Simulation Models Suggested That Optimizing Charging Levels Available to Electric Vehicles Across a Network Grid Can Be 3.7 Percent More Profitable Than Always Providing the Highest Power Level To Charge an EV Upon Arrival.

University Researchers Developed Optimization Models to Improve Electric Vehicle (EV) Charging Schedules and Site Design.

Date Posted

Electrical Vehicle Charging Infrastructure Design and Operations

Summary Information

To realize California’s goals of achieving five million zero-emission vehicles (ZEVs) on the road by 2030, and 250,000 electrical vehicle (EV) charging stations by 2025, this study developed a simulation-based system for EV charging infrastructure design and operations. The study introduced two modules, one for optimizing vehicle charging schedule, and one for infrastructure planning for the solar-powered charging station, to increase customer satisfaction, reduce the burden on the power grid, and maximize the cost effectiveness of charging stations. The study utilized the charging demands and duration data from 2019 on the California State University, Long Beach campus and the solar power and weather data from the Long Beach Airport.


This study used global optimization techniques, machine-learning-based solar power prediction, and two scheduling algorithms, including robust Model Predictive Control (MPC) and empirical rule-based methods, compared under different scenarios. A response surface methodology (RSM) was also developed for the infrastructure optimization solely based on the empirical rule. MPC assumed that one-hour ahead solar power can be known, and the operational revenue was maximized by optimizing the charging power of each EV considering the Time-Of-Use (TOU) price and the Battery Energy Storage System’s (BESS) charging/discharging ratios. The empirical rule charging policy did not use any forecast. The data collected in this study were synthesized to create simulation scenarios for the charging schedule and station infrastructure optimization.

The study tested five charging schemes, including: (i) Robust MPC (R-MPC), which optimizes solar power predictions, charging requests predictions, and grid status; (ii) greedy charging (rule-based), which always used the highest power to charge an EV; (iii) MPC-I, which uses the week-ahead average charging requests were used for demand prediction; (iv) MPC-II, which maximizes charging service revenues and solar energy sales, and minimizes electricity purchase; and (v) MPC-III which uses only three power levels—7.2 kW, 3.6 kW, and 0 kW. A ten-year total profit was calculated by counting operational revenue and infrastructure investment together to choose the optimal sizing of solar photovoltaic capacity, the number of chargers, and BESS capacity.


  • The results for the EV charging schedule optimization revealed that R-MPC algorithms were up to 3.7 percent more profitable than greedy charging depending on the scenario.
  • The results for charging site infrastructure optimization showed that MPC-based algorithms were 2.8 percent more profitable than empirical rule-based algorithms in the absence of BESS, in terms of charging site operations.
  • With the inclusion of BESS, the results for the infrastructure optimization revealed 8.6 percent higher operational profitability for the empirical rule-based algorithms compared to the MPC-based algorithms, in terms of charging site operations.
Goal Areas
Results Type
Deployment Locations