A Time- and Space-Integrated Expansion Planning Method for AC/DC Hybrid Distribution Networks

Area of Science:

• Electrical Engineering
• Power Systems Engineering
• Computational Intelligence

Background:

• Growing renewable energy and electricity demand challenge traditional power grids.
• AC/DC hybrid distribution networks (AC/DC-HDNs) offer a solution for enhanced reliability and flexibility.
• Expansion planning for AC/DC-HDNs is complex due to topology, dynamic loads, and costs.

Purpose of the Study:

• To propose a time- and space-integrated expansion planning method for AC/DC-HDNs.
• To optimize the planning process by balancing economic and reliability objectives.
• To address challenges in AC/DC-HDN expansion planning.

Main Methods:

• Graph theory-based distribution grid modeling integrating spatial layouts.
• Time-series analysis for dynamic load and renewable generation characteristics.
• Modified Graph Attention Network (MGAT)-based Deep Reinforcement Learning (DRL) for optimization.

Main Results:

• The proposed MGAT-based DRL algorithm demonstrates superior training efficiency and stability over traditional methods.
• Faster convergence and reduced fluctuation in cumulative rewards were observed.
• Consistently higher cumulative rewards indicate effective optimization of AC/DC-HDN expansion planning.

Conclusions:

• The developed method provides an effective solution for the complex expansion planning of AC/DC-HDNs.
• The MGAT-based DRL approach enhances optimization performance for future power grids.
• This research contributes to reliable and flexible grid expansion with increasing renewable energy sources.