E. Mallett, E. Fainman, T. Jin
Texas State University,
United States
Keywords: electric vehicle charging infrastructure, dynamic distributionally robust optimization, data-driven decision making
Summary:
The increasing adoption of EVs demands reliable and accessible charging infrastructure to support the transition to electrified mobility. For example, a survey in California’s Greater Bay Area revealed only 72.5% reliability for public DC fast chargers, which requires improvement to meet user expectations. Furthermore, in Texas, the anticipated growth of EVs—reaching over 12 million vehicles by 2040—will require approximately 165 GWh of electricity daily, equivalent to the output of three nuclear power plants. Establishing clean, cost-effective, and self-sustaining infrastructure is essential to meet this demand while supporting environmental goals. Problem Definition: This study focuses on the multi-period deployment of direct-current fast chargers (DCFCs) in public spaces under uncertain and time-varying demand. While public EV charging networks are critical for accelerating EV adoption, the profitability and long-term financial sustainability of these investments are key challenges. The uncertainty in charging demand—driven by factors like growing EV fleets, expanding battery capacities, and evolving consumer behavior—complicates infrastructure planning. Current studies often overlook long-term strategies that incorporate these uncertainties. Academic and Practical Relevance: This research contributes to the literature on dynamic facility allocation, particularly for EV charging infrastructure, by addressing the limitations of static models. Unlike traditional approaches, our study uses a dynamic and data-driven optimization framework to handle time-varying demand and uncertain conditions. The results are relevant for urban planners, utility providers, and policymakers, offering a roadmap for equitable, profitable, and sustainable EV infrastructure deployment. Methodology: The study is organized around three key research questions: 1. Forecasting the number of EVs utilizing public DCFCs over time through time series analysis (e.g., ARIMA) and machine learning models (e.g., regression and neural networks). 2. Developing a deterministic optimization model to identify the optimal siting and sizing of charging stations under certain demand conditions. 3. Extending the model to a dynamic distributionally robust optimization (DRO) framework to incorporate data uncertainty and pricing strategies. A novel non-parametric lifting framework is applied to expand the uncertainty space, ensuring computational tractability while improving the realism of decision-making. This approach incorporates demographic data, EV registry records, and historical parking usage data from Santa Monica, CA. Results: Preliminary findings demonstrate that DRO models outperform stochastic programming (SP) and robust optimization (RO) in maximizing profitability and computational efficiency. Simulations using real-world data from public parking lots show that dynamic deployment strategies can adapt to decreasing DCFC costs and increasing battery capacities, achieving higher financial returns compared to static methods. This study pioneers a scalable, dynamic framework for multi-period EV charging deployment, adaptable to various cities using real-world data. By integrating operations research, data science, transportation engineering, and sustainability, it fosters cross-disciplinary collaboration and innovation. The findings provide actionable insights for policymakers, utility providers, and developers to plan equitable, profitable, and sustainable charging networks, accelerating EV adoption and green mobility. Additionally, the study enhances reliability by using EVs as energy storage during adverse weather, supporting quick recovery. It also offers valuable hands-on learning opportunities in data analytics and sustainable planning, preparing students for careers in emerging industries.