Abstract
A comprehensive monitoring system was deployed at the SINTEF Oslo PV demonstration site to assess the performance of multiple photovoltaic technologies, including monocrystalline silicon, bifacial silicon, and CIGS thin-film modules. The monitoring infrastructure combines a Campbell Scientific data logger with the Tigo® platform, providing high-resolution operational data alongside environmental measurements from a pyranometer, a soiling monitoring system, and a compact weather station. Short-term datasets were analyzed in the statistics tool R to investigate the influence of environmental conditions on PV power generation. Nonlinear regression models were developed to describe electrical power output as a function of solar irradiance and module temperature, the two primary performance-driving variables. Two model structures were evaluated: a full cubic model and a power-exponential model. Both approaches demonstrated strong predictive capability, with coefficients of determination (R²) ranging from 0.93 to 0.97 across different PV module groups. Model performance was further assessed using the Akaike Information Criterion (AIC). The full cubic model achieved the highest predictive accuracy and most effectively captured the nonlinear relationships between irradiance, module temperature, and power output. However, because the improvement over the power-exponential model is relatively small, the power-exponential model is recommended for routine monitoring due to its lower complexity and better practical applicability.