Scholarlink

<p>Directional overcurrent relays (DOCRs) protection in modern microgrids is problematic, because of the massive inverter-based distributed generation penetration, the flow of power in both directions, lower fault currents, and an extremely high rate of grid-connected/islanded operational changes. Microgrids with a variety of operating cases, N-2 contingencies and relay characteristics that are user-defined cannot be coordinated using conventional methods. Also, the existing optimization methods have high computational complexity in the cases, when high number of contingencies and relay setting groups are taken into account, and they cannot be practically implemented. The purpose of this work is to create a computationally efficient and dynamic protection coordination model that can guarantee the effective DOCR coordination in microgrids dominated by inverters and preserve the security of coordination in varying conditions of operation. A hybrid optimization model that integrates reinforcement learning and metaheuristic search is presented, to jointly optimize time multiplier setting (TMS), plug setting (PS), and user-specified parameters of the relay curve of several groups of settings. A contingency reduction strategy that runs on hierarchical clustering of K-means augmented with BIRCH, is used to reduce the computational load of computing representative operating scenarios. The suggested approach is confirmed through altered IEEE 14-bus and IEEE 33-bus microgrid test systems with high substitutions of inverter-based distributed generation. The contingency reduction that is proposed reduced the operating scenarios to 124 as compared to 847 without sacrificing accuracy in coordination. This technique recorded the reduction of total operating time of the relay by 94.7, the success rate of coordination to 99.2% and an interval of coordination time that is more than 0.35&#xa0;s to fault resistance up to 40 ohms plus over 80% cut in computation time. The K-meansBIRCH approach was three times faster with lower operating time of 1.73&#xa0;s (OM1) and 5.80s (OM2) than the BRKGA-MILP method and two times quicker with better coordination performance and convergence rate, as compared to the BRKGA-MILP method. The suggested framework offers communication-independent, reliable, and quick protection coordination of microgrids based on inverters. The contingency reduction, clustering, and hybrid optimization approach allows the secure coordination of relays with high accuracy and with less computational effort, which makes the approach appropriate to a range of practical microgrid protection systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

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

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月4日

查看全文
  • Yuxuan Chen

摘要

Directional overcurrent relays (DOCRs) protection in modern microgrids is problematic, because of the massive inverter-based distributed generation penetration, the flow of power in both directions, lower fault currents, and an extremely high rate of grid-connected/islanded operational changes. Microgrids with a variety of operating cases, N-2 contingencies and relay characteristics that are user-defined cannot be coordinated using conventional methods. Also, the existing optimization methods have high computational complexity in the cases, when high number of contingencies and relay setting groups are taken into account, and they cannot be practically implemented. The purpose of this work is to create a computationally efficient and dynamic protection coordination model that can guarantee the effective DOCR coordination in microgrids dominated by inverters and preserve the security of coordination in varying conditions of operation. A hybrid optimization model that integrates reinforcement learning and metaheuristic search is presented, to jointly optimize time multiplier setting (TMS), plug setting (PS), and user-specified parameters of the relay curve of several groups of settings. A contingency reduction strategy that runs on hierarchical clustering of K-means augmented with BIRCH, is used to reduce the computational load of computing representative operating scenarios. The suggested approach is confirmed through altered IEEE 14-bus and IEEE 33-bus microgrid test systems with high substitutions of inverter-based distributed generation. The contingency reduction that is proposed reduced the operating scenarios to 124 as compared to 847 without sacrificing accuracy in coordination. This technique recorded the reduction of total operating time of the relay by 94.7, the success rate of coordination to 99.2% and an interval of coordination time that is more than 0.35 s to fault resistance up to 40 ohms plus over 80% cut in computation time. The K-meansBIRCH approach was three times faster with lower operating time of 1.73 s (OM1) and 5.80s (OM2) than the BRKGA-MILP method and two times quicker with better coordination performance and convergence rate, as compared to the BRKGA-MILP method. The suggested framework offers communication-independent, reliable, and quick protection coordination of microgrids based on inverters. The contingency reduction, clustering, and hybrid optimization approach allows the secure coordination of relays with high accuracy and with less computational effort, which makes the approach appropriate to a range of practical microgrid protection systems.