Electric / Gas / Water Eric Fox Oleg Moskatov Itron, Inc. April 17, 2008 VELCO Long-Term Demand Forecast Methodology Overview

Electric / Gas / Water

Eric Fox

Oleg Moskatov

Itron, Inc.

April 17, 2008

VELCO Long-Term Demand ForecastMethodology Overview

Knowledge to Shape Your Future Page 2

Overview

• Methodology– SAE energy model– Hourly Load and Peak Demand Model

• Assumptions– Weather data

• Normal weather

– Economic drivers– End-use saturation and efficiency trends– Price

• Preliminary Results– Recent peak and energy trends– Hourly load build-up results– Peak and energy forecast


Forecasting Approaches

• Three general approaches are used for forecasting long-term peak demand:– Load factor model

• Load factor = Average Demand / Peak Demand

• Peak Forecast = Energy Forecast / Hours * Load Factor

– Generalized econometric model• Peak Forecast = ƒ(peak-day weather, customers, economic activity)

– Build-up approach• Combine class energy forecasts with hourly profiles

• Aggregate to system load

• Find system peak

Load factors and econometric models are adequate for short-term forecasting, but can’t capture the impact of changing load diversity on long-term peak demand.


Build-up Forecast Approach

• Develop long-term energy forecasts by customer class– Residential, Commercial, Industrial, and Other

• Combine class energy forecasts with class hourly load profile models– Evaluate using end-use hourly load and energy estimates

• Aggregate class profiles to generate a long-term system forecast and extract the monthly system peak demand

• Calibrate to weather-normalized 2008 demand estimates


System Peak and Energy

Summer Winter GWh Load FactorPeak MW Peak MW

2002 1,023 1,013 6,277 0.701 2003 1,001 1,004 6,285 0.717 2004 985 1,086 6,390 0.741 2005 1,074 1,084 6,523 0.694 2006 1,118 1,060 6,473 0.661 2007 1,073 1,042 6,537 0.696

2002 - 07 1.0% 0.6% 0.8% -0.1%


Summer Peak

Date Peak (MW) AvgTemp CDD652002 3-Jul 1,023 83.5 18.5 2003 26-Jun 1,001 81.0 16.0 2004 9-Jun 985 76.5 11.5 2005 19-Jul 1,074 81.5 16.5 2006 2-Aug 1,118 84.0 19.0 2007 3-Aug 1,073 84.5 19.5 2008 Aug 1,089 82.3 17.3


System Peak Demand AnalysisDaily Peak Demand (MW) 2002 to 2007

Significantly less temperature-sensitive load than compared with

other regions

Significantly less temperature-sensitive load than compared with

other regions


Winter and Summer Monthly Peaks (MW)

But not surprisingly, peaks are driven by heating and

cooling demand

But not surprisingly, peaks are driven by heating and

cooling demand


System Peak Demand (Weekdays vs. Weekends)

Summer peak demand always falls during the week capturing

both commercial and residential cooling loads

Summer peak demand always falls during the week capturing

both commercial and residential cooling loads

Winter peaks also tend to fall during the week-days,

but winter week-end peaks can be nearly as high on cold days

Winter peaks also tend to fall during the week-days,

but winter week-end peaks can be nearly as high on cold days


Monthly peak demand (MW)

Since 2002, peak demand has been increasing roughly 1.0% per year

Since 2002, peak demand has been increasing roughly 1.0% per year

Summer peak: 10 MW per yearWinter peak: 6 MW per year

Summer peak: 10 MW per yearWinter peak: 6 MW per year


System Monthly Load Factor

Load FactorLoad Factor

Moving AverageMoving AverageTrendTrend

The load factor appears to be trending down slightly

implying peak demand is growing slightly faster than energy

The load factor appears to be trending down slightly

implying peak demand is growing slightly faster than energy


Peak-Day System Hourly Load Profile (MW)

System System

CommercialCommercial ResidentialResidential

IndustrialIndustrial

Small differences in customer class load growth can have a

significant impact on the peak and its timing

Small differences in customer class load growth can have a

significant impact on the peak and its timing


Peak-Day Residential Load Profile (MW)

ResidentialResidential

CoolingCooling

Base UseBase Use

Changes in end-use sales growth in turn impact

customer class hourly load

Changes in end-use sales growth in turn impact

customer class hourly load


Build-Up Model

HourlyAnd Peak

Forecast

Monthly/AnnualEnergy ForecastMonthly/AnnualEnergy Forecast

Class orEnd Use Profiles

Class orEnd Use Profiles

HourlySystem Forecast

HourlySystem Forecast

Long-RunLoad Shape Forecasting

System

Combine energy forecast and hourly class profiles using MetrixLT

Need class and end-use energy and profile forecasts


Long-Term Energy Forecasting

• Model that can account for economic changes as well as long term structural changes– Economic impacts – income, household size, household growth

– Price impacts

– Structural changes – saturation, efficiency, floor space, and thermal

integrity trends

– Weather impacts

– Appropriate interaction of these variables

• Approaches

– End-Use Modeling Framework – REEPS and COMMEND

– Statistically Adjusted End-Use Model – Econometric model specification


SAE Modeling Approach

• Blend end-use concepts into an econometric modeling framework:

– Average Use = Heating + Cooling + Other Use

• Define components in terms of its end use structure:

– Cooling = f (Saturation, Efficiency, Utilization)

• Utilization = g (Weather, Price, Income, Household Size)

• Leverage off of EIA census region end-use forecasts

– Adjust for known differences in service area saturations


Residential & Commercial SAE Model Regions


Statistically Adjusted End-use Modeling (cont.)

Estimate model using Ordinary Least Squares:

tt3t2t10t XOtherbXCoolbXHeatbbAvgUse


Residential Cooling End Use

m,yym,y CoolUseCoolIndexXCool

01

,

20.0

01

,

20.0

01

,

15.0

01

,, Pr

Pr

CDD

CDD

HHSize

HHSize

Income

Income

ice

iceCoolUse mymymymy

my

Type

Type

Typey

Typey

Type

Typeyy

EffSat

EffSat

UECIndexStructuralCoolIndex

01

01

01


Residential Cooling Saturation Trends

0.0%

10.0%

20.0%

30.0%

40.0%

50.0%

60.0%

70.0%

80.0%

1990 1993 1996 1999 2002 2005 2008 2011 2014 2017 2020 2023

CAC HPCool RAC CoolingNEC

0.0%

20.0%

40.0%

60.0%

80.0%

100.0%

1990 1993 1996 1999 2002 2005 2008 2011 2014 2017 2020 2023

CAC HPCool RAC CoolingBurlington


Residential Cooling Efficiency Trends

• Efficiency for cooling equipment is given for the total US• Seasonal Energy Efficiency Ratio (SEER) is defined as a ratio of the

total cooling of a central unitary air conditioner or a unitary heat pump in Btu during its normal annual usage period for cooling and the total electric energy input in watt-hours during the same period

6

8

10

12

14

1990 1993 1996 1999 2002 2005 2008 2011 2014 2017 2020 2023

CAC HPCool RAC


Residential Cooling Index (Annual kWh)

600

700

800

900

1,000

1990 1993 1996 1999 2002 2005 2008 2011 2014 2017 2020 2023


Residential XCool Variable

Monthly cooling requirements (kWh)

Average cooling use increases with increasing air conditioning saturation


Residential XHeat Variable

Monthly heating requirement (kWh)

Average heating use declines with declining heating saturation


Residential Non HVAC End-uses

mymymy OtherUsedexOtherEqpInXOther ,,,

31Pr

Pr ,

20.0

01

,

20.0

01

,

15.0

01

,,

mymymymymy

BDays

HHSize

HHSize

Income

Income

ice

iceOtherUse


Residential Non HVAC End-uses (cont.)

Typem

Type

Type

Typey

Typey

Typemy MoMult

UEC

Sat

UEC

Sat

UECOtherIndex

01

01

01,

1

1


Residential XOther Variable

Monthly base use requirement (kWh)


Impact of 2007 Energy Act - Lighting

• 2007 Energy Independence and Security Act introduces a number of new appliance efficiency standards

• New lighting standards have the most significant impact on residential load– Lighting accounts for

approximately 20% of residential “other” use

-

300

600

900

1,200

1,500

1,800

2,100

2006 2009 2012 2015 2018 2021 2024 2027 2030

Light07 Light08NEC

New England Lighting UEC (2007-2008)

New England XOther

New lighting standards

New lighting standards

Sharp drop in electric sales

results

Sharp drop in electric sales

results

Results in a sharp drop in base use

Results in a sharp drop in base use

Current lighting standards

Current lighting standards


New England Residential Forecast Comparison (GWh)

Due to the high lighting replacement rate, residential electric sales drop off quickly once the new standards go in place.

0

2,000

4,000

6,000

8,000

10,000

2001 2004 2007 2010 2013 2016 2019 2022 2025 2028

NEC07 NEC08

Residential energy use is 2.5% lower by 2013

Residential energy use is 2.5% lower by 2013


Estimated SAE Model – Residential Average Use

Variable Coefficient StdErr T-Stat P-ValueRes_StrucVars.WtXHeat 0.933 0.036 26.262 0.00%Res_StrucVars.WtXCool 0.709 0.04 17.529 0.00%Res_StrucVars.XOther 0.929 0.015 61.84 0.00%MBin.Jul99 -77.244 26.815 -2.881 0.49%MBin.Feb05 -60.864 25.474 -2.389 1.88%MBin.Feb07 -76.892 25.61 -3.002 0.34%

Regression StatisticsIterations 1Adjusted Observations 107Deg. of Freedom for Error 100R-Squared 0.914Adjusted R-Squared 0.908Durbin-Watson Statistic 1.743AIC 6.51BIC 6.685F-Statistic 151.173Prob (F-Statistic) 0Log-Likelihood -488.52Model Sum of Squares 667649Sum of Squared Errors 63092Mean Squared Error 630.92Std. Error of Regression 25.12Mean Abs. Dev. (MAD) 19.85Mean Abs. % Err. (MAPE) 3.29%Ljung-Box Statistic 57.45Prob (Ljung-Box) 0.0001Skewness 0.015Kurtosis 2.468Jarque-Bera 1.3Prob (Jarque-Bera) 0.3325


Predicted Vs. Actual Average Use


Residential Sales Forecast by End-Use (GWh)

0

50,000

100,000

150,000

200,000

250,000

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

Base UseBase Use

HeatingHeating

CoolingCooling


Residential End-Uses (EIA)

• Heating – electric resistance, heat pump• Cooling – CAC, room air conditioning, heat pump• Other Use

– Water heating– Cooking– Refrigeration– Second refrigerator– Freezer– Dishwasher– Clothes washer– Dryer– Microwave– Color TV– Lighting– Miscellaneous


Commercial End-Uses (EIA)

• Heating• Cooling• Other Use

– Ventilation– Water heating– Cooking– Refrigeration– Lighting– Miscellaneous


Vermont Monthly Sales Forecast Models

• Customer Classes– Residential

– Commercial

– Industrial

– Other

• Monthly residential and commercial class models are estimated using an SAE specification

• The industrial sales model estimated using a generalized econometric model

• We assume historical DSM activity is embedded in the sales data and thus in the estimated models


Data Sources

• Monthly Sales, Customer and Revenue Data– Energy Information Agency

• January 1999 to November 2007• Depending on system loss factor, sales data account for 95% to 97% of

delivered system energy

• Weather Data– Historical daily maximum and minimum temperatures

• Burlington Airport, 1970 to current– Evaluated other weather stations, however, there were too many holes

in the data series• Burlington-based HDD and CDD explain state-level sales well.

• Price Data– Price series was calculated from reported revenues, sales, and Vermont

CPI• Price calculated as a 12-month moving average of the prior twelve-month

average rate (real basis)• Assume constant real price in the forecast


Data Sources

• Economic Data– Fall 2007 Vermont economic forecast by Economy.com

• Population, number of households, real personal income

• Gross State Product, manufacturing output, non-manufacturing and manufacturing employment

– Final forecast will be based on Economy.com’s current state economic forecast

• End-Use Saturation and Efficiency Trends– Developed from the 2007 EIA Energy Outlook Forecast for New

England – Currently updating efficiency projections to reflect the recently

passed energy bill– End-use saturation trends will be calibrated against recent state

and Burlington Electric residential saturation surveys


Long-Term Vermont Economic Projections

Year Pop (thou) Hsehlds (thou) HHInc (thou $) GSP (mil $) Emp (thou)1998 598.95 233.52 65.98 16,204 285.041999 602.72 236.29 67.87 16,953 291.572000 607.52 239.51 70.49 17,782 298.702001 611.49 241.69 71.90 18,543 302.102002 614.69 242.95 71.76 18,910 299.292003 617.79 244.18 72.43 19,606 299.142004 620.25 245.15 74.40 20,481 302.962005 622.21 245.92 75.86 21,103 305.322006 623.93 246.89 77.59 21,365 308.212007 626.34 248.54 79.07 21,785 310.232008 629.95 250.81 80.69 22,315 312.002009 633.56 253.13 82.10 22,888 314.622010 636.42 255.26 83.27 23,432 317.062011 639.01 257.36 84.52 24,010 319.582012 642.21 259.78 85.69 24,609 322.562013 645.65 262.34 86.61 25,203 325.282014 648.65 264.69 87.46 25,787 327.922015 651.36 266.82 88.33 26,348 330.542016 653.72 268.66 89.22 26,881 333.212017 655.88 270.29 90.16 27,417 336.14% Change98-07 0.5% 0.7% 2.0% 3.3% 0.9%08-12 0.5% 0.9% 1.5% 2.5% 0.8%08-17 0.4% 0.8% 1.2% 2.3% 0.8%


Long-Term US Economic Projections

Year Pop (mil) Hsehlds (mil) HHInc (thou $) GDP (bil $) Emp (mil)1998 275.85 103.11 75.00 9,067 125.921999 279.03 104.44 76.56 9,471 128.992000 282.21 105.77 79.69 9,817 131.792001 285.22 106.88 79.95 9,891 131.832002 288.12 107.96 79.46 10,049 130.352003 290.79 108.95 79.65 10,301 129.992004 293.63 110.00 81.57 10,676 131.422005 296.5 111.07 83.11 11,004 133.702006 299.39 112.15 85.40 11,320 136.182007 301.79 113.63 87.16 11,541 137.952008 304.48 115.00 88.38 11,879 139.072009 307.16 116.42 90.10 12,234 140.622010 309.83 117.90 91.67 12,597 142.352011 312.5 119.40 93.23 12,967 144.232012 315.2 120.96 94.58 13,332 146.212013 317.89 122.53 95.81 13,690 148.132014 320.59 124.07 97.02 14,045 150.082015 323.3 125.56 98.26 14,402 152.042016 326.01 126.97 99.54 14,763 154.042017 328.71 128.34 100.81 15,131 156.04% Change98-07 1.0% 1.1% 1.7% 2.7% 1.0%08-12 0.9% 1.3% 1.7% 2.9% 1.3%08-17 0.9% 1.2% 1.5% 2.7% 1.3%


0

500,000

1,000,000

1,500,000

2,000,000

2,500,000

3,000,000

2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

Preliminary Class Energy Forecasts (MWh)

Commercial1.1%

Commercial1.1%

Residential1.1%

Residential1.1%

Industrial0.2%

Industrial0.2%

OtherNo Growth

OtherNo Growth


Preliminary Class Energy (MWh)

Year Residential Commercial Industrial Lighting Total2004 2,081,975 1,967,719 1,621,043 44,589 5,715,3252005 2,220,509 2,052,483 1,624,195 44,589 5,941,7762006 2,163,848 2,025,182 1,625,522 44,589 5,859,1412007 2,270,534 2,050,976 1,614,599 44,589 5,980,6982008 2,289,996 2,060,190 1,630,336 44,589 6,025,1112009 2,320,487 2,081,770 1,633,240 44,589 6,080,0862010 2,355,521 2,102,549 1,635,998 44,589 6,138,6572011 2,384,952 2,125,854 1,638,928 44,589 6,194,3232012 2,422,707 2,149,752 1,641,964 44,589 6,259,0132013 2,436,951 2,172,592 1,644,976 44,589 6,299,1082014 2,460,597 2,195,852 1,647,934 44,589 6,348,9712015 2,484,521 2,218,892 1,650,776 44,589 6,398,7782016 2,520,406 2,243,082 1,653,478 44,589 6,461,5552017 2,537,851 2,267,883 1,656,198 44,589 6,506,521

% Change2004 - 2007 2.9% 1.4% -0.1% 0.0% 1.5%2008 - 2012 1.4% 1.1% 0.2% 0.0% 1.0%2008 - 2017 1.1% 1.1% 0.2% 0.0% 0.9%

Class Energy


Class Hourly Profile Data Sources

• Load Data – Burlington Electric Load Research Data

• Residential

• Small General Service

• Large General Service

– Other Load Research Data• Industrial

• Street Lighting

• Daily maximum and minimum temperatures– Burlington Airport

• Daily calendar– Day of the week, month, holiday, hours of sunlight


Class Hourly Profile Models

• Class Hourly Model Structures– Twenty-four hourly regression models

• HDD and CDD

• Month, Day of the Week, Holidays

• Hours of Light

• Estimation Period– January 1, 2004 to December 31, 2006


Calculation of Daily Normal Weather

• Ten years daily average temperature for Burlington– 1998-2007

• Rank and Average approach– Daily average temperatures ranked from highest to lowest and

within each year then averaged across all 10 years

• Map daily normal ranked weather data to a typical daily weather pattern– Typical year weather pattern calculated from historical daily

weather data– Map the ranked temperature data to the typical year weather

pattern

• Used MetrixLT for calculating daily normal temperature series


Chaotic Daily Normal Weather Series

• Daily normal weather series mapped to the average ten-year weather pattern


Residential Load Model (kW per customer)


Large General Service Load Model


Residential End-Use Profiles

• Cooling, heating, and other use profiles estimated from end-use weather response models– Data is based on building simulation runs

• Models simulated for 2004 to 2007 using Burlington actual weather

• End-use profiles scaled to residential profile model


Residential Cooling Profile


Residential Heating Profile


Load Build-up Comparison

SystemSystem

SystemSystem

Build-upBuild-up

Build-upBuild-up

Uncalibrated ComparisonUncalibrated Comparison

Calibrated ComparisonCalibrated Comparison


Preliminary Forecast (No Additional DSM)

Based on EIA saturation projections

Year Energy (MWh) Summer Peak (MW) Winter Peak (MW)2008 6,025,111 1,089 1,0702009 6,080,086 1,104 1,0752010 6,138,657 1,118 1,0882011 6,194,323 1,130 1,0872012 6,259,013 1,141 1,1012013 6,299,108 1,154 1,1122014 6,348,971 1,165 1,1192015 6,398,778 1,177 1,1212016 6,461,555 1,188 1,1222017 6,506,521 1,203 1,119

% Change2008 - 2012 1.0% 1.2% 0.7%2008 - 2017 0.9% 1.1% 0.5%

Preliminary Forecast Summary


Class Coincident Demand

Year Residential Commercial Industrial Total2008 549 337 203 1,0892009 559 341 204 1,1042010 569 345 204 1,1182011 578 348 204 1,1302012 586 351 204 1,1412013 594 355 205 1,1542014 602 359 205 1,1652015 610 362 205 1,1772016 619 365 205 1,1882017 628 369 205 1,203

% Change2008 - 2012 1.7% 1.0% 0.1% 1.2%2008 - 2017 1.5% 1.0% 0.1% 1.1%

Class Coincident Summer Peak Demand (MW)


Preliminary Energy and Demand Forecast (No Additional DSM)

BEC end-use saturation projections (calibrated into state

RASS)

Year Energy Summer Peak Winter Peak2008 5,937,926 1,089 1,0702009 5,984,034 1,108 1,0762010 6,035,289 1,124 1,0882011 6,086,413 1,140 1,0892012 6,145,698 1,154 1,1022013 6,182,957 1,170 1,1132014 6,230,511 1,185 1,1212015 6,278,384 1,200 1,1242016 6,339,292 1,215 1,1282017 6,384,266 1,233 1,131

% Change2008 - 2012 0.9% 1.5% 0.7%2008 - 2017 0.8% 1.4% 0.6%

System Preliminary Forecast Summary

Documents

Electric / Gas / Water Eric Fox Oleg Moskatov Itron, Inc. April 17, 2008 VELCO Long-Term Demand Forecast Methodology Overview