TIM IMING IS IS EVERYTHING: OPTIMAL ELECTRIC VEHICLE CHARGING TO - PowerPoint PPT Presentation

TIM IMING IS IS EVERYTHING: OPTIMAL ELECTRIC VEHICLE CHARGING TO MAXIMIZE WELFARE Miguel Castro Inter-American Development Bank

Hourly private and external costs/Current charging profiles 34 45 Current EV charging 40 33 patterns (Houston & 35 32 Dallas) withdraw most USD Damages/MWh 30 31 power after owners USD/MWh 25 return to their home 30 20 (7-9 PM) 29 15 28 10 Excessive generation 27 5 cost and 26 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 environmental Marginal damages Prices damages

Hourly private and external costs/Current charging profiles 34 45 Current EV charging 40 33 patterns (Houston & 35 32 Dallas) withdraw most USD Damages/MWh 30 31 power after owners USD/MWh 25 return to their home 30 20 (7-9 PM) 29 15 28 10 Excessive generation 27 5 cost and 26 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 environmental Marginal damages Prices damages

Introduction Empirical model • Partial equilibrium models of ERCOT wholesale electricity market ( Decentralized market with invariant tariff and Social planner hourly tariff ) • Simulate how EV charging should be spread among hours to maximize welfare (charging emissions damages) and surplus (no damages) • Simulate second best private and full social costs day-night tariffs Main findings • EV charging in Texas can be met efficiently during the first hours of the day (0-4 H). First best hourly social tariff . • Even day-night and hourly tariff based on private costs (no damages) can guide users to charge EVs efficiently (overlap of low prices and low marginal carbon emissions during first hours of the day).

Empirical Model (Social Planner) 23 𝑟 𝑢 Where: 𝑁𝑏𝑦 𝒈,𝑭𝑾 𝑄 𝑢 (𝑟 𝑢 )𝑒𝑟 𝑢 − 𝐷 𝑔 𝑢 − 𝜐 𝑢 𝐹𝑊 𝑢 𝑟 𝑢 Electricity demand 0 𝑢=0 𝑥 𝑢 Wind power 23 𝑔 𝑢 Fossil generation 𝐹𝑊 𝑢 = 𝐹𝑊 𝑡. 𝑢. 𝐹𝑊 𝑢 charging demand 𝑢=0 𝐷 𝑢 𝑔 𝑢 private fossil generation costs 𝑟 𝑢 + 𝐹𝑊 𝑢 = 𝑥 𝑢 + 𝑔 𝑢 + 𝑜𝑣𝑙𝑓 𝑢 𝜐 𝑢 Marginal damages (carbon, sulfur, nitrogen oxide, and PM 2.5 emissions) and charging constraints • Estimate fossil supply curve with hourly fossil generation, heat input data (EPA), and monthly fuel costs (EIA Form 860) for ERCOT generators in 2017 • Data on actual EV mileage (EV and plug-in hybrids) by auto model in TX (2017 National Household Travel Survey) and current charging patterns (EV Project in Houston and Dallas, DOE, 2013)

Empirical Model (Decentralized market) 2) 𝐷𝑝𝑜𝑡𝑣𝑛𝑓𝑠 𝑝𝑞𝑢𝑗𝑛𝑏𝑚𝑗𝑢𝑧 𝑑𝑝𝑜𝑒: 𝑄 𝑢 𝑥 𝑢 + 𝑔 𝑢 + 𝑜𝑣𝑙𝑓 𝑢 − 𝐹𝑊 𝑢 = 𝑞 𝑒 ∀𝑢 𝑢 = 𝑞 𝑢𝑥 ∀𝑢 3) 𝐺𝑝𝑡𝑡𝑗𝑚 𝑕𝑓𝑜𝑓𝑠𝑏𝑢𝑝𝑠 𝑝𝑞𝑢𝑛 𝑑𝑝𝑜𝑒: 𝐷′ 𝑔 Wholesale cost recovery condition: Where: 23 23 5) 𝑞 𝑒𝑠 𝑂 𝑡 Number plug-in hybrid (gasoline) electric 𝑢 + 𝑜𝑣𝑙𝑓 𝑢 ) ∗ 𝑞 𝑢𝑥 (𝑟 𝑢 + 𝐹𝑊 𝑢 ) = (𝑥 𝑢 + 𝑔 vehicles and full electric vehicles by model s 𝑢=0 𝑢=0 𝑓𝑤 𝑡 daily individual charging demand based Charging constraints: on total annual miles estimated in the NHTS; 23 𝑂 𝑡 ∗ 𝑓𝑤 𝑡𝑢 = 𝑡 𝑂 𝑡 ∗ 𝑓𝑤 𝑡 = 𝑢=0 23 𝐹𝑊 𝑢 = 𝐹𝑊 6) 𝑡 𝑢=0 and using EPA fuel economy (kWh/mi) 𝑓𝑤 𝑡𝑢 hourly EV charging 𝑐𝑏𝑢𝑢𝑓𝑠𝑧 𝑡𝑗𝑨𝑓 𝑡 7) 𝑓𝑤 𝑡𝑢 ≤ 𝑀 𝑑ℎ𝑏𝑠𝑕𝑗𝑜𝑕 𝑢𝑗𝑛𝑓 𝑡 ∀𝑡 𝑀 𝑑ℎ𝑏𝑠𝑕𝑗𝑜𝑕 𝑢𝑗𝑛𝑓 𝑡 for L1 and L2 types • Demand is calibrated with a linear functional form and hourly (short run) elasticity from literature (Deryugina, 2017; Wolak, 2011)

Empirical Model (Marginal damages and emissions) 23 23 𝑛 = 𝛾 0𝑛 + 𝑍 𝛾 𝑚ℎ𝑛 𝐼𝑃𝑉𝑆 ℎ ∗ 𝐸 𝑢 + 𝛾 𝑥ℎ𝑛 𝐼𝑃𝑉𝑆 ℎ ∗ 𝑋 𝑢 + 𝜀 𝑥 + 𝛿 𝑥𝑓 + 𝜁 𝑢 𝑢 ℎ=0 ℎ=0 where: 𝑛 emissions (tCO 2 , lbs SO 2 , lbs NOx, and lbs PM2.5) and total air pollution damages (summation of 𝑍 𝑢 SO 2 , NOx, and PM2.5 damages in 2017 USD) at hour t in the entire grid, 𝑋 𝑢 , 𝐸 𝑢 are ERCOT aggregate wind power and demand (load) in MWh at hour t, 𝜀 𝑥 stands for weekly fixed effects and 𝛿 𝑥𝑓 for weekend FE, 𝜸 are regression coefficients. Average partial effects 𝛾 𝑚ℎ𝑛 give the estimate of the hourly marginal emissions and damages of increasing load in one MWh Air pollution damages using county level marginal damages (morbidity and mortality) for medium and tall stacks from AP2 Model (Holland et al., 2016)

Baseline calibration Decentralized market, one invariant daily tariff Year-round total generation Year-round prices Model reproduces fairly well the median and trends for the entire year and even for different seasons. USD/MWh Static version of startup and ramp up costs, no transmission congestion constraints, but it captures with simplicity the Year-round fossil generation Average seasonal generation main features and results Band depicts a 95% confidence interval, while the solid lines represent medians.

Results Welfare maximizing charging schedule (0-4 H) is the opposite of current patterns (18-23 H). Unconstrained first best charging has welfare gains of ~42% of wholesale prices (10.44 USD per MWh charged) Constraining power withdrawals to L1 and L2 chargers limits using energy from hours with lower prices and marginal carbon emissions reducing welfare gains Overlap of low prices and low marginal carbon emissions from 0-4H . 5-8H rapid increase in carbon emissions and air pollution. *The bands depict a 95% confidence interval, while the solid lines represent averages.

Results Second best day-night tariff: based on the optimal hours from welfare maximization problem (1-7AM, 12AM) Day-night private tariff (only generation costs) cause EV charging mostly at 4-5 AM . It captures ~93.7% of first best gains. Day-night social tariff (generation costs + emissions damages) EV charging at 3-4 AM , less emissions and larger gains ~98% of FB. Without emissions taxes , both hourly and day-night tariffs increase carbon and air pollution damages compared to current patterns charging *The bands depict a 95% confidence interval, while the solid lines represent averages.