Estimation of solar spectral irradiance using spectral skewness based on integrated sky-ground observation

Solar spectral irradiance fundamentally governs the actual efficiency of photovoltaic technologies. This study proposes a novel estimation model for solar spectral irradiance based on integrated sky-ground observations. To address the limitation of existing metrics in stably capturing spectral shape variations, spectral skewness is proposed as an innovative metric to characterize solar spectral curves. Compared to photon average energy, spectral skewness exhibits 10 % higher stability, demonstrating its superior capability in mapping solar spectral variations. The proposed model utilizes derived meteorological parameters obtained from integrated sky imaging and ground-station monitoring as inputs. These parameters are processed by an XGBoost model to estimate solar spectral metrics. Validation results confirm the model has a 20.37 % improvement in low-error estimation compared to other spectral models like SPCTRAL2. Furthermore, since solar spectral irradiance varies significantly under different sky conditions, this study classifies sky states into four categories and analyzes their spectral characteristics. Results indicate that the spectral skewness of overcast skies typically exceeds 0.06, whereas clear skies exhibit values below 0.06. These findings contribute to recalibrating the practical performance of spectrally selective materials, especially in PV applications, and provide robust guidance for their selection and fabrication tailored to local weather conditions.