Solar panel system and method for self-cleaning

Abstract

Solar panel system and method for self-cleaning The disclosure covers a solar panel system (10) comprising: at least one solar panel (1); a self-cleaning mechanism (2) for the at least one solar panel (1); one or more panel sensors (4) configured to measure at least one performance parameter linked to an energy-production efficiency of the solar panel (1); environmental sensors configured to measure environmental data; a data acquisition module (6) configured to obtain the at least one performance parameter and the environmental data as a function of time; a control system (8) for controlling the self-cleaning mechanism (2), the control system (8) comprising: one or more processor configured to: predict a future weather pattern based on the obtained environmental data; and determine, based on the at least one performance parameter and the predicted future weather pattern, a time tA at which to activate the self-cleaning mechanism (2). The present disclosure also relates to a computer-implemented method for determining when a cleaning mechanism (2) of a solar panel (1) is to be activated.

Description

[0001] Title: Solar panel system and method for self-cleaning

[0002] TECHNICAL ASPECT OF THE INVENTION

[0003] The present invention relates to a method for determining when a self-cleaning mechanism of a self-cleaning solar panel is to be activated.

[0004] The invention further covers a solar panel system comprising solar panels with selfcleaning mechanisms.

[0005] BACKGROUND OF THE INVENTION

[0006] Solar energy plays an increasingly important role in the global transition to sustainable energy sources. Solar panels are an indispensable technology for converting sunlight into electricity. Their efficiency, however, depends heavily on the condition of the surface of the solar panels. Soiling as a result of dust, sand, bird droppings, leaves and other particles can significantly reduce the amount of light reaching the solar panels, leading to lower energy output.

[0007] Particularly in regions with high levels of air pollution, low rainfall or frequent sandstorms, soiling can accumulate rapidly, resulting in a dramatic drop in energy production. In some cases, this loss can amount to 20% or more of the nominal output of the panels. In order to prevent these losses, solar panels must be cleaned regularly.

[0008] One of the current cleaning methods used is manual cleaning, which is not optimal. Manual cleaning of solar panels is often time-consuming, costly and impractical for large-scale solar farms or installations that are difficult to access.

[0009] There are also automatic cleaning systems, such as water or brushing systems. These often operate at fixed intervals without taking into account the actual degree of soiling or environmental conditions.

[0010] This leads to unnecessarily high water and energy consumption, especially at times when cleaning is not strictly necessary. These automatic systems likewise do not take into account any natural cleaning sources, such as rain, which also leads to a waste of resources and increased costs.

[0011] These shortcomings highlight the need for a more efficient and environmentally friendly system that can optimise the cleaning of solar panels while minimising the waste of resources and energy.

[0012] SUMMARY

[0013] The aim of the present disclosure is to provide a solar panel system that manages the cleaning of the solar panels in an efficient and optimised manner, thereby avoiding unnecessary cleaning cycles and waste of resources. The system aims to maximise energy output by dynamically predicting the right time for cleaning based on environmental and performance data.The present disclosure is defined in the accompanying independent claims. The dependent claims define advantageous embodiments.

[0014] According to a first aspect of the disclosure, a computer-implemented method is provided for determining when a self-cleaning mechanism of a self-cleaning solar panel is to be activated.

[0015] The method comprises performing the following steps at specified control interval: - obtaining values of at least one performance parameter associated with the energy-production efficiency of the solar panel, the values of the performance parameter are obtained during a first time period prior to the control interval;

[0016] - obtaining environmental data measured during the first time period prior to the control interval, the environmental data comprises, but is not limited to, one or more of the following environmental parameters: temperature, humidity, air pressure, wind speed, wind direction, precipitation amount, dust concentration;

[0017] - predicting a future weather pattern based on the environmental data and / or based on a weather forecast provided by an external weather forecasting service, the future weather pattern comprises a prediction of the values of one or more of the environmental parameters for a second time period after the first time period;

[0018] - predicting the change in the values of the at least one performance parameter during the second time period, the prediction process uses a machine learning model that has been trained to establish correlations between the values of the performance parameter and the values of one or more environmental parameters;

[0019] - based on the predicted values of the performance parameter in the second time period, deciding whether cleaning is necessary within the second time period, and if cleaning is necessary, determining a time tA within the second time period for activating the self-cleaning mechanism.

[0020] By using environmental data together with a performance parameter of the solar panel associated with energy-production efficiency, it is advantageously possible, when determining a cleaning time, to take into account both the reduction in energy production caused by soiling and the weather conditions, such as dust levels, wind and precipitation, that may negatively or positively affect the soiling.

[0021] Advantageously, the method can accurately predict the optimal time for cleaning, allowing the solar panels to consistently operate at their highest efficiency levels.

[0022] The advantage of the method of the disclosure is that the decision whether or not to clean is not based on a single measurement at a single time, but on a prediction of the expected performance of the solar panel over a longer period in the future, taking into account futureweather patterns. For example, if the performance parameter has dropped at a certain time, this does not necessarily mean that cleaning needs to be activated immediately if the weather forecast indicates that the performance parameter will rise again as a result of rainfall.

[0023] Advantageously, the number of cleaning sessions can be kept to a minimum, while maximising the performance of solar panels.

[0024] In general, deciding whether cleaning is required and / or determining the time tA for activating the self-cleaning mechanism can be carried out by comparing the values of one of at least one predicted performance parameter with predefined performance criteria.

[0025] The method for determining when the self-cleaning mechanism of the self-cleaning solar panel is to activated is performed at each control interval. In embodiments, the control intervals are determined by a fixed control frequency, preferably the control frequency is equal to or greater than 1 / (TP2), with TP2 being the duration of the second time period. In other embodiments, the control intervals correspond to a specific time, the value of at least one performance parameter of the solar panel or the value of another performance parameter of the solar panel has fallen below a predefined threshold.

[0026] In embodiments, the method further comprises: predicting the effectiveness of cleaning in the second time period, the effectiveness of cleaning is an expression of a relative difference between the performance of the solar panel before and after cleaning at a specific time, preferably the effectiveness of cleaning is an expression of a relative difference between at least one performance parameter, or another performance parameter of the solar panel, before and after cleaning the solar panel.

[0027] Preferably, for these embodiments the effectiveness of cleaning is also predicted, the method comprises: deciding whether cleaning is necessary within the second time period, based on the predicted values of the performance parameter in the second time period and based on the predicted effectiveness of cleaning in the second time period, and if cleaning is necessary, determining a specific time tA within the second time period for activating the selfcleaning mechanism.

[0028] To predict the effectiveness of cleaning, the aforementioned machine learning model can be used, which predicts the performance parameter and which has been further trained to establish correlations between the values of the effectiveness of cleaning and the values of one or more environmental parameters. As an alternative, predicting the effectiveness of cleaning can use a second machine learning model, which is trained to establish correlations between the values of the effectiveness of cleaning and the values of one or more environmental parameters.

[0029] In embodiments, the decision whether or not to perform cleaning during the second time period comprises:

[0030] - calculating an expected energy production during the second time period if no cleaning wereperformed, and calculating an expected energy production during the second time period if cleaning were performed at a specific time within the second time period, and, - if the energy-production gain with cleaning compared with the energy production without cleaning is less than a predefined energy threshold, deciding not to perform cleaning during the second time period.

[0031] According to a second aspect of the disclosure, a solar panel system is provided comprising: at least one solar panel; a self-cleaning mechanism for at least one solar panel; one or more panel sensors configured to measure at least one performance parameter associated with an energy-production efficiency of the solar panel; environmental sensors configured to measure environmental data, and the environmental data comprise, but are not limited to, one or more of the following environmental parameters: temperature, humidity, air pressure, wind speed, wind direction, precipitation amount, dust concentration; a data acquisition module linked to the panel sensors and the environmental sensors and configured to measure the values of at least one performance parameter and the values of the environmental data as a function of time, and the data acquisition module comprises a storage medium for storing the measured values; and a control system for monitoring the self-cleaning mechanism, and the control system comprises a processor configured to determine, using a machine learning model and using at least one performance parameter and the environmental data, a specific time tA at which the self-cleaning mechanism is to be activated, and the controller is configured to activate the selfcleaning mechanism at the determined time tA.

[0032] Advantageously, the system can adapt to changing environmental and weather conditions, such as dust storms or seasonal soiling events, making it suitable for a wide range of locations and situations. Indeed, the ML model is continuously learning using the selfrecorded local data, so that the ML model automatically adapts to the specific environment of the geographical area where the solar panels are installed.

[0033] Advantageously, the method together with the system can be integrated into large-scale commercial solar farms as well as residential installations, thereby serving a broad target group.

[0034] The self-cleaning mechanism is preferably a vibrating mechanism. This has the advantage that water or chemicals do not need to be used, and the system can therefore also be used in arid areas. It also means that the solar panel system is completely autonomous, both in terms of deciding when to clean and in terms of performing the cleaning.

[0035] In embodiments, the solar panel may be provided with a nanocoating to repel dust and dirt.

[0036] In embodiments, the control system of the solar panel system can be linked to anexternal weather forecasting service or an external weather forecasting station to obtain a future weather pattern, the future weather pattern comprises a prediction of the values of one or more of the environmental parameters.

[0037] In other embodiments, the solar panel system can comprise a weather prediction module configured to determine a future weather pattern, the future weather pattern comprises a prediction of the values of one or more of the environmental parameters, preferably in which the future weather pattern is determined based on the measured environmental parameters.

[0038] In an aspect of this disclosure there is provided a computer-implemented method for determining when a self-cleaning mechanism of a solar panel is to be activated, the method comprising, at specific control intervals: obtaining at least one performance parameter associated with the energy-production efficiency of the solar panel, where the at least one performance parameter is obtained during a first time period prior to a control interval; obtaining environmental data measured prior to the control interval; predicting a future weather pattern for a second time period after the first time period, the future weather pattern based on the obtained environmental data; predicting a change in the at least one of the performance parameter during the second time period based on the predicted weather pattern; determining whether to perform cleaning within the second time period based on the predicted change in the at least one performance parameter for the second time period; and if cleaning is to be performed in the second time period, determining a time tA within the second time period to activate the self-cleaning mechanism.

[0039] In an embodiment of the present disclosure there is provided a solar panel system comprising: at least one solar panel; a self-cleaning mechanism for the at least one solar panel; one or more panel sensors configured to measure at least one performance parameter linked to an energy-production efficiency of the solar panel; environmental sensors configured to measure environmental data; a data acquisition module configured to obtain the at least one performance parameter and the environmental data as a function of time; a control system for controlling the self-cleaning mechanism, the control system comprising: one or more processor configured to: predict a future weather pattern based on the obtained environmental data; and determine, based on the at least one performance parameter and the predicted future weather pattern, a time tA at which to activate the self-cleaning mechanism.

[0040] In a further aspect of the present disclosure there is provided a solar panel system comprising: at least one solar panel; a self-cleaning mechanism for the at least one solar panel; one or more panel sensors configured to measure at least one performance parameter linked to an energy-production efficiency of the solar panel; environmental sensors configured to measure environmental data; a data acquisition module configured to obtain the at least one performance parameter and the environmental data as a function of time; a control system for controlling the self-cleaning mechanism, the control system comprising one or more processor configured to determine, using a machine learning model, and based on at least oneperformance parameter of the solar panel and environmental data obtained over time, a time tA at which the self-cleaning mechanism is to be activated, wherein the determination of the time tA is performed autonomously by the control system.

[0041] In an aspect of this disclosure there is provided a computer-implemented method for determining when a self-cleaning mechanism of a solar panel is to be activated, the method comprising, at specific control intervals: obtaining at least one performance parameter associated with the energy-production efficiency of the solar panel, where the at least one performance parameter is obtained during a first time period prior to a control interval; obtaining environmental data measured prior to the control interval; predicting a future weather pattern for a second time period after the first time period, the future weather pattern based on the obtained environmental data; predicting a change in the at least one of the performance parameter during the second time period using a machine learning model; determining whether to perform cleaning within the second time period based on the predicted change in the at least one performance parameter for the second time period; and if cleaning is to be performed in the second time period, determining a time tA within the second time period to activate the selfcleaning mechanism.

[0042] In embodiments, the machine learning model is one of - or a combination of - the following: random forest, recurrent neural networks or gradient boosting.

[0043] BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The invention will be explained in more detail with reference to the drawings, on the basis of examples of embodiments.

[0045] Fig. 1 is a schematic representation of an example embodiment of a solar panel system according to the present invention, comprising at least one solar panel with a selfcleaning mechanism and a control system for activating the self-cleaning mechanism at a time determined by a processor.

[0046] Fig. 2 is a flow diagram illustrating the process steps of an embodiment of a method for determining when a self-cleaning mechanism of a self-cleaning solar panel is to be activated according to the present invention;

[0047] Fig. 3 is a schematic representation of an example of a timeline in which, repetitively on control intervals til, ti2, ti3, ti4, ti5, a decision is made whether or not to perform cleaning in a subsequent time period following the control interval at a predicted time tA1 , tA2, tA3;

[0048] Fig. 4 is a flowchart that provides an overview of the process steps of a further embodiment of a method according to the present invention.

[0049] Figs. 5 to 7 illustrate example solar panel systems according to this disclosure.

[0050] The drawings are not drawn to scale and are not in proportion. In general, identical parts in the drawings are indicated using the same reference numbers.DETAILED DESCRIPTION

[0051] The present disclosure is presented in terms of specific embodiments that are merely illustrative of the disclosure and are not to be construed as limiting. It will be apparent to a person skilled in the art that the present disclosure is not limited to what is specifically shown and / or described, and that alternative or modified embodiments can be developed in light of the general tenet of this disclosure. The drawings provided are only schematic and non-exhaustive.

[0052] For the sake of clarity and conciseness, elements are described as part of the same or different embodiments; however, it should be clear and understood that the scope of the invention can comprise embodiments with combinations of all or some of the elements described.

[0053] It should also be understood that the embodiments shown have the same or similar elements, except where they are described as being different.

[0054] The use of the verb ‘comprise’ and its respective conjugations does not exclude the presence of elements other than those mentioned. The use of the definite and indefinite article ‘the’ or ‘a’ before an element does not exclude the presence of multiple of such elements.

[0055] Furthermore, the terms first, second, etc. in the description and in the claims are used to distinguish between similar elements and not necessarily to describe a sequence, whether in time, space, rank, or in any other way. It is to be understood that the terms used in this way are interchangeable in appropriate circumstances and that the embodiments of the disclosure as described here are capable of functioning in series other than those described or illustrated here.

[0056] Where this specification refers to “one embodiment” or “an embodiment”, this means that a particular function, feature, structure or characteristic described in connection with the embodiments, is comprised in one or more embodiments of this disclosure. Therefore, instances of the phrases “in one embodiment” or “in an embodiment” appearing in various places in this specification do not necessarily all refer to the same embodiment, but could do so as each embodiment is provided as an example and the features should be understood as combinable.

[0057]

[0058] Fig. 1 provides a schematic representation of an example embodiment of a solar panel system 10 according to the present invention.

[0059] The solar panel system 10 comprises one or more solar panels 1, and each solar panel 1 may have a self-cleaning mechanism 2. Each solar panel may comprise at least one glass layer and one or more photovoltaic cells, without being limited thereto. In some embodiments, the self-cleaning mechanism 2 can be a vibration mechanism 2, 20, preferably an ultrasonic vibration mechanism. When the vibration mechanism 2, 20 is activated, contaminants are loosened from the surface of the solar panel without the need for water or mechanical brushes.Fig. 5 illustrates a solar system comprising a vibration mechanism 2, 20 according to this disclosure. The vibration mechanism 2, 20 may include one or more mechanical elements, which will now be described, however it should be understood that the vibration mechanism is not limited to these elements alone. The vibration mechanism 2, 20 may be mechanically coupled to the solar panel 1. The vibration mechanism may, in some cases, be removably connected to the solar panel. Such a connection may be at a front side surface, a side surface and / or a back surface of the solar panel. The vibration mechanism may be comprised of one or more of a piezoelectric actuator, a piezoelectric transducer, an electromagnetic actuator, an electromagnetic microactuator, an Eccentric Rotating Mass (ERM) motor, piezo stack, piezo ring and a flexible piezoelectric film, which in some circumstances may be mechanically bonded to the panel frame or panel itself. The vibration mechanism may further comprise an actuator, a motor, and / or stack. When a cleaning event is triggered, in other words when it is determined that cleaning is to be performed and a vibration signal is generated and / or predetermined, said signal may be sent to the vibration mechanism. Once received the vibration mechanism may be driven to ease detachment of dust or particulates. The vibration signal may be a signal that is fixed so as to cause the mechanism to vibrate at a set frequency or range of frequencies and may be determined or set at the time of installation.

[0060] The piezoelectric film 30 may be a polymer-based piezoelectric layer (like PVDF or a copolymer thereof) which has conductive electrodes. In such a configuration, the piezoelectric film may be mechanically attached along one or more edge of the panel frame. This piezoelectric film is configured to bend and slightly contract when AC voltage is applied across the film, which will transmit mechanical oscillations into the frame and panel laminate thereby losing and removing debris / material from the solar panel surface. The advantage of this technique is that such a configuration and piezoelectric film is low mass and low cost and can be implemented in various panel configurations.

[0061] Fig. 6 illustrates an example solar panel system in which a piezoelectric film 30 may be bonded to one or more frame edge of a solar panel 1. The solar panel 1 may be held by a frame (not shown) around one or more of its edges and the piezoelectric film 30. The piezoelectric film 30 may be connected in a permanent or removable manner to one or more of the frame portions at one or more of the edges of the solar panel 1. Alternatively, the piezoelectric film may be connected to the front surface, side surface of back surface of the solar panel 1 and may in some cases be adhesively connected or mechanically connected to the solar panel. In some cases the piezoelectric film is flexible and / or bonded to the edge of the frame.

[0062] Fig. 7 illustrates an example solar panel system in which a piezoelectric film 30 may be bonded to one or more frame edge of a solar panel 1 and an additional vibrational mechanism 20 may be connected, in some cases releasably, to the solar panel 1. The example of figure 7 may be a combination of the examples shown in Fig. 5 and 6. In the example of Fig. 7 the vibrationmechanism 20 may provide a vibrational transfer, in other words provide a vibration to, the piezoelectric film 30 and / or the solar panel 1. The piezoelectric film 30 may also provide vibration to the solar panel in tandem with the vibration mechanism 20.

[0063] It should be understood that the term vibration mechanism as used in this disclosure may be used to describe a piezoelectric film 30 when no additional vibration mechanism, such as an actuator and motor configuration, is provided.

[0064] In some embodiments, the solar panel 1, in combination with, for example, a vibration mechanism 2, 20, may also comprise a nanocoating 3 applied to the surface of the solar panel to repel dust and dirt. The nanocoating that will reduce adhesion of dust, particulates and other contaminants, and will facilitate cleaning. The term nanocoating used herein refers exclusively to a material layer applied to at least a portion of the surface of the panel, and expressly excludes any surface modification produced by structuring, laser processing, or other topographical alteration of the substrate. Such nanocoatings may be applied by spraying, dipping, deposition or film bonding.

[0065] The nanocoating 3 is for example a hydrophobic and oleophobic nanocoating. Preferably, this coating is UV-resistant, ensuring that it remains effective throughout the lifespan of the solar panels, even after prolonged exposure to sunlight. The coating is also transparent and designed to have minimal impact on the light-absorption efficiency of the underlying solar cells. Hydrophobic nanocoating may comprise hydrophobic silica-based nano-layers, which are SiC>2 nanoparticles spread in a polymeric binder. This configuration creates low surface energy and water-beading effects on the solar panel surface. The nanocoating may also be comprised of alumina (AI2O3) or zirconia (ZrCh) nanoparticle layers, which are durable and scratch-resistant films. Other nano-texture polymer films may also be utilised. In some cases fluoropolymer nanocoatings, such as PFTE-, PCDF-or fluorosilane-modified coatings may be applied to the solar panel. In some cases, titania (TiO2) photocatalytic nanocoatings may be applied, which will self-clean under UV exposure and which will decompose organic contaminants. Each of these coatings may be applied alone or in combination and may be applied to all or a portion of the surface of the solar panel. In some cases, different coatings may be applied to different regions of the solar panel.

[0066] Some methods for applying the nanocoating will now be described, however it should be understood that other methods of applying nanocoatings may also be utilised. Nanocoatings may be applied using one or more of the following methods: a spray coating may be applied by utilising aerosols on panel glass to achieve a coating. Alternatively, nanocoating may be applied by applying liquid to a surface, in other words applied as a dip-coat or flow-coat. This technique involves using immersion of coating solution across the surface. A further method may include the lamination of nano-textured films. In this case a pre-formed polymeric nano-film will be bondedto the surface of the panel. It should be understood that the application of nanocoatings in this disclosure are not limited to these techniques and may comprise any combination of these techniques as well as others. For each solar panel, the solar panel system comprises one or more panel sensors 4 configured to measure at least one performance parameter P1 (t), P2(t), ...Pi(t) linked to an energy-production efficiency of the solar panel. These performance parameters are measured as a function of time, and each parameter value is linked to a time t.

[0067] In the embodiments, one or more panel sensors 4 can be integrated into the solar panel.

[0068] Examples of performance parameters P1(t), P2(t), ...Pi(t) are: light transmission through the panel, the soiling level of a panel, an energy-conversion efficiency of the panel and an energy production.

[0069] One example of panel sensors are optical sensors that measure the intensity of light passing through the surface of the solar panels and reaching the photovoltaic cells. These sensors can therefore measure the light transmission through the panel, and this light transmission is a measure of the soiling level of the panels.

[0070] In another example of panel sensors are sensors that perform a direct measurement of dust accumulation. These sensors can therefore directly measure the soiling levels of the panels.

[0071] The soiling level of the solar panels can also be derived from the energy-conversion efficiency, which is the ratio between the electrical energy produced by the solar panel and the amount of solar energy that falls on the panel.

[0072] The measurement of energy production can also be used as a performance parameter; however, this can be affected by different weather factors.

[0073] The solar panel system further comprises one or more environmental sensors configured to measure environmental data E1(t), E2(t), ...Ei(t). These environmental data are measured as a function of time, and each measured parameter value is linked with a time t.

[0074] Examples of environmental data E1(t), E2(t), ...Ei(t) are one or more of the following environmental parameters: temperature, humidity, air pressure, wind speed, wind direction, precipitation amount and dust concentration. Dust concentration is, for example, the amount of particulate matter PM2.5 in the air or another parameter that indicates a dust concentration.

[0075] The environmental parameters may depend on weather conditions 15 such as storms, rain, sun, snow, hail, sandstorms and any other possible weather conditions that may occur in a particular geographical area. The environmental parameters may also depend on day or nighttime, the time of year or the seasons.In some embodiments, humidity sensors can, for example, use capacitive or resistive elements to detect humidity levels, temperature sensors can, for example, use thermocouples or thermistors to measure changes in ambient temperature, or air quality sensors can use gas detectors to measure polluting particles in the air.

[0076] The solar panel system 10 further comprises a data acquisition module 6 that is linked to the panel sensors 4 and the environmental sensors 5. The data acquisition module is configured to obtain, in some cases by measurement, the values of one or more performance parameters and the values of the environmental data as a function of time. Preferably, this measurement is made in real time. The measured values may then be stored in a storage medium 7. In some embodiments, the storage medium is a computer memory, such as RAM memory.

[0077] The solar panel system 10 further comprises a control system 8 for monitoring the self-cleaning mechanism 2. In particular, the control system 8 can activate the self-cleaning mechanism 2 to start a self-cleaning cycle. For this purpose, the control system 8 comprises a processor 9, also referred to as a central processing unit, which is configured to execute a software programme that calculates the specific time tA at which the self-cleaning mechanism 2 is to be activated.

[0078] The software algorithm uses a machine learning model, ML model, to predict a moment in time tA when the self-cleaning mechanism is to be activated during a second time period TP2 following the first time period TP1.

[0079] The ML model uses at least one performance parameter as input data, as well as the environmental data measured during a first time period and / or predicted environmental data for the second time period.

[0080] In some examples, the method may further comprise: predicting the effectiveness of cleaning in the second time period, wherein the effectiveness of cleaning is an expression of a relative difference between the performance of the solar panel before and after cleaning at a specific time, preferably wherein the effectiveness of cleaning is an expression of a relative difference between at least one performance parameter, or another performance parameter of the solar panel, before and after cleaning the solar panel. In some cases predicting the effectiveness of cleaning uses the aforementioned machine learning model that has been further trained to establish correlations between the values of the effectiveness of cleaning and the values of one or more environmental parameters, or, alternatively, predicting the effectiveness of cleaning uses a second machine learning model, which is trained to establish correlations between the values of the effectiveness of cleaning and the values of one or more environmental parameters.

[0081] In some embodiments the method may further comprise determining when toperform cleaning in the second time period based on the predicted change in the at least one performance parameter in the second time period and based on the predicted effectiveness of cleaning in the second time period, and if cleaning is to be performed, determining a time tA within the second time period for activating the self-cleaning mechanism. In some cases, an additional time At, may be determined on the basis of the predicted effectiveness of cleaning. In some cases, At is a predefined configuration value or whereby At is determined based on the weather pattern, which may be based on environmental data. In some cases, At is a predefined configuration value set during installation or commissioning and / or whereby At is determined based on a predicted evolution of at least one performance parameter after a fictitious cleaning event, rather than being determined from a predicted weather pattern. In some cases, the effectiveness of cleaning may have a value between 0 and 1, where a value of 1 indicates that cleaning will maximise the at least one performance parameter of the solar panel after cleaning, and where a value of 0 indicates that after cleaning, the value of the performance parameter will not have changed compared to the value of the at least one performance parameter before cleaning.

[0082] In some embodiments, the control system can obtain weather information from an external weather forecast service. For example, the control system of the solar panel system can be linked, preferably wirelessly, to an external weather forecasting service or an external weather forecasting station to obtain a future weather pattern, the future weather pattern comprises a prediction of the values of one or more of the aforementioned environmental parameters.

[0083] Examples of weather forecasting services are OpenWeatherMap, AccuWeather, or a regional weather service.

[0084] This external service provides data such as expected rainfall, wind speed and humidity for specific locations. This weather information is then passed on to the machine learning model as input data, as discussed further below.

[0085] In embodiments, the one or more processor can perform a weather forecasting algorithm that extrapolates from historical weather data sets to determine a future weather pattern. For example, the system can analyse trends in rainfall measured over recent days and extrapolate these to the future. This can take into account factors such as the time of year, for instance: rainy season, as well as previous patterns. The advantage of this extrapolation-based algorithm is that it may not rely on external services and / or external weather forecasts. The disadvantage is that it could be less accurate in the event of sudden weather changes.

[0086] In embodiments, the solar panel system can comprise its own weather prediction module that makes weather predictions based on locally measured environmental parameters, such as temperature, air pressure, wind speed, wind direction and humidity. This way, a future weather pattern is obtained, the future weather pattern comprises a forecast of the values for one or more of the aforementioned environmental parameters.

[0087] The weather forecast module can use the same or an additional ML model to predictweather changes. For example, a drop in air pressure can indicate an approaching rain storm. Wind speeds and directions can be combined with seasonal trends to predict rainfall. The time of year can also be taken into account in order to model seasonal influences. The advantage of using its own weather prediction module is that the system is completely independent and locally optimised, and can therefore be highly specific to the location of the solar panels.

[0088] In embodiments, the software algorithm uses the ML model or a second ML model to predict the effectiveness of cleaning, this prediction of the effectiveness of cleaning helps determine the specific time tA for activation of the self-cleaning mechanism. The effectiveness of cleaning is an expression of the relative difference between a performance of the solar panel before and after cleaning, , preferably, the effectiveness of cleaning is an expression of the relative difference between the performance parameter after and before cleaning the solar panel.

[0089] The control system 8 then activates the self-cleaning mechanism 2 using an activation signal SA at moment in time tA determined by the algorithm, as shown schematically in Fig. 1. The control system may comprise one or more processors.

[0090] In embodiments, the control system 8 comprises an internal clock, such as a real-time clock, to keep time.

[0091] The computer-implemented method performed by the software algorithm is discussed in more detail below.

[0092] Computer-implemented method

[0093] Fig. 2, shows a flowchart outlining a number of process steps of an example embodiment of a method for determining when a self-cleaning mechanism of a self-cleaning solar panel is to be activated.

[0094] The method with the various process steps is performed repeatedly at set control intervals ti.

[0095] In embodiments, the control intervals are determined by a fixed control frequency and the controls are therefore always performed after a predefined period of time. The frequency of controls is usually a configuration value. Depending on the embodiments, controls can be performed daily, weekly or at fixed intervals.

[0096] In other embodiments, the control interval ti can correspond to a time at which a performance parameter of the solar panel falls below a predefined threshold. In this embodiment, the control is still repeatedly performed at the control intervals ti, but the duration between two consecutive controls can be variable.

[0097] Fig. 3 is a schematic representation of an example of a time line in which controls are repeatedly performed at control intervals til, ti2, ti3, ti4, ti5 and a decision is made whether or not to perform cleaning at a specific predefined time tA1, tA2, tA3 in a subsequent time period following the control interval. For example, at moment in time ti 1 , a control is performed based ondata obtained in the time period TP1 for moment in time til, and a decision is made whether or not cleaning is required in the subsequent period TP2, after moment in time til. In this example, it has been decided to perform cleaning at moment in time tA1 in the period following control interval til. At time ti2, a second control is performed based on data obtained from time period TP1 for time ti2, which in this example was period TP2 for control interval til. After performing a control at moment in time ti2, it was decided not to perform cleaning in the period following moment in time ti2. For control intervals ti3 and ti4, in this example, it was decided to perform cleaning at the predicted times tA2 and tA3 that follow moments in time ti3 and ti4, respectively.

[0098] In embodiments, the repetitive execution of the cleaning controls can occur at a frequency that is equal to or greater than 1 / (TP2), with TP2 being the duration of the second time period.

[0099] A first process step 100, shown schematically in Fig. 2, relates to data input, and a second process step 200 relates to processing the input data and / or determining derived features. Depending on embodiments, this can be considered as one or two steps.

[0100] In a process step 100, the data from the panel sensors P1 (t), P2(t), ... Pi(t) and the data E1(t), E2(t), ...Ei(t) from the environmental sensors are read. These values from the panel sensors and environmental sensors are therefore values measured before the moment in time ti when the control is performed. The values are obtained, for example, in a first time period TP1 before the moment in time ti.

[0101] The method according to the present disclosure comprises obtaining values of at least one performance parameter P1(t) associated with the energy-production efficiency of the solar panel, where the values of the performance parameter are obtained during the first time period TP1 prior to control interval ti.

[0102] As discussed above, performance parameters that can directly or indirectly detect the soiling of solar panels comprise, for example: light transmission through the panel, soiling level of the panel, the energy-conversion efficiency of the panel and an energy production. Other examples of performance parameters may include a performance ratio which may define a ratio of actual energy output to theoretically expected output under given irradiance; solar panel temperature defining the temperature of one more panel; an electrical power output; an energy yield over a time interval; the conversion efficiency of the solar panel; a maximum power point current; and / or a soiling-related performance loss.

[0103] The method according to the present disclosure further comprises obtaining environmental data measured during the aforementioned first time period TP1 prior to control interval ti. This environmental data comprises, but is not limited to, one or more of the following environmental parameters: temperature, humidity, air pressure, wind speed, precipitation amount, dust concentration.

[0104] In some embodiments, the data from the panel sensors and environmental sensorscould be raw data. Depending on the embodiments, this raw data may first be converted into a modified format that can be used for input into the ML model. Optionally, this processing of the input data can be performed in step 200.

[0105] Optionally, derived features can also be determined in step 200. Optionally, for example, based on an incoming performance parameter P1 (t) of the solar panel, such as the light transmission through the panels, a solar panel feature dP1 / dt can be determined, such as the feature of the rate of soiling accumulation on the panels. This contamination rate dP1 / dt can be derived from the change in light transmission. Such derived features can also be considered as additional performance parameters.

[0106] In some embodiments, time-related features such as season, day or night-time can also be determined by using an internal clock of the control system or the processor. These time-related features can then be linked to the input data.

[0107] The method further comprises determining a future weather pattern based on the aforementioned environmental data and / or based on a weather forecast received from an external weather forecasting service or external weather station. The weather pattern comprises a prediction of values PR1 (t’), PR2(t’), ... PRi(t’) of one or more of the aforementioned environmental parameters for a second time period TP2, following the first time period TP1. The time t' refers to a time in the second time period TP2.. In some cases, the future weather pattern may be understood to be predicted environmental data which may include data obtained from a weather forecasting service or an external weather station.

[0108] In some embodiments, the environmental data is measured during a first time period TP1 of, for example, one week, and a prediction is made of the weather pattern for the second time period TP2 of, for example, a second week following the first week. In other embodiments, the first and second periods may be shorter or longer.

[0109] Process step 300 relates to the use of the ML model to predict the performance of the solar panel.

[0110] The ML model uses input data of at least one performance parameter linked to the energy-production efficiency of the solar panel as discussed above and the predicted weather pattern for time period TP2 following a control interval ti.

[0111] The ML model can also use additional parameters as input, such as a solar panel feature, such as the rate of soiling accumulation, derived on the basis of at least one performance parameter or on the basis of a second or more performance parameters obtained from panel sensors.

[0112] As specified above, the ML model can also use time-related parameters as input data, such as: seasons and day or night-time.

[0113] The ML model further uses as input data the predicted weather pattern for the second period TP2 after taking the reading at moment in time ti. As discussed above, the predictedweather pattern comprises a prediction of values PR1 (t’), PR2(t’), ... PRi(t’) of one or more of the environmental parameters. As a result of these environmental parameters, the performance parameter of the solar panel may change in the second time period TP2. As discussed above, the prediction of the environmental parameters for the upcoming period TP2 can either be obtained from an external weather forecasting service or the future environmental parameters can be determined using an internal weather forecasting module.

[0114] The method according to the present disclosure comprises predicting the change of the values of the performance parameter during the second time period TP2, after moment in time ti at which the control is performed. This prediction uses the ML model, which has been trained to establish correlations between the values of the performance parameter and the values of one or more environmental parameters.

[0115] The method further comprises deciding whether cleaning is necessary within the second time period TP2, based on the predicted values of the performance parameter P1(t’) in the second time period TP2. The time t' refers to a time in the future, after moment in time ti at which the control is performed. If cleaning is required, then, in process step 400, the method comprises determining a moment in time tA for performing cleaning during the second time period.

[0116] In other words, the decision to perform cleaning in the future, after control interval ti, is not directly based on the actual experimental performance data taken in the period before control interval ti, but is based on predicted future performance data. This allows the decision to be based on the future evolution of the solar panels' performance over a longer period of time, the change depends on future weather patterns.

[0117] For deciding whether or not to perform cleaning based on the predicted values P1(t’) of the performance parameter in the second period after control interval ti, there are various options, depending on the chosen embodiments.

[0118] These various options for deciding whether or not cleaning is necessary and / or determining the moment in time tA for activating the self-cleaning mechanism are generally based on a comparison between one of at least one predicted performance parameter with predefined performance criteria.

[0119] In an embodiment, the decision whether or not to perform cleaning can be made by comparing the predicted values of the performance parameter with predefined performance criteria. If the performance criteria are met, a decision is made not to activate the self-cleaning mechanism during the second time period. If the performance criteria are not met, a time tG is determined that lies within the second time period during which the performance criteria are no longer met.

[0120] The determination of time tA for activation of the self-cleaning mechanism can then be expressed as follows: tA = tG + At, where At is an additional period with At > 0. The additional period At can be regarded as a delay period to allow some extra time between determining thatthe performance parameter has fallen below a certain threshold and the time at which the cleaning is actually performed. In some embodiments, tA = tG.

[0121] In embodiments, the delay period At may be a predetermined configuration value, and in other embodiments, At can be determined based on the weather pattern. For example, if the weather pattern indicates that large amounts of dust are present in the second time period, and the energy production of the solar panels is zero or very low as a result, it may be decided to temporarily postpone cleaning taking into account the expected weather pattern, despite the fact that the solar panels are dirty and in urgent need of cleaning.

[0122] The delay period At can also be used to postpone cleaning until after sunset or until a period when the expected energy production is low.

[0123] In embodiments, the performance criteria for determining whether cleaning is required can comprise one or more minimum values for the performance parameter, preferably the minimum values are configuration data.

[0124] In embodiments, the determination of moment in time tA at which the cleaning is to be performed can be done in two steps. The first step examines at what point tG, the predicted values of the performance parameter no longer meet the predefined performance criteria. For example, if light transmission through the solar panel falls below a first predefined threshold, it may be decided that the cleaning mechanism needs to be activated at a time tA > tG, where tG is the time at which the predefined first threshold is reached. A second step can then determine the exact time for tA, whereby tA = tG + At.

[0125] At can be determined in various ways, depending on the embodiment.

[0126] In an embodiment, the method comprises predicting the effectiveness of cleaning the solar panel in the second time period TP2, and At can be determined based on a predicted effectiveness of cleaning.

[0127] The effectiveness of cleaning, as discussed above, indicates the extent to which cleaning improves the performance of the solar panel. The effectiveness of cleaning is therefore an expression of the relative difference between the performance of the solar panel before and after cleaning.

[0128] In embodiments, the effectiveness of cleaning can be an expression of a relative difference between performance parameters after and before cleaning the solar panel. For example, performance parameter P1 may have dropped from an initial value of 100% to 80% over time. When cleaning is performed, performance parameter P1 may have then risen back to, for example, 95%. In this case, the effectiveness of the cleaning is EF = (95%-80%) / (100%-80%) = 0.75. In another example, if the performance parameter were to return to 100% after cleaning, the effectiveness EF = 1.

[0129] In general terms, the effectiveness of cleaning can be expressed as the difference between the value of the performance parameter after and before cleaning divided by thedifference between the maximum achievable value of the performance parameter and the value of the performance parameter before cleaning.

[0130] In embodiments, the effectiveness of cleaning has a value between 0 and 1, where a value of 1 indicates that cleaning will maximise the value of the performance parameter of the solar panel after cleaning, and where a value of 0 indicates that after cleaning, the value of the performance parameter will not have changed compared to the value before cleaning.

[0131] The effectiveness of cleaning has a direct impact on the energy output of the solar panels. With an effectiveness value of 1, the maximum energy output of the solar panel can be achieved again. For example, by measuring the energy output of the solar panel before and after cleaning, the effectiveness of the cleaning can be measured.

[0132] The effectiveness of cleaning can depend on various factors, such as weather conditions. For example, the effectiveness of cleaning performed in the middle of a sandstorm will be lower than the effectiveness achieved when cleaning is performed on a sunny day.

[0133] At can be determined on the basis of the expected effectiveness of cleaning at moment in time tG and the development of the effectiveness of cleaning after moment in time tG. If the effectiveness of cleaning at a time t after time tG is greater than the effectiveness of cleaning at time tG, then cleaning can be delayed by time At so that when cleaning is performed, the effectiveness of cleaning is greater than at time tG.

[0134] The expected effectiveness of cleaning can be predicted using the same or a second, additional ML model. This ML model is trained to establish a correlation between the effectiveness of cleaning and future environmental parameters determined by weather patterns. The environmental parameters are the parameters as discussed above. The algorithm may be configured in a number of ways discussed below and may perform one or more of the following functions.

[0135] In some embodiments, the algorithm can be configured to perform cleaning only at a time at which the effectiveness of cleaning is above a predetermined minimum effectiveness threshold. If the effectiveness is very low, there is little point in performing cleaning.

[0136] In some embodiments, the algorithm can compare the predicted performance parameter with a second threshold, and as long as the performance parameter does not fall below the second threshold, cleaning can be postponed by time At if the effectiveness of cleaning at the later time is higher. However, if the predicted performance parameter falls below the second threshold, it may be decided to perform cleaning regardless of the expected effectiveness of cleaning. This is to prevent the energy production of the solar panel from becoming extremely low.

[0137] In other embodiments, another method can be used to determine At, by: - calculating a new value for the performance parameter after a fictitious cleaning performed at moment in time tG, - predicting values of the performance parameter for a third time period after the fictitiouscleaning, - calculating a rate of change of the performance parameter during the third time period, and - determining the delay period At based on the rate of change of the performance parameter in the third time period. In embodiments, the ML model can be used to predict the values of the performance parameter or more than one performance parameter for the third time period after the first fictitious cleaning.

[0138] Whichever method is used to determine At, when the moment in time tA = tG + At is reached, the control system will generate an activation signal SA at time tA, determined by the algorithm, to activate the self-cleaning mechanism (2), as shown schematically in Fig. 1.

[0139] In embodiments, the decision whether or not to perform cleaning can be made based on calculating the expected energy production during the second time period if no cleaning were performed, and calculating the expected energy production during the second time period if cleaning were performed at a specific time within the second time period. If the gain in energy production with cleaning compared to energy production without cleaning is less than a predefined energy threshold, a decision can be made not to perform cleaning during the second time period.

[0140] If cleaning results in energy production gains that are higher than or equal to the energy threshold, cleaning will be performed in the second time period and the moment in time for cleaning can be determined, for example, as the moment in time at which maximum energy production is achieved over the second time period.

[0141] The calculation of the expected energy production in the second time period depends on the predicted change of the values of the performance parameter during the second time period and / or on the predicted effectiveness of cleaning during the second time period.

[0142] In embodiments, if, when repeating the cleaning control, the frequency of controls is such that, after performing a first control at moment in time til, a new control is performed at, for example, moment in time ti2, and this moment in time ti2 is before the moment in time of cleaning tA1 as determined during the previous control til, and therefore at the moment ti2 the predicted cleaning on moment in time tA1 has not yet been performed, then the previous moment in time tA1 for cleaning is overwritten and replaced with a new predicted moment in time tA2 for cleaning as determined at moment in time ti2. This situation can occur if, for example, time period TP2 is a time period of a week and the frequency of control and determining the moment in time tA for cleaning is perhaps daily.

[0143] Fig. 4 shows a flowchart that provides an overview of process steps of a further embodiment of a method according to the present disclosure. Process step 500 corresponds to the activation of the self-cleaning mechanism at a moment in time determined by the algorithm.

[0144] In the present disclosure a ML model may be used to predict the performance parameter and / or to predict the effectiveness of cleaning, and this ML model can be selected from different types, depending on the specific embodiment and the conditions in which the solar panelsystem is used.

[0145] Some examples of suitable ML models that are well known and can be selected by the expert are: random forest (RF), recurrent neural networks (RNN) or gradient boosting (GB).

[0146] Random Forest is an ensemble learning method that combines multiple decision trees to provide a robust prediction. This model is particularly suitable for datasets with complex, nonlinear relationships between input data, such as soiling levels, weather conditions, and solar panel performance.

[0147] RNNs are designed to process time-dependent data and recognise trends in sequential data points. This model is particularly suitable for predicting changes in performance parameters over time, such as seasonal soiling or daily fluctuations in weather conditions. This type is more suitable for analysing long-term trends, such as soiling accumulation during dry seasons.

[0148] Gradient Boosting is an ensemble method that iteratively improves weak models to create a strong model. It is particularly effective in addressing minor errors in previous predictions, resulting in high accuracy. For example, optimising the prediction of effectiveness of cleaning by correcting errors from previous predictions.

[0149] In some embodiments, the ML model can use a combination of different techniques, also known as ensemble learning. This can increase the accuracy and robustness of the model and algorithm. For example, a combination of RF, RNN, and GB can be implemented.

[0150] A combined approach may utilise the strengths of multiple models to make a more accurate prediction. There are two known ways to combine models: “stacking” (stacked ensemble), in which the models work in parallel and their predictions are combined by a metamodel, and “hybrid modelling”, in which different models are used in a series and the output of one model is the input for the next. In a combined approach, RF can be used in an embodiment to determine which parameters contribute most to, for example, soiling, and an initial assessment can be made. RNN can then be applied to predict future developments in soiling levels based on historical trends and weather forecasts. And finally, GB can also be used to further refine the prediction and correct minor errors in the earlier models.

[0151] In an example scenario of a combined approach could be as follows. For example, suppose the input is as follows: a light transmission of 85%, a temperature of 30°, a rain forecast of 10 mm within 24 hours and historical pollution trends indicate an average decrease of 2% in light transmission per day. In an initial phase, RF and RNN can be applied and RF can, for example, estimate that the soiling will reach a critical threshold within two days, and RNN can predict that light transmission will improve by 5%, but will not fully recover. Then, in a second phase, GB can be applied and, based on the RF and RNN outputs, predict that, for example, cleaning will be necessary in 36 hours.The ML model can be continuously trained using historical data, namely based on the actual performance parameters and actual corresponding environmental factors that are continuously measured, this is shown schematically in Fig. 4 in step 700.

[0152] The ML model for self-cleaning solar panels can therefore be designed as a supervised learning model, whereby the system’s learning is based on historical data about soiling, weather conditions, and solar panel performance before and after cleaning cycles.

[0153] In some embodiments, a new training dataset for the ML model can be defined, whereby input and output data for the new training dataset are based on the values of the performance parameter and environmental data measured during the first time period TP1 before the control interval ti and the second time period TP2 after the control interval. This way, the ML model is continuously training, based on the actual measured performance of the solar panel and the actual measured environmental factors. This feedback to the ML model, resulting in continuous training, is shown schematically in Fig. 4 in step 700.

[0154] In some embodiments, as discussed above, the determination of the time of performing cleaning can depend on a predicted effectiveness of cleaning. In these embodiments, as schematically shown in Fig. 4, in step 600, the actual effectiveness of cleaning can be measured after cleaning based on the performance of the solar panel before and after cleaning. The ML model, used to predict the effectiveness of cleaning in the second time period TP2, can be continuously trained by defining a new training dataset for the ML model based on the actual measured effectiveness of the cleaning after cleaning and the actual environmental factors measured.

[0155] The initial training of the ML model can be done with a basic set of training data that can be determined for a number of standard weather patterns and corresponding known effects on the performance of the solar panel. This basic set can be obtained by performing simulations for different environmental factors and by determining standard patterns of soiling accumulation and effectiveness of cleaning. Then, as discussed above, the system will train itself and adapt to local conditions based on the measured data.

[0156] As indicated above, one advantage of at least some embodiments of the invention is that they provide for improved efficiency of energy generation from solar panels, as a result of reduced accumulation of dirt on the panels. A further advantage may be that the process of cleaning the panels uses fewer resources than known cleaning processes. As described herein, by selecting a suitable time to clean the panels, the cleaning process can be arranged to use less energy and / or other resources such as water or detergent or eliminate the use of water or detergent. As an example, by making use of natural precipitation to assist the cleaning process it may be possible to reduce or entirely avoid the application of other water to panels during the cleaning process.LIST OF REFERENCE NUMERALS

[0157] 1 solar panel

[0158] 2 self-cleaning mechanism

[0159] 3 nanocoating

[0160] 4 panel sensor

[0161] 5 environment sensor

[0162] 6 data-acquisition module

[0163] 7 storage medium

[0164] 8 control system

[0165] 9 processor

[0166] 10 solar panel system

[0167] 15 weather conditions

[0168] 100 input data

[0169] 200 processing data and / or determining derived features 300 prediction using ML model

[0170] 400 determining time for cleaning

[0171] 500 activating the cleaning mechanism

[0172] 600 checking effectiveness of cleaning

[0173] 700 feedback to ML model

[0174] Ei(t) Environmental parameter

[0175] Pi(t) performance parameter

[0176] tA, tA1 to tA3 time of cleaning

[0177] til to ti5 control interval

[0178] TP1, TP2 first and second time period

[0179]

Claims

AMENDED CLAIMSreceived by the International Bureau on 08 May 2026 (08.05.2026)1. A computer-implemented method for determining when a self-cleaning mechanism of a solar panel is to be activated, the method comprising, at specific control intervals:obtaining at least one performance parameter associated with the energy-production efficiency of the solar panel, where the at least one performance parameter is obtained during a first time period prior to a control interval;obtaining environmental data measured prior to the control interval;predicting a future weather pattern for a second time period after the first time period, the future weather pattern based on the obtained environmental data;predicting a change in the at least one of the performance parameter during the second time period based on the predicted weather pattern;predicting an effect of cleaning in the second time period, wherein the effect of cleaning depends on the weather pattern during the second time period and is based on the performance of the solar panel before and after cleaning;determining whether to perform cleaning within the second time period based on the predicted change in the at least one performance parameter and the predicted effect of cleaning for the second time period; andif cleaning is to be performed in the second time period, determining a time tA within the second time period to activate the self-cleaning mechanism.

2. The computer-implemented method of claim 1, wherein at least one performance parameter is selected from, but not limited to, one of the following parameters: light transmission through the panel, soiling level of the panel, the energy-conversion efficiency of the panel and an energy production.

3. The method of any preceding claim, wherein predicting a change in the at least one of the performance parameter comprises using a machine learning model that has been trained to establish correlations between the at least one performance parameter and the environmental data.

4. The method according to any preceding claim, wherein the environmental data comprises one or more environmental parameters comprised of one or more of temperature, humidity, air pressure, wind speed, wind direction, precipitation amount, dust concentration.

5. The method according to any preceding claim, wherein determining whether to perform cleaning and / or determining the time within the second time period to activate a self-cleaningmechanism comprises comparing the at least one predicted performance parameter with predetermined performance criteria; andif the predetermined performance criteria are met, not performing cleaning by activating the self-cleaning mechanism during the second time period,otherwise determining a third time tG within the second time period during which the performance criteria are no longer met, and determining the time tA for activation of the selfcleaning mechanism as being: tA = tG or tA = tG + At, whereby At is an additional time period with At > 0.

6. The method according to claim 5, wherein the performance criteria comprises one or more minimum values for the at least one performance parameter, preferably the minimum values are configuration data.

7. The method according to claim 3, or any of claims 4 to 6 when dependent on claim 3, further comprising:generating a solar panel feature based on the at least one performance parameter,using the solar panel feature as additional input data for the machine learning model, optionally the solar panel feature is the rate of soiling accumulation on the solar panel.

8. The method according to any one of the preceding claims, wherein the self-cleaning mechanism is a vibrating mechanism, preferably the vibrating mechanism is an ultrasonic vibrating mechanism.

9. The method according to claim 8, wherein the vibration mechanism comprises a piezoelectric film, preferably the piezoelectric film comprises a polymer-based piezoelectric layer.

10. The method according to one of the preceding claims, wherein the control intervals are determined by a fixed control frequency, preferably where the control frequency is equal to or greater than 1 / (TP2), with TP2 being the duration of the second time period, or, alternatively, the control intervals correspond to one or several times at which the value of at least one performance parameter of the solar panel or the value of another performance parameter of the solar panel has fallen below a predefined threshold.

11. The method according to one of the preceding claims characterised in the method further comprises:predicting the effectiveness of cleaning in the second time period, wherein the effectiveness of cleaning is an expression of a relative difference between the performance of the solar panel before and after cleaning at a specific time, preferably wherein the effectiveness of cleaning is an expression of a relative difference between at least one performance parameter, or another performance parameter of the solar panel, before and after cleaning the solar panel.

12. The method according to claim 11, predicting the effectiveness of cleaning uses a machine learning model that has been trained to establish correlations between the values of the effectiveness of cleaning and the values of one or more environmental parameters, or, alternatively, predicting the effectiveness of cleaning uses a second machine learning model, which is trained to establish correlations between the values of the effectiveness of cleaning and the values of one or more environmental parameters.

13. The method according to claim 11 or 12, comprising:deciding whether cleaning is necessary within the second time period, based on the predicted values of the performance parameter in the second time period and based on the predicted effectiveness of cleaning in the second time period, and if cleaning is necessary, determining a time tA within the second time period for activating the self-cleaning mechanism.

14. The method according to one of the claims 11 to 13, wherein deciding whether or not cleaning is necessary and / or determining the tA time for activating the self-cleaning mechanism will happen by comparing the values of one of at least one predicted performance parameter with predetermined performance criteria, and optionallywherein an additional period At is determined on the basis of the predicted effectiveness of cleaning.

15. The method according to one of the claims 11 to 14, the effectiveness of cleaning has a value between 0 and 1, where a value of 1 indicates that cleaning will maximise the value of the performance parameter of the solar panel after cleaning, and where a value of 0 indicates that after cleaning, the value of the performance parameter will not have changed compared to the value before cleaning.

16. The method according to one of the preceding claims, deciding whether or not to perform cleaning during the second time period further comprises: calculating the expected energy production during the second time period if no cleaning were performed out and calculating the expected energy production during the second time period if cleaning were performed at aspecific time within the second time period, and if gain in energy production with cleaning compared to energy production without cleaning is less than a predefined energy threshold, to decide not to perform cleaning during the second time period.

17. The method according to claim 16, wherein the calculation of the expected energy production in the second time period is dependent on the predicted change of the values of the performance parameter during the second time period and / or on the predicted effectiveness of cleaning during the second time period.

18. The method according to one of claims 1 to 10, wherein deciding whether or not cleaning is necessary and / or determining the tA time for activating the self-cleaning mechanism will happen by comparing the values of one of the at least one predicted performance parameter with predetermined performance criteria, and wherein the method further comprises:determining At by:calculating a new value for the performance parameter after a fictitious cleaning would have been performed at time tG,predicting values of the performance parameter for a third time period after the fictitious cleaning,calculating a rate of change of the performance parameter during the third time period,and determining the additional period At based on the rate of change of the at least one performance parameter in the third time period.

19. The method according to claim 18, wherein predicting the values of the at least one performance parameter for the third time period after the first fictitious cleaning uses the machine learning model.

20. The method according to one of the preceding claims, wherein, if it is determined that cleaning is to be performed, the method further comprises:generating an activation signal “SA “ to activate the self-cleaning mechanism at the specified time for cleaning.

21. A computer-readable storage medium comprising instructions which, when executed by a computer system comprising one or more processors, cause the one or more processors to perform the method according to one of claims 1 to 20.

22. A solar panel system comprising:at least one solar panel;a self-cleaning mechanism for the at least one solar panel;one or more panel sensors configured to measure at least one performance parameter linked to an energy-production efficiency of the solar panel;environmental sensors configured to measure environmental data;a data acquisition module configured to obtain the at least one performance parameter and the environmental data as a function of time;a control system for controlling the self-cleaning mechanism, the control system comprising:one or more processor configured to:predict a future weather pattern based on the obtained environmental data; andpredict an effect of cleaning in the second time period, wherein the effect of cleaning depends on the weather pattern during the second time period and is based on the performance of the solar panel before and after cleaning;determine, based on the at least one performance parameter and the predicted future weather pattern and predicted effect of cleaning, a time tA at which to activate the self-cleaning mechanism.

23. The solar panel system according to claim 22, wherein the data acquisition module comprises a storage medium for storing the obtained values.

24. The solar panel system according to claim 22 or 23, wherein the control system is configured to determine a time tA at which the self-cleaning mechanism is activated using a machine learning model.

25. The solar panel system according to any one of claims 22 to 24, wherein the self-cleaning mechanism is a vibration mechanism, preferably an ultrasonic vibration mechanism.

26. The solar panel system according to any one of claims 22 to 25, wherein the vibration mechanism comprises a piezoelectric film, preferably the piezoelectric film comprises a polymer-based piezoelectric layer.

27. The solar panel system according to any one of claims 22 to 26, wherein the solar panel comprises a nanocoating to repel dust and dirt; and / or the one or more panel sensors are integrated into the solar panel.

28. The solar panel system according to claim 27, wherein the nanocoating is comprised of one or more of: hydrophobic silica-based nano-layers, alumina or zirconia nano particle layers, nano-texture polymer films, fluoropolymer nanocoatings, titania photocatalytic nanocoatings.

29. The solar panel system according to any one of claims 22 to 28, wherein the one or more processor is configured to execute the computer-implemented method according to one of claims 1 to 21.

30. The solar panel system according to claim 29, wherein the control system is configured to activate the self-cleaning mechanism at time tA determined by executing the computer-implemented method by the processor.

31. The solar panel system according to one of claims 22 to 30, wherein the control system is linked to a weather forecasting service and / or a weather forecasting station and is configured to obtain a future weather pattern, wherein the future weather pattern comprises a prediction of the environmental data.

32. The solar panel system according to one of claims 22 to 31, wherein the one or more processor comprises a weather forecasting module configured to determine the future weather pattern, wherein the future weather pattern comprises a prediction of the environmental data, preferably in which the future weather pattern is determined based on measured environmental data.

33. The solar panel system according to claim 31 or 32, characterised in that the machine learning model is configured to receive the prediction of the environmental data as input data.

34. The solar panel system according to claim 24 or any one of claims 25 to 33, when dependent on claim 24, wherein the machine learning model or a second machine learning model used by the processor is configured to predict an effectiveness of cleaning and wherein the prediction of the effectiveness of cleaning helps determine the time tA for activation of the self-cleaning mechanism, and wherein the effectiveness of cleaning is an expression of a relative difference between a performance of the solar panel before and after cleaning, preferably, the effectiveness of cleaning is an expression of a relative difference between the at least one performance parameter, or another performance parameter of the solar panel, after and before cleaning of the solar panel.

35. The solar panel system according to one of claims 22 to 34, wherein one or more panel sensors are integrated into the solar panel.

Images