A Model Approach toward the Management of Future Uncertainties - - PowerPoint PPT Presentation

▶
a model approach toward the management of future
SMART_READER_LITE
LIVE PREVIEW

A Model Approach toward the Management of Future Uncertainties - - PowerPoint PPT Presentation

Jan 13, 2023 175 likes •421 views

A Model Approach toward the Management of Future Uncertainties Shunsuke Mori (Tokyo University of Science) ICA-RUS International Symposium, Dec.4-6, 2013, Tokyo 1. Introduction Various uncertainties in the climate decision making ->

slide-1
SLIDE 1

A Model Approach toward the Management of Future Uncertainties

Shunsuke Mori (Tokyo University of Science)

ICA-RUS International Symposium, Dec.4-6, 2013, Tokyo

slide-2
SLIDE 2
  • 1. Introduction

Various “uncertainties” in the climate decision making

  • > key barrier against policy agreement
  • Impacts of climate changes
  • Uneven societal distribution of cost and benefits
  • Deployment strategy of technology options, etc.

Existing method – basically focusing on expected utility, however

  • “Extreme (tipping) Event” : low probability and high risk
  • Long tail risk

Shutdown of thermohaline circulation (THC), collapse of west antarctic ice sheet, the collapse of Greenland ice sheet, methane outburst, increase of hurricane and cyclones, etc.

Dec ecision bas n based on ed on max aximum ex expec pected ut d utility ( (MxU xU) w woul

  • uld

d not not be be pr pref eferab able.

  • > A

Alter ernat native m e metho hod d : minimum um regret et strat ategy egy ( (MnR nR)

slide-3
SLIDE 3

“Extreme (tipping) Event” : low probability and high risk

  • Nuclear power : high technological potential but low societal

acceptance, at least in Japan. After March 11, 2012, Gigantic earthquake followed by nuclear station accident, people in Japan seriously consider the “unexpected

  • utcomes”.
  • CCS and geo-engineering : large possibility to mitigate the

global warming, but regrettable when warming damage is low.

  • Other possible tipping events: -

shutdown of thermohaline circulation (THC), collapse of west antarctic ice sheet, the collapse of Greenland ice sheet, methane outburst, increase of hurricane and cyclones, etc.

Dec ecision bas n based on ed on max aximum ex expec pected ut d utility ( (MxU xU) w woul

  • uld

d not not be be appl applicab able.

  • > A

Alter ernat native m e metho hod d : minimum um regret et strat ategy egy ( (MnR nR)

slide-4
SLIDE 4

Structure of the Integrated Research on the Development of Global Climate Risk Management Strategies Project

Theme 4: (Leader, S.Mori) Technology Development and Implementation Strategies under Uncertainties:

  • Potential contribution and constraints?
  • Potential cost?
  • Potential trade-offs and synergies among options?

→ Need for the comprehensive quantitative analysis

slide-5
SLIDE 5
  • 2. Maximum Expected Utility vs. Minimum Regret - example

Optimum fuel mix of P1(Kerosene): low expected price but large uncertainty P2(Coal based DME): high expected price but small uncertainty

Option A: Low expected cost but high uncertainty Option B: High expected cost but low uncertainty

Expected cost strategy always chooses Option A only.

  • > how about risk? Mixed-strategy focusing on regret

Option A is better than Option B in many cases. However, Option B can be better than Option A with low probability.

Realized cost

  • f Option B

Realized cost

  • f Option A

Realized cost

  • f Option B

Realized cost

  • f Option A

Realized cost

  • f Option B

Realized cost

  • f Option A

Realized cost

  • f Option A

Realized cost

  • f Option B

Mixed strategy =α×Option A + (1-α)×Option B Regret of Option A Regret of

mixed-strategy

slide-6
SLIDE 6
  • 2. Maximum Expected Utility vs. Minimum Regret – example-2

Optimum fuel mix of P1(Kerosene): low expected price but large uncertainty P2(Coal based DME): high expected price but small uncertainty

( )

) ; t ( P ) t ( 1 ) ; t ( P ) t ( ) ; t ( P

2 1 *

ω α − + ω α = ω

20 40 60 80 100 120 140 160 180

Jan-09 Jan-10 Jan-11 Jan-12 Jan-13 Jan-14 Jan-15 Jan-16 Jan-17 Jan-18 Jan-19 Jan-20 Jan-21 Jan-22 Jan-23 Jan-24 Jan-25 Jan-26 Jan-27 Jan-28 Jan-29

10 20 30 40 50 60 70 80

Jan-09 Jan-10 Jan-11 Jan-12 Jan-13 Jan-14 Jan-15 Jan-16 Jan-17 Jan-18 Jan-19 Jan-20 Jan-21 Jan-22 Jan-23 Jan-24 Jan-25 Jan-26 Jan-27 Jan-28 Jan-29

Figure 1 Kerosine price in ¥/ktoe Figure 2 DME price in ¥/ktoe Optimum fuel mix

Kerosene is always fully selected when the expected value is employed.

slide-7
SLIDE 7
  • 2. Maximum Expected Utility vs. Minimum Regret – cont.

Conventional minimizing regret of strategy α*(t)

{ }

) n ; t ( P ) n ; t ( P ), n ; t ( P ) n ; t ( P max ) n | ) t ( ( gret Re

2 * 1 *

− − = α

) n , t ( LO _ P ) n , t ( UP _ P ) n ; t ( P ) n ; t ( P ) n , t ( LO _ P ) n , t ( UP _ P ) n ; t ( P ) n ; t ( P

2 2 2 * 1 1 1 *

− = − − = −

Alternative formulation

{ }

∑ ∑

θ θ θ + t / 1 2 1 n

) n ; t ( UP _ P ) n ; t ( UP _ P ) n ( w . min

∞ → θ

the above converges to the min-max strategy.

→ Minkowski generalized distance

{ }

) n | ) t ( ( gret Re . max min ) t ( *

n

α = α

  • Normal distribution → theoretical “maximum value” ?
  • Strong correlation among variables → theoretical distribution ?
  • “Spaghetti” simulation results → 95% confidence interval ?

Problems

slide-8
SLIDE 8

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Jan-09 Jan-10 Jan-11 Jan-12 Jan-13 Jan-14 Jan-15 Jan-16 Jan-17 Jan-18 Jan-19 Jan-20 Jan-21 Jan-22 Jan-23 Jan-24 Jan-25 Jan-26 Jan-27 Jan-28 Jan-29 θ=1 θ=2 θ=4 θ=5 θ=6 θ=7 θ=8

Figure 3 Optimal fuel mix weights α*(t) corresponding to the θ changes

→ Share of synthetic fuel DME increases with larger θ while no DME was employed under MxU strategy.

slide-9
SLIDE 9
  • 3. Extension of MARIA for the regret based assessment

MARIA (Multiregional Approach for Resource and Industry Allocation)

  • an inter-temporal optimization model integrating top-down macroeconomic

activity and bottom-up technology flows

Capital Stock Thermal energy Electricity Labor Biomass Hydropower Geothermal Solar power Wind power Nuclear

  • LWR
  • LWR
  • Pu
  • FBR

Economic Activity Energy Supply Oil Coal Gas Bern carbon cycle model Ocean

  • Atmosphere

heat exchange Other GHG CO2 Climate Change Food and Feed demand Land

  • use Changes

Potential Cropland Crop production Land

  • use Change

GDP Consumption Trade Investment Forest area Max.

Figure 4 Structure of MARIA model

slide-10
SLIDE 10

3-2. Future Uncertain Scenarios

Uncertainty: 12 scenarios on loss of GDP when the global atmospheric temperature rises 3.0 Celsius degree from the pre-industry level

Scenario-1 0.6% - 0.9% Scenario-2 1.2% - 1.8% <- reference case Scenario-3 1.8% - 2.7% Scenario-4 2.4% - 3.6% Scenario-5 3.0% - 4.5% Scenario-6 3.6% - 5.4% Scenario-7 4.2% - 6.3% Scenario-8 4.8% - 7.2% Scenario-9 5.4% - 8.1% Scenario-10 6.0% - 9.0% Scenario-11 12% - 18% Scenario-12 18% - 27%

→ Conventional Min.-Max regret strategy refers only Scenario-1 and Scenario-12, two extreme cases.

slide-11
SLIDE 11
  • 4. Formulation – conventional Min-max approach

Objective function f(SCN) : discounted present value of utility under scenario SCN. Optimum solution under SCN is f*(SCN).

∑ ∑

        + =

− t t , h t , h t , h t h *

L ) SCN ( C ln L ) r 1 ( . max ) SCN ( f

X*(SCN) -- Optimal solution of control variables under scenario SCN

Regret of strategy X*(SCN) under the realized scenario scn’ is

( )

) SCN ; t ( X ) ' scn ; t ( X ; ' scn f max ) ' scn | SCN ( f

* X *

= =

) ' scn | SCN ( f ) ' scn ( f ) ' scn | SCN ( gret Re

* *

− = ∴

  • Min-Max regret solution is basically determined by the extreme

assumption regardless of the plausibility.

  • Optimal “policy mix” cannot be generated.
slide-12
SLIDE 12

4-2. Formulation – generalized distance approach

( )

( ) ( ) ( )

SCN | ) t ( X DN _ D SCN | ) t ( X UP _ D ) t ( X f ) SCN ; t ( X f ] SCN | ) t ( X [ gret Re

* * *

− = − =

( )

{ }

θ θ

∑

× − ×

/ 1 SCN t ) t ( X

SCN | ) t ( X DN _ D ) d 1 ( ) SCN ( P min P(SCN) denotes occurrence probability of scenario SCN

Expansion – ATL multi-stage decision approach

T t for ) t ( X ) t , m ( X ≤ =

m: future bifurcation possibilities

( ) ( )

) SCN | t ( X f d 1 ) SCN ( P max

t t SCN

∑ ∑

−

(single stage decision)

( ) ( )

{ }

θ θ / 1 ) , (

| ) , ( _ 1 ) , ( min

∑ ∑ ∑

× −

SCN t t m t m X

SCN t m X DN D d m SCN P (single stage decision) (multi stage decision)

Existing MxU formulation

( ) ( )

) SCN | t , m ( X f d 1 ) m , SCN ( P max

t t m SCN

∑ ∑ ∑

−

(multi stage decision)

slide-13
SLIDE 13

4-3. Decision Strategy under Future Uncertainties

A: Three decision stages Single stage decision: under 12 scenarios (SS1) Two stage decision: 3 decisions under 12 scenarios (SS2) Three stage decision 3 decisions then 12 decisions (SS3) L M H

SS3-1 SS3-12

B: Two decision basis MxU: Maximizing expected utility MnR: Minizing regret in generalized distance

slide-14
SLIDE 14
  • 5. Simulation Results

5-1 perfect information

  • In the perfect information cases, carbon control strategies bifurcate broadly.
  • Under future uncertainties, how the policy maker(s) can select single emission

path?

Figure 5 GDP without uncertainty Figure 6 CO2 emission without uncertainty in trillion US dollars in Gt-C (L-m and L-h overlap each other.)

50 100 150 200 250 1 2 3 4 5 6 7 8 9 10 11 12 5 10 15 20 25 30 1 2 3 4 5 6 7 8 9 10 11 12

slide-15
SLIDE 15
  • 5. Simulation Results

Figure 7 CCS implementation without uncertainty Figure 8 Nuclear power implementation in Gt-C without uncertainty in GTOE

5-1 perfect information -2

  • Optimal CCS implementation also distributes broadly.
  • When and how much CCS should be implemented under non-repeatble and

irreversible situation?

0.5 1 1.5 2 2.5 3 3.5 1 2 3 4 5 6 7 8 9 10 11 12 2 4 6 8 10 12 1 2 3 4 5 6 7 8 9 10 11 12

slide-16
SLIDE 16
  • 5. Simulation Results

Figure 7 Atmospheric temperature rise Figure 8 Biomass power implementation in degree in GTOE

5-1 perfect information -3

  • Biomass energy increases earlier as the climate damage costs increase due to the

high costs assumptions.

0.5 1 1.5 2 2.5 3 3.5 1 2 3 4 5 6 7 8 9 10 11 12 0.5 1 1.5 2 2.5 1 2 3 4 5 6 7 8 9 10 11 12

slide-17
SLIDE 17
  • 5. Simulation Results

5-2 Decision making under uncertainty: MnR vs.MxU and Single stage vs. Multi stage

Figure 10 CCS implementation paths of the maximum expected utility (MxU) and minimum regret (MnR) in the single stage decision (SS1)and the two stage decision (SS2) Figure 11 CCS implementation paths of the maximum expected utility (MxU) and minimum regret (MnR) in the three stage ecision (SS3)

  • CCS implementation pathways appear differenmtly between MxU and MnR.
  • CCS is implemented moderately comparing with perfect information cases.

0.5 1 1.5 2 2.5 3 3.5

1 2 3 4 5 6 7 8 9 10 11 12 SS1_MxU SS1_MnR SS2_MxU_L SS2_MxU_M SS2_MxU_H SS2_MnR_L SS2_MnR_M SS2_MnR_H 0.5 1 1.5 2 2.5 3 3.5 19972007201720272037204720572067207720872097 1 2 3 4 5 6 7 8 9 10 11 12 SS3_MxU_1 SS3_MxU_4 SS3_MxU_5 SS3_MxU_8 SS3_MxU_9 SS3_MxU_12 0.5 1 1.5 2 2.5 3 3.5 19972007201720272037204720572067207720872097 1 2 3 4 5 6 7 8 9 10 11 12 SS3_MnR_1 SS3_MnR_4 SS3_MnR_5 SS3_MnR_8 SS3_MnR_9 SS3_MnR_12

slide-18
SLIDE 18
  • 5. Simulation Results

5-2 Decision making under uncertainty: -2 MnR vs.MxU and Single stage vs. Multi stage

Figure 12 Nuclear implementation paths of the maximum expected utility (MxU) and minimum regret (MnR) in the single stage decision (SS1)and the two stage decision (SS2) Figure 13 Nuclear implementation paths

  • f the maximum expected utility (MxU)

and minimum regret (MnR) in the three stage ecision (SS3)

  • Nuclear power pathways are not so different between MxU and MnR.
  • Uncertainty consideration tends to implement nuclear power.

2 4 6 8 10 12

1 2 3 4 5 6 7 8 9 10 11 12 SS1_MxU SS1_MnR SS2_MxU_L SS2_MxU_M SS2_MxU_H SS2_MnR_L SS2_MnR_M SS2_MnR_H 2 4 6 8 10 12

1 2 3 4 5 6 7 8 9 10 11 12 SS3_MxU_1 SS3_MxU_4 SS3_MxU_5 SS3_MxU_8 SS3_MxU_9 SS3_MxU_12

2 4 6 8 10 12

1 2 3 4 5 6 7 8 9 10 11 12 SS3_MnR_1 SS3_MnR_4 SS3_MnR_5 SS3_MnR_8 SS3_MnR_9 SS3_MnR_12

slide-19
SLIDE 19
  • 5. Simulation Results

5-2 Decision making under uncertainty: -3 MnR vs.MxU and Single stage vs. Multi stage

Figure 14 CO2 emission paths

  • f

the maximum expected utility (MxU) and minimum regret (MnR) in the single stage decision (SS1)and the two stage decision (SS2) Figure 15 CO2 emission paths of the maximum expected utility (MxU) and minimum regret (MnR) in the three stage ecision (SS3)

CO2 emission pathways in MnR are apparently lower than MxU cases.

5 10 15 20 25 30

1 2 3 4 5 6 7 8 9 10 11 12 SS1_MxU SS1_MnR SS2_MxU_L SS2_MxU_M SS2_MxU_H SS2_MnR_L SS2_MnR_M SS2_MnR_H

5 10 15 20 25 30

1 2 3 4 5 6 7 8 9 10 11 12 SS3_MxU_1 SS3_MxU_4 SS3_MxU_5 SS3_MxU_8 SS3_MxU_9 SS3_MxU_12

5 10 15 20 25 30

1 2 3 4 5 6 7 8 9 10 11 12 SS3_MnR_1 SS3_MnR_4 SS3_MnR_5 SS3_MnR_8 SS3_MnR_9 SS3_MnR_12

slide-20
SLIDE 20
  • 5. Simulation Results

5-2 Decision making under uncertainty: -4 MnR vs.MxU and Single stage vs. Multi stage

Figure 16 Biomass energy paths of the maximum expected utility (MxU) and minimum regret (MnR) in the single stage decision (SS1)and the two stage decision (SS2) Figure 17 Biomass energy

  • f

the maximum expected utility (MxU) and minimum regret (MnR) in the three stage ecision (SS3)

Biomass implementation in MnR are apparently larger than MxU cases.

0.5 1 1.5 2 2.5

1 2 3 4 5 6 7 8 9 10 11 12 SS1_MxU SS1_MnR SS2_MxU_L SS2_MxU_M SS2_MxU_H SS2_MnR_L SS2_MnR_M SS2_MnR_H

0.5 1 1.5 2 2.5

1 2 3 4 5 6 7 8 9 10 11 12 SS3_MxU_1 SS3_MxU_4 SS3_MxU_5 SS3_MxU_8 SS3_MxU_9 SS3_MxU_12

0.5 1 1.5 2 2.5

1 2 3 4 5 6 7 8 9 10 11 12 SS3_MnR_1 SS3_MnR_4 SS3_MnR_5 SS3_MnR_8 SS3_MnR_9 SS3_MnR_12

slide-21
SLIDE 21
  • 5. Simulation Results

5-2 Decision making under uncertainty: -5 MnR vs.MxU and Single stage vs. Multi stage

Figure 18 Atmospheric temperature paths of the maximum expected utility (MxU) and minimum regret (MnR) in the single stage decision (SS1)and the two stage decision (SS2) Figure 19 Atmospheric temperature of the maximum expected utility (MxU) and minimum regret (MnR) in the three stage ecision (SS3)

Atmospheric temperature in MnR are apparently lower than MxU cases.

0.5 1 1.5 2 2.5 3 3.5

1 2 3 4 5 6 7 8 9 10 11 12 SS1_MxU SS1_MnR SS2_MxU_L SS2_MxU_M SS2_MxU_H SS2_MnR_L SS2_MnR_M SS2_MnR_H

0.5 1 1.5 2 2.5 3 3.5

1 2 3 4 5 6 7 8 9 10 11 12 SS3_MxU_1 SS3_MxU_4 SS3_MxU_5 SS3_MxU_8 SS3_MxU_9 SS3_MxU_12

0.5 1 1.5 2 2.5 3 3.5

1 2 3 4 5 6 7 8 9 10 11 12 SS3_MnR_1 SS3_MnR_4 SS3_MnR_5 SS3_MnR_8 SS3_MnR_9 SS3_MnR_12

slide-22
SLIDE 22
  • 5. Simulation Results

5-3 Decision making under uncertainty: MnR vs.MxU in single stage case

Figure 20 Loss of GDP from the perfect information case in the maximum utility (MxU) of the single stage decision

GDP in MnR tends to increase under the carbon control case. The difference of the property of MxU and MnR appears in consumption figure.

Figure 21 Loss of GDP from the perfect information case in the minimum regret (MnR) of the single stage decision

  • 3.000%
  • 2.000%
  • 1.000%

0.000% 1.000% 2.000% 3.000% 4.000% 1 2 3 4 5 6 7 8 9 10 11 12

  • 3.000%
  • 2.500%
  • 2.000%
  • 1.500%
  • 1.000%
  • 0.500%

0.000% 0.500% 1.000% 1 2 3 4 5 6 7 8 9 10 11 12

slide-23
SLIDE 23
  • 5. Simulation Results

5-3 Decision making under uncertainty: MnR vs.MxU in single stage case

Figure 22 Loss of consumption from the perfect information case in the maximum utility (MxU) of the single stage decision Figure 23 Loss of consumption from the perfect information case in the minimum regret (MnR) of the single stage decision

Loss of consumption in MnR is higher than that in MxU in the early stage. This comparison suggests that the investment and the capital stock in MnR strategy are larger than those in MxU strategy.

  • 1.500%
  • 1.000%
  • 0.500%

0.000% 0.500% 1.000% 1.500% 2.000% 2.500% 3.000% 3.500% 4.000% 1 2 3 4 5 6 7 8 9 10 11 12

  • 1.500%
  • 1.000%
  • 0.500%

0.000% 0.500% 1.000% 1.500% 2.000% 1 2 3 4 5 6 7 8 9 10 11 12

slide-24
SLIDE 24
  • 6. Conclusion
  • A new method to deal with the future uncertainties focusing
  • n the “regret” values.
  • When we consider the “long-tail” distribution, decision

making based on “expected utility” would underestimate the extreme case, while exaggerated “risk aversion” strategy will derive policy depending on the extreme assumptions regardless of the plausibility.

  • The minimum regret policy tends to prefer lower carbon

emission paths.

  • The approach described in this study will be useful in the

irreversible, unrepeatable and asymmetric uncertainty cases.