Abstract
Increasing the efficiency of photovoltaic panels (PV) is one of the important goals of researchers worldwide in the field of renewable resources. The important results obtained in the case of finding new materials for the manufacturing of the panels to obtain the highest possible conversion efficiency must be doubled by research for developing methods for efficient real-time monitoring of PV operation in order to rapidly or in advance identify possible failures.
This paper looks for some types of failures and how they can be identified as quickly as possible from the information coming from different sources, the most important being the PV monitored parameters, the PV control system parameters, and from different cloud services.
One way to identify different types of failures is to use machine learning (ML) methods. In applying these methods, an important thing is the availability of a great number of good training data sets. In order to obtain such data sets, this paper aims to create a model of PV using Matlab, which is fed with both real data and data synthesized using fault models. A number of four simulation cases were considered which take into account the normal operation of the photovoltaic panels, their malfunction due to a failure (two different types of failures were considered), and the malfunction of the panels due to the appearance of the two types of failures simultaneously, using input data that was partially measured and partially generated in Matlab. The outputs of these model simulations will be used for training the ML model.