Development of a data analysis methodology to assess PV system performance

Abstract

There are many possible operational problems that may occur in a PV system and the sooner they are detected and repaired, the lowest the energy loss will be. Most of the malfunctions are taking place in the inverter of a PV system [1]. The result of the majority of the inverter malfunctions is not a standard reduction on the output of the system but a variable energy loss [2]. These malfunctions, like bad Maximum Power Point Tracking MPPT or inverter outage, are very difficult to be detected using only the so-called Performance Ratio of the system, which is standard in PV performance analyses [1]. During this study, a new monitoring method was proposed in which energy yield is related to reference energy yield as scatterplot using the mean square error (MSE). Any unusual change of the linear regression implies the presence of a malfunction, and in fact our MSE method showed that inverter malfunctions and cases of partially shading can be detected, where other methods fail.