SQSL method for soiling detection
The Stochastic Quantifying Soiling Loss (SQSL) method, developed by CEA-INES, offers a way to measure soiling, based on usual monitoring data from a PV plant. It does not require environmental data or information about artificial cleaning operations. Instead, it calculates the soiling impact directly from the electrical data measured in solar power plants. Unlike other approaches, SQSL does not look for change points or clean-only days. It considers a performance metrics profile, classifying days into cleaning days, stable periods, and soiling periods based on the performance metrics.
SQSL considers a performance metrics (PM) profile. The PM is a ratio between relative current (DC current / STC DC current) and relative irradiance (G/ STC G). A PM profile should be composed of different periods, divided into 3 types: cleaning days, stable Period and soiling periods. Once each period family has been characterized, the Monte Carlo method is used to generate many random PM profiles based on the probability distributions of different parameters such as frequency and duration distributions, soiling rate distributions, etc. Statistical analysis of the distribution of Performance Metric profiles makes it possible to then estimate the most likely average soiling ratio as well as the uncertainty interval.
The following Figure shows the sequence of these different periods. The curve represents the evolution of the parameter that characterizes soiling. The circles indicate the days on which a cleaning effect was detected, with the diameter of each circle being proportional to the cleaning efficiency. The dots mark the days and the corresponding rainfall. The dotted line represents the sliding average of this rainfall over a 28-day period.
Another result is a mapping of the plant with different grey levels indicating the soiling rate: the darker, the dirtier:
