Scholarlink

<p>The increasing penetration of inverter-interfaced renewable sources in modern AC microgrids has altered conventional fault characteristics, leading to limited fault current levels and challenging protection performance. In addition, fault datasets are often imbalanced because normal operating conditions occur more frequently than abnormal events, which can degrade the performance of conventional classifiers. To address these challenges, this paper proposes an intelligent hybrid framework based on the Synthetic Minority Oversampling Technique (SMOTE) and Radial Basis Function Neural Network (RBFNN) for accurate and fast fault detection and classification in inverter-based hybrid AC microgrids. A detailed MATLAB/Simulink model consisting of solar photovoltaic generation, PMSG-based wind turbine, PEM fuel cell, battery energy storage system, utility grid, and dynamic loads is developed to generate multi-class fault datasets under unbalanced fault conditions. Measured three-phase voltage and current signals are preprocessed and normalized, followed by SMOTE-based dataset balancing to improve minority fault representation. The balanced dataset is then used to train the proposed RBFNN classifier for multi-class fault recognition. To validate practical applicability, the method is further tested under measurement noise, load variation, and communication delay scenarios. Simulation results demonstrate that the proposed framework achieves high classification accuracy, strong robustness under uncertain operating conditions, and fast computational performance suitable for protection applications. The low inference time and reliable fault recognition capability indicate that the proposed method is suitable for real-time intelligent protection of renewable-based AC microgrids. Furthermore, the effectiveness of the proposed framework is validated through real-time Hardware-in-the-Loop (HIL) implementation using the OPAL-RT platform. The proposed framework improves the reliability and resilience of renewable energy microgrids by enabling fast and accurate fault detection. Therefore, the proposed approach significantly supports secure renewable energy integration and contributes to sustainable and low-carbon power 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

An intelligent SMOTE–RBFNN framework for robust fault detection and classification in inverter-based renewable microgrids

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年8月4日

查看全文
  • M Prabhavathi,
  • R Sitharthan

摘要

The increasing penetration of inverter-interfaced renewable sources in modern AC microgrids has altered conventional fault characteristics, leading to limited fault current levels and challenging protection performance. In addition, fault datasets are often imbalanced because normal operating conditions occur more frequently than abnormal events, which can degrade the performance of conventional classifiers. To address these challenges, this paper proposes an intelligent hybrid framework based on the Synthetic Minority Oversampling Technique (SMOTE) and Radial Basis Function Neural Network (RBFNN) for accurate and fast fault detection and classification in inverter-based hybrid AC microgrids. A detailed MATLAB/Simulink model consisting of solar photovoltaic generation, PMSG-based wind turbine, PEM fuel cell, battery energy storage system, utility grid, and dynamic loads is developed to generate multi-class fault datasets under unbalanced fault conditions. Measured three-phase voltage and current signals are preprocessed and normalized, followed by SMOTE-based dataset balancing to improve minority fault representation. The balanced dataset is then used to train the proposed RBFNN classifier for multi-class fault recognition. To validate practical applicability, the method is further tested under measurement noise, load variation, and communication delay scenarios. Simulation results demonstrate that the proposed framework achieves high classification accuracy, strong robustness under uncertain operating conditions, and fast computational performance suitable for protection applications. The low inference time and reliable fault recognition capability indicate that the proposed method is suitable for real-time intelligent protection of renewable-based AC microgrids. Furthermore, the effectiveness of the proposed framework is validated through real-time Hardware-in-the-Loop (HIL) implementation using the OPAL-RT platform. The proposed framework improves the reliability and resilience of renewable energy microgrids by enabling fast and accurate fault detection. Therefore, the proposed approach significantly supports secure renewable energy integration and contributes to sustainable and low-carbon power systems.