Automatic failure detection in photovoltaic systems

The present work introduces a new method for the automatic detection of misbehaviours in photovoltaic systems, minimizing the amount of data to be sensed. Different anomalous situations, including frame (ground) derivations, highly resistive connections, battery or panel short circuits, etc. are parameterised based on a model of the PV system under study. The same characteristic parameters are extracted from a reduced set of measures and, through a statistical analysis, a correspondence can be established which indicates the state of the physical system. In an experiment, several failures were introduced in a real system, including series resistances in connections, current leak to ground, short circuit of a battery vessel, disconnection of a branch of panels, and they were accurately detected by the algorithm.