“7ge
SYSTEM DYNAMICS AND UNCERTAINTY:
Results of Two Applications of Formalized Sensitivity Analyses
with System Dynamics Models of the Electric Utility Industry
Andrew Ford
Systems Science Department
Institute of Safety & Systems Management
University of Southern California
Well structured system dynamics models are often quite
useful in the analysis of policy impacts in the face of multiple
sources of uncertainty. Simulation searches for a "robust"
policy that performs well under widely varying conditions are
often the most rewarding portion of a system dynamics study.
This paper reports the results of two studies where the analysis
of uncertainty is carried a step further. Here, we are
interested not only in policy impacts under widely varying
conditions but in whether a policy can reduce the uncertainty of
the system.
The paper begins with an important example from the electric
utility industry. Utility planners are interested in learning
the extent to which efficiency standards for new homes and
businesses lead to an important reduction in the uncertainty of
the electric utility system. The planners generally agree that
uncertainty in the number of new homes and businesses translates
into less uncertainty in electric load if the new buildings are
more efficient in their use of electricity. And many planners
feel that reduced uncertainty in electric load growth will lead
to reduced uncertainty in other variables like the average price
of electricity.
Two recent studies have been completed which combine system
dynamics models of electric utility systems with a formalized
statistical analysis techniques described at the 1983
International System Dynamics Conference. One study was
performed for the California Energy Commission for a hypothetical
California utility; the second was performed for the Bonneville
Power Administration for the Pacific Northwest electric system.
(The Bonneville model is explained ‘in papers at the 1985 and 1986
conferences. )
The paper provides a short review of how utility planners
commonly represent the long term. uncertainty in system
performance. Key differences between the system
dynamics/statistical analysis approach and the more common
methods are identified. Selected results .are presented to
illustrate the usefulness of the method. We conclude with a
discussion of several highly unusual findings from the Bonneville
study. The discussion of the "counter intuitive" results focuses
on the key role of information feedback in the Bonneville model.
-79-
THE IMPACT OF PERFORMANCE STANDARDS
ON THE UNCERTAINTY OF THE
PACIFIC NORTHWEST ELECTRIC SYSTEM
A FINAL REPORT ON THE HYPERSENS ANALYSIS OF CPAM
by
Andrew Ford
Systems Center
Institute of Safety and Systems Management
University of Southern California
Los Angeles, California 90089-0021
and
Jay Geinzer
Energy Planning Services Department
Applied Energy Services, Inc.
1925 North Lynn St., Suite 1200
Arlington, Virginia 22209
Work Performed Under Contract No. DE-AC79-85BP24760 by the University of Southern California and
Applied Energy Services, Inc.
BONNEVILLE POWER ADMINISTRATION
OFFICE OF CONSERVATION
-80-
CONSERVATION AND UNCERTAINTY:
AN ILLUSTRATIVE ANALYSIS FOR THE
CALIFORNIA ENERGY COMMISSION
Prepared by:
ANDREW FORD
901 18th Street, Suite 111
Los Alamos, New Mexico 87544
Prepared for:
ASSESSMENT DIVISION
CALIFORNIA ENERGY COMMISSION
1516 Ninth Street
Sacramento, California 95814
March 1987
-81-
A SIMPLE NUMERAL EXAMPLE
OF UNCERTAINTY REDUCTION
REGIONAL
(av. Gw)
Zs NO PERFORMANCE
STANDARDS
1994
UNCERTAINTY
INTERVAL
=10Gw
—_— HIGH BY 3Gw
+
1999
+
2004
= 22
UNCERTAINTY
_—__ INTERVAL
tH =8Gw
i =18
REGIONAL
ELECTRIC
DEMAND
(av. Gw)
PERFORMANCE STANDARDS REDUCE:
TOLERANCE INTERVAL
LOW BY 1Gw } BY 2Gw
| 1994 1999 2004
-82-
ANALYSIS WITHOUT
PRICE FEEDBACK
BUILDING
STOCKS
DEMAND SIDE
PARAMETERS
ELECTRIC
DEMAND
ELECTRIC
RATES
-83-
ANALYSIS WITH
PRICE FEEDBACK
BUILDING
STOCKS DEMAND SIDE
e PARAMETERS
ELECTRIC
DEMAND
[ (+)
+ ELECTRIC
ELECTRIC RATES
SALES *
+
POWER
PLANT
CONSTRUCTION (-)
ALLOWED
REVENUES
+
SUPPLY SIDE °
PARAMETERS. |
~84-
USING HYPERSENS WITH CPAM
Demand-Side Uncertainty
Market Shares
Program Compliance
Economic Growth
pae
Gas Prices
Potential Savings
Construction Costs
Intertie Operations
Supply-Side Uncertainty
Rate Making
‘Aluminum Industry
HYPERSEN
@ Randomly samples values from
each distribution for each
variable
© Produces a rerun‘ile with a user-
Specified number of "tests".
YNAMO
© Runs CPAM from the rerun file.
© Puts the output into a file for
further analysis.
HYPERSENS
© Analyzes the results from the
CPAM runs.
® Produces a variety of reports.
Off-line comparison of results
with the results of other runs.
THE
START
-85-
ITERATIVE APPLICATION
OF HYPERSENS
RANGE OF PLAUSIBILITY ON.
ORIGINAL SET OF INPUTS TO
THE ELECTRIC UTILITY MODEL
[| USING LATIN HYPERCUBE
DESIGN SET OF RERUNS
PERFORM RERUNS
[OOF «ELECTRIC UTILITY
AND THEIR RANGES OF
PLAUSIBILITY FOR
ALTERED ELECTRIC
UTILITY MODEL
PROCEDURES MODEL
NEW SET OF PARAMETERS CALCULATE PARTIAL
CORRELATION COEFFICIENTS.
‘TO SELECT MOST IMPORTANT
INPUTS TO ELECTRIC UTILITY
MODEL
Y
CALCULATE
TOLERANCE
INTERVALS
ALTER THE ELECTRIC
UTILITY MODEL TO
REMOVE THE CORRELATION
AMONG TOP INPUTS,
‘ARE THE
TOP INPUTS
INDEPENDENT,
?
Y
INTERPRET
TOLERANCE
INTERVALS
25.
19.
13:
00
. 99777777.
- 9777777 MMMMMMMMMM
TITTTTTIVTTITVITTTITATAT
HYPERSENS TOLERANCE
INTERVALS FOR
—~86-
REGIONAL DEMAND
7MMMMMMMMMMMMMM
$999999999999999999999999999999999999999999999999999999999999
. MMMMMMMM
9999
99999 77777
9999.77777
999977777
999977777 MMMMMMMM:
7777777777
hud
90 PERCENT COVERAGE
75 PERCENT COVERAGE
MEAN
1984.0
1989.0
1994.0
1999.0
2004.0
-87-
STANDARDS' IMPACT ON
REGIONAL DEMAND WITH THE
INITIAL AND FINAL VIEWPOINTS
UNCERTAINTY fe
_ REDUCTION 7
= (Gw) <
1.0 s / e
oe /
eS
oe i
we
ey e@
/ +
7 (1994) (2004)
00 | | |
0.0 0.5 1.0 1.5
MEAN REDUCTION (Gw)
@ FINAL VIEWPOINT
+ INITIAL EXAMPLE
-88-
STANDARDS' IMPACT ON
BONNEVILLE LOAD WITH THE
INITIAL AND FINAL VIEWPOINTS
2.0 =
UNCERTAINTY ®
REDUCTION
(Gw)
1.5; y,
7
7
_
a
et
Rey Y (2004)
i 7
gY
0.5 }—
Sy
7
yo e
1994
y, 1 (1994)
0.0% | | |
0.0 0.5 1.0 15
MEAN REDUCTION (Gw)
@ FINAL VIEWPOINT
+ INITIAL EXAMPLE
RELATIVE IMPORTANCE OF THE
REDUCED OPTIONS COSTS
-89-
MADE POSSIBLE BY THE REDUCTION IN
DEMAND UNCERTAINTY
IN UNCERTAINTY
DISCOUNTED DISCOUNTED AVERAGE
UTILITY ENERGY RETAIL
REVENUES SERVICE ELECTRIC
. costs RATE
SIMULATED IMPACT OF BENEFIT OF PENALTY OF BENEFIT OF
PERFORMANCE $2.835 $1,262 0.014
STANDARDS BILLION BILLION milis/kwh
UNDER BASE CASE
CONDITIONS
EXTRA BENEFIT FROM $0.177 $0,177 0.12
THE REDUCTION IN BILLION BILLION mills/kwh
OPTIONS COSTS
RELATIVE IMPORTANCE 6% 14% VERY
OF THE REDUCTION LARGE
