Monitoring of Photovoltaic Systems Using Improved Kernel-Based Learning Schemes

Data-based procedures for monitoring the operating performance of a PV system are proposed in this article. The only information required to apply is availability measurements, which routinely on-line collected via sensors. Here, kernel-based machine learning methods, including support vector regression (SVR) and Gaussian process (GPR), used model multivariate data from fault detection because their flexibility capability nonlinear approximation. Essentially, SVR GPR models adopted obtain residuals detect identify occurred faults. Then, passed through an exponential smoothing filter reduce noise improve quality. In work, scheme based on kernel density estimation sense faults by examining generated residuals. Several different scenarios were considered study, string fault, partial shading, modules short-circuited, module degradation, line-line array. Using 20 MWp grid-connected system, successfully traced using developed procedures. Also, it has been demonstrated that GPR-based achieve better over SVRs monitor systems.