MULTI-ATTRIBUTE DECISION ANALYSIS: AN ILLUSTRATION
• Overall focus: environmental impacts
• Decision environment
- Options (Coal, nuclear)
- Impacts
- Conflicting objectives
- Uncertainty
- Several decision makers
- -
EVALUATION OF STRATEGIES FOR ELECTRICAL ENERGY
PRODUCTION IN WISCONSIN
(Ref: Buehring, Foell and Keeney 1978; Watson and Beude 1987)
MADAALTERNATIVE POLICIES FOR WISCONSIN ELECTRICAL GENERATION: 1970 TO 2000
Policy 1
• • • •
Electrical generation increase at an average annual growth rate of 4.7%.
Almost all new plants are fuelled with coal.
SO2 emissions controlled by using stack gas removal systems and low-sulfur coal.
99% particular control.
Policy 2
• • • •
Electrical generation – same as Policy 1.
Almost all new plants are fuelled with coal.
No SO2 stack gas removal systems and same amount of low-sulfur coal as policy 1.
89% particulate control.
Policy 3
• • •
Electrical generation – same as Policy 1.
Almost all new plants are nuclear.
Emission controls for SO2 and particulate same as Policy 1.
Policy 4
• • • • •
Electrical generation – same as Policy 1.
After 1975, all coal is low-sulfur from distant mines in Western states.
New capacity after 1982 is 50% coal and 50 % nuclear.
After 1975, all coal and uranium are obtained from surface mines.
Emission controls for SO2 and particulate same as Policy 1.
Policy 5
• • •
Electrical generation increases at an average annual growth rate of 2.8%.
Almost all new plants are fuelled with coal.
Emission controls for SO2 and particulate same as Policy 1.
Policy 6
• • •
Electrical generation – same as Policy 5.
Almost all new plants are nuclear.
Emission controls for SO2 and particulate same as Policy 1.ATTRIBUTES FOR MULTIATTRIBUTE DECISION
ANALYSIS
ATTRIBUTE UNIT
1. Total quantified fatalities
2. Permanent land use
3. Temporary land use
4. Water evaporated
5. SO2 emissions
6. Particulate emissions
7. Thermal energy needed
8. Radioactive waste
9. Nuclear safeguards
10. Health effects of chronic air pollution
11. Electricity generated
Deaths
Acres
Acres
10¹² Gallons
10⁶ Tons
10⁶ Tons
10¹² KWh (t)
Metric tons
Tons puf produced
Tons lead emitted
10¹² Kwh (e)CUMULATIVE IMPACTS OF ALTERNATIVE PROLICIES
ATTRIBUTES Policy 1 Policy 2 Policy 3 Policy 4 Policy 5 Policy 6
1. Fatalities (deaths) 380 380 240 680 280 210
2. Permanent Land Use (acres) 420 420 1,100 770 380 730
3. Temporary Land Use (acres) 140,000 137,000 85,000 43,000 99,000 71,000
4. Water Evaporated (10¹²gallons) 0.72 0.71 0.84 0.79 0.55 0.61
5. SO2 Pollution (10⁶tons) 12 23 8.0 8.6 9.5 7.4
6. Particulate Pollution (10⁶tons) 0.69 6.2 0.40 0.56 0.51 0.37
7. Energy Needed (10¹²kWh) 4.9 4.9 5.2 5.3 3.8 3.9
8. Radioactive Waste (MT fissile
products)
61 61 160 110 54 105
9. Nuclear Safeguards (tons fissile
plutonium)
11 11 30 21 10 19
10. Chronic Health Effects (tons lead
emitted)
124 1,110 71 100 92 66
11. Electricity Generated
(10¹²kWh nuclear)
(10¹²kWh coal)
(10¹²kWh coal + nuclear)
0.36
1.37
1.73
0.36
1.37
1.73
0.99
0.74
1.73
0.68
1.05
1.73
0.33
0.99
1.32
0.64
0.62
1.32
Polices
1. Mostly coal, SO2 removal, 99% particulate control.
2. Mostly coal, SO2 removal, 89% particulate control.
3. Mostly nuclear.
4. Mostly western coal, more coal than nuclear.
5. Mostly coal, less energy.
6. Mostly nuclear, less energy.MADA: MAIN STEPS
• Introducing terminology & ideas
• Determining general preference structure
- preference independence
- utility independence
• Assessing single attribute utility functions
• Evaluating scaling constants
• Evaluating alternativesDETERMINING GENERAL PREFERENCE STRUCTURE
• Preferential independence
Preferences for consequences involving any two
attributes taken at a time do not depend on the levels at
which the remaining attribute are fixed
Decision maker’s trade-off between these two attributes
do not depend on the levels at which other attributes are
fixed (example: next page)• Utility independence
DETERMINING GENERAL PREFERENCE STRUCTURE (CONTD.)
UTILITY INDEPENDENCE FOR AN ATTRIBUTE
Preference order for lotteries involving only changes in the level of that
particular attribute does not depend on the levels at which other
attributes are held.Selected Single-Attribute Utility Functions for Two
Decision Makers
A
B
A
B
A
B
A
B
A
B
Individuals
and
A
B
1 U1
U
9
U
11
U
10
U
3
U
1
680 X
1 210
X
11
X
X2 10
X
3
X
9
1,110
1.32
1,100
140
30
1.73
66
10
380
43
(Fatalities) (Tons Plutonium)
(1012 KWH Electricity)
(Acres) (Tons Lead)
(103 Acres)
0
0 0
0
0
0
1 1 1
1 1Utility Function for Decision-Maker A
Attribute Conditional Utility Function (Ui)
X
1 (700 - X1)) / 600
X
2 (1- exp (- .004(2000-X2))) / 0.550
X
3 (1- exp (-3.5x 10 6 (200,000-X2))) / 0.486
X
4 (1- exp (- .825 (1.5-X4))) / 0.562
X
5 (1- exp (- .011 (80-X5))) / 0.562
X
6 (10 – X6) / 9.8
X
7 (10 – X7) / 3
X
8 (200– X8) / 199
X
9 (exp ( .12(50-X9)) – 1) / 357
X
10 (1- exp (- .0005(2000-X10))) / 0.632
X
11 (X11-0.5)0.2(1-(X11-1.5)2/2.5)/(1+(X11-1.5)2)0.5Utility Function for Decision-Maker B
Attribute Conditional Utility Function (Ui)
X
1 (700 - X1)) / 600
X
2 (2000-X2)/ 1999
X
3 (1- exp (-5.8x 10 6 (200,000-X2))) / 0.668
X
4 (1- exp (- .825 (1.5-X4))) / 0.562
X
5 (1- exp (- .011 (80-X5))) / 0.562
X
6 (10 – X6) / 9.8
X
7 (1- exp (- .1 (6-X7))) / 0.259
X
8 (200– X8) / 199
X
9 (1- exp ( -.035(50-X9))) / 0.820
X
10 (1- exp (- .0002(2000-X10))) / 0.330
X
11 (1- exp (- 1.31(X11 -0.5))) / 0.962EVALUATING SCALING CONSTANTS
TWO STEP PROCESS
a) Ranking in order of importance (for specified ranges)
b) Quantifying each scaling constant
- Develop relationship between various scaling constants
- Determine value of a specific scaling constant
- Calculate remaining scaling constants
Decision maker is indifferent
between the consequences
represented by point A&DRelationship Between Various Scaling Constants:
Decision-Maker B
K
i Relationship to K10
k
10
K
1 K1 = k10u10 (500)
K
9 K9 = k10u10 (1,200)
K
5 K5 = k10u10 (1,700)
K
8 K8 = k5u5 (10) = k10u10 (1,110)u5 (10)
K
11 K11= k5u5 (20) = k10u10 (1,700)u5 (20)
K
3 K3= k5u5 (60) = k10u10 (1,700)u5 (60)
K
2 K2= k3u3 (50) = k10u10 (1,700)u5 (60)u3 (50)
K
4 K4= k3u3 (75) = k10u10 (1,700)u5 (60)u3 (75)
K
7 K7= k3u3 (125) = k10u10 (1,700)u5 (60)u3 (125)
K
6 K6= k3u3 (150) = k10u10 (1,700)u5 (60)ASSESSING THE INDIFFERENCE PROBABILITY FOR DECISION
MAKER BSolving for Scaling Constants: DecisionMaker B
K
i Ki equals Value of Ki
k
10 0.339
K
1 0.768 K10 0.266
K
9 0.449 K10 0.152
K
5 0.177 K10 0.060
K
8 0.169 K10 0.057
K
11 0.152 K10 0.051
K
3 0.062 K10 0.021
K
2 0.054 K10 0.018
K
4 0.048 K10 0.016
K
7 0.033 K10 0.011
K
6 0.023 K10 0.008
Σ= 2.953 K
10 Σ=1.0
Note: Solving Σ Ki = 2.953 k10 =1 yields k10 = 0.339, from which other kis are
evaluated.Scaling Constants
Attribute Individual A Individual B
X
1 = total quantified fatalities 0.354 0.267
X
2 = permanent land use 0.004 0.018
X
3 = temporary land use 0.033 0.021
X
4 =water evaporated 0.083 0.016
X
5 = SO2 pollution 0.008 0.060
X
6 = particulate pollution 0.008 0.008
X
7 = thermal energy needed 0.017 0.011
X
8 = radioactive waste 0.132 0.057
X
9 = nuclear safeguards 0.177 0.052
X
10 = health effects of chronic air pollution exposure 0.118 0.339
X
11 = electricity generated 0.066 0.051
Σ=1.0 Σ=1.0Utilities for the Six Scenarios
Individual A Individual B
1. Mostly coal, SO2 removal, 99% particulate
control
0.616 0.798
2. Mostly coal, no SO2 removal, 89% particulate
control
0.561 0.632
3. Mostly nuclear 0.588 0.793
4. More coal than nuclear, mostly western coal 0.379 0.638
5. Mostly coal, less energy 0.712 0.856
6. Mostly nuclear, less energy 0.680 0.857Relationship Between Various Scaling Constants:
Decision-Maker B
K
i Relationship to K10
k
10
K
1 K1 = k10u10 (500)
K
9 K9 = k10u10 (1,200)
K
5 K5 = k10u10 (1,700)
K
8 K8 = k5u5 (10) = k10u10 (1,110)u5 (10)
K
11 K11= k5u5 (20) = k10u10 (1,700)u5 (20)
K
3 K3= k5u5 (60) = k10u10 (1,700)u5 (60)
K
2 K2= k3u3 (50) = k10u10 (1,700)u5 (60)u3 (50)
K
4 K4= k3u3 (75) = k10u10 (1,700)u5 (60)u3 (75)
K
7 K7= k3u3 (125) = k10u10 (1,700)u5 (60)u3 (125)
K
6 K6= k3u3 (150) = k10u10 (1,700)u5 (60)CUMULATIVE IMPACTS OF ALTERNATIVE PROLICIES
ATTRIBUTES Policy 1 Policy 2 Policy 3 Policy 4 Policy 5 Policy 6
1. Fatalities (deaths) 380 380 240 680 280 210
1. Permanent Land Use (acres) 420 420 1,100 770 380 730
1. Temporary Land Use (acres) 140,000 137,000 85,000 43,000 99,000 71,000
1. Water Evaporated (10¹²gallons) 0.72 0.71 0.84 0.79 0.55 0.61
1. SO2 Pollution (10⁶tons) 12 23 8.0 8.6 9.5 7.4
1. Particulate Pollution (10⁶tons) 0.69 6.2 0.40 0.56 0.51 0.37
1. Energy Needed (10¹²kWh) 4.9 4.9 5.2 5.3 3.8 3.9
1. Radioactive Waste (MT fissile
products)
61 61 160 110 54 105
1. Nuclear Safeguards (tons fissile
plutonium)
11 11 30 21 10 19
1. Chronic Health Effects (tons lead
emitted)
124 1,110 71 100 92 66
1. Electricity Generated
(10¹²kWh nuclear)
(10¹²kWh coal)
(10¹²kWh coal + nuclear)
0.36
1.37
1.73
0.36
1.37
1.73
0.99
0.74
1.73
0.68
1.05
1.73
0.33
0.99
1.32
0.64
0.62
1.32
Expected Utilities:
Individual A 0.620 0.569 0.589 0.383 0.711 0.680
Individual B 0.789 0.624 0.785 0.631 0.846 0.849