INVESTIGATIONS ON THE IMPACT OF ELECTRIC VEHICLE CHARGING ON POWER DISTRIBUTION NETWORKS

Abstract

Continuously depleting fossil fuels reserves, rising fuel costs and environmental concerns have led to the emergence of EVs as a popular choice of transportation. The rapid growth of EV adoption presents significant operational challenges for power distribution networks due to the stochastic, time-varying, and unbalanced nature of EV charging demand as it significantly affects power distribution network performance, causing distortions in the daily demand profile, increased power distribution network losses, and greater voltage magnitude deviations across the system. The concentrated and often uncoordinated charging load can place additional stress on network components, leading to reduced voltage stability and overall operational challenges. Accurate modeling of EV charging behavior and the development of effective mitigation strategies are therefore essential to ensure reliable and efficient grid operation under high EV penetration scenarios. The work in this thesis investigates the adverse effects of both slow residential and DC fast charging at public fast charging stations on power distribution network performance in detail and ways to improve the same. Realistic models of slow residential charging and fast public charging are developed using travel behavior data from the 2017 National Household Travel Survey and operational data from a operator based in the US which provides fast charging services at the charging stations. The proposed models capture key characteristics such as the type of EV, charging start times, daily distance travelled, state-of-charge dependence, and charging duration. The impacts of EV charging demand are investigated on a real-world 240-bus radial power distribution network located in the Midwest United States. A key distinction of this distribution grid is the inclusion of one year’s worth of smart meter data (hourly real and reactive power demands of all customers)for all nodes, thereby enabling realistic time-series analyses and facilitating more comprehensive planning and operational studies. The smart meter data spans from January 2017 vi to December 2017 which includes hourly energy consumption (kWh) for 1,079 customers. The network has 139 single-phase and 55 three-phase load buses. The customers allotted to each three-phase bus are distributed equally among the three single phases (A, B and C), in order to minimize the phase unbalance. Unlike standard IEEE and existing power distribution test systems, the 240-bus network provides the 24-hour real and reactive power demands of all the 1079 customers in the network. Simulation results demonstrate that slow residential charging increases network demand during afternoon and evening hours, while fast charging causes sharp morning demand spikes. Although slow charging leads to relatively smaller voltage magnitude deviations, fast charging results in severe voltage dips and increased network losses. This underscores the need for practical demand-side management policy design, particularly time-of-use–based mechanisms, that can operate alongside realistic EV charging behavior while preserving system reliability. To mitigate the adverse effects, state-of-charge-based smart residential EV charging policies are proposed and evaluated by distributing EV loads across three network phases. The results show that coordinated charging significantly improves daily load profiles, reduces the network power losses, and enhances bus voltage profiles. Beyond demand-side mitigation, the thesis explores network-side solutions through the optimal allocation of Distributed Generation resources. Appropriately sized distributed generation resources placed at suitable locations in the power distribution network have been reported to substantially improve the daily demand profile, bus voltage magnitude profile and also reduce distribution network losses. A novel human-based metaheuristic optimization algorithm, Driving Training-Based Optimization, is applied for the optimal siting and sizing of wind-based Distributed Generation units. The performance of driving training-based optimization is benchmarked against several established metaheuristic algorithms, demonstrating superior convergence behavior and solution quality. With single-objective driving training-based optimization formulation, the maximum reduction in the network real power loss is 45.65% whereas with multi-objective driving training-based optimization formulation, the maximum reductions in the network real power loss and voltage magnitude deviation are around 45.62% and 64.71% respectively.