Evaluation of photovoltaic system power rating methods for a Cadmium Telluride array

A variety of metrics are commonly used to assess whether or not a photovoltaic (“PV”) system is operating as expected, but to date no standard metric has been accepted. Three commonly used metrics for assessing PV system power performance are the Power Performance Index (“PPI”), PVUSA rating as contemplated in ASTM WK22009 (“ASTM”), and Performance Energy Ratio (“PER”). This paper evaluates the suitability of each of the three metrics for use with large Cadmium Telluride (CdTe) arrays. Of particular interest is the uncertainty and stability of each result and relative differences between their magnitudes. Two different approaches for propagating measurement uncertainty to final metric uncertainty are discussed: (1) analytical and, (2) bootstrapping (similar to a Monte Carlo method). Additionally, best practices to achieve low uncertainty and high stability of a metric are addressed including choice of regression method, reference conditions and filtering range. Iteratively reweighted least squares regression methods were found to improve the stability of metrics in cloudy climates relative to ordinary least squares methods. Choosing irradiance filtering ranges that are sufficiently large and asymmetrical about the chosen reference condition was found to bias the metrics on the order of 0.6%. Final PPI uncertainty was found to be most sensitive to irradiance and power measurement errors and ranged from +/- 3% to +/- 8% for typical ranges of sensor accuracies.