Optimal protection coordination of DOCRs for microgrids using a hybrid deep reinforcement learning and metaheuristic...

Area of Science:

• Electrical Engineering
• Power Systems Protection
• Smart Grids

Background:

• Modern microgrids face significant challenges in directional overcurrent relays (DOCRs) protection due to high penetration of inverter-based distributed generation, bidirectional power flow, and frequent changes in grid-connected/islanded states.
• Conventional protection coordination methods are inadequate for microgrids with diverse operating scenarios, N-2 contingencies, and user-defined relay characteristics, leading to high computational complexity and practical implementation issues.

Purpose of the Study:

• To develop a computationally efficient and dynamic protection coordination model for effective DOCR coordination in inverter-dominated microgrids.
• To ensure the security and reliability of protection coordination under varying operational conditions and contingencies.

Main Methods:

• A hybrid optimization model integrating reinforcement learning and metaheuristic search is proposed to optimize time multiplier setting (TMS), plug setting (PS), and user-specified relay curve parameters.
• A contingency reduction strategy using hierarchical clustering (K-means augmented with BIRCH) is employed to decrease the computational load by identifying representative operating scenarios.

Main Results:

• The proposed contingency reduction method decreased operating scenarios from 847 to 124 without compromising coordination accuracy.
• Achieved a 94.7% reduction in total relay operating time, a 99.2% coordination success rate, and maintained coordination intervals exceeding 0.35s for fault resistances up to 40 ohms.
• The K-means-BIRCH approach demonstrated significantly faster computation times (e.g., 1.73s and 5.80s) compared to the BRKGA-MILP method, with improved coordination performance and convergence.

Conclusions:

• The developed framework offers communication-independent, reliable, and rapid protection coordination for inverter-based microgrids.
• The integration of contingency reduction, clustering, and hybrid optimization enables secure and accurate relay coordination with reduced computational effort, making it suitable for practical microgrid protection systems.