Estimation of Electric Vehicle Adoption Rates Using Sequential Generative Adversarial Networks

In the rapidly evolving domain of electric vehicle (EV) usage and infrastructure development, the paucity of comprehensive and detailed data poses significant challenges for effective planning and implementation. This study proposes an innovative framework for synthesizing and fusing data to construct comprehensive synthetic datasets that can assist with policy-making at different stages of EV adoption. The framework comprises a Data Generation module, leveraging Sequential Generative Adversarial Networks to generate survey and travel sequence data, and a Data Fusion module to integrate the synthetic data, resulting in profiles that combine socio-demographic attributes with travel behaviors. By utilizing data from Indiana state, USA, as a case study, the research generates synthetic datasets that reflect realistic household and travel behaviors. The usefulness of the dataset is demonstrated by one sample application that estimates EV penetration rates in the next few years under a range of future scenarios. This innovative method bridges the existing gap in data availability, significantly enhancing our capacity to analyze EV usage patterns. The application of this framework is instrumental for policymakers and urban planners, offering a sophisticated tool to guide the strategic development of EV infrastructure and support the transition towards sustainable mobility.

I. Introduction

Millions of electric vehicles (EVs) are expected to be on the road within the next few decades ([1]). The rapid expansion of the EV market underscores the need for an in-depth network-level analysis to grasp the implications of EV adoption on energy systems and transportation infrastructure. However, the scarcity of comprehensive EV datasets poses a significant challenge, primarily due to the current low penetration of EVs and the sensitive nature of the data involved. This lack of data is especially noticeable in the area of detailed, geo-spatial travel behaviors, such as the need for charging stations ([2], [3]). Therefore, robust datasets are crucial for a thorough understanding of EV usage patterns and their associated charging demands. Overcoming these data-related obstacles is not just a technical challenge; it's a fundamental step in informing policy decisions and developing the necessary infrastructure to support a future dominated by EVs.