Method for calculating dust accumulation in photovoltaic power station
By constructing a physics-based dynamic model of dust accumulation, and combining PM2.5 concentration and environmental data, the pollution sensitivity coefficient k is updated through self-learning, which solves the problem of real-time monitoring and calculation of dust accumulation in photovoltaic power plants, thereby improving power generation efficiency and operational benefits.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- KGE
- Filing Date
- 2025-12-26
- Publication Date
- 2026-05-12
Smart Images
Figure CN122019938A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of photovoltaic system data processing technology, and in particular to a method for calculating dust accumulation in a photovoltaic power station. Background Technology
[0002] With the acceleration of the global energy transition, photovoltaic power generation, due to its clean and renewable characteristics, has seen its share in the global energy structure continue to grow rapidly. Large-scale photovoltaic power plants have become the norm for grid connection.
[0003] Photovoltaic power generation is characterized by significant intermittency and fluctuation, and its output power is directly affected by real-time changes in meteorological factors such as solar irradiance, ambient temperature, cloud cover, humidity, wind speed, and air pollution levels.
[0004] Dust accumulation on photovoltaic (PV) modules can block sunlight, preventing the panels from fully absorbing solar energy and causing a shadow effect, which significantly reduces power generation. Studies show that severe dust accumulation can lead to power generation losses of up to 15%-30%, and even higher in areas prone to sandstorms. By accurately calculating the efficiency loss caused by dust accumulation, the operation and maintenance team can scientifically determine the optimal cleaning time. This avoids wasting water and manpower by cleaning too early, or losing revenue by cleaning too late. Accurate calculation and monitoring of dust accumulation is a crucial step in improving the profitability, ensuring safety, and achieving efficient operation of PV power plants.
[0005] In traditional models, the dust occlusion coefficient is a static constant that needs to be manually set periodically. This means that the model completely ignores the continuous accumulation of dust in the environment between cleaning cycles, leading to a linear increase in prediction error over time. The assessment and management of dust accumulation in photovoltaic power plants is mostly in a passive, lagging, and rudimentary state. It relies either on expensive hardware, manual experience, or rigid statistical models, making it difficult to achieve accurate, real-time, and low-cost perception and prediction. Summary of the Invention
[0006] Based on this, the purpose of this invention is to solve the above-mentioned technical problems and to accurately, in real time and at low cost calculate and monitor dust accumulation in photovoltaic power plants.
[0007] To achieve the above-mentioned objective, this application provides a method for calculating dust accumulation in a photovoltaic power station, comprising: Obtain the PM2.5 concentration, ambient temperature, and ambient humidity of the photovoltaic power station at the current time; Obtain the time of the most recent cleaning of the photovoltaic power station; A dynamic model of dust accumulation is constructed. The dynamic model of dust accumulation is a physics-based differential equation model used to describe the dynamic change of the dust occlusion coefficient over time. Based on the PM2.5 concentration, ambient temperature, ambient humidity, and cleaning time, the dust obstruction coefficient at the current moment is calculated using the dust accumulation dynamic model.
[0008] Preferably, the functional expression of the dust accumulation dynamic model is:
[0009] Where p_soiling(t) represents the dust occlusion coefficient at time t; p_clean is the baseline value of the dust occlusion coefficient; k is the site pollution sensitivity coefficient; PM2.5(τ) is the PM2.5 concentration value at time τ; α is the nonlinear accumulation exponent; λ() is the dust deposition decay constant that varies with time; and t0 is the time of the most recent cleaning.
[0010] Preferably, the dust deposition attenuation constant λ() is a function of ambient temperature T and ambient humidity RH, and the relationship is as follows:
[0011] Wherein, λ0 is the basic attenuation constant of dust deposition, representing the proportion of settled dust blown away by natural wind each day under standard conditions; β T β is the temperature compensation coefficient. RH T is the humidity compensation coefficient; ref and RH ref These are the reference temperature and reference humidity, respectively.
[0012] Preferably, the value of the nonlinear cumulative exponent α is greater than 1.
[0013] Preferably, the method further includes the step of: performing a self-learning update on the pollution sensitivity coefficient k of the site, wherein the self-learning update is triggered after the photovoltaic power station is cleaned.
[0014] Preferably, the self-learning update specifically refers to: Based on the measured shading coefficient p of the photovoltaic module after cleaning actual The predicted value p of the dust obstruction coefficient calculated by the dynamic model of dust accumulation after cleaning. clean And the dynamic model of dust accumulation is based on k before cleaning. n The predicted value p of the calculated dust blocking coefficient predicted The gradient descent method is used to update the value of k according to the following formula:
[0015] Where, k n k represents the site's pollution sensitivity coefficient before the update. n+1 η represents the updated site contamination sensitivity coefficient, and η is the learning rate.
[0016] Preferably, the loss function for self-learning and updating the site pollution sensitivity coefficient k is:
[0017] In the self-learning update process of the site pollution sensitivity coefficient k, to minimize To achieve this, the new k value is shifted along the negative gradient direction of the loss function.
[0018] Preferably, the pollution sensitivity coefficient k of the site ranges from 0.01 to 0.05 (μg / m³). - ¹.
[0019] Preferably, after each cleaning, the dust blocking coefficient p_soiling(t) is reset to the dust blocking coefficient baseline value p_clean, and the cleaning time t0 of the most recent cleaning is updated to the cleaning time of this cleaning.
[0020] Preferably, the dust shading coefficient p_soiling(t) is used to determine the optimal cleaning time for the photovoltaic power station and / or to assess the power generation loss caused by dust accumulation.
[0021] Compared with the prior art, the beneficial effects of this invention are: This invention views dust accumulation as a dynamic process, the rate of which is driven by the concentration of PM2.5 in the environment, and its state is affected by natural removal mechanisms such as wind and rain, and environmental conditions such as temperature and humidity. By establishing a physics-based differential equation model to estimate this process in real time, dust accumulation is transformed from a black-box empirical parameter into a dynamic state variable with clear physical meaning driven by environmental data. This enables accurate, real-time, and low-cost calculation and monitoring of dust accumulation in photovoltaic power plants. Attached Figure Description
[0022] Figure 1 A schematic diagram illustrating the steps of a method for calculating dust accumulation in a photovoltaic power plant; Figure 2 A schematic diagram of the system processing flow for a dynamic model of dust accumulation; Figure 3 This is a flowchart illustrating the k-value self-learning mechanism. Detailed Implementation
[0023] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of the invention. The following embodiments are used to illustrate the invention but are not intended to limit its scope.
[0024] Example 1 Embodiment 1 of this application provides a method for calculating dust accumulation in a photovoltaic power station, such as... Figure 1 As shown, it includes the following steps: S1: Obtain the PM2.5 concentration, ambient temperature, and ambient humidity of the photovoltaic power station at the current time; S2: Obtain the time of the most recent cleaning of the photovoltaic power station; S3: Construct a dynamic model of dust accumulation, which is a physics-based differential equation model used to describe the dynamic change of the dust occlusion coefficient over time. S4: Based on the PM2.5 concentration value, ambient temperature, ambient humidity, and cleaning time, the dust obstruction coefficient at the current moment is calculated using the dust accumulation dynamic model.
[0025] It should be noted that in this invention, dust refers to PM2.5 particles.
[0026] The functional expression of the dynamic model of dust accumulation constructed in step S3 is as follows:
[0027] Where p_soiling(t) represents the dust occlusion coefficient at time t; p_clean is the baseline value of the dust occlusion coefficient; k is the site pollution sensitivity coefficient; PM2.5(τ) is the PM2.5 concentration value at time τ; α is the nonlinear accumulation exponent; λ() is the dust deposition decay constant that varies with time; and t0 is the time of the most recent cleaning.
[0028] The dust deposition attenuation constant λ is not a fixed value; it reflects the environment's ability to remove settled dust and is a function of ambient temperature T and ambient humidity RH. The relationship is as follows:
[0029] Wherein, λ0 is the basic attenuation constant of dust deposition (basic value 0.3 days). - ¹). This indicates that under standard conditions, approximately 30% of settled dust may be blown away by natural winds each day; β T λ is the temperature compensation coefficient (base value 0.03). The higher the temperature, the stronger the air convection, and the more significant the wind scavenging effect (λ increases). RH (Base value 0.02) is the humidity compensation coefficient. The higher the humidity, the stronger the adhesion of dust particles after they absorb moisture, and the less likely they are to be blown away (λ decreases); T ref and RH ref These represent the reference temperature and reference humidity, respectively. The dynamic dust deposition attenuation constant λ( ) allows the model to adapt to changes in weather conditions.
[0030] The system processing flow of the dynamic model of dust accumulation is as follows: Figure 2 As shown, the process includes: initially acquiring real-time PM2.5 concentration data, ambient temperature, and relative humidity at the photovoltaic power station site; calculating the dynamic decay constant (i.e., dust deposition decay constant λ()) of the current environment based on the function expression of the dust accumulation dynamic model, using the above data; and then performing an exponentially weighted integral on the historical PM2.5 data since the last cleaning to obtain the cumulative pollution index. The process stops after calculating the current dust obstruction coefficient based on the cumulative pollution index.
[0031] Example 2 This application's Embodiment 2, based on Embodiment 1, further explains the site pollution sensitivity coefficient k, specifically including: The k (site pollution sensitivity coefficient) is the most critical yet most uncertain parameter in the model. It quantifies how much power generation loss is caused by a unit of PM2.5 exposure. Numerous factors influence the k value, and these factors vary depending on the site. For example, the surface material of the components, glass texture, and dust-proof coatings (hydrophobicity) significantly affect dust adhesion rates. Furthermore, flat surfaces with smaller inclination angles are more prone to dust accumulation, while steeper or vertical surfaces are more easily washed away by rainwater. It is also related to the local microenvironment; whether the site is near farmland (dust), roads (vehicle exhaust particles), factories (industrial emissions), or ocean (salt spray), the dust composition and physical properties are completely different. Other factors include the frequency and intensity of rainfall in the area, and the strength of the prevailing winds. Therefore, accurately pre-setting a k value for a new power plant is impossible or inaccurate. Traditional methods involve inverting and fitting data from long-term historical data or relying on expert experience for guesswork, which are costly and inaccurate. The k value updated using a self-learning mechanism in this invention aims to automatically and online find the unique k value most suitable for the specific site.
[0032] The self-learning update is triggered within 24 hours after the photovoltaic power station is cleaned. The process for self-learning and updating the pollution sensitivity coefficient k of the site is as follows: Figure 3 As shown, specifically: Based on the measured shading coefficient p of the photovoltaic module after cleaning actual The predicted value p of the dust obstruction coefficient calculated by the dynamic model of dust accumulation after cleaning. clean And the dynamic model of dust accumulation is based on k before cleaning. n The predicted value p of the calculated dust blocking coefficient predicted The gradient descent method is used to update the value of k according to the following formula:
[0033] Where, k n k represents the site's pollution sensitivity coefficient before the update. n+1The updated site contamination sensitivity coefficient, η is the learning rate (default 0.001).
[0034] Define the error function: After a cleaning event, the occlusion coefficient predicted by the model should be p. clean However, using the old k value k n The calculated predicted value is p predicted The error between the predicted and actual values is the prediction error function.
[0035] Define the loss function for self-learning updates based on the value of k:
[0036] In the self-learning update process of the site pollution sensitivity coefficient k, to minimize To achieve this, the new k value is shifted along the negative gradient direction of the loss function.
[0037] The k-value self-learning mechanism enables the model to self-diagnose, self-optimize, and adaptively grow, eliminating the need for years of historical meteorological and power generation data to fit the k-value. It self-calibrates through discrete events (cleaning) that occur during operation, allowing the model to automatically converge to the k-value best suited for the specific power plant. For example, the k-values of a power plant located on the edge of a desert and one located in a humid mountainous area will automatically differentiate after several cleaning events. Furthermore, if the surrounding environment of the power plant changes (e.g., new construction begins nearby), causing changes in pollution characteristics, the old k-values will become invalid. However, the self-learning mechanism can detect these changes in subsequent cleaning events and automatically adjust the k-value, ensuring the model maintains high accuracy.
[0038] Example 3 This embodiment 3 is based on embodiments 1 and 2, and describes some parameters in the dynamic model of dust accumulation in a specific embodiment. It should be noted that the value range of each parameter in Table 1 can be changed according to the actual situation.
[0039] Table 1. Explanation of parameters in the dynamic model of dust accumulation.
[0040] The parameters are shown in Table 1. Dust accumulation is an integral process related to historical PM2.5 concentrations, with recent PM2.5 having a greater impact (reflected in the exponential decay term). Simultaneously, due to α>1, the model exhibits a nonlinear cumulative effect (i.e., prolonged periods of moderate pollution may accumulate more dust than short periods of high pollution). The dust shading coefficient p_soiling(t) is used to determine the optimal cleaning timing for photovoltaic power plants and / or assess power generation losses due to dust accumulation.
[0041] Additionally, after the power plant is cleaned, the model is reset: t0 is set to 0, and p_soiling(t) is reset to p_clean (i.e., 0.98). Simultaneously, the k value is calibrated using calculations based on changes in power generation before and after the cleanup.
[0042] In summary, this invention provides a method for calculating dust accumulation in photovoltaic power plants. Dust accumulation is viewed as a dynamic process, driven by ambient PM2.5 concentration, and its state is influenced by natural removal mechanisms such as wind and rain, and environmental conditions such as temperature and humidity. By establishing a physics-based differential equation model to estimate this process in real time, dust accumulation is transformed from a black-box empirical parameter into a dynamic state variable with clear physical meaning, driven by environmental data. This enables accurate, real-time, and low-cost calculation and monitoring of dust accumulation. Furthermore, it solves the problem of models relying too heavily on historical data, providing a method that is completely independent of historical data. It physically maps environmental data to system performance in real time, rather than relying on historical statistics. The model exhibits strong generalization ability and adaptability to changes in power plant characteristics.
Claims
1. A method for calculating dust accumulation in a photovoltaic power station, characterized in that, Includes the following steps: Obtain the PM2.5 concentration, ambient temperature, and ambient humidity of the photovoltaic power station at the current time; Obtain the time of the most recent cleaning of the photovoltaic power station; A dynamic model of dust accumulation is constructed. The dynamic model of dust accumulation is a physics-based differential equation model used to describe the dynamic change of the dust occlusion coefficient over time. Based on the PM2.5 concentration, ambient temperature, ambient humidity, and cleaning time, the dust obstruction coefficient at the current moment is calculated using the dust accumulation dynamic model.
2. The method according to claim 1, characterized in that, The functional expression of the dynamic model for dust accumulation is: Where p_soiling(t) represents the dust occlusion coefficient at time t; p_clean is the baseline value of the dust occlusion coefficient; k is the site pollution sensitivity coefficient; PM2.5(τ) is the PM2.5 concentration value at time τ; α is the nonlinear accumulation exponent; λ() is the dust deposition decay constant that varies with time; and t0 is the time of the most recent cleaning.
3. The method according to claim 2, characterized in that, The dust deposition attenuation constant λ is a function of ambient temperature T and ambient humidity RH, and its relationship is as follows: Wherein, λ0 is the basic attenuation constant of dust deposition, representing the proportion of settled dust blown away by natural wind each day under standard conditions; β T β is the temperature compensation coefficient. RH T is the humidity compensation coefficient; ref and RH ref These are the reference temperature and reference humidity, respectively.
4. The method according to claim 2, characterized in that, The value of the nonlinear cumulative exponent α is greater than 1.
5. The method according to claim 4, characterized in that, The method further includes the following steps: The pollution sensitivity coefficient k of the site is updated through self-learning, and the self-learning update is triggered after the photovoltaic power station is cleaned.
6. The method according to claim 5, characterized in that, The self-learning update specifically refers to: Based on the measured shading coefficient p of the photovoltaic module after cleaning actual The predicted value p of the dust obstruction coefficient calculated by the dynamic model of dust accumulation after cleaning. clean And the dynamic model of dust accumulation is based on k before cleaning. n The predicted value p of the calculated dust blocking coefficient predicted The gradient descent method is used to update the value of k according to the following formula: Where, k n k represents the site's pollution sensitivity coefficient before the update. n+1 η represents the updated site contamination sensitivity coefficient, and η is the learning rate.
7. The method according to claim 6, characterized in that, The loss function for self-learning and updating the site contamination sensitivity coefficient k is: In the self-learning update process of the site pollution sensitivity coefficient k, to minimize To achieve this, the new k value is shifted along the negative gradient direction of the loss function.
8. The method according to claim 7, characterized in that, The pollution sensitivity coefficient k of the site ranges from 0.01 to 0.05 (μg / m³). - ¹.
9. The method according to claim 2, characterized in that, After each cleaning, the dust occlusion coefficient p_soiling(t) is reset to the dust occlusion coefficient baseline value p_clean, and the cleaning time t0 of the most recent cleaning is updated to the cleaning time of this cleaning.
10. The method according to claim 2, characterized in that, The dust smearing coefficient p_soiling(t) is used to determine the optimal cleaning time for photovoltaic power plants and / or assess the power generation loss caused by dust accumulation.