基于多模态深度学习的电力设备故障诊断系统  

Power Equipment Fault Diagnosis System Based on Multimodal Deep Learning

作　　者：周元峰 ZHOU Yuanfeng(China Electronic Systems Technology Co.,Ltd.,Beijing 100141)

机构地区：[1]中国电子系统技术有限公司,北京100141

出　　处：《软件》2024年第12期171-173,共3页Software

摘　　要：随着电力系统复杂程度的提高,电力设备故障诊断面临严峻挑战。本文设计了一种基于多模态深度学习的电力设备故障诊断系统,提出多模态数据融合策略,利用门控循环单元、卷积神经网络和注意力机制实现多源数据特征提取与动态融合,并结合变分自编码器与极限学习机完成故障类型识别与程度评估。系统通过主动学习与增量学习提升鲁棒性,功能性测试实验验证了系统的高准确性与实时性。With the increasing complexity of the power system,fault diagnosis of power equipment is facing severe challenges.This article designs a power equipment fault diagnosis system based on multimodal deep learning,proposes a multimodal data fusion strategy,uses gate controlled recurrent units,convolutional neural networks,and attention mechanisms to achieve multi-source data feature extraction and dynamic fusion,and combines variational autoencoder and extreme learning machine to complete fault type recognition and degree evaluation.The system improves robustness through active learning and incremental learning,and functional testing experiments have verified the high accuracy and real-time performance of the system.

关 键 词：多模态深度学习 诊断系统 电力设备 

分 类 号：TN219[电子电信—物理电子学] TP277[自动化与计算机技术—检测技术与自动化装置]