Data2Charge: An End-to-End Framework for Synthetic EV Charging Data and Fleet Optimization via an Interactive Web Platform
2026 International Conference on Smart Energy Systems and Technologies, SEST 2026, Ciudad Real, İspanya, 2 - 04 Eylül 2026, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1109/sest67798.2026.11712349
- Basıldığı Şehir: Ciudad Real
- Basıldığı Ülke: İspanya
- Anahtar Kelimeler: Electric vehicle charging, fleet scheduling, generative adversarial network, smart charging, synthetic data generation
- Yıldız Teknik Üniversitesi Adresli: Evet
Özet
Research on electric vehicle (EV) charging behavior is persistently constrained by the limited availability of high-fidelity data due to scarcity and privacy restrictions. Existing simulation tools partially address this gap but typically rely on historical data replay or rule-based heuristics, and remain decoupled from fleet-level operational optimization. This paper proposes Data2Charge, an end-to-end framework that, unlike existing tools, integrates generative data synthesis directly with operational optimization, enabling on-demand scenario analysis for fleets not represented in recorded traces. A two-stage architecture is adopted: a conditional tabular generative adversarial network (CTGAN) learns temporal patterns of arrivals and session durations, while a k-nearest neighbor (kNN) module assigns corresponding energy demands to preserve the physical consistency between duration and delivered energy. The synthesized fleet is scheduled through a linear programming (LP) formulation that minimizes charging cost under day-ahead pricing, subject to each vehicle's required energy, availability window, and maximum charging power. An uncontrolled first-in-first-out (FIFO) scheme serves as a baseline to quantify the gains of the proposed schedule. Validated on two real-world datasets and scaling up to 50,000 vehicles, the framework is delivered through an interactive web platform that enables on-demand fleet configuration and comparison of FIFO and cost-optimized profiles. Data2Charge thereby provides a reproducible testbed for fleet operations, tariff evaluation, and infrastructure studies, independently of proprietary data traces.