Multi Modal Data Fusion using Fusion Convolution Network for High Voltage Insulator Condition Monitoring in Power Grid

This paper proposes a novel approach for condition power grid monitoring of high voltage insulators using a multi-modal data fusion (MMIF) system with deep neural networks. The proposed system addresses the challenge of accurately assessing insulator surface conditions, which is crucial for preventing electrical failures. Traditional methods often rely solely on visual inspection, which can be limited by factors like resolution, background clutter, and illumination variations. Our MMIF system, termed the Fusion Convolutional Network (FCN), incorporates both image data and leakage current data to achieve a more robust and comprehensive analysis. The FCN leverages the strengths of convolutional neural networks (CNNs) for image feature extraction and a binary multilayer neural network (MNN) for sub-classification. The results demonstrate that the FCN achieves superior classification accuracy of 98.6% compared to using image data or leakage current data alone. This paves the way for a more reliable and automated approach to high voltage insulator condition monitoring.