CDT-RiskNet: A Risk-Aware Copula-Diffusion-Transformer Network for Stochastic EV Charging Station Optimization in Energy Market

The growing penetration of photovoltaic (PV) systems and electric vehicles (EVs) poses systemic uncertainty for EV charging stations (EVCSs) participating in day-ahead energy market. These uncertainties are characterized by complex dependencies and time-varying tail risks that challenge traditional scenario generation and risk management techniques. This paper presents CDT-RiskNet, a Copula-Diffusion-Transformer framework for risk-aware stochastic optimization. A copula-enhanced diffusion model is developed to generate realistic joint scenarios of PV generation and EV charging demand. To manage time-varying tail risk, a Transformer-based risk module predicts dynamic CVaR weights from both historical and forecasted features, enabling coordinated evaluation across time. Simulation resultson both small-scale and large-scale EVCSs using real-world data demonstrate that CDT-RiskNet improves scenario generation quality, risk control, and adaptability to varying market conditions, leading to better economic performance under uncertainty.