.png){kind=icon}
.png){kind=icon}
Return
Deep learning-driven fault detection and classification in microgrids using Temporal Convolutional Network
.png){kind=icon}
PRE
AI
DOI：10.1016/j.compeleceng.2025.110777.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The proliferation of distributed generations is transforming electric distribution networks into microgrids (MGs). MGs can operate in both grid-tied and standalone modes. These varying modes of operation complicate the fault protection in MGs. Existing fault detection methods rely heavily on signal processing techniques to extract fault features. However, this becomes a challenging task due to the multi-mode operation of MGs. This paper introduces a new intelligent fault detection and classification scheme (FDCS) for MGs based on Temporal Convolutional Network (TCN). The proposed FDCS can efficiently capture low-level fault features and high-level temporal dependencies without the need for an external feature extractor. In the proposed FDCS, the three-phase (TP) current signals acquired at relay points are used as inputs to two different TCNs. These TCNs employ a series of dilated causal convolution-based residual blocks to extract low-level features and high-level dependencies from the TP current signals to generate fault type and phase information. Comprehensive simulations on a standard MG system validate the successful detection and classification of all fault types across various operating scenarios of MGs. The results indicate that the FDCS achieves a fault detection accuracy of 100% in grid-tied mode and 99.8% in standalone mode. The unbalanced fault classification accuracy of the TCN-based FDCS is 100% for both modes of operation. Furthermore, a Comparative analysis with established intelligent protection methods, such as support vector machines, decision trees, convolutional neural networks, and long short-term memory networks, highlights the superior performance of the proposed FDCS in terms of accuracy, dependability, security, operation time, and robustness under noisy measurement conditions.
Keywords:
Fault classification
Fault detection
Feature extraction
Microgrid protection
Temporal Convolutional Network
Journal
C
IF:
4.9
Papers:
6.7K
Citations:
1.3W