Fault detection in trackers for PV systems based on a pattern recognition approach

In many photovoltaic (PV) power plants, the PV modules are installed in trackers. In these systems, the PV modules are fixed in a mobile structure to always maintain a perpendicular position to the brightest point in the sky, obtaining in this way the maximum power from the sun, during the all day. Nevertheless, these systems are subject to problems that reduce their efficiency. Thus, visual inspection or complex methods can be used to detect this problem. However, these systems normally result in delays or are expensive. To overcome these problems, this paper proposes a new method for that detection. This, method is based on the pattern recognition analysis. Thus, through the analysis of the images of the several solar panels, the PV module that presents a problem in the tracker will be detected. The orientation of the PV modules is determined using the centroid of the PV cells after applying an image pre-processing stage. The angle is calculated using the statistical moments or by the slope of the line joining two centroids of the PV cells that are located at the vertices of the PV module. Several test cases are presented to verify the efficiency of the proposed approach.