Case Study on Photovoltaic System: Impact of Solarimetric Stations on Simulations and Anomaly Detection

Local solarimetric stations (LSS) are essential for collecting data to evaluate photovoltaic (PV) plant performance and improve simulation accuracy. When unavailable, commercial solar databases (CSD), typically derived from satellite-based typical years, are used. This study compared the impact of LSS and CSD data on PV simulations and explored the use of LSS data for anomaly detection. Two LSS were analyzed: one at a small-scale PV system (minigeneration in Brazil) and another at a large-scale PV plant. The minigeneration station measured irradiance components to calculate Plane of Array (POA) irradiance, while the large-scale station directly measured POA. For the minigeneration system, simulations using LSS data showed a lower discrepancy (0.21%) compared to CSD (3.13%). For the large-scale plant, a -5.99% discrepancy using LSS revealed anomalies in energy generation. MAE and RMSE improved significantly with LSS for the large-scale system, with MAE decreasing from 660.25 MWh (CSD) to 348.08 MWh. Additionally, an unsupervised anomaly detection flagged 2.88% and 4.47% of data for two inverters, showcasing LSS potential for predictive models. These findings suggest that while LSS data are valuable for PV plant performance analysis, their effectiveness may depend on spectral range, averaging intervals, and irradiance transposition in simulations.