Scholarlink

Abstract <p>To enhance the effectiveness of solar power (SP) prediction, particularly in the face of variability and uncertainty, this study proposes an advanced prediction approach focusing on three key areas: feature extraction, data training, and online learning. The feature extraction process involves clustering historical data using a self-organizing map network and reducing dimensionality through kernel principal component analysis (KPCA) to extract the most relevant features from input-output pairs. The data training phase employs interval prediction methods, which address uncertainty and variability more comprehensively than conventional point prediction methods. Finally, the proposed method incorporates online learning, which is not widely used in SP prediction. This method enhances adaptability to real-time weather uncertainties by continuously validating and updating models with new data, thereby improving SP interval prediction accuracy. Simulation results, based on a set of comparisons using real-world SP datasets, demonstrate the effectiveness of the proposed method.</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

Solar Energy Interval Prediction Based on Online Learning and Kernel Density Estimation

  • 期刊论文

  • 发表日期：

    2025年10月1日

查看全文
  • Maoxin He,
  • Chixin Xiao

摘要

Abstract

To enhance the effectiveness of solar power (SP) prediction, particularly in the face of variability and uncertainty, this study proposes an advanced prediction approach focusing on three key areas: feature extraction, data training, and online learning. The feature extraction process involves clustering historical data using a self-organizing map network and reducing dimensionality through kernel principal component analysis (KPCA) to extract the most relevant features from input-output pairs. The data training phase employs interval prediction methods, which address uncertainty and variability more comprehensively than conventional point prediction methods. Finally, the proposed method incorporates online learning, which is not widely used in SP prediction. This method enhances adaptability to real-time weather uncertainties by continuously validating and updating models with new data, thereby improving SP interval prediction accuracy. Simulation results, based on a set of comparisons using real-world SP datasets, demonstrate the effectiveness of the proposed method.