Resilient Optimal Defensive Strategy of Micro-Grids System via Distributed Deep Reinforcement Learning Approach...

Area of Science:

• Electrical Engineering
• Computer Science
• Control Systems

Background:

• Interconnected microgrids face significant challenges in energy management due to increasing false data injection (FDI) attacks on the demand side.
• Existing defense mechanisms may not adequately address the complexities of FDI attacks in dynamic microgrid environments.

Purpose of the Study:

• To propose a resilient optimal defensive strategy against FDI attacks in interconnected microgrids.
• To develop an online evaluation method for assessing the impact of FDI attacks on microgrid security and stability.
• To enhance the autonomy and efficiency of microgrid defense through a distributed deep reinforcement learning (DRL) approach.

Main Methods:

• An online evaluation approach using the recursive least-square (RLS) method to assess FDI attack impact on supply security and voltage stability.
• A distributed actor network learning approach based on evaluated security confidence to derive optimal network weights.
• Implementation of a resilient optimal defensive strategy using distributed deep reinforcement learning (DRL).

Main Results:

• The proposed RLS method effectively evaluates the impact of FDI attacks on microgrid systems.
• The distributed DRL approach successfully deduces optimal network weights for generating effective defensive schemes.
• Simulation results demonstrate the proposed method's capability in evaluating FDI attack impact and providing optimal defense.

Conclusions:

• The developed resilient optimal defensive strategy effectively safeguards microgrids against FDI attacks.
• The improved distributed DRL approach offers a viable and promising solution for enhancing microgrid energy management and security.
• The methodology enhances microgrid autonomy and accelerates DRL efficiency in defense applications.