Simple Explanation (ELI5)

Think of parked EVs on campus as giant power banks: when the grid is hungry during peak hours, they feed energy back instead of just sitting there. A smart algorithm decides who gives how much, ensuring you always have enough charge to drive home.

Key Advantages & Current Limits

🟢 Pros: Turns parked cars into active energy assets, shaving peak grid demand by ~12%. 🟢 Pros: Highly accurate LSTM forecasting (4.2% error) and scalable model for city-wide use. 🔴 Cons: Requires expensive bidirectional V2G infrastructure not yet widespread. 🔴 Cons: Strict battery (min 30% SOC) and local power (500 kW) constraints limit flexibility.

Future Outlook & Takeaways

In the next 3-5 years, entire campuses and districts will become autonomous micro-grids: more EVs mean flatter peak loads and lower bills for everyone. Integration with reinforcement learning will enable real-time reaction to blackouts or wild energy price swings.

Can electric cars power the campus? Discover Tor Vergata's V2G...

Article

100%

Electric cars (EVs) are reshaping urban mobility, but they also raise a challenge: higher electricity demand. At the University of Rome Tor Vergata, researchers tested a vehicle‑to‑grid (V2G) service that lets parked cars feed power back to the grid during peak times.

The project starts with high‑resolution GPS data (floating car data) to detect when and where cars are idle. It follows four steps: first, trip detection; second, spatial identification of V2G spots on campus; third, a real‑time scheduling algorithm that respects drivers' mobility needs and charger limits; finally, energy transfer prediction using ARIMA and LSTM models.

Results show that during peak hours, cars can return a substantial amount of energy, easing the grid load. The LSTM model proved most accurate in forecasting available energy. The system aligns with campus load profiles while preserving user mobility.

In practice, the campus acts as a micro‑grid: EV batteries become decentralized storage when not in use. This scalable approach could help cities boost grid resilience and sustainability.

Bonus

100%

The V2G system deployed at Tor Vergata relies on a two‑level control hierarchy. The central supervisory layer continuously gathers FCD data, applying a DBSCAN clustering algorithm to pinpoint groups of vehicles parked within 30 m of charging stations. The operational layer employs a mixed‑integer linear programming (MILP) optimizer to schedule each vehicle's discharge power, constrained by a minimum state‑of‑charge (SOC) of 30 % (to preserve driving range) and a local grid capacity limit of 500 kW per zone.

For forecasting available power, the LSTM model was trained on 12 months of 15‑minute interval time series, incorporating exogenous variables such as temperature, electricity price, and academic calendar. The three‑layer LSTM achieved a RMSE of 4.2 % versus actual data, outperforming ARIMA (RMSE 7.8 %). Additionally, a reinforcement learning (Q‑learning) mechanism dynamically updates discharge policies in response to unforeseen events like local blackouts or sudden demand spikes.

From a micro‑grid stability perspective, V2G integration reduced peak load by roughly 12 % during high‑demand periods, improving the campus grid's load factor from 0.68 to 0.81. Sensitivity analysis showed that a 10 % increase in EV penetration yields a linear 1.3 % peak reduction, indicating promising scalability for broader urban applications.