Scholarlink

<p>The exponential growth of Internet information has posed significant challenges in efficiently acquiring valid knowledge within complex and dynamic information environments. Particularly in hydropower station management, substantial volumes of monitoring data remain underutilized, resulting in considerable resource wastage. This study addresses these challenges through the integration of natural language processing and deep learning technologies to effectively collect and classify unstructured data from the Jiangya Hydropower Station, ultimately establishing a knowledge graph and intelligent question answering system. The principal research contributions are fourfold: (1) A comprehensive data acquisition framework combined with deep learning techniques enables efficient knowledge extraction and construction of triple structures; (2) Comparative analysis of five distinct named entity recognition (NER) models reveals the superior performance of the RoBERTa-BiLSTM-CRF architecture (F1-score: 0.923), demonstrating 5.7–12.4% improvement over baseline models; (3) A novel knowledge fusion methodology effectively resolves data redundancy and inaccuracy issues, while the proposed ALBERT-CNN hybrid model achieves 89.6% accuracy in multi-label intent classification, outperforming conventional approaches by 7.3–15.8% in controlled experiments; (4) Implementation of these technical innovations in system development successfully addresses core challenges in text parsing and answer retrieval, establishing a functional intelligent Q&amp;A platform. The research outcomes not only provide new momentum for information management modernization at Jiangya Hydropower Station but also offer valuable practical experience and technical references for knowledge graph applications in related engineering domains.</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

A deep learning and knowledge graph integration framework for intelligent inspection in hydropower stations: a case study of Jiangya hydropower station

  • 期刊论文

  • 发表日期：

    2025年9月24日

查看全文
  • Jinhui Hu,
  • Changtao Deng,
  • Ziyi Wang,
  • Aoxuan Pang,
  • Qiuwen Zhang

摘要

The exponential growth of Internet information has posed significant challenges in efficiently acquiring valid knowledge within complex and dynamic information environments. Particularly in hydropower station management, substantial volumes of monitoring data remain underutilized, resulting in considerable resource wastage. This study addresses these challenges through the integration of natural language processing and deep learning technologies to effectively collect and classify unstructured data from the Jiangya Hydropower Station, ultimately establishing a knowledge graph and intelligent question answering system. The principal research contributions are fourfold: (1) A comprehensive data acquisition framework combined with deep learning techniques enables efficient knowledge extraction and construction of triple structures; (2) Comparative analysis of five distinct named entity recognition (NER) models reveals the superior performance of the RoBERTa-BiLSTM-CRF architecture (F1-score: 0.923), demonstrating 5.7–12.4% improvement over baseline models; (3) A novel knowledge fusion methodology effectively resolves data redundancy and inaccuracy issues, while the proposed ALBERT-CNN hybrid model achieves 89.6% accuracy in multi-label intent classification, outperforming conventional approaches by 7.3–15.8% in controlled experiments; (4) Implementation of these technical innovations in system development successfully addresses core challenges in text parsing and answer retrieval, establishing a functional intelligent Q&A platform. The research outcomes not only provide new momentum for information management modernization at Jiangya Hydropower Station but also offer valuable practical experience and technical references for knowledge graph applications in related engineering domains.