ML20031A423
| ML20031A423 | |
| Person / Time | |
|---|---|
| Site: | Susquehanna  |
| Issue date: | 09/15/1981 |
| From: | Mcnair G ALLEGHENY ELECTRIC COOPERATIVE, INC., PENNSYLVANIA POWER & LIGHT CO. |
| To: | |
| Shared Package | |
| ML20031A420 | List: |
| References | |
| NUDOCS 8109230506 | |
| Download: ML20031A423 (31) | |
Text
~
e e
UNITED STATES OF AMERICA NUCLEAR REGULATORY COMMISSION REl NIED CORllESI'ONDENCF BEFORE THE ATOMIC SAFETY AND LICENSING BOARD In the Matter of
)
)
PENNSYLVANIA POWER & LIGHT COMPANY
)
)
and
)
Docket Nos. 50-387
)
50-388 ALLEGHENY ELECTRIC COOPERATIVE, INC. )
)
/.f ~ ; R (Susquehanna Steam Electric Station, )
4 Units 1 and 2)
)
9 ~
Eusw g
- j v3tmc SEP171980 P' C
\\cp r"
- cf Be Se:rety j APFLICANTS' TESTIMONY OF GRAYSON E. McNAIR ON CONTENTION 4a AND '4b :(LOAD FORECASTING) w Septembcr 15, 1981 4
8109230506 010915' PDR ADOCK 05000387:
r 0
CONTENTS I
Page Introduction 1
The Forecast -- Its Purpose, Development, and Design 2
Evaluation of the Future by the Econometric Method 4
Evaluation of the Future by the Traditional Method 12 Derivation of the Probability Band Forecast 20 1980 Short-Term Forecast: Methodology and Rer.ults 24 Peak Load Forecast 27 4
INTRODUCTION One of the elements used in development of PP&L's long-range capacity requirement plans is a forecast of the future peak demand of our customers. The current demand forecast was prepared in October 1980. This testimony discusses the development of the sales and peak load forecast levt 3 The first section discusses the preparation of forecasts, the setting of corporate goals, and the incorporation of uncertainty. This is followed by the methodology, assumptions, and results of the econometric analysis.
The traditional or judgment evaluation is also explained.
In addition to the long-range analysis that goes into the prepara-tion of the projected annual sales, there is also a,short-range analysis to provide monthly sales detail needed for budget purposes. A section detailing its development and the blending of the long and short-range outlooks into a peak load forecast is provided.
4 l
l l
l i
1 l
l
-_.-_--,-m--
i THE FORECAST -- ITS PURPOSE, DEVELOPMENT, AND DESIGN 1
\\
The planning needs of the company dictate that forecasts of sales and peak loads be prepared for both the short-term and the long-term. The short-term forecast, usually for two years into the future, is available by t
months for sales to individual SIC (Standard Industrial Classification) customer classes and for peak loads.
This amocnt of detail is necessary for a short-term forecast because -it is used in preparing the company's operating budget, which in turn is a decision-making tool when developing strategies concerniag cash flow, rate filings, and security of ferings. The long-term forecast extends the forecast period to as much.as 20 years. Sales and peaks 1
are forecasted for each year and are used in making decisions about capacity
~
expansion.
1 There are numerous factors which influence the level of sales and peak load, One phase of forecasting is to determine those key factors and l
measure their effects. These factors include the level of the economy, fuel prices, population level, technological and social changes -- along with ranges of values they may exhibit in the future. Then, one or more fore-l casting methods are employed to determine the collective effect of these events and the variations that our sales and peaks may exhibit due to vari-l ations in the influencing factors. The result is a band-forecast that encompasseo the most probable levels in which the future level of sales and loads will fall. Since the uncertainty increases as the forecast period is l
extended, the band is normally an ever-widening one. Although a single-line forecast that falls within the band forecast may be selected for planning s
l purposes, the band serves the purpose of showing the probsble range of outcomes as a basis for decision-making.
The approach PP&L has taken to position the lower limit of the band is to link together assumptions about l
tbc future that tera to reduce sales.
An effort must be made to be sure they are internally consistent and are at reasonable levels. A similar approach with assumptions that tend to increase sales is used to place the uppar band.
An example for the low side is assuming simultaneous occurrence cf low' eco-l nomic activity, conservation, and high interest rates -- a majot influence on new construction of dwellings and places of business. This approach reduces the risk of future loads falling outside the stated band.
The resulting product is a band forecast about one percent on either side of the single-line forecast.
There are a variety of forecasting methodologies from which to choose today.
In what might be called the tradit.ianal methodology, a com-bination of marketing, engineering, and end-use data is analyzed and tempered with judgment to produce a forecast. The.results are highly dependent on the forecaster's experience, knowledge, and ability. A detailed forecast using this traditional method is reviewed in a later section of this paper.
A second methodology which has gained favor during the past decade is econometrics -- representing economic behavior in terms of mathematical equa tions. This approach to forecasting aims to quantify past relationships between causes and effects which in turn provides a starting point for appli-cation of judgment. The goal is to quantify the ef fects that the economy, s
energy prices, population, and other factors have hcd on sales of electri-city. A detailed description of this methodology starts on page 4.
2
A common concern to all forecasting methods is capturing fully the effect of new significant factors that are likely to alter historical growth pa tterns. Conservation of resources n,nd subsequent new energy technologies l
certainly fall in this category. This concern has been subject to constant L
review since the oil embargo -- in the beginning making broad estimates about what might happen and more recently refining this broad estimate to specific 4
items. Thy current forecast includes an extensive list of conservation /new technology events that are likely to occur in t'e next 20 years. Some will n
reduce substantially the sales of electricity, but others in the long run will foster the substitution of electricity for fossil fuels.
Specific items included because of their significant impact on loads in the next 15 to 20 years form the following list.
In the residential market, the heat pump will be installed in an expanding number of new homes; older homes will be retrofitted with significant amounts of insulating materials; homes will be smaller and household appliances will be more efficient.
Offsetting some of these reductions are conversion of fossil-fuel heated homes to electric heat in response to rising fossil-fuel prices; the use of portable electric space heaters to maintain the temperature in only one room of fossil-fuel heated homes; and the proposed introduction of the massed produced electric car in the middle of this decade.
In the commercial and industrial market, space conditioning and lighting equipment have under-gone extensive redesign to minimize the use of electricity. The same is true for production processes, i.e. cooking, motors, assembly lines, furnaces, 4
pumps, and compressors.
Due to conservation and new energy technologies developing between now and the year 2000, the expected net reduction is 5 to 6 billion kilowatt-hours or 1,000 mW of load. PP&L has also assumed a reduction of another 400 mW of loads that will be derived by shifting on peak loads to off peak.
PP&L's forecasting approach provides the following benefits:
e Risks inherent to the future are clearly defined.
e Monitoring is easier and more realistic. For example, when actual values begin falling outside of the band, this signals that a review of the forecast is needed.
I e Forecasts are valid over a longer period of time thus improving decision-making. 'The lengthening of time between forecasts allows time for extending the scope of planning efforts.
i 3
EVALUATION OF THE FUTURE BY THE ECONOMETRIC METHOD METHODOLOGY Introduction The PP&L econometric model forecasts long-range electricity sales for the residential, commercial, and industrial sectors in the PP&L service area. By identifying key variables affecting salas, the model utilizes historic values to measure interrelationships that existed between these variables and kWh sales. The crucial linkages captured by the model equa-tions provide the user with a frLmework which enables him to produce "first cut" forecasts of sales consistent with various econ,mic, energy, weather, and policy outlooks.
In the following sections, the methodology used in the forecasting of each market sector is described, e.g.,
residential, commercial. Also included are the assumptions developed by Data Resources, Inc., (DRI) and selected by PP&L for the twenty-five year macroeconomic outlook, since this DP.I/PP&L forecast forms the basis for the regional economic outlook of the base case kWh sales forecast.
The forecast results using the base case assumptions are provided along with the forecast results basad on alternative assumptions to indicate the sensitivity of kWh sales to changes in the U.S.
economy, particularly employment, production, income, etc., and to relative fuel price projections of alternative forms of energy, particularly oil and natural gas.
It should be emphasized that it is not the intent of this analysis to arrive at a point estimate of future sales. These forecasts were i
developed to reveal a broad outlook required to expose the risks the future may hold for planners and decision-makers.
l l
Econometric Methodology in the Residential Sector j
In the PP&L econometric model, residential sales are forecasted as I
the product of the number of PP&L residential customers and the estimated average electric use per customer.
Because usage characteristics differ between electrically heated (EHH) and general (GRS) residential service cus-tomers, sales to these two classes are forecasted separately.
The estimate of the number of future customers in the electric heat and general residential classes is made by starting with a known fact, the l
existing number of customers in each class. Next, a forecast of the number of new dwelling units to be added in each future year must be determined.
These new dwelling units are forecasted using the current housing stock, demo-graphic housing demand (i.e., the population in our service area over age 20), the real per capita disposable income in the service area, and the real dollar amount of U.S. mortgage commitments. The percentaga of new dwelling units which are electrically heated is a function of the marginal price of electricity relative to the price of fuel oil and a binary variable to repre-sent the market penetration of electric heat. The depreciation and loss of i
i 4
l
i existing homes as a result of demolition, fire, etc., is taken into account with a depreciation factor.
Finally, the conversion of general residential i
customers to electric heating is a function of real per capita disposable i
income in the PP&L service area and a binary variable for the termination of PP&L's promotional electric heat rate.
Average use per customer forecasts rise out of estimates of appli-ance saturation and appliance usage. Each appliance saturation represents j
the number of appliancas per 100 customers and is a function of the unde-preciated appliance stock and a corresponding real per capita disposable income term.
Fixed kWh weights are assigned to each major appliance (these may change over time due to conservation, etc.) reflecting reported annual kWh usage per appliance. Special account is taken of the lower kWh usage of second and third appliances, such as televisions and refrigerators. Each appliance saturation is multiplied by its corresponding kWh usage and then aggregated to form CRS and EHH indexes of appliance usage.
The GRS and EHH average usage equations are functions of their re-spective appliance indexes, adjusted for past temperature and humidity vari-ations since forecasts are made assuming normal weather, and the real price j
of electricity. GR5 average usage is also dependent on the number of persons j
per household and the percentage of income available to households after
]
food, clothing, and shelter expenditures.
New Residential Technologies - Residential sector methodology has recently been elaborated. The model now accommodates assumptions concerning the electric vehicle, future appliance efficiencies, and the extent of kWh savings resulting from smaller energy-efficient homes, retrofitting of in-sulation, heat pumps, and solar water heaters.
Econometric Methodology in the Commercial Sector' In our model, the commere: 21 sector is divided into four parts:
Wholesale and Retail Trade, Financial and Personal Services, Other Commer-cial, and Small Commercial. The Other Commercial category consists of state and federal government, agriculture, construction, and transportation and public utilities. Small Commercial is a conglomeration of businesses pres-ently not classified by SIC Code, such as beauty shops, service garages, doctors' offices, etc.
The foundation of most commercial sales forecasting equations is employment, since employment growth has played a critical role in explaining past growth in the commercial sector.
In general, service area commercial employment is linked to U.S. employment and the relationship of real per capita income in PP&L's service area to that in the U.S.
The PP&L sales forecasting equations are, in turn, dependent on PP&L customers per employee in our service area, megawatt hours per employee, and the average price of electricity in each sector relative to oil prices.
Small Commercial sales are forecasted differently from the other three groups, mainly because the paucity of data makes attribution to specific factors difficult. Small Commercial sales forecasts are based mainly on real per capita d sposable income, weather variation, and prices for other energy supplies.
5
1 In each equation, special variables are incorporated to account for jolts in past commercial sales.
Among the most important adjustments are binary variables for the energy crisis, Hurricane Agnes, and data classifi-cation shifts.
i Commercial Conservation - Af ter the basic projection of commercial sales is made using the historic relationships of economic and price vari-ables to sales, adjustments are made to incorporate conservation. Sales, before conservation, are separated into sales by end-use, such as lighting, space conditioning, and water heating. Assumptions are made for the per-centage reduction in sales that can be attained in each end-use through
'onservation measures. The percentages are not constant but increase over timV.
These percentage reductions are applied to the original sales levels to obtain sales levels reflecting conservation.
Econometric Methodology in the Industrial Sector Specific sales equations are created for Steel, Chemicals, Coal Mining, and Small Industrial as data exists which permits the individual j
estimation. For the emainder of the industrial sector, sales are forecasted
~
in aggregate, then broken out according to the proportion each industry con-1 tributes to an estimated production of goods and services in Central Eastern Pennsylvania.
As in commercial sales forecasting, service area employment in each industry is a critical element explaining sales growth. Most of our indus-trial employment equations are linked to PP&L service area employments of the preceding quarter, corresponding U.S. industrial production indexes, a linear time trend which picks up changing technology, and, in some cases, binary variables for strikes and the exit or entry of large plants in our service area.
Employment is then linked to an index represeating the production the industry is generating in the PP&L territory. Each PP&L production index is the product of our service area employment and expected production per employee in the U.S.
Sales are directly dependent on these PP&L production l
indexes.
Another major determinant of sales is price. Most equations incorporate the real electricity price relative to oil and/or natural gas.
Some also include real average electricity prices.
In the coal mining and steel industries, real marginal prices play an important role.
Industrial Conservation - As with the commercial class, after the basic projection of industrial sales is made using the historic relationships of economic and price variables to sales, adjustments are made to incorporate conservation. Sales, before conservation, are separated into sales by end-use, such as pumps, furnaces, packing, process heating, etc.
Assumptions are made for the percentage reduction in sales that can be attained in each end-use through conservation measures. The percentages are not constant but increase over time. These percentage reductions are applied to the original sales levels to obtain sales levels reflecting conservation.
6
l ASSUMPTIONS Macroeconomic Assumptions of CYCLELONG2004 Data Resources, Inc., of Lexington, Massachusetts, regularly pro-duces multiple sets of long-term foreensts of the U.S. economy. Each fore-cast is based on varying macroeconomic assumptions and, therefore, represents varying outlooks on the U.S. economy.
The DRI forecast selacted by PP&L to produce the PP&L base case evaluation is titled CYCLELONG2004.
It envisions -a period of moderate real national output growth. The projected average annual rate of real GNP growth is 2.3%, a considerable slowdown frcu the higher than average 3.9% growth rates of the 60's, and slower than the growth we experienced from 1965 to 1980 (3.2%).
1
(
The major factor causing this slowdown is the predicted decline in labor force growth. Over the past twenty years, the percentage of women entering the labor force was phenomenal. When combined with the increasing large numbers of " baby boom" children who began working-during this earlier period, the labor force grew dramatically.
It appears that the labor force participation rate for women will increase more slowly in the future.
Furthermore, growth in the number of entrants into the work world will taper off as the age composition of the population changes.
Because this slower labor force growth has important implicationa for the amount of output this country is capable of producing, it is one of the most important considerations in the national forecast.
I Like labor force growth, investment will proceed at a lower rate than we witnessed in the 60's, but unlike the energy and interest rate jolts 5
that plagued investment growth during the 70's, we foresee a pickup in the pace of capital formation in the next few decades as some of our recent j
problems are sorted out.
Although the personal savings rate will remain at relatively low rates, government policies will be designed to encourage investment.
l l
Over the twenty-five year forecast period, productivity growth will also recover from its post oil-embargo levels.
Several factors contribute to this conclusion including a more mature labor force, replacement of older less-efficient machinery, and a proportional decline in ncn-goods producing investments such as pollution abatement equipment. This moderate future i
productivity growth is one of the factors contributing to the gradual unwind-ing of the high inflation rates we've been witnessing in recent years.
Energy supply considerations are an important aspect of the l
forecast.
In the short-term, disruptions of energy supplies can still greatly affect our customers since they are limited by the flexibility of
[
equipment currently in place.
In the long-run, customers will become less l
susceptible to energy price shocks as substitution of other factor inputs --
j such as energy-efficient machinery -- takes the place of energy.
l In the next few decades, consumers will be spendir3 their incomes differently than they have in the past. Although the share of income spent on consumption remains almost constant at 63%, the portion spent on services 7
l
_=. - -
= -
i
{
e
=
is expected to climb. Durable purchases such as furniture and automobiles will become relatively more important, especially in the mid 80's, as a con-siderable portion of the population reaches home-buying age.
On the other hand, nondurable purchases, e.g., lood and gas, should become less important
-in the : future because of the decline in population growth and the continued r
trend-toward small, fuel-efficient cars.
Government expenditures on the state and national level are pre-dicted to decrease as a percentage of total output in the next twenty years.
The continued decrease in the growth of tax burdens is largely the result of l
demographics, e.g.,
a continuation of the declining trend in school age population, and growing consumer dissatisfaction with government spending.
j This forecast has taken into account the recent manifestations of the economy's vulnerability to shocks. The increased likelihood that our j
economic future will b2 characterized by stubborn inflation, low productivity l
growth, and tight energy supplies has resulted in a less optimistic outlook l
than some earlier outlooks.
Since our forecast philosophy is based on the idea of producing not
{
just a one-line forecast but a band that has a reasonable probability of
]
covering the eventual outcome (actual), other DRI scenarios with different i
assumptions were evaluated and used. HIGHTREND2004 and LOWTREND2004 were j'
used to quantify the outside risks due to the economy.
TRENDLONG2004 was j
used to project sales based on an economy absent of shocks and lost economic opportunities.
This study provided another rough estimate at a high case.
l The impact that changes in macroeconomic assumptions had on PP&L sales is illustrated in Table 1 on page 10.
Each economic outlook was used with an 4
identical set of pricing assumptions and produced forecasts on line 3 that ranged from 42,011 GWh to 48,538 GWh.
Pricing Assumptions -
1 The price assumptions for fuels used in the PP&L base case econo-l metric forecast were derived from discussions with various departments at j
PP&L. The resulting consensus electricity price forecast was that electric prices would range between -3% and 1% real average annual growth through 2000 i
(
with.5% av'erage annual real increase as the expected ' value. This real price growth primarily reflects the expected increases in plant and fuel costs and varying plant capacity utilization rates.
For alternate fuels, oil i
(
prices were assumed to increase one to three percentage points above the real average annual price increases for coal -- the main fuel at PP&L's power plants.
Coal had a consensus range in real average annual growth of 1% to 3%.
The consensus rarge for average real growth in oil prices was between 2%
and 6%.
Natural gas prices were assumed to increase more than oil by an average of two percentage points a year, with a consensus range of real average annual growth of 4% to 8%.
The expected values for increases were 2%
for coal, 4% for oil, and 6% for natural gas.
Given the erratic decisions of OPEC, the uncertainty of deccetrol timetables, economic health, etc., the risks associated with adhering to just one price forecast are apparent. Therefore, several model_ runs were made by altering energy price assumptions within their respective ranges. The table of results on page 10 shows that the choice of pricing assumptions greatly influences the results.
8
_. _. -.. ~. _ _, - _. _ _ _ _ _ _. _.... _ _ _. _ _, _ _ _. _ _ _ _. _ _. -
RESULTS The economic simulation is based on the CYCLELONG2004 scenario adjusted for expected prices and conservatinn assumptions. The result of this simulation, which is found in Graph 1 on page 11, is our point-estimate econometric forecast.
Table 1 lists the forecast results developed by altering price and macroeconomic assumptions.
For the year 2000, one can see that altering the economic assumptions alone can result in a sales range between 42,000 and 48,500 GWh when expected prices are assumed in each case. The basis for the outer points of this band lies in DRI's alternative assumptions regarding potential U.S. economic growth.
The basis for the upper band limit (HIGHTREND2004) envisions more employment of labor and capital, resulting in higher production levels and a lower inflation rate (higher real incomes).
Since PP&L sales are linked to the macroeconomic variables, it is not sur-prising that HIGHTREND2004 has the highest predicted sales, all else con-stant. The lower band limit based on LOWTREND2004 is nearly a mirror image.
KWh sales are also sensitive to fuel and energy price changes. On a percentage change basis, prices generate an even wider range of forecast re-sults than economic activity. The variation in forecasts based on TRENDLONG2004 but with alternative price assumptions attests to this. With oil cnd gao real price increases kept constant and varying electric real price increases from 1% to -3%, sales range between 39,700 GWh and 56,100 GWh.
l 9
e TABLE 1 ECON 0HETRIC MET 110D TOTAL GWil SALES IN 2000 l
DRI HACROECONOMIC FORECAST 4
llICilTREND2004 TRENDLONG2004 CYCLELONC2004 LOWTREND2004 Price Acsumptions (Real Average Annual Percent Increases) 4 i
Elec-Natural tricity Oil Gas
-3 4
6 50,122
,,o
.5 4.9 5.7 46,688 44,693 t
.5 4
6 48,538 45,325 43,433 42,011 1
4 6
39,667 i
1 1
1 36,012
-3 6
8 62,714
-3 2
4 55,671 '
1 2
4 34,563 1
1 6
8 38,814
e e
J 8
6,+
\\
N N
s i
r
/\\
8
\\
5 x
0 C
U LL.I C
\\
O
\\
s L
C
[U 8C a-
\\
m tu hC b
w LLI f
I
/
O Z
O O
\\
(LI K
\\
8
\\
'N N,
~\\
8.
8 8
8 8
8 8
8 m
m m
~
~
SWn0H-116M0 LIM JO SNCIllIW 11 i
l i
I EVALUATION OF THE FUTURE BY THE TRADITIONAL METHOD INTRODUCTION As mentioned previously, forecasting methods at PP&L fall into two general categories, the traditional method and the econometric approach. The traditional or judgment method allows the forecast team a freer hand to apply many implicit sales relationships which cannot be accommodated when employing the econometric method because of the difficulty in stating the relationships as equations.
Several studies have helped uncover and provide much insight into the connection between sales and recent energy developments. These studies have reviewed conservation, throwover (substitution of fuel sources),
and residential conversions.1/ As a part of the forecast, special emphasis was directed toward the natural gas market, the electric vehicle, and the historical relationship between Gross National Product and energy consump-tion.
METHODOLOGY The long-term judgment forecast has been cor.structed as a band fore-cast to quantify a probable range of kilowatt-hour sales. The band forecast is composed of an upper bound (JFMAX) that is based upon a set of assumptions that may not be 100% consistent with one another, but all have the effect of pushing electric sales to an upper limit. The lower bound (JFMIN) is simi-larly constructed with the exception that all assumptions hold electric sales down.
These bands are further adjusted to include conservation (JFMAXC and JFMINC), throwover from natural gas and oil, and cogeneration. Each of these variables was reviewed with the most recent data and forecasted separately for the upper and lower bands of the forecast.
Traditional Methodology in the Residential Sector The general residential and electrically heated homes sales fore-casts were prepared by determining the levels of sales that each end-use of 1/ Conservation and throwover analyses were based upon a report submitted to PP&L solely for the PP&L system:
Hamel, Bernard B. and Brown, Harry L.,
}
Pennsylvania Power and Light Co. Alternative Fuel Evaluations and Energy Projections for Major Industrial Groups, General Energy Associaces, Inc.,
Cherry Hill, NJ 08003, September 7, 1979.
Further conservation data was based upon a report by Conservation Services, PP&L: Analysis of Industrial & Commercial Customer Peak Demand & Load l
Management Opportunities, Conservation Services, Industrial 4 Commercial Section. December 1979.
Residential Conversion analysir is based upon a repart:
Frazier, Donald N.,
i The Residential Conversion Market 1980-2000, Market Research -- Pennsylvania Power and Light Co., April 1980, 12
electricity in homes would contribute to the total. These end-use categories included 15 major appliances, heating, lighting, miscellaneous small appli-ances, and future appliances that may be marketed. Annual sales to a partic-ular end-use in any year is a product of the number of customers in that year, the percent saturation of the end-use, and its average annual kWh usage. Since altering any of these three variables will change the outcome, a band forecast of sales can be prepared by combining band forecasts of these three variables.
As with the econometric method, the starting point for the customer forecast of the electric heat and general residential classes is the existing customer count.
To this is added the increase due to new dwelling units (NDU).
The new dwelling unit forecast is a function of the population growth in the twenty and over age group and the breakdown of households into one, two, and three adult homes, In addition -- based on historical observation
-- new dwelling units are included to account for replacement homes and q&
seasonal homes. A forecast of the saturation of electric heat is made to break new dwelling units into the two residential classes. The final step is to account for the depreciation of the housing stock due to fire, abandon-ment, ate., and the conversion of customers from general residential service to electric heat.
t In order to reflect a future without an additional increment of conservation, the appliance, heating, and lighting average ueos factored into the sales calculations are those that are experienced currently.
The final variable in the calculation of sales to each end-use is the saturation of each appliance. Current appliance saturations for general residential and electrically heated customers are available through recent appliance saturation surveys, along with saturations in new dwelling units.
These surveys are conducted every few years and were used to determine the starting point for developing the band of saturation for each appliance. To determine a maximum level of saturation, factors that would tend to increase saturations were developed.
Some examples are energy prices, change in percentage of apartments, size of home, convenience and savings, appliance becoming standard in NDU, recent trend in saturation, renovations, higher food prices (for freezers), and the price of the appliance.
Estimates were made of how these factors would af fect the saturation of appliances in new dwelling units and the future purchase rate of those who currently do not have the applimnces. These, in conjunction with the current number of appliances, yielded the maximum saturations. The low band was similarly constructed, only taking into account factors that would minimize the saturations.
The sales forecast bands are a product of the 1980 average uses, the high (low) level of customers and the high (low) level of saturations.
Residential Conservation - To produce the band forecast with con-servation, the calculation above was repeated using average appliance uses that reflected some conservation for the high band and a high level of conservation for the low band.
It is expected that some appliances will exhibit a decrease in average use. As these new, more-efficient appliances are added to the stock in PP&L's service territory, the average uses will drop.
Sy 2000 most, if 13
_ _, _ _. _ _ _.. ~
i
)
not all, appliances are expected to te the more-efficient ones.
The water j
i heater is one appliance where average use can be dropped on both new and i
3 existing appliances, simply by insulating the tank and reducing the i
thermostat setting. Varying the assumption on the percentage of people who will take these steps produced average uses at two levels of conservation.
The average use for space heating has been dropping and is expected to continue to drop.
The factors whose effect we have quantified are higher insulation levels on NDU, insulation retrofit, smaller home size, and heat 4
pumps.
i j
Electric Vehicle - Speculation and skepticism have surrounded the mass production of the electric vehicle (EV) for decades. Engineers have explored the wide variety of electric battery systems in their attempt to find an electric storage system capable of powering a vehicle several hundred miles at current highway speeds.
It appears that Gulf and Western (G&W) and General Motors Corporation (GM) now lead in the race toward the creation of an electric vehicle that is worthy of mase production. General Motors is planning to mass produce its first electric vehicle in 1984.
]
Qur analysis of the electric vehicle and its impact on the PP&L system is dependent upon the number of vehicles in operation, the average l
mileage per vehicle, and the kilowatt-hour per mile. PP5L's service area population is approximately 1% of the U.S. population and we assume that PP&L customers will. purchase 1% of the electric vehicles sold in the U.S.
Our forecast is produced in a band to accommadate the most ambitious electric vehicle activity we can imagine and the least amount of electric vehicle activity possible. The average number of miles traveled per vehicle is expected to be less than the average for a gasoline-driven vehicle due to constraints on range and speed. As for the number of kWh consumed per mile, j
current technology has reached a rough average of 2.3 kWh per mile. Assuming technology may make improvements on this rate of consumption, a band of this variable's values was made.
l Traditional Methodology In The Commercial Sector In preparing the commercial forecast by the traditional method, the
{
four class breakdown used in the econometric method is not employed. This is l
due to the fact that the historical and forecasted data for the factors used in producing sales to this class by this method are avaiJable only for com-i l
mercial as a whole and not for individual classes.
I l
The methodology is somewhat of a bailding-block process. Each l
year's forecast is equal to the previous year's commercial sales plus the increase (or decrease) due to any changes in the existing customers' average 1
use plus the kilowatt-hours from any new customers (or lost from old ones).
1 The kilowatt-hours from new customers is broken down into contributions from electrically heated'and non-electrically heated customers. Each is a func-tion of the average number of kWh used per square foot of building (which is different for electrically heated and non-electrically heated buildings) and j
the amount of new square footage added to each heating type. A variable was derived using data accumulated.over the 1975 to 1979 period to link square footage to the number of commercial customers, which in turn is related to the number of residential customers.
14
Commerical Conservation - The application of conservation by com-mercial classes is centered around each of the four major end-uses:
space conditionin Utilizing a number of studies /, g, water heating, lighting, and other.
2 commercial sales is disaggregated into these four end-uses and a percent reduction due to conservation is applied to each. The percent reductions were based on technological advances, substitution of more effi-cient systems, (e.g., HVAC systems), and design changes.
Commercial Throwover - With rising fuel prices and questionable supply sources, throwover (conversion) to electricity becomes an important consideration for long-term planning. Using a survey of fuel type within the commercial sectot / and assuming PP&L cuntomers correspond to those sur-3 t
veyed, it is possible to derive the amount of oil and gas consumed by PP&L commercial customers, which is a measure of the maximum amount of throwover possible. From this point, assumptions are made on the percent of oil and natural gas switchable to other fuels, the percent switchable to electricity, and, of this potential figure, the projected amount that will switch.
i Traditional Methodol'ogy In The Industrial Se.for i
The band forecast for the industrial sector is composed of a forecast for each SIC class. The research that goes into preparing these forecasts deals with the historical relationship between real GNP (Gross National Product) and the associated national production level for each SIC class. These relationships are used with a band forecast of GNP to determine future national production levels. Local production levels take into account PP&L's share of the national market and the economic health of the regional industry.
The electric sales forecast is the amount required for the forecasted level of production.
Industrial Conservation - The application of conservation is centered around five major end-uses:
lights, space conditioning, pumps and compressors, mechanical work form, and assembly and packaging. Each SIC class's sales are disaggregated into these ead-uses and a percent reduction due to conservation is applied to each. The major contributors to the conservation accomplishments are technological advances (particularly improved electric motor application) and more-efficient lighting equipment and design.
Industrial Throwover - As with the commercial sector, there is a potential for customers to switch from oil and gas to electricity. The 2/ Energy Users News, "How the Commercial Sector Uses its Energy," Volume 3, Number 2, January 9, 1978.
Commerical Energy Use: A Disaggregation by Fuel, Building Type, and End-Use by J. R. Jackson and W. S. Johnson, Oak Ridge National Laboratory Oak Ridge, Tennessee, February 1978.
f 3/ Ibid.
l 15 l
throwover forecast of gas to elegtricity is based on an industrial fuel study which outlines gas availabilityi' and a study on the amount of gas suitable 1' for switching to other fuels.
Analyzing oil throwover began with the Hamel & Brown figure of current oil consumption. Assumptions from this point had to be made on:
the percent of oil switchable to other fuels, of this the amount switchable to electricity, and of this potential figure the projected amount that will switch from oil to electric. These studies served as a basis for the maximum amount of throwover. The minimum was assumed to be zero.
Industrial Cogeneration - The ability to produce electricity on-site from waste steam has become alluring to many large industrial customers.
Hamel & Brown reviewed each SIC for cogeneration potential to the year 2000 with their findings applied to the JFMIN band.
The current level of cogen-eration is not expected to expand in the JFMAX band based on the assumptf.on that the current cogeneration facilities would expire without replacement.
ASSUMPTIONS Economic Assumptions The assumptions within the long-term judgment forecast are its cornerstone. The forecast group spent a great deal of time and deliberation in formulating these items that will have an ef fect on the nation's economy as well as the economy and sales picture of the PP&L system.
In keeping with the idea of forecasting in a band, the assumption was made that real GNP per capita will increase at an average annual rate of between $92 and $102 per person per year. The $102 per person per year level is found to be lower than the majority of forecasts, but is equivalent to the historical level since World War II.
Population growth has a substantial influence on the nation's economic health. Although population growth typically produces growth in our economy, it does not necessarily produce an increase in GNP per capita. The U.S. population growth rate is expected to approach zero by 2000, producing a stable population level by 2010. A potential short-term upswing in the birth rate is expected as the post-war baby boom children enter the prime child-bearing years during the 1980's.
The need for a second family income, which forces more women into the workforce, may limit this growth. As population growth slows, the average age of our population will rise.
This will produce a larger pool of experienced workers, which will aid productivity, and a larger pool of Social Security recipients. Those of retirement age may find it necessary to remain in the workforce longer due to financial need.
$/ Hamel & Brown, Alternate Fuel Evaluations, September 7, 1979.
5/ Ozarks Regional Commission, Ozarks Regicnal Energy Alternatives Study, Missouri Summary, Little Rock, Arkansas, August 1977, page 22.
16
i Productivity is expected to increase, but at a decreasing rate. A continuing shift of employment from the manufacturing sector to service-oriented positions contributes to this trend.
Energy issues are an important facet of any kWh sales forecast.
In the long term, oil producers can be expected to cut production to keep supply just short of demand. The probability of embargoes of oil, natural gas, and valuable raw materials increases as we continue to import greater quantities frort less reliable sources.
Pricing Assumptions 4
A key assumption within the traditional long-term forecast concerns i
fuel prices.
Prices for the substitute fuels of oil, gas, and coal are a vital ingredient in a forecast which helps outline the market for electricity in the home, office, and industrial plant. The electricity market is ex-
]
pected to grow due to the attractiveness of its cost, versatility, and i.
resource security in comparison to other fuels. Table 2 shows the band of assumptions for all fuel prices.
For JFMIN the assumption was made that there would 'be positive real price increaset for all fuels. The percentage increases for electricity and coal are expected to be less than those for oil and natural gats.
i For JFMAX the assumption was made that the real price of electricity would decrease in the 1980-2000 period while all other fuels would exhibit i
increases in real prices. Again, oil and natural gas show the largest increases.
f I
1
(
17
(
.._..,.....,,.-~.-_.-___m.,.
TABLE 2 ASSUMPTIONS OF REAL ENERGY PRICE GROWTH 1980 - 2000 Electricity Coal Oil Natural Gas Inflation i
JFMIN 0.0 - 1.0%
0.5 - 1.0%
1.0 - 2.0%
2.0 - 3.0%
8.0 - 9.0%
JFMAX (0.5) - 0.0%
1.0 - 2.0%
3.0 - 4.0%
5.0 - 6.0%
6.0 - 8.0%
i i
r p
b r
18
GRRPH 2 LONG-TERM JUDGMENT FORECAST 70000 Co cc 60000 a
59,09@
I d
53,654 H
8+
C h!
2 50000 O
/
J O
r g+
O 40000 2
l B
O
,183 30000 r
7,346 r
JFMINC
+-
l 20000 1980 1985 1990 1995 2000 YERR 19
_ =, _.
- _ =
DERIVATION OF THE PROBABILITY BAND FORECAST 4
THEORY To improve the applicability of the long-term judgment forecast for planning, it is important to produce an estimate that proves durable and long lasting and spans the most probable outcomes.
The intent is to produce a i
band forecast with as narrow a range as possible but wide enough, so that with proper placement, it has a high probability (80%) of covering the re-sulting actual sales level.
METHODOLOGY It must be noted that the probability band forecast is formed using consistent sets of assumptions and is an outgrowth of an initial analysis which displays the absolute maximum and minimum sales levels (JFMAX, JFMIN) that can occur 1980-2000. Adding conservation, throwover, and cogeneration l
estimates to the maximum and minimum values provides the forecast team with a l
refined set of outer bounds. The probability band refines this initial estimate to capture the best estimate available.
Suggesting a band that increases significantly less than plus or i
minus (+) one percent per year around a midpoint significantly increases the probability of the band being too narrow to contain the eventual, actual level of sales.
The method for setting a narrower forecast band would include i
analyzing each of the input variables within caen separate class for a i
potential range, statistical distribution, and/or probability of occurring.
l Computer costs and time constraints limit this approach and strongly suggest the development of the narrower forecast for total sales be approximated using an inherent understanding of each of the individual variables that influence total sales.
It is f ait by the fotecaster that making the estimate l
for the total rather than each class has detracted little from the end t
result.
i The initial placement of the probability band about a center point of the outer bounds produces a 16,000 GWh range of sales estimates. Implicit i
knowledge of the actual distributions of the variables led to the conclusion that the distributions of each variable were not actually symmetrical or uniform but were skewed toward the lower side.
This observation formed the backdrop in determining each side of the band which in tura also determined the band width.
The selected upper bound is premised upon a mean level of industrial activity as well as a mean level of the ratio of industrial activity to elec-tric energy consumed. The full conservation effort / as well as the maximum 1
1/ A full conservation effort by customers is most likely to occur when economic conditions are good and when money is available to invest in new improved equipment either for replacement or for expansion.
20
__,__.m.-.__
level of fuel substitution to electricity is also included. Finally, 70 percent of the electric vehicle potential pushes the upper astimate to 44,000 GWh in the year 2000.
The bottom side estimate takes into consideration the technically minimum annual increaae -- based on historical observation in each sales cla s s.
This estimate is applied as a guideline in the fornation of the actual lower bound.
Incorporating a minimal amount of conservation as well as a modest addition for the electric vehicle and a low estimate for fuel substitution to electric produces a level of 31,500 GWh.
As a check the econometric analysis implies the following:
e Fluctuations due to uncertainty in economic activity, all else constant, indicates that the band width cannot be less than 6,500 GWh in the year 2000.
e Similarly, the amount of uncertainty surrounding future anergy prices, all else constant, suggests the band width must be 16,500 GWh for the year 2000.
RESULTS In 2000 the probability band is 12,500 GWh wide (+ 0.86% annual rate from mid point), symmetrically located around a mid point. With respect to capacity planning, enough uncertainty exists within this baad to provide the demand for one additional 900 mW generating unit or one less unit from the needs of the band mid point.
A major observation which surtaced from this analysis suggests threa eras of electric energy usage which have been embedded within the forecast.
The analysis predicts the continuation of the conservation era to approxi-mately 1986. This era is characterized by relatively low year-to-vear gains due to various conservation / energy management programs. A second more dynamic era of throwaver from oil and perhaps natural gas to coal and nuclear *aased electric is expected to follow. The period ending near 1997 is characterized by continued high oil and natural gas energy prices which persuade energy consumers to switch to electric power. Deregulation of oil and natural gas spurs the switch by homeowner and manufacturer alike. This period portends relatively high year-to-year gains for electric sales. The final three years to 2000 will be dominated by the maturation of alternate renewable fuels.
Included in this category are wind, cogeneration, biomass, solar, low-head hydro, geothermal, and synthetic fuels.
These three rather distinct phases with varying growth rates result in the band of expected sales shown on Graph 3.
l l
21 l
L
GRAFH 3 THREE ERAS OF ELECTRIC ENERGY CONSUMPTION LONG-TERM JUDGMENT FORECAST: PROBABILITY BAND 1980 TO 2000 50000 -
9 1
i l
5998-200d l
ALT $RNIITE' F0ELS 4s000 l ---
i ERA i
l
!~ W z
r d
goooo 1987-97____
g.
~~ lllg g
l TH'ROWOVER
/
!lf" l
/
l ll 1 x
S di[ i g
3s000 gf N!:
llli!
u_
a l
/
iiri
!';i 1980-86 l
/
iIL_._
N 30000 -CON $ER'VATION-l f
/
/pe'j205/YERR_
sg O
ERA yl l
.u ii4ii e
=
a J
As i
/
4Ng6 N
!!I!!..........
61 2sooo 3311 j.....f j
.g p 7
/-
310/ TERR l
l 2o000 1980 1985 1990 1995 2000-TEB16hB
1 GRAPH 4 PROBABILITY BAND IN COMPARISON TO LONG-TERM JUDGMENT FORECAST 70000 (n
OC 60000 O
JFMAX 9
/
l-I-
E Z
50000 o
___I w
r o
40000
/
glijji{
1 k.PROBABILIT'
- .rei
!!llf-BAND (n
Z i+:s
/
m a
7 y
J 30000 JFMINC y
=
20000 1980 1985 1990 1995 2000 YEAR 9
23
1980 SHORT-TERM FORECAST: METHODOLOGY AND RESULTS INTRODUCTION The 1980 short-term judgment forecast was developed within the normal annual forecast cycle. Historically the forecast period is 18 months but this year's analysis was extended to 1986. The 1980 long-term judgment forecast provided the additional coverage required to 1986. Although minor differences surface during the common six-year period because the forecasts were done at slightly different times, they have essentially been constructed around the same assumptions.
METHODOLOGY The development of PP&L's short-term forecast is based upon the data supplied by the Company's respective field divisions. Each of the six divi-sions seeks out the most recent expectations of the local home builders, commercial operations, and industrial customers with respect to new construc-tion, additions and/or layoffs of workers, production increases, and conser-vation accomplishments. Each of these changes is recorded qualitatively (what change to customer operations has occurred and why) as well as quanti-tatively (the effect expected on kWh sales) for the next 18 months.
This analysis divides each of the four major classes (Residential, Commercial, Industrial, and Other) into sub-categeries or Standard Industrial Classification (SIC) codes to improve the data collection, the knowledge of customer energy characteristics, and the forecasting task. Historically it can be demonstrated thai customers' expectations of their kWh consumption for the succeeding 18-24 month period are usually too optimistic. To counter this optimism the forecast team tempers these estimates.
Since this year s projection time frame was extended four years to cover through 1986, our method of analysis had to be altered and expanded.
No customer expectations from division sources were available from 1982-1986.
Each of the sub-categories and SIC codes was reviewed in depth with emphasis placed upon the recent historical data which displayed annual movements in kWh sales, the overall strength (weakness) found in each industry locally as well as nationally, and the cyclical nature of the industry with respect to economic conditions.
Two final adjustments were made. Due to the uncertainty and difficulty of forecasting economic activity, it was concluded that placing a recession precisely in 1985 (see CYCLELONG 2004) could not be defended -- it may not start until 1986 or 1987. The cyclical nature of the economy is an accepted event and the economic losses associated with recessions have been factored in over a longer time period. A second refinement increased the projected average use of residential customers in the early years of the 1980's by reducing the decreasing trend.
It was concluded that the rapid retardation of average use would slow up - probably the result of decreasing opportunities for conservation and relatively smaller increases in prices for electricity.
24
ASSUMPTIONS A brief review of the critical assumptions used in the 1980 forecast is recorded below.
Real GNP to be tracked at $100 per capita per year oc average to set e
the 1986 real GNP value.
e Interest rates to be on the general rise in the short term and higher than historic levels for the long run.
e Inflation rate expected to be high by historic levels, near double-digit figures in the short run.
e Foreign sources continue to exert control over the supply of valuable raw materials.
o Conservation will reduce some economic activity even though it is theoretically possible to continue production at current levels with less energy. Maintaining the current standard of living is likely to be challenging.
RESULTS The short-term forecast was constructed to produce point estisates in contrast to the band format of the long-term forecast.
This approach coincides with the various needs of the Company in supplying a specific value for each year. Table 3 contains the resultant forecast with a breakdown by major class. Levels both with and without sales to neighboring utilities of UGI-Luzerne Electric and Atlantic Electric are shown.
f
{
l l
l I
l 25
. ~. -
TABLE 3
~
1980-86 SHORT-TERM FORECAST KWH x 10 Including Sales to Excluding Saler, to UGI-Luzerne Elec.
UGI-Luzerne Elec.
& Atlantic Electric /
& Atlantit Electric 2
1 Residential 8,088 8,088 9
Commercial 5,660 5,660 8
Industrial 7,984 7,984 1
Other 1,056 729 Total 1/
23,073 22,746 1
Residential 8,405 8,405 9
Commercial 5,990 5,990 8
Industrial 8,313 8,313 2
Other 1,299 742 Total 1/
24,303 23,746 1
Residential 8,717 8,717 9
Commercial 6,240 G,240 8
Industrial 8,520 8,520 3
Other 1,636 752 Total 1/
25,416 24,532 1
Residential 9,046 9,044 9
Commercial 6,490 6,490 8
Industrial 8,730 8,730 4
Other 1,870 762 Total 1/
26,'43 25,335 4
1 Residential 9,374 9,374 9
Conmercial 6,780 6,780 8
Industrial 8,950 8,950 5
Other 1,913 769 Total 1/
27,334 26,190 1
Residential 9,728 9,728 l
9 Commercial 6,990 6,990 i
8 Industrial 9,140 9,140 6
Other 1,932 777 Total 1/
28,115 26,960 l
1/ PP&L merged with Hershey Electric Company effective March 1, 1980.
Sales
(-
to customers in this portion of our service area were estimated separately and are incorporated in the Total category but not the individual Residential, j
Commercial, Industrial, and Other pieces.
2/ PP&L has contracted to sell electricity to two neighboring utilities, UGI-l Luzerne Electric and Atlantic Electric Company. Sales to them are put into l
the Other category.
l 26
-. - ~.
PEAK LOAD FORECAST INTRODUCTION In conjunction with the preparation of the sales forecast, a peak load forecast is made.
In order to adequately provide for our customerr' demands for electricity, generating capacity must be available to serve the highest demands placed on the system. The peak load forecast provides an estimate of the magnitude of the max 12um hourly demand PP&L will experience in the future.
Its development is grounded on the sales forecast through the extensive load research program maintained by the Company. Tne assumptions on the level of the economy, fuel price levels, conservation, and new technologies that are incorporated into a sales t'orecast are automatically incorporated into the corresponding peak load forecast. The peak load fore-cast serves as a guide for capacity planning.
METHODOLOGY PP&L's peck load forecasting procedura produces summer and winter peak loads by developing the contribution to peak made by each rate class.
The term " rate class" means all customers served under similar rate schedules.
Starting in the 1950's Pennsylvania Power & Light Company inter-mittently conducted studies measuring the load characteristics of major y
groups of customers.
Since 1977 PP&L has continuously collected load data from a permanent sample of more than 1,300 customers that is designed to cover all major classes.
Half-hour loads are gathered from each test cus-tomer monthly usina speciallzed metering and are translated onto a computer tape. The data is then audited, edited, and summarized with the aid of an IBM 360/370 computer system. This data provides great flexibility and opportunity to analyze and deteunine the' contribution each customer class is making to Company system peaks with some precision.
l The Company's load research program plays a major role in developing the historical load characteristics that form the basis of the load forecast.
Our continuous load study program is designed to determine the load charac-l teristics of the customers in each of our rate classes. Since the load l
characteristics of a customer at a 'nigh-usage level may be different than l
those of a low-usage customer in the same rate class, load charactaristics l
are initially developed for customers at different ranges of usage within a l
rate class.
For those classes billed with watt-hour meters, stratified tandom samples of customers within kilowatt-hour usage ranges are selected f
and studied.
In the case of most general service rate classes (up te 7000 kw l
with demand meter billing), random samples of customers in load factor ranges l
are used. The largest commercial and industrial customers are studied indi-vidually.
The data collected through this load research program enables PP&L to construct a load curve for an average customer in each rate class stratum
[
for the days of the Company's monthly peaks. The next step is to expand 27
~.
these average customer load curves to ones that represent all the customers in that usage range.
For customers studied by kilowatt-hour ranges, demand ner customer data for each stratum of each rate class are multiplied by the number of customers in that stratum to obtain a universe daily load curve.
Average customer daily load curves for load factor strata are stated as ratios to customer monthly maximum demand. These ratios are applied to the sum of total customer demand in each load factor stratum as determined from an hours-use distribution to obtain the universe daily load curve.
For s historical year the strata of a rate class are added together to form the daily load curve for the entire rate class. The rate class load curves, which at this point are at the sales level and do not include the line losses between the generating plant and the customer, are corrected for losses to put them at the generation level.
The sum of rate class load curves represerts the system load curve and is checked against actual peak loads.
Using these techniques we have developed rate class contributions to summer and winter system peaks historically for selected hours of the day.
The ratio between class contribution to system peak and annual sales to that class is calculated for each rate class at the time of summer and winter system peak, for every historical period analyzed. The trend of this ratio for either a summer or a winter system peak is fairly constant over time.
For a given rate class the trcud of this ratio for both summer and winter system peak is projected through time.
By applying the appropriate ratios to the predicted annual sales of a class in any future year, that c'2ss' con-tribution to summer and winter peak is forecasted. The system peak for a specific time period is obtained by adding together the projected class contributions to system peak.
As is in the case of the sales forecast, judgment of experienced planners and forecasters is applied throughout this process.
RESULTS Table 4 sucmarizes the resulting forecast. The peaks determined are the summer seasonal peak, the December peak, and the January peak (winter seasonal peak). The sales are expected to grow at a slightly higher rate than the annual peak as evidenced by the increasing load factor. Due to the increase in the amount of heating load served, the winter peaks are moving from the 10-12 AM time period to the 8-9 AM period. Of the two seasonal peaks, the forecasted winter peaks have a higher growth rate than the summer peaks, which has also been true historically.
In addition to the conservation that is incorporated into the peak load forecast as a reflection of the conservation exhibited in the sales levels, the effect of demand management programs, such as demand controllers and off peak heating, is also included. These programs would focus on reducing daily peaks in the winter.
The 7,020 MW winter peak forecast for 1995 represents a compound annual growth rate of 2.5 percent over the 1977 weather-adjusted peak of 4,500 MW.
The corresponding growth rate for sales is 2.8 percent. The 28 1
e-m.,..,,. _.
extensive work in the areas of econometric forecasting, judgment fo recasting,
and the development of the probability band as described earlier in this testimony confirms that the 2.5 percent peak load growth and the 2.8 percent sales growth are most reasonable. 'Jraph 5 exhibits that over the long-term,
.this level corresponds both with the econometric forecast and with the center of the probability band.
For planning purposes, in addition to studying future capacity needs under the 2.5 percent load growth (1977-1995), 1.5 percent growth and 3.5 percent growth were also investigated. A plus or minus one percent deviation from the 2.5 percent growth was chosen for the band because using a smaller value would increase the probability that the band would be too narrow to cover the eventual actual level of peak while using a wider band would also decrease the band's practical value by encompassing too broad an outlook.
The support for these views is based on the analysis conducted to set the bands for the sales forecast. The 1.5 to 3.5 percent band covers the sets of assumptions that combine factors which cause sales and peak loads to decrease and combinations of assumptions that generate increased sales and peak loads.
Only by assuming unlikely circumstances can the forecast for the period be driven outside this band in any significant way.
t i
l 29 i
I
TABLE 4 l
Pennsylvania Power & Light Co.
i Forecast (10/80) Without UCI l
l Peaks January Subsequent Load Sales Output Summer December Year Factor MWh MWh MW MW MW 1980 22,458 24,367 3946 4290 4740 59.2 l
1981 22,746 24,710 3990 4440 4900 59.4 1982 23,746 25,775 4100 4550 5060 60.0 1983 24,532 26,628 4170 4710 5240 50.0 1984 25,335 27,561 4310 4880 5430 59.9 8
1985 26,190 28,416 4420 5010 5580 59.7 1986 26,960 29,252 4570 5210 5800 59.8 1987 27,980 30,358 4690 5430 6030 59.8 l
1988 29,070 31,541 4780 5520 6160 59.7 1989 30,400 32,984 4870 5640 6290 61.1 1990 31,100 33,744 4970 5770 6420 61.2 1991 31,800 34,503 5060 5870 6540 61.4 1992 32,500 35,263 5160 5980 6660 61.6 1993 33,100 35,914 5250 6090 6780 61.6 l
1994 33,800 36,673 5340 6200 6900 61.7 1995 34,400 37,324 5430 6300 7020 61.7 These sales and peaks exclude contract sales of electricity to a j
neighboring utility, UGI-Luzerne Electric Company.
c GRAPH 5 COMPARISON OF RESULTS 70000 FORECAST (10/80)
ECONOMETRIC FORECAST CO cc PROBRBILITY BAND 3
60000 O
I I
F-l-
C Z
50000 O
__I se r
u_
O L10000 p','
':: iE
~
27' l
CO z
.. iii:;jF.:e- ~
~
~
O
,# :s
. 3::-
f
- 3::
,.-s,-
- im
_I 30000
~
,/~~I......'..:
7
/
20000 19t0 1965 1990 1995 2000 YEAR 31
_