Soiling determination for parabolic trough collectors based on operational data analysis and machine learning - research findings

Aug 5, 2026

Advanced cleaning strategies for parabolic trough collectors at concentrated solar power plants maximize energy yield and minimize cleaning costs, but require knowing each collector’s current soiling level. This study develops a novel data-driven machine learning method for soiling estimation using gloss values as a surrogate. Operational and meteorological data from Andasol-3 with multiple time horizons are used to estimate soiling for every collector. Decision Tree delivers best performance (R²=0.77, MSE=6.14) and improves cleaning decisions versus a fixed schedule, increasing detected necessary cleanings and reducing unnecessary ones.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages20

Reading Time50 min

AccessFree to download

Introduction

Concentrated solar power and soiling impact

Need for efficient, data-based cleaning strategies

Data-driven soiling estimation approach

Solar Energy 259 (2023) 257–276 | Soiling determination for parabolic trough collectors based on operational data analysis and machine learning
Alex Brenner a,b,∗, James Kahn c,d, Tobias Hirsch a, Marc Rögere, Robert Pitz-Paalf,ba German Aerospace Center (DLR), Institute of Solar Research, Wankelstrasse 5, 70563 Stuttgart, Germany
b RWTH Aachen University, Chair of Solar Technology, Germany c Helmholtz AI, Germany
d Karlsruhe Institute of Technology (KIT), Steinbuch Centre for Computing, Hermann-von-Helmholtz Platz 1, 76344 Eggenstein-Leopoldshafen, Germany e German Aerospace Center (DLR), Institute of Solar Research, Paseo de Almería 73, E-04001 Almería, Spain
f German Aerospace Center (DLR), Institute of Solar Research, Linder Höhe, 51147 Cologne, Germany | | | | | --- | --- | --- | --- | | A R T I C L E I N F O | | A B S T R A C T | | | Keywords:
Parabolic trough Soiling
Machine learning Artificial neural network Concentrated solar power | | Advanced cleaning strategies for parabolic trough collectors at concentrated solar power plants maximize the yield and minimize the costs for cleaning activities. However, they require information about the current soiling level of each collector. In this work, a novel, data-driven method for soiling estimation with machine learning for parabolic trough collectors is developed using gloss values as a surrogate for soiling values. Operational data and meteorological data from the solar field Andasol-3 with changing time horizons are used together with various Machine Learning techniques to estimate the soiling of every collector in the field. The best results were achieved with a Decision Tree model, with a coefficient of determination of 􀁒2 = 0 .77 from the maximum value of 1 and a mean squared error of 􀁍􀁓􀁅 = 6 .14 for the determination of specific soiling values. A second metric to evaluate the quality of soiling predictions from the models classifies whether soiling is above or below a cleaning threshold was also investigated. Model results are compared to soiling measurements that indicate the need for cleanings. Cleaning recommendations are derived and compared with the current fixed-time cleaning schedule of Andasol-3. All models show an improvement over the cleaning schedule currently in use. The use of a Decision Tree model increases the detected necessary cleanings by 12.2 %, while the number of unnecessary cleanings are reduced by 14.3 %. This has the potential to reduce operational costs and increase the solar field yield. The dataset used in this work is made publicly available [https://doi.org/10.5281/zenodo.7061913](https://doi.org/10.5281/zenodo.7061913), along with the code to reproduce all results, which can be found at [https://doi.org/10.5281/zenodo.7554806](https://doi.org/10.5281/zenodo.7554806) . | | 1. Introduction An established source of renewable energy with the capability to deliver dispatchable electricity is concentrated solar power (CSP). CSP plants use direct solar irradiation, transforming it into thermal energy. The thermal energy can then be used as process heat or to run a power cycle and produce electricity directly. Most CSP sites are located in the sun belt region with high direct normal irradiance (DNI). However, these regions often have arid climates and a high dust load potential. High dust loads may lead to dust deposition on the CSP plant mirrors, an effect known as soiling. Soiling is a major source of performance loss in CSP solar fields, with 3%–4% reduction of solar power production, causing annual revenue losses of 3 to 5 billion € [1]. In comparison to photovoltaic (PV) systems, the losses due to soiling in CSP systems can be 8 to 14 times higher [2]. In general, CSP systems are directly exposed to the harsh environmental conditions and therefore soiling is always present. Since cleaning is costly, has a high water consumption, and speeds up the degradation of the mirrors, it should be reduced toa minimum. Y

FAQs

Why is soiling a critical problem for parabolic trough CSP plants?

Soiling causes dust deposition on mirrors, leading to performance losses and reduced solar power production. Because plants operate under harsh environmental conditions, soiling is continuously present.

What data and modeling strategy are used to estimate soiling levels?

The approach uses operational data recorded at the power plant and meteorological data from local sources. Machine learning models are trained using engineered features derived from measurement instrumentation.

How does the proposed model improve cleaning decisions compared with a fixed schedule?

Model-based recommendations outperform the current fixed-time cleaning schedule. The Decision Tree model increases detected necessary cleanings by 12.2% and reduces unnecessary cleanings by 14.3%, with the goal of lowering costs and increasing yield.

Aug 5, 2026

Upload by Olivia Brown