Soiling forecasting of solar plants: A combined heuristic approach and autoregressive model

• Published in: Energy (Oxford) Vol. 239; p. 122442
• Main Authors: Ballestrín, Jesús, Polo, Jesús, Martín-Chivelet, Nuria, Barbero, Javier, Carra, Elena, Alonso-Montesinos, Joaquín, Marzo, Aitor
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier Ltd 15.01.2022; Elsevier BV
• Subjects: Autoregressive model; Autoregressive models; Autoregressive processes; Complexity; Concentrating solar power (CSP); Energy loss; Forecasting; Grid integration; Heliostat; Heuristic methods; Heuristic model; Modules; Photovoltaic cells; Photovoltaic module; Photovoltaics; Power plants; Problem solving; Soiling; Soils; Solar power; Thermal power; Time series; Energy loss; Heuristic model; Soiling; Autoregressive model; Photovoltaic module; Grid integration; Heliostat; Forecasting; Concentrating solar power (CSP)
• ISSN: 0360-5442; 1873-6785
• DOI: 10.1016/j.energy.2021.122442

The soiling process of photovoltaic devices and heliostats is considered an important phenomenon to take into account in the design and operation of commercial Photovoltaic (PV) and Concentrating Solar-Thermal Power (CSP) plants, since in both cases the efficiency of these surfaces for solar use presents unexpected fluctuations. Many magnitudes and parameters influence in a complex manner the soiling process of an outdoor surface. In this work, it is assumed that a random disturbing source acts continuously on the surface causing its degree of soiling. Based on this assumption, a heuristic approach based on an analogous electrical model is proposed to simplify the complexity of the phenomenon. This model justifies the time series analysis of electrical losses due to soiling of a PV module by fitting an autoregressive-moving-average (ARMA) model to recorded time series at CIEMAT. This ARMA(1, 1) model allows predicting the average efficiency of this PV module in a period of 38 days with a Normalized Mean Bias Deviation (NMBD) of 1.35% and a mean relative error of 3.12%. This result can be improved by applying ordinary least squares to the model to minimize the NMBD obtaining a value of 0.17% and a mean relative error of 0.07%. •Soiling forecasting of solar plants.•ARMA models for soiling forecasting.•Heuristic approach to modeling the soiling process.