Assessing the performance of physical transposition models in photovoltaic power forecasting: A comprehensive micro and macro accuracy analysis

Accurate power prediction is crucial for the design, simulation, and performance assessment of Photovoltaic (PV) systems. Achieving this accuracy relies significantly on the precise plane-of-array (POA) irradiance data, which is seldom directly measured. To mitigate this limitation, transposition models have been developed to estimate POA irradiance. Numerous studies have evaluated the performance of these models using horizontal irradiances and POA measurements from specific locations. However, the absence of POA measurements poses a significant challenge to using the proposed methods in transposition model evaluation. Moreover, these studies have overlooked the critical role of POA data in predicting PV power. This study addresses these challenges by assessing the performance of thirty two transposition models within a detailed PV power forecasting framework. A comprehensive micro- and macro-level accuracy evaluation is conducted using predicted and ground-measured energy data from three distinct PV systems in different countries. The evaluation is carried out across hourly, daily, monthly, and annual time scales using four distinct statistical metrics under all-sky and the four sky-conditions. The simulation results validate the effectiveness of the proposed approach in assessing the transposition models in the absence of POA irradiance data. Consistent with previous studies, the findings reaffirm the dependence of transposition model performance on location, climate, time, and cloud cover. Furthermore, the statistical analyses identify the best-performing transposition models for each case study under all-sky and sky-conditions scenarios, providing practical guidelines for model selection. •Without POA data, transposition models were evaluated in the PV power forecasting.•Macro and micro accuracy evaluations were conducted for all sky and sky conditions.•Location-specific and climate dependencies of transposition models were examined.•Best-performing transposition models for PV power forecasting were identified.•Perez 1988 identified as best all-sky model with nMBE of −0.02, nRM SE of 9.77, t-stat of 0.11, and R2 of 0.97.