Fault detection and classification in solar photovoltaic system using graph base semi-supervised learning and support vector machine
Vitthal S. Sagde,  Nitin J. Phadkule
The behaviour of solar photovoltaic systems is different than conventional power sources as regard to fault detection and classification(FDC). In solar photovoltaic (PV) systems FDC is an ultimate need for increasing safety and reliability in PV systems. A various type of faults may become difficult to detect by conventional protection devices because nonlinear characteristics of PV system, causes safety issues and fire risk in PV system. The machine learning methods such as supervised, unsupervised, and semi-supervised are widely used in various applications. This paper focus on machine learning methods such as graph based semi-supervised learning (GBSSL) and support vector machine (SVM) to mitigate the protection issues. The GBSSL algorithm have been proposed for FDC using measurements, such as PV system voltage, current, irradiance, and temperature and the result obtained is compare with SVM (supervised learning models), which are trained by large amount of labelled data and therefore, have drawbacks: the labelled PV data are difficult or expensive to obtain, the trained model is not easy to update and the model is difficult to visualize. To mitigate these issues, this paper proposes a GBSSL algorithm only using a few labelled data that are normalized by reference values for better visualization. The feature of GBSSL model is to not only detects the fault, but also identifies the possible type of fault in order to get easier system recovery. The model can learn the all the status of the PV systems under various changing weather conditions. The simulation results and their analysis show the effectiveness of fault detection and classification of the proposed GBSSL method and drawbacks of SVM.
Keywords- Fault detection, GBSSL, SVM, PV arrays, Semi-supervised learning, etc.
Cite this Article
Vitthal S. Sagde,  Nitin J. Phadkule,   "Fault detection and classification in solar photovoltaic system using graph base semi-supervised learning and support vector machine"
, International Journal of Engineering Development and Research (IJEDR), ISSN:2321-9939, Volume.7, Issue 2, pp.341-352, May 2019, Available at :http://www.ijedr.org/papers/IJEDR1902066.pdf