Assessment of Plug-in Electric Vehicles Charging on Distribution - PowerPoint PPT Presentation

Assessment of Plug-in Electric Vehicles Charging on Distribution Networks Master Thesis Defense - Tsz Kin (Marco) Au Committee Chair: Dr. M. Ortega-Vazquez Committee Co-Chair: Dr. M. El-Sharkawi Committee Member: Dr. D. Kirschen

Presentation Outline I. Introduction of PEV II. The developed tool for investigating the impact of PEV III. Test system characteristic IV. Test result V. Conclusion Electrical Engineering Department - University of Washington 6/5/2012 2 Master Thesis Defense - Tsz Kin (Marco) Au

Presentation Outline I. Introduction of PEV II. The developed tool for investigating the impact of PEV III. Test system characteristic IV. Test result V. Conclusion Electrical Engineering Department - University of Washington 6/5/2012 3 Master Thesis Defense - Tsz Kin (Marco) Au

Technological Impacts of PEVs Potential benefits: • Lower operating cost than combustion engine vehicles: 3.7 vs. 16.7 cents • On road CO2 emission will be lower • V2G and ancillary services provide business opportunities Problems: • 10% penetration = additional 300 GWh per day in the U.S. • Increase grid losses • Reduce system spare generation and harder to schedule maintenance • Poorer voltage profile and transformer overloading in weakly meshed distribution networks Electrical Engineering Department - University of Washington 6/5/2012 4 Master Thesis Defense - Tsz Kin (Marco) Au

What causes poor voltage profile and transformer overloading? Line impedance Poor voltage profile and Coincidence between PEV charge time and system peak load overload transformer Lack of interconnection Electrical Engineering Department - University of Washington 6/5/2012 5 Master Thesis Defense - Tsz Kin (Marco) Au

Presentation Outline I. Introduction of PEV II. The developed tool for investigating the impact of PEV III. Test system characteristic IV. Test result V. Conclusion Electrical Engineering Department - University of Washington 6/5/2012 6 Master Thesis Defense - Tsz Kin (Marco) Au

Monte Carlo Simulation • Suitable for analysis when Read data and uncertainties present initialize parameters • 4 uncertainties needed to be address: – Charging time Generate random – Battery state of charge (SOC) scenarios – Charging method – Customer load variation Run deterministic • 7 major functional blocks system • Each trail represent 24 hours Electrical Engineering Department - University of Washington 6/5/2012 7 Master Thesis Defense - Tsz Kin (Marco) Au

1. Data Processing and Initialization • 34,000+ drivers’ behavior from CMAP, which consists of their to -work and to-home arrival times. • Electric vehicle parameters – Battery capacity – Energy consumption per unit distance • Distribution network conductor parameters • Average power consumption and load type at each node – Residential area – Commercial area Electrical Engineering Department - University of Washington 6/5/2012 8 Master Thesis Defense - Tsz Kin (Marco) Au

2. PEV Penetration and Charging Points 𝑈𝑝𝑢𝑏𝑚 𝑜𝑣𝑛𝑐𝑓𝑠 𝑝𝑔 𝑞𝑏𝑡𝑡𝑓𝑜𝑕𝑓𝑠 𝑄𝐹𝑊 𝑄𝐹𝑊 𝑄𝑓𝑜𝑓𝑢𝑠𝑏𝑢𝑗𝑝𝑜 = 𝑈𝑝𝑢𝑏𝑚 𝑜𝑣𝑛𝑐𝑓𝑠 𝑝𝑔 𝑞𝑏𝑡𝑡𝑓𝑜𝑕𝑓𝑠 𝑤𝑓𝑖𝑗𝑑𝑚𝑓𝑡 Charge at home or at work? Type 1: Charge at home only Type 2: Charge at home and work 33.33% 66.67% Electrical Engineering Department - University of Washington 6/5/2012 9 Master Thesis Defense - Tsz Kin (Marco) Au

3. PEV’s Arrival Time • PEV drivers will charge their vehicles anytime at their convenience • Their arrival times affect the charge profile • Drivers’ behaviors varies from day to day, which creates uncertainty • Must model the uncertainty in order to simulate its effect to the power system Electrical Engineering Department - University of Washington 6/5/2012 10 Master Thesis Defense - Tsz Kin (Marco) Au

3. PEV’s Arrival Time Inverse transformation for random number generation • Map rand(0,1) → actual distribution Electrical Engineering Department - University of Washington 6/5/2012 11 Master Thesis Defense - Tsz Kin (Marco) Au

4. PEV’s Battery State of Charge • Commute distance have an effect on the battery state of charge • A driver’s commute distance although is similar everyday, it may vary sometime, which causes uncertainty • Must model this uncertainty in order to simulate its effect to the power system • Convert commute distance to battery state of charge (SOC) 𝑇𝑃𝐷 = 𝐶𝑏𝑢𝑢𝑓𝑠𝑧 𝐷𝑏𝑞. (𝑙𝑋𝑖) − 𝐷𝑝𝑛𝑛𝑣𝑢𝑓 𝐸𝑗𝑡𝑢. (𝑛𝑗𝑚𝑓) × 0.34 𝑙𝑋𝑖/𝑛𝑗𝑚𝑓 Electrical Engineering Department - University of Washington 6/5/2012 12 Master Thesis Defense - Tsz Kin (Marco) Au

4. PEV’s Battery State of Charge Commute distance (miles) Percentage (%) 0 – 4.0 19.19 4.1 – 8.0 22.95 8.1 – 12.0 16.67 12.1 – 16.0 13.77 16.1 – 20.6 9.37 Quantile Function of Commute Distance 20.1 – 24.0 6.07 35 24.1 – 28.0 4.59 Commute Distance (Mile) y = 353.04x 5 - 725.13x 4 + 526.87x 3 - 140.15x 2 + 30 28.1 – 32.0 2.69 22.691x - 0.0038 32.1 + 4.70 25 R² = 0.9997 20 15 Commute Distance Distribution 10 25 5 20 Percentage (%) 0 0 0.2 0.4 0.6 0.8 1 -5 15 Probability 10 5 0 Commute Distance (Mile) Electrical Engineering Department - University of Washington 6/5/2012 13 Master Thesis Defense - Tsz Kin (Marco) Au

5. PEV Charge Profile • Computed individually based on arrival time, battery state of charge, and charging method # 𝑝𝑔 𝑄𝐹𝑊 𝑈𝑝𝑢𝑏𝑚 𝐷𝑖𝑏𝑠𝑕𝑓 𝑄𝑠𝑝𝑔𝑗𝑚𝑓 𝑖𝑠 = 𝑄 𝑗,𝑖𝑠 𝑗 8 6 Power (kW) 4 2 0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Hour 8 6 Power (kW) 4 2 0 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Hour Electrical Engineering Department - University of Washington 6/5/2012 14 Master Thesis Defense - Tsz Kin (Marco) Au

6. Customer Load Profile • Varies from day to day • The variation is assumed to be Gaussian distributed: 2∙(𝑄 𝑐𝑣𝑡,𝑢𝑗 −𝐵𝑤𝑕𝑄 𝑐𝑣𝑡,𝑢𝑗 ) −1 1 2 𝜏 𝑐𝑣𝑡,𝑢𝑗 𝑔 𝑄 𝑐𝑣𝑡,𝑢𝑗 = 𝑓 𝜏 𝑐𝑣𝑡,𝑢𝑗 2𝜌 𝑜𝑝𝑠𝑛 × 𝐵𝑤𝑕𝑄 𝑐𝑣𝑡 𝐵𝑤𝑕𝑄 𝑐𝑣𝑡,𝑢𝑗 = 𝑄 𝑢𝑧𝑞𝑓,𝑢𝑗 Electrical Engineering Department - University of Washington 6/5/2012 15 Master Thesis Defense - Tsz Kin (Marco) Au

7. Running Power Flow Analysis for the Distribution System • Cannot use Newton-Raphson based methods • Distribution networks characteristic: – High R/X ratio → Decoupled and fast decoupled methods won’t work – Weakly meshed, sparse network → Newton -Raphson method won’t work • Forward-backward sweep method is used Electrical Engineering Department - University of Washington 6/5/2012 16 Master Thesis Defense - Tsz Kin (Marco) Au

7. Running Power Flow Analysis for the Distribution System Forward-backward sweep method example: 1 2 3 7200V 3000’ 4 000’ S2 S3 𝑨 12 = 0.1705 + 𝑘0.3409 Ω 𝑨 = 0.3 + 𝑘0.6 Ω/𝑛𝑗𝑚𝑓 𝑨 23 = 0.2273 + 𝑘0.4545 Ω 𝑡 2 = 1500 + 𝑘750 𝑙𝑋 + 𝑘𝑙𝑊𝑏𝑠 𝑡 3 = 900 + 𝑘500 𝑙𝑋 + 𝑘𝑙𝑊𝑏𝑠 Electrical Engineering Department - University of Washington 6/5/2012 17 Master Thesis Defense - Tsz Kin (Marco) Au

7. Running Power Flow Analysis for the Distribution System Forward-backward sweep method example: 𝑊 1 = 7376.2∠0.97 𝑊 𝑊 2 = 7260∠0.23 𝑊 𝑊 3 = 7200 𝑊 1 2 3 7200V 3000’ 4 000’ 𝐽 12 𝐽 23 S2 S3 𝐽 3 𝐽 2 Forward sweep: 5) Compute 𝐽 2 1) Assume voltage at node 3 is 7200V ∗ 𝑡 2 𝐽 2 = = 231.0∠ − 26.3 𝐵 𝑊 2) Compute 𝐽 3 2 ∗ 𝑡 3 𝐽 3 = = 143.0∠ − 29.0 𝐵 6) Compute 𝐽 12 𝑊 3 𝐽 12 = 𝐽 23 + 𝐽 2 = 373.9∠ − 27.3 𝐵 3) Compute 𝐽 23 7) Compute 𝑊 1 𝐽 23 = 𝐽 3 = 143.0∠ − 29.0 𝐵 𝑊 1 = 𝑊 2 + 𝑎 12 ∙ 𝐽 12 = 7376.2∠0.97 𝑊 4) Compute 𝑊 2 8) Compute mismatch between 𝑊 1 and 𝑊 𝑡 𝑊 2 = 𝑊 3 + 𝑎 23 ∙ 𝐽 23 = 7260.1∠0.23 𝑊 Not satisfy! 𝑁𝑗𝑡𝑛𝑏𝑢𝑑𝑖 = 𝑊 𝑡 − 𝑊 = 176.2 𝑊 1 Electrical Engineering Department - University of Washington 6/5/2012 18 Master Thesis Defense - Tsz Kin (Marco) Au

7. Running Power Flow Analysis for the Distribution System Forward-backward sweep method example: 𝑊 1 = 7200 𝑊 𝑊 2 = 7085.4∠ − 0.68 𝑊 𝑊 3 = 7026.0∠ − 1.02 𝑊 1 2 3 7200V 3000’ 4 000’ 𝐽 12 𝐽 23 S2 S3 Backward sweep: 1) Assume voltage at node 1 is 7200V, and use the line currents computed from forward sweep 2) Compute 𝑊 2 𝑊 2 = 𝑊 1 − 𝑎 12 ∙ 𝐽 12 = 7085.4∠ − 0.68 𝑊 3) Compute 𝑊 3 𝑊 3 = 𝑊 2 − 𝑎 23 ∙ 𝐽 23 = 7026.0∠ − 1.02 𝑊 Electrical Engineering Department - University of Washington 6/5/2012 19 Master Thesis Defense - Tsz Kin (Marco) Au