Scholarlink

<p>Accurate prediction of wind power is crucial for grid scheduling and the integration of renewable energy, given its significant temporal variability and nonlinear characteristics. This study proposed a multi-module integrated model for wind power forecasting based on time–frequency domain analysis, aiming to enhance prediction accuracy and reliability. The mode9l combined several advanced techniques, including Wavelet Convolutions (WTC), Long Short-Term Memory Networks (LSTM), Time Series Lightweight Adaptive Network (TSLANet), Frequency Enhanced Channel Attention Mechanism (FECAM), and Fast Kolmogorov-Arnold Networks (FastKAN). Each module was designed to capture distinct characteristics in wind power data, such as local frequency features, temporal dependencies, global contextual information, frequency-domain features, and complex nonlinear relationships. Through the integration of these modules, the model achieved high-precision predictions in multi-scale and dynamic environments. Additionally, the Least Squares Support Vector Machine (LSSVM) was employed for error correction, further reducing prediction errors. Experimental results showed that the model delivered exceptional performance across various test scenarios, significantly improving the handling of multi-scale, complex nonlinear, and global dependency issues in wind power forecasting, demonstrating considerable application potential.</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

Improving wind power prediction with advanced temporal and frequency domain processing combined with error correction

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月8日

查看全文
  • Jinming Gao,
  • Yixin Sun,
  • Hankil Kim,
  • Changsu Kim,
  • Hoekyung Jung

摘要

Accurate prediction of wind power is crucial for grid scheduling and the integration of renewable energy, given its significant temporal variability and nonlinear characteristics. This study proposed a multi-module integrated model for wind power forecasting based on time–frequency domain analysis, aiming to enhance prediction accuracy and reliability. The mode9l combined several advanced techniques, including Wavelet Convolutions (WTC), Long Short-Term Memory Networks (LSTM), Time Series Lightweight Adaptive Network (TSLANet), Frequency Enhanced Channel Attention Mechanism (FECAM), and Fast Kolmogorov-Arnold Networks (FastKAN). Each module was designed to capture distinct characteristics in wind power data, such as local frequency features, temporal dependencies, global contextual information, frequency-domain features, and complex nonlinear relationships. Through the integration of these modules, the model achieved high-precision predictions in multi-scale and dynamic environments. Additionally, the Least Squares Support Vector Machine (LSSVM) was employed for error correction, further reducing prediction errors. Experimental results showed that the model delivered exceptional performance across various test scenarios, significantly improving the handling of multi-scale, complex nonlinear, and global dependency issues in wind power forecasting, demonstrating considerable application potential.