Using On-Site Measurements to Identify and Adjust PV Single-Diode Model Parameters for Real Operating Conditions

Photovoltaic (PV) module manufacturers specify their products' performance under the standard test condition. However, such a condition is hardly found outdoors, that is, during actual operation of photovoltaic systems. Despite that, many modeling methods proposed in the literature rely on standard test condition data for parametric identification. This paper presents an experimental study concerning different modeling approaches for photovoltaic modules, focusing on assessing model performance in different scenarios. The goal is to assess the influence of the data used for the model parametric adjustment when predicting the maximum power point of two photovoltaic modules, allowing the most advantageous methods to be identified. By using an experimental outdoor photovoltaic platform, over 30,000 I-V curves referring to two photovoltaic modules were measured for 16 months. Different models were applied for cases where their parameters were identified using field measurements. It was found that using measurements obtained under real operating conditions to adjust the model parameters provides significantly better prediction of the maximum power in comparison to cases employing data sheet-extracted information. For the all methods and cases studied, using in-field measurements to adjust the model parameters provided average root-mean-square error of 3.4% for power predictions, whereas the average error level found when using data sheet information was 8.8%. An error level as low as 2.83% was reached by adopting the most complex model, although using a significantly simpler modeling approach provided error of 3.25%. Therefore, using field-measured data to adjust model parameters produced the best results, but increasing method complexity did not increase performance in the same scale.

Automated summary

This study investigates different modeling approaches for photovoltaic modules and evaluates their performance in various scenarios. The results show that using field-measured data to adjust the model parameters improves the prediction of the maximum power point, leading to lower average error levels compared to using data sheet information.