Study on the cleaning and cooling of solar photovoltaic panels using compressed airflow

Solar photovoltaics (PV) are becoming one of the main sources of renewable energy to reduce carbon emissions of electricity supply. It is well recognised that dust accumulation and high temperatures result in a dramatic reduction in the performance of PV panels. To improve the efficiency of solar PV panels, a compressed air-based regulation method which can simultaneously clean and cool PV panels is studied and tested. A modelling study of the dust adhesion and detachment mechanism is conducted and the temperature variation caused by the air blowing process is analysed. Dynamic models of the compressed air release are derived which can be used to guide the design of the regulation system for increasing PV power output. A test system is developed for verifying various design and system parameters. The test results are used to validate the suitability of the modelling and illustrate how the inefficiency arising from soiling and high temperatures can be mitigated with the regulated compressed airflow. PV arrays serving in an arid region are adopted for this study and the increased energy yield arising from the cleaning and cooling effects is evaluated via the experimental test. The relationship between the airflow duration, various sizes of particles cleaning from the surface and power generation efficiency improvement is investigated to maximise the net power output increase from the PV panel. The results of this study can contribute to improving PV efficiency and help to realise decarbonisation in energy industry.


Introduction
Global solar power capacity increased from 25 GW at the beginning of 2010 to nearly 618 GW in 2019, and the overall investment in the solar energy sector within the Middle East and North Africa (MENA) region, which is ideal for photovoltaics (PV) installation, could reach $1 trillion between 2019 and 2023 (Middle east solar industry association, 2020). However, dust accumulation and high panel temperatures considerably reduce the performance of the solar panels, making them a less effective alternative energy source. For PV modules serving in the Eastern part of Saudi Arabia, the output power decreases by as much as 50% if uncleaned over a six-month period (Adinoyi and Said, 2013). In Kuwait, a 60% power reduction arising from panel soiling has been reported for the same duration (Sayigh et al., 1985). Additionally, PV modules convert only a small fraction of the solar irradiation into the electricity, the unconverted irradiation increases the temperature of the panel. When the temperature elevates, PV performance degrades because of the decrease in cell band gap (Said et al., 2018). For example, when heated from standard test temperature (around 20 ℃) to 64 ℃, an efficiency reduction of 69% occurs in monocrystalline PV (Malik and Damit, 2003). Therefore, to improve the efficiency of the PV panels, it is critical to mitigate the combined effect of soiling and heating.
Various methods have been adopted to clean the surface of PV panels. Washing with water is a traditional method that removes dust and also cools the panel (Moharram et al., 2013). Despite the effectiveness, water cleaning is not suitable for arid desert regions for largescale solar PV farms because of local water scarcity. Another basic practice for PV cleaning is manual or mechanically aided brushing (Al-Housani et al., 2019). However, rough brush cleaning can cause damage to the panel surface and lead to a reduction in efficiency and service life of the PV modules. Technologies such as surface vibration (Williams et al., 2007), acoustic waves (Alagoz and Apak, 2020), and electrodynamic dust shield (Mazumder et al., 2007;Kawamoto and Guo, 2018;Chesnutt et al., 2017) have been utilised to remove or prevent the dust falling on the solar cell and improve lifetime performance. However, relatively high costs and complex structures of these technologies limit their application in large scale, rural facilities. Additionally, a cooling effect is not available in these technologies and safety concerns in humid climates must also be considered. Therefore, a simple, anhydrous, and Cunningham correction factor C D drag coefficient C f skin friction coefficient c s specific heat of the panel surface (kJ kg − 1 K − 1 ) D pipe diameter of the pipeline (m) E abs absorption efficiency (m 2 g − 1 ) E scat scattering efficiency (m 2 g − 1 ) f near wall effect correction factor f m wall effect correction factor F ad adhesion force (N)  a  ambient  abs  absorption  ad  adhesion  air  air  back  back  C  capillary  clean  clean  cleaned cleaned  comp  compressor  cons  consumption  Cu  Cunningham  D  drag  dust  dust  E  electrostatic  G  gravitational  gain  gain  L  lift  motor  motor  p  particle  pipe  pipe  s  surface  scat  scattering  tank  tank  Top Top low-cost solution should be investigated to effectively mitigate the soiling and heating impacts on the PV panels. Using the turbulent airflow generated from the compressed air which neither consumes water nor makes physical contact with surface is an attractive PV cleaning method (Du et al., 2019). In addition to removing accumulated dust on the cell surface, the air can also help dissipate heat to keep the panel cool and thus increase the PV power output. This technology has a simple structure and relatively low cost, therefore, is suitable for a bulk solar PV farm in the MENA region. To achieve this, the theoretical guidance and an experimental basis for the design and control of the compressed air-based regulation system should be provided. However, few prior investigations have been conducted considering the simultaneous effect of cleaning and cooling using compressed airflow, and the gap between theoretical attempts and demonstration studies remains. Additionally, the aspects that influence the energy profit of this technology also need to be addressed to operate the system efficiently.
In this paper, mathematical modelling of the dust adhesion to the PV panel surfaces and the detachment under the turbulent airflow is studied. The temperature drop accompanying the cleaning process is analysed and the dynamic model of the compressed air discharging process is established. A test system of the regulation mechanism is developed to validate the proposed modelling and show the effectiveness of cleaning and cooling on the increase of PV power. Design of a regulation system for the PV arrays serving in arid regions is conducted and the improvement of the panel's performance is evaluated. In addition, the influence of the blowing duration and size of particles cleaned from the surface on the energy profit is investigated. After Conclusion, future work is suggested for further improvement of solar PV performance using compressed air-based regulation system.

Theoretical analysis and mathematical modelling
As depicted in Fig. 1, the compressed air-based regulation system has a simple structure, mainly composed of a compressed-air unit (a compressor, an air tank, and an air flow regulation valve) and nozzles. In a real application, a compressed-air unit can be designed for a group of PV arrays. The compressor is directly powered by the PV panels and the release of the compressed air from the tank is regulated by the valve to meet the mass flow requirements of cleaning and cooling. The spreading air from the nozzles installed at the edge of panels overlaps and forms a flake shape airflow, then carries away the dust and heat from the panel surface. Components used in the system are low-cost and highly-reliable standardised products. The system can be constructed using a fixed pipe assembly to transmit air to the panel and may be made as a mobile set of equipment to move to the location where cleaning and cooling is needed. This depends on applications and cost of the whole system.

Modelling of cleaning process
To facilitate the modelling work, the following assumptions are made first: (1). The dust particles are spherical, and the adhesion occurs between the dust and panel surface which are both made of SiO 2 .
(2). To fully utilise the gravitational force contributing to the dust removal from the tilting solar panel, the air blows over the panel surface in a tangential direction from top to bottom. (3). The adhesion between dust particles and the gravity effect on airflow are neglected.

Adhesion model
Referring to Fig. 2, the main adhesion forces subject by the dust particles deposited on the PV panel surface include Vander Waals (VdW) forces (F VdW ), Electrostatic forces (F E ), and Capillary forces (F C ). To receive more solar radiation, the PV panel inclined degree θ according to the located latitude (Shariah et al., 2002).

(a) Van der Waals forces
Arising from the interacting dipoles between two contacted bodies, VdW forces occur at the contact boundary of the dust particle and panel surface. In the low Relative Humidity (RH < 30%) and neutral electrostatic environment, this force can be regarded as the dominant force in adhesion (Li et al., 2006;Moutinho et al., 2017). When the surfaces of the particle and panel are smooth, VdW force can be written as (Hamaker, 1937) where A is Hamaker constant (=7 × 10 − 20 J for interaction between SiO 2 and SiO 2 ), R is the particle radius, and H 0 is the closest distance between the surfaces (=0.3 nm for minimum intermolecular distance).
In the real situation, the surface of dust particle and solar panel have some roughness. The existence of nanoscale roughness results in adhesion reduction because of a decrease in the actual contact area and an increase in the distance between the bulk surfaces. When a spherical particle interacting with a much smaller scale of asperity, the VdW force becomes (Rumpf, 1990): where RMS is the root-mean-square of the roughness.

(b) Electrostatic forces
The electrostatic forces acting on a charged dust particle can be either attractive or repulsive, depending on the conductivity of the panel surface and the permittivity of medium between the bulk surfaces (Brambilla et al., 2017). The force can be presented as (Crowley, 2008): where q is the particle charge, ε 0 is the vacuum permittivity (=8.854 × 10 − 12 C 2 N − 1 m − 2 ), ξ = H 0 /R, and K = 1 for the conductive plane.
(c) Capillary forces Capillary forces arise between the hydrophilic surface of the panel and the particle owing to water condensation. As shown in Fig. 2, the concave meniscus of water pulls the particle towards the panel. According to (Ilse et al, 2018;Figgis et al, 2018), the force can present even at lower RH (30-40%) and is represented by where ψ is the liquid's surface tension (=0.07275 N m − 1 ), and θ 1 and θ 2 are the contact angle.
In arid and semi-arid climates, the diameter of dust settling on the PV panel surface is in the range of 2-64 µm (Ilse et al., 2018). To have an intuitive view of the particle adhesion, the variation of the forces with different particle size was studied according to Eqs. (1)-(4). As depicted in Fig. 3, the adhesion forces rise with the increased diameters and the magnitude under a certain dimension falls in the following relationship: In a real situation, not all the above forces play important roles in the adhesion at the same time. When the humidity of surroundings increases, the VdW forces and electrostatic forces will considerably decrease because of the reduction of the Hamaker constant and the charge compensation (Ilse et al., 2018). On the other hand, calculated by Eqs. (1)-(2), the roughness of the panel surface induces a reduction of the VdW force. Taking the particle diameter 20 µm as an example, the VdW force decreased to 1.5% of that in the smooth condition at 100 nm RMS roughness. It is worth noting that for a very small or very large surface roughness, the situation is similar to a particle adhering to a smooth surface and the VdW force is close to the value calculated by Eq. (1).
Because of the repeated humid/dry cycles, needle-like fibrous structures form between particles and panel surface and result in cementation of dust (Ilse et al, 2016). In this case, mechanical brushing with water will come to the most feasible cleaning way (Gupta et al, 2019). Therefore, to prevent cementation, it is necessary to implement the cleaning before the ambient temperature becomes lower than the dew point at night (Kawamoto, 2020).

Detachment model
As shown in Fig. 2, when the air flows over the panel surface in a tangential direction, the drag force (F D ), rolling moment (M R ), and lift force (F L ) will exert on the dust particle. Referring to , the air drag force can be expressed as where C D is the drag coefficient, f = 1.7009 is the correction factor for the near wall effect (Goldman et al., 1967), ρ air is the air density, V m is the mean air velocity at the particle centre, and C Cu is the Cunningham correction factor. Among them: In Eqs. (6)- (8), V s is the shear velocity, v is the kinetic viscosity of air, λ is the molecular mean free path in the gas (=6.9 × 10 − 8 m), Γ is the coefficient of wall condition, and Re p is the Reynolds number of the particle which can be presented as (Burdick et al., 2005) Additionally, the rolling moment acting on the dust is given by : where f m = 0.94399 is the wall effect correction factor (Goldman et al., 1967). In addition, the lift force acting on the dust particle by the airflow can be obtained using the following equation : Apart from the adhesion forces, dust particles also suffer from the gravitational force which is where ρ is the dust density. Under the removing function of the forces and moments, three detachment modes for the dust particle are available: where F ad is the adhesion force, R r is the contact radius, and µ is the friction coefficient for the particle-surface interface.
To reach the shear velocity to remove the dust particle on the panel surface, the required air velocity needs to be determined. According to the air dynamics, the shear velocity and shear stress can be shown as: In the above equation, the skin friction coefficient is (Jiang et al., 2018) where the Reynolds number Re x = V air x/v. Combining Eqs. (14)-(16), the air velocity can be shown as In the arid environment, VdW force and electrostatic force are determined as the main adhesion forces. To evaluate the ideal cleaning mechanism with less air consumption, the required shear and air velocity for each detachment mode for the dust on the smooth surface were obtained and compared (e.g. Γ = 1.84; µ=0.2; x = L = 1.211 m, which was the length of the panel surface along the direction of the airflow). As shown in Figs. 4 and 5, the rolling mode was the favoured detachment mechanism for the dust which needed a lower airflow rate. On the other hand, the required velocity decreased with the increase of the particle size in each detachment mode. This reflects the fact that large size particles are easy to be removed than the smaller ones.
The influence of the dust deposition on the PV performance can be estimated using the reduction rate of the transmittance of visible solar energy to the PV module under per unit of dust mass (Bergin et al., 2017): Fig. 3. Variation of adhesion forces with different particle diameter (particle charge q = R × 2 × 10 − 12 C (Hinds, 2012), θ 1 = 60 • , θ 2 = 45 • ).
D. Li et al. where ΔS is the change of solar transmittance, m dust is the mass of the dust on the panel surface, E abs and E scat are the matter mass absorption and scattering efficiencies, respectively, and β is the particle upscatter fraction. Then, power improvement by the cleaning effect can be calculated as: where m cleand and P clean are the dust mass on the cleaned panel surface and power output from the clean solar panel, respectively. Cleaning rate by the air blowing can be evaluated as: where P cleand and P soiled are the power output from the cleaned and soiled solar panel, respectively.

Modelling of cooling process
When the airflow blows off the dust, it also takes away the heat from the panel. The cooling process can be expressed by the following equation: where m s and c s are the mass and specific heat of the panel surface; h is the average convective heat transfer coefficient; A s and T s are the heat transfer area and temperature of the panel surface, respectively; and T air is the airflow temperature. The convective heat transfer coefficient is presented as follows (Incropera et al., 2006): where k is the thermal conductivity and the Nu, Re, and Pr are the Nusselt number, Reynolds number, and Prandtl number, respectively. Under the initial condition t = 0 and T s = T 0 s , Eq. (21) can be obtained as follows at any moment during the cooling process: The PV power output improves with the decrease of the temperature and the magnitude can be quantified as where T a is the ambient temperature and η is the power decrease coefficient per temperature unit (Assi et al., 2012).

Modelling of compressed air release
Assuming no heat transfer occurs between the air tank and atmosphere, then the dynamic change of the pressure in the tank during discharging is: where γ is isentropic coefficient, T tank and V tank are the temperature and volume of the air tank, respectively, R g is the universal gas constant, N PV is the number of PV modules in the arrays, and ṁ air is the mass flow rate of the air. Considering the pressure loss in the pipeline, the initial air pressure in the tank is where Δt is the air blowing time and P noz is the working pressure of the nozzle. The pressure loss in the pipeline is represented as (Carello et al., 1998) P loss = 1.6 × 10 3 Q 1.85 air N PV L pipe D 5 pipe ⋅(P noz + P loss ) where Q air is the flow rate of the air, L pipe and D pipe are the length and diameter of the transmission pipeline, respectively. Assuming an isentropic compression process, power consumption to obtain the compressed air is (Hartmann et al., 2012): where m air is the mass of air, R A is individual gas constant for air, P a is the ambient pressure, and η comp and η motor are efficiencies of the compressor and motor, respectively.

Experimental setting up
To validate the effectiveness of the proposed modelling and mitigation of soiling and heating by the compressed airflow, as shown in Fig. 6   (a) and 6(b), a demonstration regulation system was developed and tested for monocrystalline PV panels operating in an arid climate. The light panel consisting of bulbs with different spectrums can simulate the radiation from the sunlight to the PV module. The tilting angle of the solar panel can be regulated by the adjustable frames A, B, and C, and the panel surface was always keeping the same parallel distance to the light panel. Frame D was used to fix solar panels and the total dimension of panel surface that can be tested was 1.3 m × 1.2 m. The main function of the measurement and control platform was to learn the variation of the power output and temperature of PV modules and regulate the airflow from the tank by controlling a solenoid valve. Referring to Fig. 6 (c), 8 thermocouples (A-H) were installed at the top and back side of the PV panel respectively to measure the average temperature of the module, which is The specific parameters related to the test rig were given in Table. 1.

Results and discussion
Design and control of a compressed air system for the PV arrays (composed of 12 panels described in the test rig) serving in an arid region of northwestern India was conducted. Referring to Fig. 7, the average size of the dust deposited on the panel surface was 20 µm and almost 90% of particles had diameters less than 30 µm (Nahar and Gupta, 1990;Bergin et al., 2017). The tilting angle of the panel was set to 30 • and the average temperature of the surface can reach up to 333 K.

Validation
The components influenced the experiment results including the uncertainties in the measurement of PV voltage (U 1 ) and current (U 2 ), panel temperature (U 3 ), air flowrate (U 4 ), and time (U 5 ). As shown in Table 2, the standard uncertainties related to the components were obtained according to the verification report of the sensors and measuring devices.

(a) Cleaning
As shown in Table. 3, talcum mainly composed of SiO 2 with an average diameter of 20 µm was adopted and deposited evenly on the panel through a sieve. The deposition rate of the particle was set as 0.5 × 10 − 3 kg m − 2 per day, and as shown in Fig. 8(a), the panel surface was covered by 5.3 g dust after two week-period operation. This led to a power drop of the tested PV module from 42.50 to 37.50 W at 303 K (PV load = 90 Ω). Referring to Fig. 8(b), two fan nozzles were installed on the top edge of the panel aiming to clean the whole panel surface using an airflow with an average thickness of 5 mm. To avoid the re-deposition of dust after the detachment, air blowing was implemented 5 s to remove the detached dust out of the panel surface under a flowrate of 1,370 L min − 1 which was determined by the modelling work presented in Section 2.1. Then the power output of PV was measured again at 303 K after   Fig. 7. Mass size distribution of the dust on the PV panel.
cleaning. To mitigate the effect of remaining dust on the evaluation of cleaning rate, the residual dust was cleaned completely by wiping after each test. The test was repeated three times and the results were shown in Table. 4. After cleaning, the power output of the PV module recovered to an average of 41.82 W and,according to Eq. (20), the cleaning rate reached 86.4%. The reasons for not achieving full power recovery can be illustrated in four aspects. Firstly, although most areas of the panel surface can be cleaned by the airflow, dead ends depicted in Part A, Fig. 8(b) exist at current nozzle arrangement. Secondly, arising from the air expansion along the panel, the air thickness in Part B, Fig. 8(b) was larger than the average value determined by the design parameters of the nozzle. It resulted in the velocity in this part was not high enough to detach the particles from the panel surface. Thirdly, the actual particles smaller than 20 µm in the sample may not have been removed. Last but not least, underestimation of the VdW force may happen for some particles on the surface where a smaller roughness was superimposed on the larger-hemisphere roughness (Rabinovich et al., 2000a(Rabinovich et al., , 2000b. Dust particles that were not removed by these four causes blocked the irradiation and continued to reduce the power output of the PV module. It is worth noting that, many finer particles will be settled on the panel surface in aerosol deposition while larger particle size can be found in the sieving method owing to the agglomerate formulation . Therefore, a slight decrease of the cleaning rate may be predicted in the field application comparing with that in the laboratory test using sieve deposition.

(b) Cooling
Referring to Fig. 9(a), when the clean panel was heated up from 303 to 333 K, the power output of the module decreased from 42.50 to 28.24 W. Then an airflow at 1,370 L min − 1 (297 K) was released to cool down the panel surface. From the cooling results shown in Fig. 9(b), after 130second cooling, the average panel temperature dropped to 315 K and the power output increased to 32.42 W. R-square value of fitting the measured temperature and simulated one obtained by Eq. (24) was 0.978.

Design and control
The parameters related to the dust on the panel surface in the arid region of northwestern India were: E abs = E scat = 0.02 m 2 g − 1 and β = 0.02 (Bergin et al., 2017). According to Eq. (18), the reduction rate of Table 2 Standard uncertainties for the components.   the transmittance of the solar energy caused by the dust deposition was − 4% g − 1 m − 2 . Taking the deposition rate of the dust as 0.5 × 10 − 3 kg m − 2 per day, after two week-operation, the power output of the PV module dropped by 28% compared to that from the clean panel. On the other hand, when the temperature of the surface reached up to 333 K, the corresponding output power of the clean PV module dropped to 60 W (η = 0.6% K − 1 ) (Assi et al., 2012). Using the same nozzle mechanism shown in Fig. 8 for each solar panel, to clean up dust particles from the size of 2 µm, the required initial pressure in the air tank under different blowing time and tank volume was obtained to inform the design of the compressed air system. Referring to Table 5, it can be learned that more initial air pressure was needed with the increase of blowing time and the reduction of tank volume. Taking the cleaning rate as 86.4% based on the experiment results, the performance improvement of a solar PV panel was studied and depicted in Fig. 10. After 10-second air blowing, the power output from the PV arrays increased from 567.4 to 741.5 W where the contribution of cleaning and cooling was 75.7% and 24.3% respectively. When the blowing time extended to 15 s and 20 s, the PV power improved to 758.2 W and 772.5 W, and the contribution of the cooling increased to 30.9% and 35.7%.
From the energy perspective, power consumption for producing the compressed air needs to be compared to the energy gain from the PV modules by the cleaning and cooling effects. Two following aspects influencing the energy Return of Investment (ROI) which can be calculated by the Eq. (30) were investigated to inform the formulation of an optimal control strategy.
where W gain is the energy gain from cleaning and cooling effects. To facilitate the analysis, W gain was referring to additional energy obtained before the reduction rate of the PV output by the dust deposition back to 28%.
(a). Blowing duration When the tank volume was set as 1 m 3 , and η comp = 0.8 and η motor = 0.98, the comparison of the energy consumption and gain from compressed-air regulation was conducted under different air blowing duration to clean up 2 µm-dust particles. A linear increase of dust deposition on the panel surface after cleaning was assumed and the reheating rate of the panel by the solar radiation (twice the radiation from the light panel in the test rig) after cooling was set as 0.5 K min − 1 according to the test results shown in Fig. 9(a). Referring to Fig. 11(a), the power improvement of the PV arrays increased from 0.17 kW to 0.19 and 0.21 kW when the blowing time extended from 10 s to 15 and 20 s because of the strengthening of cooling effect. The corresponding Fig. 9. PV performance in the process of heating and cooling.  temperature of the solar panel after cooling was 325.0, 321.8, and 319.1 K. However, as shown in 11(b), the energy ROI decreased from 9.8 to 8.3 and 7.2. The reason for this situation was the maintaining time of the cooling effect (i.e. tens of mins) had a much lower order of magnitude than that of the cleaning effect (i.e. more than 10 days) and the energy gain from the extension of the cooling process was much lower than the energy consumption for producing required airflow.
(b). Size of cleaning dust Referring to Fig. 5, cleaning larger dust particles needs lower air velocity which leads to less air consumption. When the air blowing time and tank volume was set as 10 s and 1 m 3 , the energy consumption of the compressor and improvement of PV performance from removing different size of dust and accompanying cooling were compared. As shown in Fig. 12(a), the more dust on the panel surface was cleaned, the more PV power output can be achieved. However, it can be learned from Fig. 12(b), removing dust from the size of 10 µm would lead to a better energy ROI comparing with cleaning the smaller particles at the same time. With further increase of the initial cleaning size, the ROI decreased gradually, and no energy profit was available when the cleaning started from dust larger than 40 µm.

Conclusion
Cleaning and cooling of a solar Photovoltaic (PV) panel using compressed airflow was studied and tested in this paper for the improvement of PV performance. Modelling work of the dust adhesion and detachment was conducted first to obtain the airflow rate to clean the dust particles. Then the temperature variation of the panel surface during the cleaning process was analysed to evaluate the accompanying power increase by the cooling effect. Dynamic modelling of the tank pressure in the discharging process was established to decide the tank volume and operation time of the compressed air system.
To demonstrate the effectiveness of the proposed modelling and system, an experimental rig was established for the monocrystalline PV modules operating in the arid climate. The power improvement of the PV panel through cleaning and cooling was validated. The cleaning rate from the current nozzle structure could reach 86.4% and R-square value for the temperature modelling fitting was 0.978. Based on the test results, the design and control of a regulation system for the PV arrays (12 panels) operating in an arid region of northwestern India were investigated. When the blowing time was set as 10, 15, 20 s, the power output of the PV arrays increased from 567.4 W by 30.7, 33.6, and 36.1% respectively. To develop an optimal control strategy for maximising energy profit, the influence of the air blowing duration and dust cleaning size on the energy Return of Investment (ROI) was investigated. On the one hand, although the release of airflow could continuously cool down (a) Power improvement (b) Energy ROI the panel surface and increase the PV power after cleaning, the ROI would decrease because the cooling effect can only maintain a relatively short time comparing to the cleaning effect and therefore the energy used for producing required airflow was much higher than the energy benefit from an increased cooling duration. On the other hand, cleaning particles including those with small size and mass would contribute to higher PV power output, however, may decrease system ROI because of the increase of energy consumption for the required detachment flowrate. Therefore, the blowing time and specific particle size for removals need to be determined considering the optimal balance between energy consumption in compressing air and energy gain from PV performance improvement for the application scenario studied. This study could provide theoretical guidance and an experimental basis to design a regulation system using compressed air for boosting the performance of solar PV installations and develop an optimal control strategy to maximise energy profit. The results would help improve the efficiency of renewable energy utilisation and reduce the emission of greenhouse gas from the energy industry.

Prospective
Based on the current research progress, future work for further improvement of solar PV performance using compressed air-based regulation system are as follows: • The mechanism of the nozzle (number, position, angle, etc) needs to be updated to increase the cleaning rate of airflow. • Maximum Power Point Tracking (MPPT) under different dust deposition rate and temperature of the PV panel needs to be integrated to further improve the PV efficiency. • For the long-term operation of PV arrays, the frequency and time of the regulation should be optimised considering the variation of solar radiation and load demand, and weather conditions, etc, to achieve more energy benefit and economic gain.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.