Scholarlink

<p>Efficient analysis of solar photovoltaic (PV) system performance demands processing large-scale environmental data while preserving critical trends for energy prediction. This study proposes Graph-Oriented Information Fusion (GOIF), a novel data reduction framework that employs graph-based community detection to identify representative days for solar PV performance analysis. GOIF constructs a graph with days as nodes and Euclidean-based similarities as edges, integrating daily average irradiance and temperature to capture their combined impact on PV energy output. Using Louvain modularity, it clusters days into communities and applies PageRank to select one representative day per community. GOIF represents annual data using a few days with a 1.5% error in energy yield approximation versus 7.31% for k-means while improving cluster stability (measured by standard deviation) and reproducibility. This approach reduces computational complexity without sacrificing accuracy, achieving a robust representation of yearly PV performance. This study establishes GOIF as a robust and efficient data reduction tool for PV performance analysis, enhancing computational efficiency and decision making. Future work could focus on refining GOIF’s ability to optimize data storage and retrieval, further improving its utility for long-term solar energy applications.</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

Scalable weather data reduction for solar PV analysis using graph-based approach

  • 期刊论文

  • 发表日期：

    2025年9月16日

查看全文
  • Srijani Mukherjee,
  • Laurent Vuillon,
  • Denys Dutykh,
  • Ioannis Tsanakas

摘要

Efficient analysis of solar photovoltaic (PV) system performance demands processing large-scale environmental data while preserving critical trends for energy prediction. This study proposes Graph-Oriented Information Fusion (GOIF), a novel data reduction framework that employs graph-based community detection to identify representative days for solar PV performance analysis. GOIF constructs a graph with days as nodes and Euclidean-based similarities as edges, integrating daily average irradiance and temperature to capture their combined impact on PV energy output. Using Louvain modularity, it clusters days into communities and applies PageRank to select one representative day per community. GOIF represents annual data using a few days with a 1.5% error in energy yield approximation versus 7.31% for k-means while improving cluster stability (measured by standard deviation) and reproducibility. This approach reduces computational complexity without sacrificing accuracy, achieving a robust representation of yearly PV performance. This study establishes GOIF as a robust and efficient data reduction tool for PV performance analysis, enhancing computational efficiency and decision making. Future work could focus on refining GOIF’s ability to optimize data storage and retrieval, further improving its utility for long-term solar energy applications.