Scholarlink

This paper proposes a hybrid forecasting framework that combines unsupervised clustering and supervised machine learning. The proposed method first uses PCA to reduce the dimension of wind power characteristics, extract the main change patterns, and remove redundant information. Then, K-means is adopted to group similar time series, thereby achieving local modeling of subsequences. An XGBoost model based on a multivariate sliding window is constructed for each cluster subset to capture the temporal dependence and the interaction between variables. The experimental results show that this method has significantly better accuracy and generalization ability than traditional forecasting models on real wind power datasets.

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

Wind Power Forecasting via PCA-K-Means Clustering and Multivariate Sliding Window XGBoost

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ran Wei,
  • Yuchao Gao

摘要

This paper proposes a hybrid forecasting framework that combines unsupervised clustering and supervised machine learning. The proposed method first uses PCA to reduce the dimension of wind power characteristics, extract the main change patterns, and remove redundant information. Then, K-means is adopted to group similar time series, thereby achieving local modeling of subsequences. An XGBoost model based on a multivariate sliding window is constructed for each cluster subset to capture the temporal dependence and the interaction between variables. The experimental results show that this method has significantly better accuracy and generalization ability than traditional forecasting models on real wind power datasets.