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+
172
 Base Period Model Model Equation
November Linear +256.76 LTEQ(WD)+ 265.67LTEQ(WE)
WBE*=‐0.43 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐
0.009)++0.94LTEQ(WD)+0.57LTEQ(WE)
October- 2P HW* =29331.22 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +
December Linear 183.13 LTEQ(WD)+ ‐320.97LTEQ(WE)
WBE*= 7.82‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+0.56LTEQ(WD)+‐0.24LTEQ(WE)
2P HW* =19723.37 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +
November Linear 521.79 LTEQ(WD)+ 664.70LTEQ(WE)
WBE*=3.58 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+0.79LTEQ(WD)+0.35LTEQ(WE)
2P HW* =29391.72‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+
Nov-Dec Linear +239.11 LTEQ(WD)+ ‐492.47LTEQ(WE)
WBE*=‐0.56 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐
0.009)++0.98LTEQ(WD)+0.83LTEQ(WE)
2P HW* =25934.02 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +
Nov-Jan Linear 364.98 LTEQ(WD)+ ‐167.17LTEQ(WE)
WBE*= ‐1.90‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.09LTEQ(WD)+1.00LTEQ(WE)
2P HW* =42478.22 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ + ‐
December Linear 149.12 LTEQ(WD)+ ‐1920.61LTEQ(WE)
WBE*=‐3.58 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐0.009)+
+1.16LTEQ(WD)+1.21LTEQ(WE)
December ‐ 2P HW* =27219.71‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+
Jan Linear +367.53 LTEQ(WD)+ ‐364.90LTEQ(WE)
WBE*= ‐3.53 ‐ 0.00 (34.71 ‐ DBTk)+ + 0.22(DBTk ‐ 34.71)+ + 777.72 (wk‐
0.009)++1.13LTEQ(WD)+1.17LTEQ(WE)
December‐ 2P HW* =28768.46 ‐‐337.56 (57.70 ‐ DBTk)+ + 0.00(DBTk ‐ 57.70)+ + ‐45822.31(wk‐0.009)+ +
Feb Linear 279.35 LTEQ(WD)+ ‐439.65LTEQ(WE)

C3. Hybrid Inverse Modeling Approach-Method 2

Method 2 looks at predicting building performance at daily timescales using utility

bills only.

C3.1. Large Hotel at Chicago, Illinois

Table C3.1.1

Regression Equations derived from Utility Bills Only for Large Hotel at Chicago, IL

Model Model Equation R^2 RMSE CV‐RMSE

WBEk = 945.70 ‐ 0.00 (44.32 ‐ DBTk)+ + 7.07(DBTk ‐ 44.32)+ + 35757.29 (wk‐0.009)+ 1 5.49 0.50%
4P + + +
HWk = 664.71 ‐‐226.29 (61.40 ‐ DBTk) + 0.00(DBTk ‐ 61.40) +‐ 72833.71 (wk‐0.009) 0.98 532.01 13.40%
Model + + +
CHWk = 469.37 ‐ 0.00 (50.33 ‐ DBTk) + 156.92(DBTk ‐ 50.33) + 767034.10 (wk‐0.009) 0.99 192.24 8.70%

173
C3.2. Office Building at Albuquerque, New Mexico

Table C3.2.1

Regression Equations derived from Utility Bills Only for Office, Albuquerque, NM.

Model
Model Equation R^2 RMSE CV‐RMSE
Type

+ + +
4P WBEk = 21.57 ‐ 0.00 (34.71 ‐ DBTk) + 0.22(DBTk ‐ 34.71) + 777.72 (wk‐0.009) 0.97 0.93 3.5
Model + + +
HWk =31447.21 ‐‐337.56 (57.70 ‐ DBTk) + 0.00(DBTk ‐ 57.70) + ‐45822.31(wk‐0.009) 0.98 490.83 1.40%

C3.3 Full Service Hotel, Washington D.C. Area

Table C3.3.1

One Year of Monthly Utility History (Daily Average Values) for Full Service Hotel,

Washington D.C. Area.

Monthly Utility History

S.No. Year Month Days DBTk EWB (kWh)

1 2009 1 31 31.3 1046.85
2 2009 2 28 39.3 811.77
3 2009 3 31 44.8 634.78
4 2009 4 30 56.7 497.06
5 2009 5 31 65.2 506.37
6 2009 6 30 73.1 608.27
7 2009 7 31 76.4 653.07
8 2009 8 31 77.2 661.67
9 2009 9 30 73.6 612.08
10 2009 10 31 56.9 556.62
11 2009 11 30 46.3 637.00
12 2009 12 31 39.3 813.78
Table C3.3.2

Regression Equation derived from Utility Bills Only for Full Service Hotel, Washington

D.C. Area.

Model Type Model Equation Derived from Utility Bills R^2 RMSE CV‐RMSE
4P WBEk = 472.63 - - 28.55(51.28-TDB)+ + 6.55 (TDB - 51.28)+ 0.91 1029.2 8.79%

174