SUSTAINABILITY MCDM MODEL COMPARISONS Yuan-Sheng Lee, Tamkang University Hsu-Shih Shih, Tamkang University David L. Olson, University of Nebraska European DSI 2014, Kolding, Denmark SUSTAINABILITY Tzeng et al. [2005] Energy Policy • DECISION: select bus type from 12 choices • Eleven criteria • Our use: • Demonstration of features of various multi-criteria methods European DSI 2014, Kolding, Denmark Multi-Criteria Models of Sustainability • Non-dominated Identification • Lotov et al. [2004]; Bouchery et al. [2012] • Cardinal weighting • Equal weights; Tchebychev; Ordinal; SMART; AHP • Outranking • ELECTRE; PROMETHEE • TOPSIS (Technique for Preference by Similarity to the Ideal Solution) • Min distance to ideal while Max distance from nadir • Hwang & Yoon [1981] • TODIM • From cumulative prospect theory, S-shaped value function • Gomes & Lima [1992] European DSI 2014, Kolding, Denmark Urban Transportation Selection Decision Select a bus type – CRITERIA (Tzeng et al., 2005) • • • • • • • • • • • Energy supply Energy efficiency Air pollution Noise pollution Industrial relations Employment cost Maintenance cost Capability of vehicle Road facility Speed of traffic Sense of comfort European DSI 2014, Kolding, Denmark TODIM • Classify multiple criteria into benefits, costs • • • • • STEP 1: DM constructs normalized decision matrix (see next slide) STEP 2: Value alternatives on each criterion with 0 the worst and 1 the best STEP 3: Compute matrix of relative dominance STEP 4: Calculate global measure for each alternative STEP 5: Rank alternatives by global measures European DSI 2014, Kolding, Denmark Part 1: Bus Type Energy Supply Energy Efficiency Air Pollution Noise Pollution Industrial Relations Employ Cost A1 Diesel 0.82 0.59 0.18 0.42 0.58 0.36 A2 CNG 0.77 0.70 0.73 0.55 0.55 0.52 A3 LPG 0.79 0.70 0.73 0.55 0.55 0.52 A4 Hydrogen 0.36 0.63 0.86 0.58 0.51 0.59 A5 Methanol 0.40 0.54 0.69 0.58 0.51 0.52 A6 Elec OpC 0.69 0.76 0.89 0.60 0.72 0.80 A7 Elec Dir 0.77 0.79 0.89 0.59 0.73 0.80 !8 Elec Bat 0.77 0.79 0.89 0.59 0.73 0.80 A9 HybGas 0.77 0.63 0.63 0.52 0.66 0.63 A10 HybDies 0.77 0.63 0.51 0.58 0.66 0.63 A11 HybCNG 0.77 0.73 0.80 0.48 0.63 0.66 A12 HybLPG 0.77 0.73 0.80 0.48 0.63 0.66 European DSI 2014, Kolding, Denmark Part II Bus Type Maintenance cost Vehicle capability Roads Traffic speed Comfort A1 Diesel 0.40 0.79 0.81 0.82 0.56 A2 CNG 0.53 0.73 0.78 0.66 0.67 A3 LPG 0.53 0.73 0.78 0.66 0.67 A4 Hydrogen 0.74 0.56 0.63 0.53 0.70 A5 Methanol 0.68 0.52 0.63 0.60 0.70 A6 Elec OpC 0.72 0.54 0.35 0.79 0.73 A7 Elec Dir 0.72 0.47 0.44 0.87 0.75 A8 Elec Bat 0.72 0.51 0.48 0.87 0.75 A9 HybGas 0.65 0.67 0.70 0.80 0.74 A10 HybDies 0.65 0.67 0.70 0.80 0.74 A11 HybCNG 0.65 0.67 0.71 0.62 0.78 A12 HybLPG 0.65 0.67 0.71 0.62 0.78 European DSI 2014, Kolding, Denmark NON-DOMINANCE • • • • • • • • A1 (Diesel Bus) A3 (LPG Bus) {> A2 on energy supply, = on all others} A8 (Electric bus with exchangeable batteries) {>A7 on capability, roads} A6 (Electric bus with opportunity charging) A9 (Hybrid electric bus with gasoline engine) A10 (Hybrid electric bus with diesel engine) A11 (Hybrid electric bus with CNG engine) A12 (Hybrid electric bus with LPG engine) identical ratings to A11 • A4, A5 dominated by combinations European DSI 2014, Kolding, Denmark WEIGHTING • EQUAL WEIGHTING (LaPlace) • A8 Electric bus with exchange batteries wins • A7 a very close second • PROVIDES FULL RANKING • Uses cardinal (continuous?) numbers • TCHEBYCHEV WEIGHTS • Maximize worst rating – A2 (CNG – dominated by A3), A3(LPG), A9 (Hybrid) • ORDINAL WEIGHTS (centroid) • A8 Electric bus with exchange batteries wins • A7 a very close second • CARDINAL WEIGHTS (from Tzeng et al. - AHP) • A8 Electric bus with exchange batteries wins • A7 a very close second European DSI 2014, Kolding, Denmark Simulation Bus Type (nondominated) Proportion Won A1 Diesel 0.005 A3 LPG 0.110 A6 Electric optional charging A8 Electric battery 0.625 A9 Hybrid gas 0.110 A10 Hybrid diesel 0.045 A11 Hybrid CNG or LPG 0.205 European DSI 2014, Kolding, Denmark 0 PROMETHEE European DSI 2014, Kolding, Denmark Distance methods • TOPSIS • A8 Electric exchange batteries • A6 Electric optional charge close behind • A7 Electric direct exchange (dominated solution) close behind • TODIM • A8 Electric exchange batteries • A7 Electric direct exchange (dominated solution) second • A11/A12 Hybrid CNG or LPG third European DSI 2014, Kolding, Denmark Rankings Bus Type A1 Diesel = wgt Tcheb centroid AHP PROM TOPSIS TODIM 10 12 11 11 10 12 12 A2 CNG 8 2- 8 8 9 10 9 A3 LPG 6.5 2 7 7 8 11 8 A4 Hydrogen 11 10 10 10 11 8 10 A5 Methanol 12 9 12 12 12 9 11 A6 Elec OpC 6.5 11 3 3 3 2 6 A7 Elec Dir 2 8 2 2 2 3 2 A8 Elec Bat 1 6 1 1 1 1 1 A9 HybGas 5 2 4 6 7 4 7 A10 HybDies 9 4 9 9 6 7 5 A11 HybCNG 3.5 6 5.5 4.5 4.5 5.5 3.5 A12 HybLPG 3.5 6 5.5 4.5 4.5 5.5 3.5 European DSI 2014, Kolding, Denmark SELECTION Bus Type Dominance Simulation A1 Diesel A2 CNG A3 LPG A4 Hydrogen Dominated A5 Methanol Dominated A6 Electric optional charge A7 Electical direct A8 Electrical battery 0.625 A9 Hybrid gas 0.010 A10 Hybrid diesel 0.045 A11 Hybrid CNG 0.205 A12 Hybrid LPB A8 picked A3 picked A9 picked 0.005 Dominated 0.110 Tchebychef 0 Dominated All others Duplicate European DSI 2014, Kolding, Denmark Tchebychef DISCUSSION • Fair consistency in rankings • No two identical • Continuous allows close second to be ranked even if dominated (A7) • Tchebychef the most extreme • Only looks at worst • Thus is sensitive to scale • A2 considered, though dominated European DSI 2014, Kolding, Denmark CONCLUSIONS • Many multiple criteria methods • All valuable to some degree • more • SIMULATION preferred by author 1. Nondominance might be useful in selection, not in ranking • You can always come up with another criterion 2. Accuracy of data critical • A11/A12 identical, but might vary on some additional factor 3. Outranking methods help explore 4. PREFERENCE important • Machine-methods {omit preference as much as possible} (TOPSIS) • Individual preference well-studied • Group preference problematic European DSI 2014, Kolding, Denmark