-85—
The Importance of Feedback in the Pacific
Northwest Electric Conservation Planning Model
Mike Bull
Bonneville Power Administration
Andrew Ford
Los Alamos National Laboratory
Roger Naill
Applied Energy Services Inc.
ABSTRACT
This paper describes the importance of feedback loops included in a policy
model constructed for the Office of Conservation of the Bonneville Power
Administration (BPA). First there is a description of the region and the
responsibilities for conservation planning at the BPA, and then a
description of the purpose, structure, and use of the policy model. Several
feedback loops involving customer response to higher electric rates are
selected for our discussion of feedback. The system dynamics treatment of
these feedback loops is contrasted with the treatment found in most electric
utility planning models used in the USA. The paper concludes with an
assessment of whether the inclusion of feedback has been important in BPA's
application of the model.
INTRODUCTION
Electric utility companies in the USA have become increasingly interested in
information and subsidy programs to encourage their customers to invest in
conservation. Conservation programs are viewed as necessary to overcome
market obstacles that limit customer investment to improve the efficiency of
electricity use. Utility programs include general information such as
advertising, specific information such as audits, and direct financial
subsidies such as zero interest loans. Utility conservation programs are
often viewed as a better use of company funds than investment in
conventional coal or nuclear power plants (Bryson and Elliott 1981).
Nowhere is interest in conservation programs stronger than in the Pacific
Northwest. This region is unique because of its vast hydro-electric
resource which permits the lowest electric rates in the country... Because of
historically low rates, the region's homes and businesses have not made the
same level of investment in conservation as in other parts of the country.
Thus, the potential conservation savings available at relatively attractive
costs is quite large (Council Plan 1983). The Pacific Northwest is also
uniquely organized to plan for the orderly development of the large
conservation resource. With the passage of the Pacific Northwest Electric
Power Planning and Conservation Act in 1980, this region organized itself to
plan the development of all the region's electric resources in a
-86-
co-ordinated manner. The Act created a Planning Council with members from
different states and the responsibility for setting broad policies. The Act
created and directed substantial new responsibilities for the BPA to help in
the implementation of the Council's policy. The Act also called for the
Council and the BPA to give highest priority to the acquisition of
conservation savings in planning for the electricity needs of the region.
During the period from 1981 to 1983 BPA greatly expanded its conservation
planning and program implementation capabilities. At the same time the
Council launched a two-year planning process culminating in the adoption of
the Regional Plan in early 1983. Due to the challenges involved in these
new responsibilities, both entities spent considerable effort on developing
resource assessments and tools to characterize the effect of conservation
and other resources for power planning purposes. For conservation planning
at BPA the work was concentrated in two primary areas of development: (1)
building program offerings based on the experience of utilities in the
region with residential retrofit programs; and (2) using preliminary
assessment data to represent the regional conservation potential within the
evolving corporate planning models. Conservation supply curves were
developed based on existing end-use assessments and structured so they could
be reconciled with both the end-use load forecasts and the resource
acquisition model for system expansion planning.
These modeling efforts generated three difficulties for conservation
planning. First, the conservation models were based on detailed end-use
assessments and hence were very cumbersome to use. Also, existing corporate
end-use demand forecasting models were not suited to retrieve the effects of
alternative conservation programs and policies. Second, none of the initial
conservation modeling had the capability to easily or practically model the
effects of BPA conservation subsidy strategies or program timing decisions.
Finally, the desk top analysis that was done for early program designs was
inadequate to answer questions about the ultimate system impacts of
programs, or potential tradeoffs among programs.
Therefore, in 1983 the BPA Office of Conservation initiated a study to
improve its ability in modeling the effects of its conservation programs and
consumer subsidy designs for the Pacific Northwest regional electric power
system. The model was to provide ready access for program planners and
analysts alike, build from the results of running existing models and
databases, and provide quick analysis of many scenarios, while preserving
general consistency with actual system planning and operations.
THE MODEL
The model is known as CPAM, the Conservation Policy Analysis Model. CPAM is
a system dynamics model with around 2,400 variables, about a third of which
must be specified by the user. BPA staff operate CPAM on the Dartmouth
College computer with the help of a user interface program which provides
English language prompts about the variety of policies that may be tested,
scenarios that may be assumed, and outputs that may be requested. The first
version of CPAM is known as the Regional Model because the loads and
resources of the Pacific Northwest are treated as if they were under the
control of a single utility. Further information on CPAM is given in the
summary paper by Bull (1984), the working notebooks prepared for the Office
-87-
of Conservation (Ford, Martinez, Naill, Geinzer, and Wood 1984) and in a
recent analysis of conservation policy in the Pacific Northwest (Ford and
Naill 1985).
Figure 1 shows the Regional Model along with the special programs that have
been constructed to facilitate the model's use at the BPA. These include
the “user interface" which assists the user in setting up new simulations,
the "LOTUS pre-processor" which translates cost and savings information on
thousands of individual conservation measures into conservation cost curves
for different end uses, and the “documentor" which generates a documented
listing of the equations and variable definitions. CPAM is comprised of 5
sectors which represent different aspects of the region's electric system.
Each of these sectors is designed to “mimic" on a very simplified basis the
existing corporate models which BPA uses to do its overall resource
acquisition planning and financial analysis. The most important sector is
the conservation and electricity demand sector which is highlighted in
Figure 1.
This sector simulates the utility customer's investments in conservation
measures (and the electricity sales and conservation savings resulting from
those investments) both with and without conservation programs. Here the
model's forecasts have been calibrated to BPA's annual load forecast, using
relatively gross assumptions about sectoral growth rates. Simulating
projected electricity sales with and without a specified conservation
program determines the net conservation savings from it, taking into account
all the effects of the system's feedback loops on electricity demand and
conservation. The demand sector has a great deal of structure that allows
the testing of different types of program designs, and different subsidy
levels, for 13 different end-use service categories.
The price of electricity needed in the conservation calculations is
generated in the price regulation and construction financing sector. This
sector mimics a simplified version of the ratemaking practices of the
region. Rates are based on the annual costs of operating the hydro-thermal
system and the return allowed on the utility company's investment.
Operating costs are tracked in the hydro-thermal system operation sector
which dispatches the region's thermal power plants and hydro-electric units,
keeps track of the amount of interruptible load to be served, and keeps
track of the secondary sales to utilities in California. The utility's
assets are represented in the capacity and assets accumulation sector which
performs most of the bookkeeping functions of the model. Decisions on the
timing, magnitude, and type of generating capacity to add are represented in
the capacity expansion sector. Investment in new generating units is
determined endogenously, based on an internal forecast of demand growth and
a comparison of the levelized bus-bar costs of the generating options
available.
Figure 2 shows the model's base case projection of regional electric load
used in a recent analysis of conservation policy in the region (Ford and
Naill 1985). This projection combines the effects of the assumed growth in
the region's economy, the customers' response to price changes, and the
customers’ additional response to BPA's current program to subsidize
customer improvements in the efficiency of electric space heating. Figure 3
shows the model's base case projection of capacity additions needed to keep
USERS
BPA OFFICE OF CONSERVATION, LOS ALAMOS NATIONAL LABORATORY, APPLIED ENERGY SERVICES
DOCUMENTED
: MENUS,
SELECTION OF: RESULTS OF SIMULATIONS, AND
POLICIES LIST OF PARAMETER CHANGES
SCENARIOS
ouTPUT
MEASURES DATA
RESIDENTIAL (LBL)
COMMERCIAL (PNL) USER INTERFACE
INDUSTRIAL (SRC) AND CHARTMAKER
(BASIC LANGUAGE}
COST, SAVINGS
OF EACH MEASURE
wv INSTRUCTIONS RESULTS
‘CONSERVATION vy
MEASURES
PREPROCESSOR THE REGIONAL MODEL
(Lotus)
DYNAMO LANGUAGE, 2400 VARIABLES, 800 INPUTS, $2 PER RUN
60K CORE ON DARTMOUTH COMPUTER, TURNAROUND TIME IN SECONDS.
CONSERVATION PRICE REGULATION
INVESTMENTS AND AND CONSTRUCTION
ELECTRICITY DEMAND FINANCING
CONSERVATION COST FEEDBACK LOOPS
CURVES FOR EACH ‘ARE ACTIVE
END USE (OURING EACH TIME
STEP OF THE |
OSAP HYDRO-THERMAL
SYSTEM OPERATION
CAPACITY
EXPANSION
PLANNING NL
CAPACITY AND ASSETS
ACCUMULATION.
(BOOKKEEPING)
LISTING OF
THE MODEL
DOCUMENTOR
(BASIC LANGUAGE)
2400 VARIABLE
DEFINITIONS
EQUATIONS
Figure 1. The Regional Modeling System.
-9g-
~89-
12
INDUSTRIAL
COMMERCIAL
6 - 4
RESIDENTIAL
i 4 i
1984 1994 2004
Figure 2. Base Case Projection of Regional Electric Load.
T
GENERATING CAPACITY
.p. GW, hydro in av.
(POW. Hyco MTBY.GM) COAL-FIRED CAPACITY
4e —_- oe =
NUCLEAR CAPACITY
2 SMALLHYDRO —
—— —
“7 -
ee Se a 1
1984 1994 2004
Figure 3. Base Case Projection of Capacity Additions.
pace with the load growth. The discrete changes in nuclear capacity occur
at the user specified dates for completion and retirement of the region's
nuclear units. Small hydro and coal capacity are determined endogenously,
based on resource assessment data from BPA. The small hydro resource is
developed first because the base case assumptions allow for about 1.3 GW of
small hydro capacity that is more attractive than new coal plants.
In a hydro dominated system, such as that in the Northwest, the relative
balance between the loads and resources is often summarized by showing the
surplus that would exist under critical hydro-electric conditions. Figure 4
shows the model projections of a regional surplus to remain until around
1996 under the base case assumptions. The long duration of the surplus is
responsible for the model's reflection of a long period of low avoided costs
shown for the base case in Figure 5. During the first 8 years of the
planning period, for example, the avoided cost is equal to the estimated
secondary sales rate charged to utilities in California. (If electricity
demand were to be reduced by 1 kwhr, it is expected that the region would
continue to operate its hydro and nuclear units as before and sell the extra
1 kwhr to California. Total production costs would be unchanged; net
production costs would decline by the secondary sales rate.) Figure 5 shows
that the avoided cost would increase later in the simulation as growing
loads force the region to first invest in small hydro facilities, and then
in new coal-fired capacity.
The base case results shown in Figures 2-5 provide a point of departure for
the analysis of conservation subsidy programs. Any one of a variety of
proposed conservation strategies may be tested through direct simulation,
and the overall effects are summarized in terms of simulated changes in the
resource plan, electric rates, financial indicators, and the region's "total
system cost." The most extensive analysis to date has shown that new
programs are likely to be successful in reducing the "total system cost”
(where the cost of the customer's investments in conservation measures are
combined with the monthly bills from the electric utility). This is a key
finding because of the importance placed on “total system cost" by the Act,
the Regional Planning Council, and the BPA. A problem with introducing
large new programs, however, is that they tend to increase the average
electric rate over the planning period. Previous analysis has shown that
conservation planning is made difficult by the conflict between two worthy
goals--reducing total system costs and avoiding electric rate increases
(Ford and Naill 1985). Depending on how the rate increases on different
groups are weighted, one might adopt a variety of conservation strategies
ranging from discontinuing current programs to the initiation of new
programs which could acquire five times as much electricity savings as the
current weatherization program.
ANALYSIS OF CONSERVATION PROGRAMS
The principal application of CPAM is to test the effectiveness of
conservation programs in the region or utility service area.
Figure 6 shows how conservation is calculated in the Regional model. With
no conservation programs, the model projects conservation induced by price
effects alone; the addition of conservation programs enhances price-induced
conservation. The model therefore can be used to project net conservation
2r ADDITIONAL COAL AND SMALL HYDRO CAPACITY =“ NJ
a DEFICIT (av. GW), 7 i
1984 1994 2004
48
36
24
12
-91-
qT
SURPLUS (av. GW)
BASE CASE SURPLUS
t +
SURPLUS IF ONE IGNORES
Figure 4. Base Case Projection of Capacity Surplus.
(mills/kwhr, 1980 $) /
y a
i.
LEVELIZED COST OF A ]
|. NEWCOAL PLANT ’
LEVELIZED COST OF NEXT AVOIDED COST
SMALL HYDRO FACILITY
oxsmmu em ° cme a
$e +——~+
SECONDARY SALES RATE
| AVOIDED COST
1984 1994 2004
Figure 5. Base Case Projection of Avoided Cost
of Electricity Generation.
-92-
BASE USE
PER HOUSE
DEMAND
NUMBER OF PER HOUSE
HOUSES CONSERVATION/
HOUSE
ELECTRICITY
Began PARTICIPATION
DELAY
SUPPLY AND INDICATED
PRICING CONSERVAT1ON/
ALGORITHM HOUSE
BPA
CONSERVATION
PRISE PROGRAMS.
Figure 6. How Conservation Is Calculated in the Regional Model.
Conservation
Savings Structure
average
Price of
Hlectricity Information
Programs
ca
toaidanea __tgitet
Conservation “~~ “ractar
Invastaant Without
Subsidies
nM)
uedtiey
i subsidies
(57RM)
Savinge
(nw unite)
Investment ($/KW)
todilared
Conservation
Savings Before
Indidated em Hunt)
Conservation
‘Savings
oruroate
Figure 7. How Conservation Programs Are Included in the Model.
-93-
savings, measured by savings from conservation programs in addition to those
savings that might occur without the programs (from price alone). For
conservation programs, net conservation savings are an appropriate measure.
This measure takes into account a program's redundancy: the fact that
customers may purchase some conservation measures subsidized by a
conservation program even if the program never existed.
The rest of the structure in Figure 6 accounts for the dynamics of
conservation savings -- how conservation savings change through time. For
retrofit programs in existing buildings, there is a delay or lag between
indicated conservation savings (which is determined by price and
conservation programs) and actual conservation per house. This delay
represents the rate at which customers participate in the proposed programs
(it takes time to achieve significant participation rates for any program).
With no programs, it is assumed that consumers participate in purchasing
conservation measures by replacing old equipment with new equipment as the
old equipment wears out. The model calculates total demand and conservation
by multiplying conservation and electricity use per house by the total
number of houses in the region.
Figure 7 shows in more detail how indicated conservation is determined, and
how conservation policies and programs are accounted for in the Regional
Model. Consumers are assumed to move toward an indicated level of
conservation investment, measured in first-cost dollars per average kilowatt
saved, according to the current price of electricity and their financial
criteria for such investments (as determined by the capital recovery factor
based on their presumed discount rate). Consumers are assumed to make
investments until the annual cost of these savings (the cost/kWh of the
savings from the conservation measure) equals the price of electricity.
Data on costs and savings for conservation measures are used to calculate
the amount of savings that can be purchased from a given amount of
conservation investment.
Figure 7 also shows how conservation programs affect savings in the Regional
Model. Three generic types of programs can be represented in the model:
information programs (audits for example), subsidies, or performance
standards. Information programs affect the decision criteria consumers use
to make their investments, but not the costs of the investments. For
example, a good information program might lower the risk and uncertainty
associated with conservation investments, thereby lowering the hurdle rate
for such investments. Subsidies add directly to consumer investments (or
subtract from the purchase price of conservation measures), increasing the
amount each consumer is willing to invest. Performance standards determine
a minimum level of conservation measures that must be purchased by consumers.
The Regional Planning Council's 1983 Northwest Conservation and Electric
Power Plan estimated the economic potential for conservation savings in the
Northwest region, based on conservation savings costing less than 4
cents/kWh (in 1980 dollars according to utility cost accounting). For
residential space heating in existing houses, this economic potential was
estimated at 615 average Megawatts (Mw) at 4 cents/kWh. Newer data on the
potential amount of conservation in the Region indicates that this economic
potential may be larger than originally thought in 1983--as much as 1,220
average Mw, almost twice the Regional Plan's initial estimate of 615 Mw.
-94-
2004
CONSERVATION
SAVINGS
INDICATED SAVINGS 1075 Mw
(Base Case; 3.6 cents/kWh)
REDUCTION IN SAVINGS DUE TO:
PARTICIPATION DELAY ~-340 Mw
REDUNDANT SAVINGS (SAVING WITH NO PROGRAMS) -336 Mw
SECONDARY EFFECTS
Rebound Effect *
Price Feedback -6 Mw
Cost Escalation *
Interfuel Substitution *
Worn-Out Measures --
NET CONSERVATION 393 Mw
(* these effects are not included in the Base Case projection.)
Figure 8: Summary of Conservation Savings in the
Base Case (Residential Space Heating in
Existing Homes)
The Regional Model starts with the costs and savings of individual
conservation measures that add up to this 1.2 Gigawatts as input data, and
attempts to calculate how much of this conservation potential will be
realized with the different conservation programs. Figure 8 summarizes the
Regional Model's projection of conservation from the Base Case conservation
programs (a projection of conservation from the Base Case program, i.e., a
residential weatherization program paying 75 percent of the cost of
qualifying measures). With Base Case programs, conservation measures
costing up to about 3.6 cents/kWh might be purchased, according to the
model's calculations. These measures would result in 1,075 Megawatts of
conservation savings by 2004 (88 percent of the economic potential at 4
cents/kWh).
Yet not all of these indicated savings are realized as conservation by
2004. First of all, the program is not projected to result in full
participation by 2004--not all homes are expected to participate in the
program and install the measures. The model projects about a 70 percent
participation rate for the Base Case program by 2004, resulting in a
reduction of 340 Megawatts from potential or indicated savings. Second,
some of these savings--the Regional Model says about one-third of indicated
savings (336 Mw)--might occur anyway even without the conservation program.
These 336 Mw are referred to as redundant savings in Figure 8 (savings that
might occur with no programs). Finally, secondary effects--for example, the
rebound effect, price feedback, cost escalation of conservation measures,
interfuel substitution or the effect of worn-out measures--can further
~95-
reduce net savings from a conservation program. These secondary effects are
reactions caused by the savings themselves that work to erase some of the
savings. For example, conservation programs might cause customers to turn
their thermostats up because they can now afford more comfort. This
reaction offsets some of the program's conservation savings.
Figure 8 shows that these secondary effects are either small or excluded in
the Base Case projection. For example, price feedback reduced net savings
by only 6 Mw (1 percent) in the Base Case, because the price effects of the
conservation program were small (about 2 percent at most). The potential
impacts of the other secondary effects were either purposely turned off (the
rebound effect was turned off; it was assumed that worn-out measures were
replaced) or ignored (cost escalation, interfuel substitution). Later
versions of the Regional Model will explore the importance of these
secondary effects more completely.
Of the 1075 Megawatts of potential conservation savings from the
weatherization programs in the Base Case, Figure 8 shows that only about 40
percent--or about 400 Megawatts-~is realized as net conservation savings by
the year 2004. The Regional Model helps sort out the effects of a complex
number of factors on net costs and savings of utility conservation programs.
FEEDBACK
We now turn our attention to information feedback in the Pacific Northwest
electric system and in the CPAM representation of that system. As noted in
Figure 1, the different sectors of CPAM are tied together through feedback
loops which are active during each and every time step of a simulation.
Although this may appear as standard practice to participants in the
Keystone Conference, it is not the approach typically taken by electric
utility modelers in general.
To present an application of the model that may be of general interest to
the participants in the Keystone Conference, we illustrate the contrasting
approaches by focusing on the so-called "spiral of impossibility" in which
higher electric rates lead to lower sales, lower sales force the utility to
request further rate increases to cover fixed costs, and the new rate
increases lead to still further reductions in sales. We emphasize that this
application is presented merely for illustrative purposes because the focus
of the CPAM project is conservation programs and not the analysis of the
“spiral of impossibility."
The “spiral of impossibility," sometimes called the "death spiral," has been
studied for utilities with widely different characteristics by Ford and
Youngblood (1983) and for utilities in the Pacific Northwest by Moorlan
(1984). These and other studies indicate that the "spiral" could be
particularly bothersome when the utility customers react strongly and
quickly to changes in the price of electricity. Since the region's large
aluminum industry fits this description, it is natural that there should be
so much concern about the "death spiral" in the Pacific Northwest (Northwest
Energy News 1985).
In CPAM the "spiral" is represented as the first of 5 feedback loops in the
causal loop diagram given in Figure 9. To understand the effect of this
-96-
positive feedback loop, consider the likely effect of an outside disturbance
like a drop in the market price of aluminum. This change would lead to less
profitability, closures of some aluminum plants, less regional sales, an
increase in the average electric rate, an increase in the rate charged the
aluminum companies, an increase in the variable costs of aluminum
production, and a further decline in profitability. Loop #2 in Figure 9
provides some negative feedback to counter the effect of the "spiral." Here,
we assume that the least efficient aluminum plants will be the first to
close down in a depressed market. The improved efficiency of electricity
use of the plants remaining in operation reduces the importance of the rate
increases from the "spiral."
The first two loops tell only part of the story, however. A more complete
picture is provided by considering the effect of three additional loops,
each of which involve the estimated effects on revenues earned from the sale
of secondary power to California. We call loop #3 the “more secondary
sales" loop because it leads to greater sales of secondary energy over the
intertie to California when there is a loss of aluminum load in the region.
The third loop provides negative feedback which partially offsets the
severity of the "spiral" and is often cited by analysts who feel that the
“death spiral" is not a serious problem for the region. More secondary
sales are only possible, however, if there is room on the intertie to
California. As the intertie becomes more congested, the California
utilities are more effective in reducing the secondary rate. The effect of
intertie congestion is represented by the fourth loop in Figure 9.
A similar effect is represented by the fifth loop which involves the
calculation of secondary rates based on the cost of the generating resources
used to produce the secondary power for California. This loop leads to
lower secondary rates when there is a loss of aluminum load in the region.
With lower regional load, the fraction of low cost generation (mostly hydro)
that is available for secondary generation is greater, and "cost-based"
secondary rates would decline. The effect of both the fourth and fifth
loops is to lead to lower secondary rates with reductions in the aluminum
industry load. If the intertie happens to be full, the reductions in
secondary rates will lead to lower secondary revenues, higher revenue
requirements, higher electric rates, higher variable costs of aluminum
production, less profitability, and still further closures of aluminum
plants. Thus, the fourth and fifth loops in Figure 9 are similar to the
"spiral" -- they act to amplify the effects of disturbances introduced from
outside the Pacific Northwest electric system.
To show the relative importance of the internal forces (the 5 loops in
Figure 9) and the external forces of the world wide aluminum market, we use
CPAM to simulate the likely closures in aluminum plants under different
assumptions on the aluminum marketplace. As a simple example, we assume a 4
cents/lb downward movement in the world price of aluminum. If this outside
disturbance is superimposed on a base case projection of aluminum prices, we
project temporary closures of almost all the region's aluminum plants
followed by a return to operation of roughly half the region's capacity. We
obtain a rough indication of the importance of the internal forces by
examining the rate increases during the years of greatest closures. These
rate increases, when multiplied by the average electricity requirement per
pound of aluminum, amount to a 1 cent/1b additional movement against the
‘ALUMINUM
DS! <—————_ CAPACITY
UTILIZATION
LoaD +
_— / (ev.6W) ALUMINUM 4
POTENTIAL PRICE +
(¢flo) {
+
(-)
SECONDARY —
FRACTION OF SALES
SECONDARY SALES (av. GW) OAT
FROMLOWER > (+) MARGIN LOOP:
COST RESOURCES Loo 1. lem) OTTER
(-) THE SPIRAL _ EFFICIENCY
- INTER-TIE
LOOP 3. 4
SECONDARY RATE GARAGIY: MORE SECONDARY +
BASED ON COSTS + (v.aW) SALES REGIONAL ee
OF SECONDARY ELECTRIC +, costs AVERAGE
CENERATON INTER-TIE We ees (¢f) ELECTRIC
(mills/kewhr) CONGESTION > ev.ow) NON- + REQUIRE-
(+) (%) ACTUAL ELECTRIC 4
SECONDARY VARIABLE
Loops. ‘SALES costs ELECTRIC
LOWER (av.GW) (€/lb) costs +
SECONDARY (€/b)
RATES - (+)
SECONDARYRATE top 4 r
BASED ON WHAT TEaTE
CALIF. WOULD GonceeHon _ psi
PAY (millsfkwhr) AVERAGE ELECTRIC
+ + NET ELECTRIC _» RATE
ANNUAL + RATE (mills pewhe)
SECONDARY + SECONDARY COST OF ALLOWED AP (ilsyewh)
RATE REVENUES “BP system _:b REVENUES
+ (miltisfkwhr) (milion OPERATION {mittion
syn) (million Siyn)
$n
Five Feedbacks Loops Controlling the Aluminum Industry Capacity Utilization.
Figure 9.
ALUMINUM
MENT
(kwhr jib)
=6=
-98-
industry. Thus, the model shows that the region's internal forces would
amplify a 4 cents/lb problem into a 5 cents/1lb problem. This 25%
amplification may not be a major problem relative to the great uncertainties
in the worldwide market for aluminum.
OTHER ELECTRIC UTILITY PLANNING MODELS
Surveys and conferences on models used in the US electric utility industry
indicate that feedback loops such as the "death spiral" are usually left out
of most planning models. In the Electric Power Research Institute's forum
on utility corporate models, for example, only one of a dozen corporate
models provided a direct representation of price feedback (EPRI 1981).
Forum participants examined this particular model (a system dynamics model
used by the Florida Power and Light Company) and found that it contained far
less detail than the other 11 corporate models but that it exhibited unique
and interesting patterns of behavior. When the reasons for the unique
behavior were uncovered, the Florida model was judged to be highly
intuitive, and the forum participants concluded by recommending that more
corporate modeling groups attempt to "close the loop" in future modeling
efforts.
Several of the forum participants met again with other analysts at a Los
Alamos workshop on regulatory-financial models used in the US electric
utility industry (Ford and Mann 1983). A dozen models were represented by
this group of analysts from industry, state agencies, and universities. A
survey of the models' treatment of feedback loops such as the "death spiral"
showed that only one of the dozen provided for an explicit representation of
feedback loops involving customer response to higher prices and involving
the financial community's response to company performance. Workshop
participants cited a variety of reasons why they chose to leave key feedback
loops out of their models. One group, for example, cited problems in
justifying the parameter values used in closing feedback loops in advocacy
hearings before state public service commissions. Another group recounted
their unsuccessful experiences to apply statistical procedures to quantify
the appropriate parameters needed to close the feedback loops. And still
another group offered the opinion that closing the loops made model results
too confusing for upper level management. It was suggested, for example,
that upper level management would lose confidence in a model whose
projection of electricity demand changed every time a new scenario was
devised and the price of electricity was different. Representatives from
the Florida Power and Light Company agreed with the difficulty in obtaining
parameter estimates, but they discounted the view that upper level
management would be confused by projections from a model with information
feedback. The Florida planners emphasized that utility companies use some
models to generate "numbers" and other models to generate "insights," and
that their purpose in adding feedback to their corporate model was to gain
insights.
THE BPA CORPORATE MODELS
BPA maintains a collection of detailed models to assist in analysis of
policies for the Pacific Northwest region. The models are constructed and
updated in different departments, and they are coded in different languages
to suit the particular needs of each department. This collection of highly
-99-
detailed models is referred to as the corporate models. Price feedback is
represented in the corporate models as shown in Figure 10. This diagram
shows three of the corporate models and the iterative process used to obtain
consistency on price projections. The process starts with BPA's best
estimate of the likely price of electricity over the planning period. The
estimated values are used as input to a demand model which provides a
projection of the electric load for each year of the 20 year planning
period. The load projection is then used as input to a capacity expansion
planning model which determines the amount, mix, and timing of generating
unit additions. Generation unit additions are then used by a cost model
which finds the annual revenue requirements and the price of electricity
that must be charged to meet the revenue targets. The price obtained at the
end of this sequence of model calculations is compared with the starting
price, and the sequence is repeated until a consistent set of prices is
obtained.
The lower half of Figure 10 illustrates the recursive approach used in the
system dynamics CPAM model. Here we show eight causal influences involving
different CPAM variables that would also appear in the corporate models.
With the recursive approach, all the interactions are part of one model, and
the feedback loops are active throughout the 20 year simulation.
THE IMPORTANCE OF FEEDBACK
Including feedback in the CPAM model has proved valuable in BPA's analysis
of conservation programs because of the increased understanding afforded by
the direct simulation of information feedback. The CPAM representation of
price feedback, for example, has greatly increased the understanding of the
likely difference between gross savings and net savings of alternative
conservation programs.
In reviewing the practical benefits of CPAM's use at the Office of
Conservation, however, a more important advantage of direct simulation of
feedback is the ease with which multiple simulations can be performed to
test the effect of conservation policies under a wide range of different
conditions. If, for example, BPA wishes to test the effect of a
conservation program under a variety of corporate planning assuimptions,
e.g, with more rapid growth in the region's economy, with cancellations of
nuclear units under construction, with a larger intertie to California, or
with a myriad of other changes, a dozen or so parameters are changed and a
new set of simulation results are obtained in rapid order. The new
simulations are obtained without extensive repreparation of the inputs
because of the many feedback loops which automatically adjust for changing
circumstances in the new simulation. In the event of cancellations of
nuclear units and more rapid growth in the region's economy, for example,
CPAM automatically adjusts the development of the region's small hydro
resource and the investment in new coal plants in a manner that mimics the
likely reaction of utility planners to the increased need for generating
capacity. The higher electric rates needed to pay for all the new capacity,
in turn, are calculated internally and used in the new calculation of the
likely customer investment in conservation and the likely operation of the
region's aluminum plants.
-100-
ITERATIVE INTEGRATION
pnice toap
°
.
START ecccee DEMAND —}> a”
MODEL of
YRS.
PRICE / \
END ee REPEAT wwe, Y
STARTING PRICE CAPACITY
waa ENDING PRICE ADDITIONS
AGREE MODEL
YAS, XN. —_
COAL CAPACITY
STARTS
cost
MODEL
.
.
= °
Lesees® ec vas,
RECURSIVE INTEGRATION
CONSERVATION ——————B
INVESTMENT ‘
ACTUAL EACH OF THESE 8 CASUAL
PRICE INFLUENCES IS EXERTED IN
EACH AND EVERY TIME STEP
+h (OF THE SIMULATION
INDICATED
PRICE
NN
ALLOWED RATE
Revenue BASE
INITIATIONS
Figure 10. The Two Approaches to Model Integration.
-101-
BPA views CPAM as a "screening tool" which can be used to provide rapid
turnaround analysis of a wide variety of policies. By screening through
many different policy proposals with CPAM, the Conservation Office can
narrow the number of proposals to be studied with the more detailed
corporate models down to a manageable number. Our principal conclusion,
therefore, is that the primary benefit of the inclusion of feedback in the
CPAM model is to facilitate the model's use in "screening studies."
REFERENCES
Bryson J. E. and J.F. Blliott, "California's Best Energy Supply Investment:
Interest-Free Loans for Conservation," Public Utility Fortnightly, November
5, 1981, pp. 19-23.
Bull, M., "Testing Conservation Incentive Levels Through System Simulation,"
Proceedings of the ACEEE 1984 Summer Study on Energy Efficiency in
Buildings, August 1984.
Northwest Power Planning Council, “Northwest Conservation and Electric Power
Plan, 1983" 700 S.W. Taylor Street, Portland, Oregon 97207, April 1983.
Electric Power Research Institute, "Case Study Comparison of Utility
Corporate Models: Utility Modeling Forum Second Working Group," report no.
EA-65, October 1981.
Ford, A. and G. Mann, "Summary of the Workshop on Regulatory- Financial
Models of the US Electric Utility Industry," Los Alamos National Laboratory
report no. LA~15-C, June 1983.
Ford, A-, Martinez, A., Naill, R., Geinzer, J. and F. Wood, “Briefing Notes
for the Regional Model" and "Users' Guide to the Regional Model," prepared
for the Office of Conservation, Bonneville Power Administration, Portland,
Oregon, March 1984.
Ford, A. and R. Naill, "Conservation Policy in the Pacific Northwest,"
report available from the Office of Conservation, Bonneville Power
Administration, Portland, Oregon, April, 1985.
Ford, A. and A. Youngblood, "Simulating the Spiral of Impossibility in the
Electric Utility Industry," Energy Policy, March 1983.
Moorlan, T., "The Utility Death Spiral," Northwest Power Planning Council
issue briefing, 700 S.W. Taylor Street, Portland, Oregon 97207, April 1984.
Northwest Energy News, "Industry in Crisis: Northwest Aluminum Companies,”
700 S. W. Taylor Street, Portland, Oregon 97207, Jan/Feb 1985, pp. 10-15.
CC BY-NC-SA 4.0