Model‐based maximum power point tracking for photovoltaic panels: parameters identification and training database collection

Abstract

Module-level distributed maximum power point tracking (MPPT) represents an attractive solution for photovoltaic systems installed in dense urban areas, where panels are often subject to different solar irradiance levels. Model-based MPPT algorithms are particularly suitable for the purpose: they enable good steady-state accuracy and fast dynamics thanks to an underlying parametric model of the panel. The target of the present study is deeply investigating the estimation of the model parameters, and the collection of the training database, since they heavily affect overall performance. In this work, parameter values result by maximising energy production considering the training database; under some simplifications, it leads to a weighted least squares problem that can be easily solved. One of the main advantages is the robustness in the presence of some identification data that have been collected under partially shadowed conditions. Moreover, the possibility to gather the training database by running a perturb and observe MPPT is investigated and tested for the first time. Energy production is allowed also during this stage, thus opening the way to a periodic update of the parameters to follow degradation and time drift of the module. Experimental results show that performance is virtually the same as that obtained by computing parameters from a large set of volt-ampere characteristics.