Combining physical models to estimate PV power: evaluation and optimal modeling in the solar resource-rich semi-arid Brazilian region

Abstract

Accurate estimation of energy production in photovoltaic power plants is crucial for project feasibility assessment and O&M practices. This study evaluates and analyzes the impact of combining different physical models for PV power modeling, varying different techniques for global horizontal irradiance (GHI) separation, irradiance transposition, and optical, thermal and electrical modeling. High-resolution data collected at 1-min intervals from a 2.5 MWp PV plant located in the Brazilian semi-arid region are used. The PV generation is examined and modeled based on ground-measured GHI, considering a total of 11,340 possible combinations, through seven separation models, nine transposition models, four optical models, nine thermal models, and five electrical models. It is observed that the selection of physical models significantly impacts the estimation, when adopting inaccurate physical models relative differences of 49 % in nMAE and 26 % in nRMSE were evidenced. The models which achieved the best results among the top performers were Starke2 separation model, Perez's transposition model, Martin-Ruiz's optical model, Sandia or Mattei's thermal model and De Soto's electrical model. Additionally, selecting adequate models based on the literature proved to be a good choice for modeling, almost achieving the optimal performance of the best combinations.