Scholarlink

Accurate solar power forecasting is essential for grid reliability, energy planning, and smart grid integration because of the variability of solar energy. This study compared 11 machine learning models for photovoltaic (PV) power prediction using a real-world dataset of 118,865 records. The models include traditional regressors, ensemble methods, deep learning architectures, and a hybrid combining Random Forest and XGBoost. All models were trained under identical conditions and evaluated using MAE, RMSE, and R2. The ensemble models outperformed the others, with the hybrid and Random Forest achieving the lowest prediction errors (R2 ≈ 0.9996), followed by LightGBM (R2 = 0.9995). Deep learning models, particularly CNN and LSTM-MLP, underperformed with an R2 as low as 0.94.

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

Benchmarking Machine Learning Models for PV Forecasting: Performance Insights and a Novel Hybrid Ensemble Approach

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ahmed M. Abdelrhman,
  • Sayed Hashim Mahmood Salman,
  • Syed Asad Imam,
  • Iftikhar Ahmad

摘要

Accurate solar power forecasting is essential for grid reliability, energy planning, and smart grid integration because of the variability of solar energy. This study compared 11 machine learning models for photovoltaic (PV) power prediction using a real-world dataset of 118,865 records. The models include traditional regressors, ensemble methods, deep learning architectures, and a hybrid combining Random Forest and XGBoost. All models were trained under identical conditions and evaluated using MAE, RMSE, and R2. The ensemble models outperformed the others, with the hybrid and Random Forest achieving the lowest prediction errors (R2 ≈ 0.9996), followed by LightGBM (R2 = 0.9995). Deep learning models, particularly CNN and LSTM-MLP, underperformed with an R2 as low as 0.94.