Clustering framework with integrated IEC metrics for long-term PV performance assessment in Nordic conditions

This paper examines long-term PV system performance in harsh climates by reviewing the IEC 61724 series standards and relevant literature, focusing on extreme Norwegian weather. To address low irradiance and the non-linear behavior of PV modules, inverters, and other converters, rule-based and machine learning (ML)–based clustering techniques are introduced for enhanced performance assessment. The methodology was applied to data from two Norwegian sites. Results show that half-cell modules generally outperform full-cell modules, with an average gain of 2–3%. Clustering indicates that shading during low sun angles is likely the main cause of reduced performance in one half-cell array. Degradation analysis over three years highlights performance trends and maintenance impacts. Although IEC standards provide rigorous guidance for long-term PV evaluation, their application in the literature often lacks clarity on data collection, preprocessing, and converter operating regions. Rule-based clustering is easy to interpret, while ML-based clustering enables automatic pattern recognition and identifies closely aligned operating regions. The proposed hybrid framework, combining IEC-standard parameters with clustering-based analysis, offers an interpretable and robust approach for cross-array comparison, long-term monitoring, and benchmarking under challenging climatic conditions. • IEC 61724 standards and literature reviewed with focus on applicability in harsh Norwegian climates. • Long-term performance of half-cell and full-cell PV modules assessed under low irradiance and variable climate conditions. • Rule-based and machine learning clustering proposed for enhanced performance assessment and trend detection. • Hybrid framework applied for cross-array comparison and trend analysis.