Scholarlink

<p>Coastal regions are vital for global renewable energy, but wind speed variability, influenced by water vapor and land-sea interactions, challenges ultra-short-term wind power forecasting. This study introduces the VMD-WOA-TCN-LSTM method (Variational Mode Decomposition-Whale Optimization Algorithm-Temporal Convolutional Network-Long Short-Term Memory) to improve prediction accuracy in coastal areas. Using historical wind speed and power data from two wind farms during Typhoons DANAS and LEKIMA (2019) along China’s Yellow Sea coast, the method’s performance was evaluated against other models. Key findings include: (i) The VMD-WOA-TCN-LSTM model outperformed others, achieving MAE = 45 and <i>R</i><sup>2</sup> = 0.8 for coastal turbines, and MAE = 0.8 and <i>R</i><sup>2</sup> = 0.95 for offshore turbines. (ii) Turning points in wind power time series were identified as major error sources. (iii) Wind turbine performance parameters were determined: cut-in wind speed of 1.6&#xa0;m/s, cut-out wind speed of 24.7&#xa0;m/s, and rated power output at 15.9&#xa0;m/s. Accurate ultra-short-term forecasting is crucial for optimizing power systems, supporting sustainable growth, and advancing global carbon neutrality goals.</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

Optimized deep learning for ultra-short-term wind power prediction in coastal and offshore wind farms

  • 期刊论文

  • 发表日期：

    2025年11月3日

查看全文
  • Aodi Fu,
  • Chang Xu,
  • Karam Alsafadi,
  • Jinjun Gu,
  • Wenzheng Yu,
  • Feifei Xue,
  • Liyuan Deng,
  • Haibo Shen,
  • Huijun Wu,
  • Lingzi Wang,
  • Junwei Jia,
  • Wenzhi Cao

摘要

Coastal regions are vital for global renewable energy, but wind speed variability, influenced by water vapor and land-sea interactions, challenges ultra-short-term wind power forecasting. This study introduces the VMD-WOA-TCN-LSTM method (Variational Mode Decomposition-Whale Optimization Algorithm-Temporal Convolutional Network-Long Short-Term Memory) to improve prediction accuracy in coastal areas. Using historical wind speed and power data from two wind farms during Typhoons DANAS and LEKIMA (2019) along China’s Yellow Sea coast, the method’s performance was evaluated against other models. Key findings include: (i) The VMD-WOA-TCN-LSTM model outperformed others, achieving MAE = 45 and R2 = 0.8 for coastal turbines, and MAE = 0.8 and R2 = 0.95 for offshore turbines. (ii) Turning points in wind power time series were identified as major error sources. (iii) Wind turbine performance parameters were determined: cut-in wind speed of 1.6 m/s, cut-out wind speed of 24.7 m/s, and rated power output at 15.9 m/s. Accurate ultra-short-term forecasting is crucial for optimizing power systems, supporting sustainable growth, and advancing global carbon neutrality goals.