Endogenous Soiling Rate Determination and Detection of Cleaning Events in Utility-Scale PV Plants

Abstract

As the deployment rate of PV power plants continues to soar, the need for robust, scalable methods for performance analytics increases. In this paper, we demonstrate the usefulness of one approach for quantifying soiling rates in utility-scale PV power plants endogenously, i.e., directly from the production data. The temperature corrected performance ratio, normalized to a clean state, is used to derive the soiling ratio (SR). Cleaning events, caused by either rain or manual cleaning, are automatically detected by positive shifts in the running median of the SR time series. Soiling rates are then estimated by the rate of change of the SR between the cleaning events, which is determined by linear regression. The method is validated on data from three utility-scale PV power plants in the Middle East, yielding soiling rates that are in the range 0%-0.18%/day at least 50% of the time, with a median of 0.1%/day.