Studi Evaluasi Perbedaan Iradiasi Prediksi dengan Aktual pada Sistem PLTS

This study investigates the causes of discrepancies between predicted irradiation both for Global Horizontal Irradiation (GHI) and Global Tilted Irradiation (GTI) towards measured irradiation data from photovoltaic (PV) systems located in Jakarta, Bekasi, Karawang, Jombang, and Pasuruan. The analysis focuses on geographically proximate systems to minimize external variability and enable assessment of local atmospheric influences and system-specific operational characteristics. Weather variability associated with aerosol dynamics, caused by industrial activity and fossil-fuel-based transportation, is identified as a potential source of increased uncertainty, although it is not regarded as a primary conclusion. The inherent complexity of tropical meteorological conditions further contributes to irradiation variability and challenges the stability of prediction models. An evaluation of meteorological database quality was conducted through PVsyst simulation to assess the impact of estimated data accuracy, comparing two widely used data providers, Solargis and Meteonorm, against field measurements for the February–March 2025 period. The results demonstrate that Solargis consistently provides more accurate irradiation estimates than Meteonorm, although both dataset issue overestimation tendency according to Relative Mean Bias Error (rMBE). Additionally, the study highlights the further degradation of prediction reliability mainly caused by system management, particularly sensor failures that resulted in partial data loss. Such data gaps led to error accumulation, reflected in elevated Root Mean Square Error (rMSE) values, as evidently shown in nearby sites, Pasuruan and Jombang PV systems, where Solargis-based GHI rMSE values in February respectively reached 7.47 kWh/m² and 9.51 kWh/m², whereas for GTI were at 0.41 kWh/m2 and 1.4 kWh/m2, respectively. These values correspond to longer sensor fault durations. These findings emphasize that, beyond the selection of meteorological databases, data recording quality and system management practices can substantially amplify prediction errors in tropical PV systems.