Methods for Performance Evaluation of Photovoltaic Plants With Reduced Instrumentation
Abstract
Accurate performance assessment of photovoltaic power plants (PVPPs) typically relies on high-quality infield measurements, which can be impractical for small-scale systems or arrays with varied orientations. As an alternative, the use of satellite-based datasets (SBDs) combined with physical models (PhM) offers a costeffective and scalable approach for estimating meteorological conditions and predicting PV performance. This study proposes a reproducible framework for PV system evaluation that integrates open-access SBDs and deterministic PhM within the pvlib-python environment. The framework estimates plane-of-array (POA) irradiance, module temperature, and AC energy yield while minimizing instrumentation requirements to only two irradiance sensors. It includes a systematic method for selecting the most suitable GHI source and decomposition models for POA irradiance estimation. The model chain employs the DIRINT and Erbs models for decomposition, the Perez model for transposition, the Faiman model for PV module temperature, and PVWatts for energy simulation. Validation using six 3.74 MWp single-axis tracking PV plants in southeastern Brazil showed monthly POA irradiation deviations below 3 % (except May, 6.5%) and AC energy yield nRMSE between 3.05% and 6.6%, with MAPE from 2.42% to 5.01% and annual bias under 1.2%. The framework demonstrated high accuracy and robustness, providing a practical and sensor-efficient approach for PV performance assessment using open-access data and tools.
Keywords
pvlib, Satellite-Based Database, Performance Ratio, Model Chain, Solarimetric Station, IEC 61724 1:2021