AI-BASED DESIGN OPTIMIZATION AND GRID IMPACT MITIGATION OF DYNAMIC WIRELESS POWER TRANSFER SYSTEMS FOR ELECTRICAL VEHICLE CHARGING APPLICATIONS

Abstract

There is growing interest in electrifying the transportation sector to reduce fossil fuel use. Electric vehicles (EVs) offer a promising alternative but face challenges like limited driving range, range anxiety, and long charging times. Dynamic Wireless Power Transfer (DWPT) systems, which charge EVs while in motion, have emerged to address these issues. The widespread adoption of DWPT systems could be delayed by several challenges, including high costs, system efficiency, safety concerns, ensuring maximum power delivery, and accommodating tolerable misalignment offset, among others. To address these challenges, this thesis develops a design optimization environment for the DWPT charging systems composed of two main modules: a Characterization Module (identifier) and a Design Optimization Module. Designing and optimizing DWPT systems could be challenging due to their complex electromagnetic interactions, magnetic materials' nonlinearities, the switching effects of power electronics converters employed in these systems, and their dynamic operational conditions. To manage these challenges, an electromagnetic Finite Element–State Space (FE-SS) based characterization module is first developed to accurately predict DWPT system performance in terms of magnetic coupling, output power, system efficiency, and electromagnetic field (EMF) leakage under various misalignment conditions. However, directly using FE-SS models as an identifier in the design optimization requires numerous iterations and extensive computational time. To overcome this limitation, this work integrates the FE-SS characterization environment with an AI-based predictive model and Taguchi algorithm to reduce the computational requirements for the characterization module. The Optimization Module utilizes Particle Swarm Optimization (PSO) in conjunction with the AI-based identifier (Characterization Module) to rapidly predict DWPT performance indicators. The developed approach is applied to different classes of EVs, including passenger cars and heavy-duty trucks, and demonstrated through two case studies: stationary and dynamic EV wireless power transfer systems. These case studies adopt the proposed multi-objective design optimization environment to optimize system efficiency, output power, transmitted energy, charging system material cost, and EMF leakage under a range of misalignment conditions. Beyond the device design stage, large-scale deployment of DWPT might introduce new challenges for electrical grids. The highly variable, high-power charging demand of EV-DWPT can cause voltage deviations and increased power losses. To mitigate these impacts, this thesis develops a planning framework that utilizes distributed energy resources (DERs). A Mixed-Integer Nonlinear Programming (MINLP) multi-objective optimization model is developed to determine the optimal placement and sizing of DERs across several EV-DWPT load-penetration scenarios. Two solution approaches are investigated: a cost-based aggregate method (CB) and a Chebyshev goal programming (GP) approach to balance trade-offs across conflicting objectives, including cost, power losses, voltage deviation, and load curtailment. Finally, this thesis contributes to the advancement of EV-DWPT research by (i) developing an AI-driven design optimization environment suitable for DWPT systems, and (ii) proposing a DER planning optimization framework to mitigate EV-DWPT charging load grid impacts on a distribution grid. These contributions enable more efficient DWPT system design, optimization, and integration into distribution networks, supporting the deployment of roadway electrification.