Outlier identification in outdoor measurement data: effects of different strategies on the performance descriptors of photovoltaic modules

Outdoor measurement campaigns of PV module performance are normally affected by a relatively large number of outliers. The aim of this paper is to develop a statistically sound approach of obtaining a dataset that allows one to analyse continuously monitored devices for use of high volume data measurements. This paper uses ISC as a self-reference parameter to measure the incident irradiance on the module, which largely eliminates the error due to differences in spectral and angular response between test and reference detector. The outlier identification procedure is based on statistical distribution analysis of different performance descriptors and it assures 0.99 confidence level and the same skewness for the remaining data. This approach can be applied to whole datasets as well as for data in specific irradiance-temperature bins. The developed methodology will be used to analyze outdoor data from different devices at different locations with reduced uncertainty. It is shown that this approach is particularly useful for obtaining lower uncertainties in low irradiance measurements.