Scholarlink

<p>The combined method for interval forecasting (CMIF) is proposed for improved real-time prediction of wind speed uncertainty to facilitate wind turbine operation and power grid dispatching. Time-varying filtering for empirical mode decomposition and phase space reconstruction are used to decompose and reconstruct the original wind speed sequence to solve chaotic phenomena and eliminate noise. Statistical and machine learning models are considered as candidates, and models with excellent performances are selected. Finally, the selected models are combined by a multi-objective optimizer to obtain the final prediction. Experiments were performed using data from the Gansu wind tower, and the results showed that CMIF improved the accuracy of the predicted wind speed interval by 1.07–55.37% compared with single models. The prediction interval had a narrow width while maintaining a high coverage rate, which facilitated accurate quantification of the wind speed uncertainty.</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 novel interval prediction method in wind speed based on deep learning and combination prediction

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月2日

查看全文
  • XueJun Chen,
  • Tao Han,
  • Peng Cheng,
  • Xuanfang Da

摘要

The combined method for interval forecasting (CMIF) is proposed for improved real-time prediction of wind speed uncertainty to facilitate wind turbine operation and power grid dispatching. Time-varying filtering for empirical mode decomposition and phase space reconstruction are used to decompose and reconstruct the original wind speed sequence to solve chaotic phenomena and eliminate noise. Statistical and machine learning models are considered as candidates, and models with excellent performances are selected. Finally, the selected models are combined by a multi-objective optimizer to obtain the final prediction. Experiments were performed using data from the Gansu wind tower, and the results showed that CMIF improved the accuracy of the predicted wind speed interval by 1.07–55.37% compared with single models. The prediction interval had a narrow width while maintaining a high coverage rate, which facilitated accurate quantification of the wind speed uncertainty.