Development and experimental validation of a physical model for the soiling of mirrors for CSP industry applications

Highlights

• Development of a physical model that allows to predict soiling (reflectance) losses.

• The size distribution of airborne dust influences significantly its deposition.

• The concentration of airborne dust influences significantly the soiling losses.

• Model prediction of the reflectance trend validated with experimental data.

Abstract

Addressing soiling-related losses of glazed surfaces in solar power technologies is one of the most critical issues to improve the competitiveness of solar power plants in the current energy market. Reliable predictions of soiling rates and performance losses would enable optimisation of cleaning strategies and add a valuable assessment criterion for the evaluation of potential future power plant sites. However, currently available soiling models are data-driven and thus specific to the site and do not allow for reliable extrapolation to different sites and environmental conditions. To address this gap, this paper details the development of a physical soiling model for the prediction of the deposition of airborne dust onto the surface of solar collectors and the subsequent loss of performance (i.e. reflectance in the specific case of heliostats). The model inputs are the measured airborne dust concentration and estimated size distribution, the position of the mirrors, and the recorded wind speed and air temperature. The outputs of the model have been validated using experiments performed at the Queensland University of Technology. A reflectometer was used to measure the reflectance of the mirrors on an almost-daily basis while meteorological and dust data were obtained through a dust sensor and a weather station. The agreement between the experimental data and the simulated results demonstrates the effectiveness of the model in dry conditions. For example, for a mirror tilted at 45° with a daily loss of reflectance between 0.5%/day and 2.1%/day, the model reflectance loss predictions have an average relative error of 14%. Furthermore, the visual trends of the model predictions agree well with those observed in the experimental data. Finally, the model is used to investigate scenarios at three candidate sites to assess reflectance losses in different conditions.

Introduction

The exploitation of ‘clean’ or ‘green’ power is gaining global relevance as renewable energies are achieving a paramount importance in the energy supply sector of many of the most developed economies of the world (El-Katiri, 2014, International Energy Agency, 2017a, International Energy Agency, 2017b). New technologies have been developed in the last years and among them the harvesting of solar power is one of the most promising and available for improvements. Within the field of solar energy, photovoltaic (PV) technologies are achieving a great importance in the decentralized power generation while concentrated solar power (CSP) plants with storage systems are considered among the most suitable solutions for electricity generation on large scales for base load production (El-Katiri, 2014, Kolb et al., 2011, Simbolotti, 2013).

The trend towards the exploitation of energy sources different from traditional fossil fuels boosts the research in this field in an attempt to enhance these plants’ efficiency and make them competitive into the current energy market (Cohen et al., 1999). The optimization of the performance of a plant is obtained through both an optimal design and operation of the plant. In this context operation and maintenance (O&M) costs have a considerable impact, around 14–17% of the total levelized cost of electricity (LCOE) in a CSP plant (IRENA, 2012, Kutscher et al., 2010, Simbolotti, 2013). A key factor to lower the losses and hence to improve the power production is the cleanliness of the solar collectors, which has to be obtained through their artificial cleaning, since naturally soiled surfaces suffer weekly or monthly drop of reflectance of some percent (Mastekbayeva and Kumar, 2000, Roth and Anaya, 1980, Said, 1990), depending on their location and the environmental characteristics of the area. The loss of performance of heliostats is commonly expressed in terms of average reflectance losses mainly due to soiling of their surface. Studies have been performed on the topic by many authors in the last few years, also proving the growing relevance of these issues. Exhaustive literature reviews have been published recently on soiling related losses and mechanisms (Figgis et al., 2017, Ghazi et al., 2014, Maghami et al., 2016, Mani and Pillai, 2010, Picotti et al., 2018, Said et al., 2018, Sarver et al., 2013, Sayyah et al., 2014). Given the broader diffusion, researchers focused mainly on PV, conducting experimental studies in different sites, often trying to obtain semi-empirical models or to extract correlations between the measured soiling (either in terms of deposited dust or loss of performance) and a number of environmental factors, among which the most common are dust concentration (often PM₁₀), temperature, relative humidity, and wind speed (Adinoyi and Said, 2013, Boyle et al., 2015, Hammad et al., 2017, Javed et al., 2017, Lopez-Garcia et al., 2016, Mani et al., 2016, Paudyal and Shakya, 2016, Urrejola et al., 2016). Lu and Hajimirza (2017) also developed a model to compute the optimal tilt angle that, although losing some percentage of the incoming DNI, would allow one to obtain cleaner mirror thanks to higher removal of particles due to gravity. Figgis et al. (2016) developed an outdoor soiling microscope in order to better observe the phenomenon and obtain more reliable correlations. Studies concerning CSP technologies are less abundant and often related to pure experimental analysis (Bouaddi et al., 2017, Griffith et al., 2013, Merrouni et al., 2015) or modelling soiling losses based on experimental data through statistical analysis (Bouaddi et al., 2015).

To mitigate the detrimental effect of soiling, artificial cleaning is required. However, cleaning operations are expensive, require labour and water which may be scarce in the typical locations suitable for CSP application. Thus, they have a relevant impact on the O&M costs, accounting for about 20% (Kutscher et al., 2010). However, nowadays cleaning schedules are either time-dependent or planned regardless of the actual (or predicted) reflectance of the mirrors (Fernández-García et al., 2014, Hammad et al., 2017, Lopez-Garcia et al., 2016). Truong Ba et al. (2017) developed a condition-based method to optimise the cleaning of a solar field based on reflectance measurements and a statistical model of reflectance losses. The evaluation of the actual reflectance of the mirrors and its contribution in the decreasing of the solar field performance is then fundamental to optimising cleaning schedules to properly balance the energy output and cleaning costs.

Although addressing mainly PV technologies, Boyle et al. (2016) studied the deposition of dust onto glass plates, focusing on the estimation of the so-called effective deposition velocity through a straightforward comparison between airborne dust concentration and amount of deposited dust, and then through a more detailed theoretical model, described by Zhang et al. (2001) and similar to the approach adopted in this study. They conducted experiments to measure the amount of dust deposited on the surfaces and simultaneously collecting the PM₁₀ (particulate matter whose dimension in terms of diameter is smaller than 10 µm) concentration in air, together with meteorological data obtained from other organizations. Despite giving results in good agreement with the experiments related to horizontally deployed plates, the model still lacks precise accuracy for short term soiling assessments and it has not been compared with soiling of tilted plates. Moreover, considering PM₁₀ only rather than TSP (total suspended particles), the model neglects a significant percentage of airborne dust, hence biasing the prediction of the actual soiling. Furthermore, their comparison between observations and model is made considering a mass accumulation rate rather than studying the actual trend in time.

The main purpose of this study is then to describe the development of a physical soiling model which simulates the deposition of dust onto solar collectors’ surface and assesses the related losses of performance. The model differs significantly from the models available in literature since the reflectance losses are estimated through physical laws that describe and link the phenomena involved in the process. Since each parameter has physical meaning, they can be estimated from knowledge of the site and refined from some limited historical data once it is available. Previous models instead use historical data to derive correlations between input parameters and output values. Such correlations require extensive data and will likely extrapolate poorly to other sites since they are representative only of a precise location and its environmental conditions. The contribution of the study is therefore a model for the accurate prediction of soiling losses, which is validated using experimental data from naturally soiled outdoor-exposed mirrors at the Queensland University of Technology (QUT).

The first section of this paper describes the physical model developed to predict the net amount of dust particles deposited on the solar collectors in dry conditions, and its effect on their performance. The net accumulation of dust on the collectors’ surface will be a balance of a positive flow of airborne particles towards the surface and a negative flow of particles falling off the surface. Gravitational forces are the main drivers for both flows, and their magnitude relative to resistance forces (e.g. adhesion) determines the soiling rate. This balance will highly depend on particle dimensions, modelled by means of airborne particle size distributions.

The second section deals with the validation of the model, which for the first time is performed comparing directly the simulated reflectance trends with the actual reflectance trends measured among four different experimental time intervals, on five different sample mirrors. For all the aforementioned cases, simulations and experiments are in strong agreement, proving the effectiveness of the developed model.

The paper is concluded with a sensitivity analysis and a scenario analysis aimed at identifying typical ranges of expected soiling rates in different sites and environmental conditions.

Finally, the conclusion of the paper highlights the results obtained and discuss further possible studies and improvements.