Scholarlink

<p>With the rapid development of renewable energy, wind power, as a key component, has become an important pillar of the global energy transition. To improve the accuracy of wind power forecasting while reducing computational costs, this study proposes a wind power data processing method that integrates frequency-based clustering with intelligent optimization. The method first performs time–frequency decomposition on wind power sequences to obtain intrinsic mode functions (IMFs), and then applies the K-means algorithm based on sample entropy to cluster the IMFs, reconstructing components with similar frequency characteristics into new components. This effectively addresses the issues of excessive IMF numbers and signal feature loss while enhancing information representation. For the new components with high sample entropy, the Polar Light Optimization (PLO) algorithm is employed to optimize model parameters, significantly improving denoising capability and adaptability while reducing computational costs. In a case study at a wind farm in Inner Mongolia, a Bidirectional Long Short-Term Memory (BiLSTM) network was used as the forecasting model for validation. The results show that, compared with the unoptimized dual-decomposition model, the proposed method reduces the mean absolute error (MAE) and root mean square error (RMSE) by 58.33% and 47.77%, respectively, and increases the coefficient of determination (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(R^2\)</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mi>R</mi> <mn>2</mn> </msup> </math></EquationSource> </InlineEquation>) by 1.30%, demonstrating its significant potential in wind power forecasting.</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

Frequency-Based Clustering and Intelligent Optimization for Enhanced Wind Power Forecasting

  • 期刊论文

  • 发表日期：

    2026年2月20日

查看全文
  • Haoqiang Zhang,
  • Xiaopeng Wang,
  • Hongliang Tian,
  • Biqiong Wu,
  • Dongjie Zhang,
  • Yuan Yao

摘要

With the rapid development of renewable energy, wind power, as a key component, has become an important pillar of the global energy transition. To improve the accuracy of wind power forecasting while reducing computational costs, this study proposes a wind power data processing method that integrates frequency-based clustering with intelligent optimization. The method first performs time–frequency decomposition on wind power sequences to obtain intrinsic mode functions (IMFs), and then applies the K-means algorithm based on sample entropy to cluster the IMFs, reconstructing components with similar frequency characteristics into new components. This effectively addresses the issues of excessive IMF numbers and signal feature loss while enhancing information representation. For the new components with high sample entropy, the Polar Light Optimization (PLO) algorithm is employed to optimize model parameters, significantly improving denoising capability and adaptability while reducing computational costs. In a case study at a wind farm in Inner Mongolia, a Bidirectional Long Short-Term Memory (BiLSTM) network was used as the forecasting model for validation. The results show that, compared with the unoptimized dual-decomposition model, the proposed method reduces the mean absolute error (MAE) and root mean square error (RMSE) by 58.33% and 47.77%, respectively, and increases the coefficient of determination ( \(R^2\) R 2 ) by 1.30%, demonstrating its significant potential in wind power forecasting.