Abstract
Vehicle-to-Grid (V2G) technology offers a promising solution for balancing grid loads by coordinating the charging and discharging of electric vehicle (EV) clusters. However, the inherent randomness of human travel patterns and decision-making poses a significant challenge to effective dispatching. This paper proposes a multi-objective optimization strategy that integrates a realistic user behavior model. By combining survey-based behavioral data with Monte Carlo simulations and a genetic algorithm, the model identifies Pareto optimal solutions that balance grid stability with user economic benefits. Results indicate a 9.52% reduction in grid load variance and a 2.99% decrease in the peak-to-valley ratio, demonstrating that behavior-aware dispatching can effectively align grid needs with user incentives.
Introduction: The Human Element in V2G
Vehicle-to-Grid (V2G) technology is often envisioned as a massive, distributed battery system capable of "peak shaving" and "valley filling." While technically sound, most existing models overlook the most volatile component: the human driver. Real-world EV usage is dictated by erratic schedules and psychological factors like "mileage anxiety." If a dispatch strategy fails to account for a driver's need for a specific State of Charge (SOC) for unplanned trips, participation rates will plummet. This research bridges the gap between technical grid requirements and the behavioral realities of EV users.
Methodology: Modeling Behavior and Optimization
The researchers developed a three-stage pipeline to create a behaviorally realistic dispatch model, as illustrated in the system architecture below.
1. User Behavior Modeling
The foundation of the model is a questionnaire-based survey of 320 valid respondents from major Chinese cities. This survey quantified "human factor variables," such as the SOC levels at which users feel anxious versus comfortable enough to discharge their batteries to the grid.
2. Monte Carlo Simulation and Data Flow
Using the survey data, Monte Carlo simulations were conducted to account for the randomness in daily mileage, departure times, and parking durations. This allowed for the simulation of an aggregate load for a cluster of 300,000 EVs. The methodological flow, from data collection to optimization constraints, is shown in the following flowchart.
3. Multi-Objective Genetic Algorithm
A Genetic Algorithm (GA) was employed to find the Pareto optimal solution. The model optimized for two competing objectives:
- Grid Stability: Minimizing the variance of the total grid load.
- User Benefit: Maximizing the financial returns for EV owners through intelligent participation in Time-of-Use (TOU) pricing schemes.
Key Insights: The Psychological "Zones"
The study identified critical psychological thresholds that dictate user participation:
- Mileage Anxiety: On average, users begin to feel anxious when their SOC drops below 31.14%.
- Discharge Willingness: Users generally only feel comfortable discharging to the grid when their SOC exceeds 72.87%.
- Neutral Point: A 50% SOC was identified as the critical tipping point for user sentiment.
Furthermore, the model highlighted that V2G is more effective on workdays than weekends. While weekend travel is more extensive (averaging 126km vs. 35km), the predictability and higher "discharge margins" of weekday commuting provide a more stable basis for grid response.
Results and Impact
Simulation results for a Shanghai-based case study demonstrated significant improvements:
- Grid Performance: The peak-to-valley load ratio was reduced by 2.99%, and overall load variance decreased by 9.52%.
- Economic Gains: Under subsidized pricing models, user income increased by over 50% compared to standard charging scenarios, reaching approximately 1,630 yuan.
- Load Smoothing: The strategy successfully shifted load from daytime peaks to nighttime valleys, effectively smoothing the "Duck Curve" and reducing peak load by 0.7 MW.
Conclusion
This research demonstrates that grid stability and user profitability are not mutually exclusive. By incorporating human psychology and travel patterns into dispatching algorithms, V2G systems can achieve a "win-win" scenario. Future work will focus on scaling these models with larger datasets and refining optimization algorithms to handle even more complex, real-time variables in massive vehicle clusters.
Summary updated