The accelerating deployment of electric vehicles (EVs) alongside the rapid expansion of renewable energy sources (RES) presents both transformative opportunities and complex operational challenges for modern power systems. Existing policy mechanisms, such as time-of-use (ToU) tariffs, renewable portfolio standards (RPS), net metering credits, and vehicle-to-grid (V2G) incentives, are typically designed and evaluated in isolation from real-time grid dynamics, leaving a critical gap between regulatory intent and operational effectiveness. This paper proposes a novel deep learning-based analytical framework—designated PolicyRE-EV—designed to bridge this gap by coupling renewable energy availability forecasting with policy-sensitive EV charging demand modeling. The proposed framework integrates a Transformer-based multivariate forecasting backbone with a dedicated Policy Encoding Module that represents regulatory parameters as differentiable soft constraints within the training objective. Three operational policy scenarios are investigated: (1) unregulated baseline EV charging, (2) ToU tariff-regulated smart charging, and (3) full renewable-aligned V2G dispatch. Experimental benchmarks against Long Short-Term Memory (LSTM), Bidirectional LSTM (BiLSTM), and Random Forest (RF) baseline models demonstrate that PolicyRE-EV achieves superior performance across Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), Renewable Utilization Rate (RUR), and CO₂ Displacement Index (CDI). The results confirm that embedding policy parameters within deep learning architectures significantly enhances alignment between EV charging patterns and renewable generation windows, offering a scalable and policy-compliant tool for distribution network operators and energy planners.

Index Terms—Renewable Energy Integration, Electric Vehicle Charging, Policy Framework, Transformer Architecture, Vehicle-to-Grid (V2G), Time-of-Use Tariffs, Smart Grid, Deep Learning Forecasting, Demand-Side Management.