Assessment of photovoltaic performance models for system simulation

Abstract

An essential stage in assessing the feasibility of a PV project is the energy yield prediction, which estimates the total energy production of a PV system at a specific site. Photovoltaic (PV) performance models are mathematical representations used to estimate the energy yield of power systems based on PV technology. The PV performance models are subjected to a series of errors derived from the different steps in the modeling chain of the PV system. Although some studies have been conducted to assess the accuracy of these models, limited research have focused on studying the accuracy of the individual submodels that comprise the PV performance model. The main objective of this paper is to assess the performance of different combinations of the most cited models aiming to find a PV performance model with good accuracy. There were studied a total of 20 PV performance models derived from the combination of four plane-of-the-array (POA) irradiance models, five PV module models and two inverter models. All the PV performance models were implemented computationally and their performance was compared with measurements collected by a data acquisition system in a real 2.2 kWp photovoltaic system. The best PV performance model presents an accuracy of −0.201% (rMBE) and 15.099% (rRMSE) with respect to the measured AC power output, which is in line with the values reported in the literature. Several sources of error were identified, which can greatly influence PV system energy yield estimation. Among them, the uncertainty in the derating factors which represent all the non-temperature dependent losses present in the PV system is the most critical.