Daily electric vehicle charging dataset for training reinforcement learning algorithms

Area of Science:

• Power Systems Engineering
• Artificial Intelligence
• Data Science

Background:

• Reinforcement learning (RL) algorithms are crucial for power system optimization.
• Acquiring large, realistic datasets is a significant challenge for training RL algorithms.
• Effective RL deployment requires extensive, iterative training on high-fidelity data.

Purpose of the Study:

• To create a comprehensive and realistic dataset for training offline reinforcement learning algorithms in power systems.
• To address the data acquisition bottleneck in developing robust RL applications for power systems.
• To generate synthetic EV charging data that accurately reflects real-world usage patterns.

Main Methods:

• Utilized conditional tabular generative adversarial networks (CTGAN) to synthesize data from an initial EV charging dataset.
• Applied post-processing techniques to ensure synthetic data adheres to charging station capacity constraints.
• Employed kernel density estimation (KDE) to replicate the distributional characteristics of historical EV connection times.
• Simulated 29,600 days of electric vehicle charging data within a parking facility context.

Main Results:

• Successfully generated a large-scale synthetic dataset of electric vehicle charging behaviors.
• Ensured the synthetic dataset maintains realistic daily demand profiles and respects capacity limits.
• Validated the replication of key distributional features, such as connection timing, using KDE.
• Developed a dataset specifically tailored for offline reinforcement learning algorithm training.

Conclusions:

• The generated synthetic dataset effectively overcomes the challenge of limited real-world data for RL training in power systems.
• The methodology provides a viable approach for creating realistic datasets for complex systems like EV charging infrastructure.
• This resource facilitates the advancement and deployment of RL-based solutions in the power sector.