Finding a Faulty/Partially Shaded Panel in an Solar Photovoltaic Array: A Novel and Instantaneous Additional Hardware Module-Based Technique

Solar Photovoltaic (SPV) systems can exhibit multiple local maxima in their current-voltage (I-V) and power-voltage (P-V) characteristics due to partial shading at the panel or string level. The frequent occurrence and prolonged operation of panels under partial shading conditions may lead to a reduction in power output and hot-spot formation on the panels. Early detection and exact identification of these shaded panels can significantly reduce power losses and prevent hot spots. In this regard, this article proposes a novel method that utilizes multi-capacitorbased I-V and P-V techniques to determine the exact location of the shaded panels. The proposed I-V and P-V technique includes an additional hardware module (AHM) that generates the I-V and P-V curves of the SPV string, which are fed to an artificial intelligence (AI) based classification model through the sensor and control unit. The dual-phase classification algorithm is designed with a convolution neural network (CNN) to classify between normal and faulty panels. CNN model sends the relevant information about the faulty panel to the plant controller, which remotely bypasses the panel to prevent any physical damage and losses in the SPV system. The proposed method is realized in real-time by developing an experimental setup with a solar farm and AHM module. The IV and PV curves of the proposed system under different shading scenarios are explored.