Scholarlink

<p>To improve the operation safety and efficiency of hydropower units, this paper takes hydropower units as the research object and studies an intelligent early warning system based on real-time multi-point time series data, based on integrated machine learning models. Firstly, multiple sensors are installed on the hydroelectric generating unit, and commonly used data mining algorithms, such as the boosting tree algorithm, autoencoder algorithm, and long short-term memory (LSTM) model, are utilized to develop anomaly detection models for single measurement point output and overall equipment health prediction models for multiple measurement point output. These models are integrated and optimized, and then a trend prediction model is developed for measurement points based on time series analysis. A comprehensive early warning system is constructed by setting reasonable alarm thresholds. The experimental results show that the developed intelligent early warning system can accurately identify equipment status and issue early warnings in most cases. The accuracy of the system’s early warning remains between 98 and 100%, with a false alarm rate ranging from 1.22 to 2.64% and a missed alarm rate between 0.24 and 0.54%. This effectively improves the operational safety and efficiency of hydroelectric generating units.</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

Real-time early warning of multi-measurement point time series of hydropower generating units based on an integrated machine learning model

  • 期刊论文

  • 发表日期：

    2026年6月27日

查看全文
  • Dong Liu,
  • Lijun Kong,
  • Wenfeng Ren,
  • Yuxiang Guo

摘要

To improve the operation safety and efficiency of hydropower units, this paper takes hydropower units as the research object and studies an intelligent early warning system based on real-time multi-point time series data, based on integrated machine learning models. Firstly, multiple sensors are installed on the hydroelectric generating unit, and commonly used data mining algorithms, such as the boosting tree algorithm, autoencoder algorithm, and long short-term memory (LSTM) model, are utilized to develop anomaly detection models for single measurement point output and overall equipment health prediction models for multiple measurement point output. These models are integrated and optimized, and then a trend prediction model is developed for measurement points based on time series analysis. A comprehensive early warning system is constructed by setting reasonable alarm thresholds. The experimental results show that the developed intelligent early warning system can accurately identify equipment status and issue early warnings in most cases. The accuracy of the system’s early warning remains between 98 and 100%, with a false alarm rate ranging from 1.22 to 2.64% and a missed alarm rate between 0.24 and 0.54%. This effectively improves the operational safety and efficiency of hydroelectric generating units.