Multi-Agent DDPG-Based Multi-Device Charging Scheduling for IIoT Smart Grids.

Area of Science:

• Electrical Engineering
• Artificial Intelligence
• Smart Grid Technology

Background:

• Widespread adoption of electric vehicles (EVs) and Industrial Internet of Things (IIoT) smart grids necessitates coordinated charging strategies.
• Traditional algorithms face scalability challenges in multi-device environments and limitations in continuous control scenarios.

Purpose of the Study:

• To propose a dynamic charging scheduling algorithm for EVs using Multi-Agent Deep Deterministic Policy Gradient (MADDPG).
• To optimize charging and discharging strategies for multiple EVs in continuous action spaces, reducing costs and balancing grid load.

Main Methods:

• Developed a MADDPG-based algorithm integrating real-time electricity prices, battery status, and sensor data.
• Employed coordinated multi-agent learning for dynamic optimization of EV charging and discharging.
• Utilized Vehicle-to-Grid (V2G) technology for adaptation to price fluctuations and user demand.

Main Results:

• Achieved a 41.12% reduction in charging costs over a 30-day evaluation period compared to baseline methods.
• Demonstrated effective adaptation to electricity price fluctuations and user demand variations.
• Showcased enhanced grid stability through optimized charging time allocation.

Conclusions:

• The proposed MADDPG algorithm offers a scalable and effective solution for dynamic EV charging scheduling in IIoT smart grids.
• Coordinated multi-agent learning with V2G technology significantly improves economic efficiency and grid stability.
• This approach addresses limitations of traditional methods in managing complex, continuous control scenarios.