Faults detection and identification in PV array using kernel principal components analysis

The exponential growth of the photovoltaic system installations also requires an adequate maintenance and supervision system to ensure the continuity of service of the system. Conventional protection systems for electrical systems have shown their shortcomings for protecting photovoltaic systems. In this article, a statistical approach based on principal component analysis and its variants is used to detect and identify faults in a photovoltaic array. This involves analysing the variations of the data of the current–voltage and voltage–power characteristics. Subsequently, the calculation of the contributions is applied to the SPE index for the identification of faults. By employing the intermediate value theorem, six different operating states have been identified. The various results obtained first from the simulation model from the Simulink environment and then from a real system of 18 PV show that the kernel principal component analysis allows defect detection with a better precision.