Comparative Assessment of PV Energy Forecasting Tools Using Empirical Data from a Plant in Northeast Brazil
Abstract
This study presents an evaluation of photovoltaic energy forecasting tools through a 24-month empirical analysis of a 36.3 kWp solar power plant in Northeast Brazil. The plant operated continuously from May 2023 to April 2025, with 99.59% availability. Using computational tools, PVsyst simulations and Solcast satellite-based data analyses were compared with real generation data, system’s performance was evaluated using MAE, RMSE, and MAPE metrics. PVsyst was configured with site-specific parameters and loss modeling; Solcast data were processed via a data-mining workflow using module efficiency of 21.3% and loss assumptions comparable to PVsyst. PVsyst more closely matched observations (MAE 292.7 kWh/month; RMSE 403.15 kWh/month) than Solcast (MAE 355.16 kWh/month; RMSE 495.44 kWh/month). The results demonstrate PVsyst's superior accuracy (6.18%), attributed to its detailed system loss modeling, although it is a paid service. In contrast, Solcast data analyses showed a small difference (7.24%) and offers a viable open-access alternative for research. The findings reveal that, while PVsyst remains one of the widely used tools for technical designs, Solcast’s database provides sufficient accuracy for feasibility studies when using a data mining algorithm to extract specific information complemented by equipment datasheets for forecasting the energy generation of the PV system. This contribution is a rigorously monitored agricultural case in Brazil’s Northeast with a side-by-side comparison of a design-grade simulator and a satellite-based dataset.
Keywords
Photovoltaic Energy Forecasting, PVsyst, Solcast, Empirical Performance Data