Scholarlink

<p>In modern society, electrical power generates all requirements. Besides, rising demand for fossil fuel reliance generates extreme climate change and long-lasting sustainable energy concerns. Solar photovoltaics is an environmentally sustainable solution at scale, yet its integration into energy systems presumes exact predictions of power for minimizing variability and grid stability. This work experimentally simulates two state-of-the-art machine learning models, Histogram Gradient Boosting Regressor and Extreme Gradient Boosting, for the estimation of photovoltaic output. The models are optimized with hyperparameter tuning using Firefly Optimization Algorithm, Particle Swarm Optimization, and Moth Flame Optimization. Six forecast schemes, single and hybrid, are tested using data from Mildura, Australia, over the year span 1990–2014. The results indicate Extreme Gradient Boosting to be superior to single models, and maximum accuracy and least error to be generated by the hybrid Histogram Gradient Boosting Regressor optimized using Moth Flame Optimization. These reports make the presented method the most significant predictor to be global horizontal and direct normal irradiance.</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

Predictive analytics for sustainable energy: an in-depth assessment of HGBoost and XGBoost models in photovoltaic energy systems

  • 期刊论文

  • 发表日期：

    2026年3月19日

查看全文
  • Wei Huang

摘要

In modern society, electrical power generates all requirements. Besides, rising demand for fossil fuel reliance generates extreme climate change and long-lasting sustainable energy concerns. Solar photovoltaics is an environmentally sustainable solution at scale, yet its integration into energy systems presumes exact predictions of power for minimizing variability and grid stability. This work experimentally simulates two state-of-the-art machine learning models, Histogram Gradient Boosting Regressor and Extreme Gradient Boosting, for the estimation of photovoltaic output. The models are optimized with hyperparameter tuning using Firefly Optimization Algorithm, Particle Swarm Optimization, and Moth Flame Optimization. Six forecast schemes, single and hybrid, are tested using data from Mildura, Australia, over the year span 1990–2014. The results indicate Extreme Gradient Boosting to be superior to single models, and maximum accuracy and least error to be generated by the hybrid Histogram Gradient Boosting Regressor optimized using Moth Flame Optimization. These reports make the presented method the most significant predictor to be global horizontal and direct normal irradiance.