Communication Dans Un Congrès Année : 2024

Dimensionality Reduction of Environmental Data for Long-Term PV Performance Analysis Using Graph Based Methods

Résumé

Accurately analyzing performance of photovoltaic (PV) systems requires analyzing large volumes of environmental data spanning entire years. Processing such year-long datasets can be computationally intensive, potentially obscuring critical temporal trends. This study proposes a novel, data-driven approach GOIF (Graph Oriented Information Fusion) that leverages graph based community detection and explainable artificial intelligence (XAI) for efficient and insightful analysis of PV systems’ performance. In this preliminary work, we have analyzed year-long temperature and solar irradiance data, key factors impacting solar energy production. For this, we leverage Louvain Modularity optimization to partition the data into distinct weather communities. Subsequently, XAI techniques select representative days from each cluster, and this process is repeated across various data normalization methods. We evaluate cluster quality with standard deviation as a metric. Comparing the distribution of representative days, we demonstrate the method's effectiveness in capturing the data's inherent trends and seasonal variations. This approach offers a valuable tool for advanced analysis in solar PV systems, leading to efficient data analysis, improved decision-making, and optimized resource management.
Fichier principal
Vignette du fichier
4CV.1.27.pdf (301) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)

Dates et versions

cea-04806089 , version 1 (26-11-2024)

Identifiants

Citer

Srijani Mukherjee, Laurent Vuillon, Denys Dutykh, Ioannis Tsanakas. Dimensionality Reduction of Environmental Data for Long-Term PV Performance Analysis Using Graph Based Methods. EU PVSEC 2024, Sep 2024, Vienna, Austria. pp.020411-001 - 020411-004, ⟨10.4229/EUPVSEC2024/4CV.1.27⟩. ⟨cea-04806089⟩
23 Consultations
26 Téléchargements

Altmetric

Partager

More