The current–voltage curve of a photovoltaic module affected by mismatching exhibits several local maximum power points, in such a way the rightmost one exhibit a different shape to the one without mismatch. A small set of points around the maximum power point can be enough to detect the presence of mismatching, in such a way it is not necessary to perform a full curve scan, avoiding power losses. In this paper, a support vector machine is used to classify each measurement into mismatched or non–mismatched. However, the performance of this type of models strongly depends on the kernel function used by the classifier: linear, polynomial, sigmoidal or Gaussian. All of these approaches are tested and compared in this work, not only with the synthetic set of simulated curves used to train the models, but also with real measurements from commercial solar modules under different levels of mismatching. From the results, it can be seen that the minimum errors are achieved when a cubic polynomial kernel is selected. Moreover, to improve these results, we propose a novel procedure to select the samples of the training set using a self–organizing map, that increases the weight of the non–frequent cases.

Support Vector Classifiers with Different Kernel Functions to Detect Mismatching Conditions Affecting Photovoltaic Arrays

Piliougine Rocha, Michel
;
Spagnuolo, Giovanni
2023-01-01

Abstract

The current–voltage curve of a photovoltaic module affected by mismatching exhibits several local maximum power points, in such a way the rightmost one exhibit a different shape to the one without mismatch. A small set of points around the maximum power point can be enough to detect the presence of mismatching, in such a way it is not necessary to perform a full curve scan, avoiding power losses. In this paper, a support vector machine is used to classify each measurement into mismatched or non–mismatched. However, the performance of this type of models strongly depends on the kernel function used by the classifier: linear, polynomial, sigmoidal or Gaussian. All of these approaches are tested and compared in this work, not only with the synthetic set of simulated curves used to train the models, but also with real measurements from commercial solar modules under different levels of mismatching. From the results, it can be seen that the minimum errors are achieved when a cubic polynomial kernel is selected. Moreover, to improve these results, we propose a novel procedure to select the samples of the training set using a self–organizing map, that increases the weight of the non–frequent cases.
2023
979-8-3503-4837-8
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11386/4854812
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact