PV System Performance Evaluation by Clustering Production Data to Normal and Non-Normal Operation

Abstract

The most common method for assessment of a photovoltaic (PV) system performance is by comparing its energy production to reference data (irradiance or neighboring PV system). Ideally, at normal operation, the compared sets of data tend to show a linear relationship. Deviations from this linearity are mainly due to malfunctions occurring in the PV system or data input anomalies: a significant number of measurements (named as outliers) may not fulfill this, and complicate a proper performance evaluation. In this paper a new data analysis method is introduced which allows to automatically distinguish the measurements that fit to a near-linear relationship from those which do not (outliers). Although it can be applied to any scatter-plot, where the sets of data tend to be linear, it is specifically used here for two different purposes in PV system monitoring: (1) to detect and exclude any data input anomalies; and (2) to detect and separate measurements where the PV system is functioning properly from the measurements characteristic for malfunctioning. Finally, the data analysis method is applied in four different cases, either with precise reference data (pyranometer and neighboring PV system) or with scattered reference data (in plane irradiance obtained from application of solar models on satellite observations).