Digital operation and maintenance system of photovoltaic power station

Abstract

The invention discloses a digital operation and maintenance system for a photovoltaic power station, and the system comprises an integrated interaction platform and a data collection system, and the integrated interaction platform and the data collection system achieve the cooperative linkage through a data link and an instruction link. The integrated interaction platform comprises a cloud side service assembly, a side service assembly and an end side service assembly. The cloud side service assembly is provided with a photovoltaic cleaning operation digital twin model, an upper control program, a power station digital detection model and a meteorological data reading program. The side service component is configured to acquire and process real-time power generation data of the photovoltaic string, the photovoltaic sub-array and the photovoltaic power station, receive an instruction of the cloud side service component, forward the instruction to the end side service component and upload storage data to the cloud side; the end side service assembly is configured to obtain real-time power generation data of the single photovoltaic assembly; through the targeted operation and maintenance instruction, invalid operation is reduced, the operation and maintenance efficiency is improved, and the power generation loss and the economic cost caused by abrasion of the antireflection film are reduced.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power generation technology, and more specifically to a digital operation and maintenance system for photovoltaic power plants. Background Technology

[0002] In the construction and operation and maintenance of large-scale ground-mounted photovoltaic (PV) power plants, equipping them with single-axis tracking brackets and cleaning robots has become standard practice in the industry to improve power generation efficiency and module cleanliness, especially in power plants owned by companies like ACWAPower. Currently, publicly available patents in PV power plant related technologies mainly fall into two categories: one is patents disclosed by testing equipment manufacturers, primarily covering static and dynamic mechanical load equipment for PV testing and certification laboratories, focusing on performance verification during module testing; the other is related to routine PV power plant operation and maintenance technologies, but no specific technical solutions addressing the correlation between cleaning operations and module degradation have been found. Furthermore, cleaning robot manufacturers and PV power plant operation and maintenance companies have not published any patents related to monitoring wear and tear on the anti-reflective coating on module surfaces and decoupling faults during the cleaning process; existing technologies mostly revolve around the structural design of the equipment itself or general operation and maintenance procedures.

[0003] Although the application of cleaning robots has solved the problem of dust accumulation on photovoltaic modules affecting power generation efficiency, existing technologies have significant shortcomings: First, the cleaning devices of the cleaning robots (such as brushes and scrapers) rub against the anti-reflective coating on the surface of the photovoltaic modules during operation, causing the anti-reflective coating to thin or even be damaged, which in turn leads to a decrease in the maximum output power of the photovoltaic modules. It is estimated that this decrease can accumulate to 1.5% over the entire life cycle of the power station, resulting in huge losses in power generation and economy. Second, existing technologies lack effective monitoring and quantification methods for this type of decrease, and cannot accurately capture the correlation between cleaning operations and anti-reflective coating wear and power decrease. Third, in large-scale ground-mounted photovoltaic power stations, problems such as uniform decrease caused by the aging of the photovoltaic modules themselves and residual dust blockage caused by poor cleaning effects are coupled with power decrease caused by the thinning of the anti-reflective coating. Existing technologies do not provide effective fault decoupling methods, making it difficult to accurately identify the root cause of the fault, which directly affects the accuracy of operation and maintenance decisions and the improvement of operation and maintenance efficiency.

[0004] Therefore, a new technological solution is needed. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a digital operation and maintenance system for photovoltaic power plants to at least solve the problems existing in the prior art.

[0006] The embodiments of the present invention provide the following technical solutions: This invention provides a digital operation and maintenance system for photovoltaic power plants, including an integrated interactive platform and a data acquisition system. The integrated interactive platform and the data acquisition system work together through data links and command links to monitor and identify wear faults in the anti-reflective film of photovoltaic modules caused by cleaning operations. The integrated interactive platform includes cloud-side service components, edge-side service components, and terminal-side service components, wherein: The cloud-side service component is equipped with a digital twin model of photovoltaic cleaning operations, a host control program, a digital detection model of the power station, and a meteorological data reading program. The digital twin model is used to decouple the anti-reflective film wear fault of photovoltaic modules caused by cleaning operations from other types of faults, and to generate standard power generation performance benchmark data and fault simulation characteristic data of photovoltaic modules and strings. The edge service component is configured to acquire and process real-time power generation data of photovoltaic strings, photovoltaic subarrays and photovoltaic power plants, and can receive instructions from the cloud service component, forward them to the edge service component and upload stored data to the cloud. The end-side service component is configured to acquire real-time power generation data of a single photovoltaic module; The data acquisition system includes a photovoltaic inverter, a module-level power electronic MLPE device, and a meteorological data source, used to acquire power generation performance data and meteorological data at the photovoltaic module level, string level, subarray level, and power station level, and upload the acquired data to the cloud-side service component. The cloud-side service component performs simulation calculations and comparative analysis on the uploaded data through the digital twin model and the power station digital detection model. Combining the location information of the photovoltaic modules with the simulated feature data, it completes fault decoupling and wear location. Finally, the upper-level control program issues operation and maintenance instructions to realize targeted operation and maintenance of the photovoltaic power station.

[0007] Preferably, the system is configured to operate in model-triggered mode, wherein: The photovoltaic cleaning operation digital twin model is configured to run autonomously in real time. Based on real-time meteorological data, historical meteorological data and predicted meteorological data collected by the data acquisition system, it simulates the power generation, power output and cleaning operation effect of the photovoltaic power station. The digital twin model is further configured to autonomously judge based on the simulation results and trigger the upper-level control program to issue instructions to control the data acquisition system to collect power generation performance data of the photovoltaic power station, so as to perform predictive maintenance on the wear of the anti-reflective coating on the glass surface of the photovoltaic modules; or The system is configured to operate in a detection-triggered mode, wherein: The digital detection model of the power station is configured to operate in a manual or semi-automatic manner according to the preset photovoltaic power station cleaning work schedule and technical requirements, and to control the data acquisition system to collect the power generation performance data of photovoltaic modules, photovoltaic strings and photovoltaic subarrays in the photovoltaic power station. The digital detection model of the power station is further configured to transmit the collected data back to the digital twin model of the photovoltaic cleaning operation for interactive judgment, so as to perform planned maintenance on the wear of the anti-reflective film on the glass surface of the photovoltaic module.

[0008] Preferably, the data acquisition system includes: A photovoltaic inverter is configured to collect current and voltage data of the photovoltaic string in response to instructions from the host control program. A module-level power electronic MLPE device is configured to work with the photovoltaic inverter to collect current and voltage data at the photovoltaic module level.

[0009] Preferably, the digital detection model for the power plant is configured to perform the following detection process: Based on the digital twin model, a standard clear-sky power generation performance benchmark dataset of photovoltaic modules and photovoltaic strings under fault-free conditions and a standard power location marker map containing the location information of photovoltaic modules are generated. The data acquisition system is controlled to collect measured power generation performance data of the photovoltaic strings and photovoltaic modules after the cleaning operation; The measured power generation data of each photovoltaic module is compared with the corresponding standard clear-sky power generation in the benchmark dataset, and the photovoltaic modules are classified into usable datasets, verification datasets, or excluded datasets based on the comparison deviation value. Based on the available dataset, the power generation attenuation value and location information of each photovoltaic module are calculated, which serve as input parameters for interactive judgment with the digital twin model.

[0010] Preferably, the classification based on the comparison deviation value includes: Photovoltaic modules with a comparison deviation less than a first predetermined threshold are classified into the available dataset; Photovoltaic modules with a comparison deviation between the first predetermined threshold and the second predetermined threshold are classified into the verification dataset, and their data is verified to eliminate interference factors. After the verification is passed, they are incorporated into the available dataset. Photovoltaic modules whose comparison deviation is greater than the second predetermined threshold and which have not improved after data review are classified as the excluded dataset.

[0011] Preferably, the digital twin model of the photovoltaic cleaning operation includes: A 3D optical model of a photovoltaic module is configured to calculate the photon absorption power of photovoltaic cells and photovoltaic modules based on a 3D optical calculation algorithm. The photovoltaic module electrical model is configured to be based on the equivalent diode model to calculate the electrical power of photovoltaic cells and photovoltaic modules; A 3D optical model of a photovoltaic string is configured to calculate the optical loss between the photovoltaic string and the surrounding physical structure based on the 3D optical calculation algorithm. The electrical model of the photovoltaic string is configured to calculate the electrical power of the photovoltaic string based on Ohm's law; The photovoltaic power calculation model is configured to use the photovoltaic module electrical model and meteorological input data to calculate the power generation of photovoltaic modules, photovoltaic strings, photovoltaic subarrays and photovoltaic power stations; The photovoltaic power prediction model is configured to calculate the predicted power generation of the photovoltaic power generation units at different levels using the photovoltaic module electrical model and predicted meteorological data, or by adopting a data-driven method. The solar irradiance prediction model is configured to use a data-driven approach, combining ground-based sensor data and satellite inversion data, to predict the temporal solar irradiance at a target location.

[0012] Preferably, the 3D optical calculation algorithm is configured as follows: Energy integration is performed on incident light with wavelengths ranging from 300 nm to 1200 nm. Its calculation range and the area irradiated by incident photon flux can be customized, and it can be gradually expanded from the photovoltaic cell level to the photovoltaic module, photovoltaic string, photovoltaic sub-array and photovoltaic power station level; In the calculation, for the response of each beam of light on the photovoltaic material, at least the following factors are considered: the photon flux corresponding to different wavelengths, the absorption probability of the photovoltaic cell material for photons of different wavelengths, and the quantum efficiency of generating charge carriers under a specific absorption probability.

[0013] Preferably, the simulation of the digital twin model has a configurable simulation granularity, which includes: simulation of the power generation performance of a single photovoltaic module, and / or simulation of the power generation performance of photovoltaic strings, photovoltaic subarrays, and photovoltaic power plants.

[0014] Preferably, the digital twin model is configured to perform time-series process simulation to decouple multiple coexisting faults of the photovoltaic module; specifically configured as follows: For the N cleaning operations planned for a photovoltaic power station throughout its entire life cycle, the time node (t ≤ N) corresponding to the t-th cleaning operation is simulated in the digital twin model to generate the corresponding simulation results; Based on the generated time-series simulation results, feature data for identifying the relationships between different faults are extracted.

[0015] Preferably, the digital twin model is configured to generate a power generation performance spectrum of a single photovoltaic module, wherein: the spectrum is time-series characteristic data characterizing the main faults of the photovoltaic module, generated by simulation using the 3D optical and electrical models of the photovoltaic module; the spectrum includes at least simulated photocurrent (I-module-simulated) and simulated power generation (P-module-simulated) indicators; the spectrum generates simulation results for the following fault types: uniform aging fault of the photovoltaic cell itself, residual dust shading fault caused by poor cleaning effect, and thinning fault of the anti-reflective film of the photovoltaic module glass caused by the cleaning process.

[0016] Preferably, the digital twin model is further configured to extend the power generation performance map of the single photovoltaic module to the power generation performance maps of photovoltaic strings, photovoltaic subarrays, and photovoltaic power plants, wherein the extension method includes: Extract the characteristic value Ex of the uniformity of the photocurrent in the power generation performance spectrum of the single photovoltaic module. The characteristic value Ex is formed into an array based on the coordinates of the cell with the largest photocurrent in the module and the current uniformity (E1, E2, E3, E4) in its four quadrants. For the three fault types, respectively generate photocurrent uniformity characteristic value arrays Ex1_t, Ex2_t, and Ex3_t at each simulation time node t; Based on the location marking map and electrical topology of the photovoltaic power station, the feature value array of each photovoltaic module is classified and aggregated to generate a station-wide simulated uniformity feature distribution of photovoltaic current. According to the station-wide feature distribution, the simulated power attenuation value, the corresponding module location information and the feature value array are associated to form a fault judgment basis that can be used for cross-level comparison.

[0017] Preferably, the power generation performance map is constructed based on any of the following meteorological datasets: the typical meteorological year TMY dataset, or the predicted meteorological year FMY dataset; and the photovoltaic module electrical model, photovoltaic string electrical model and photovoltaic power calculation model include an algorithm for decomposing solar irradiance in time-series meteorological data into direct irradiance and diffuse irradiance, and inputting it into the photovoltaic module 3D optical model to generate photocurrent results corresponding to different fault types.

[0018] Preferably, the system further includes a fault detection module, configured as follows: Obtain the measured data array output from the digital detection process of the power plant, wherein the array contains the measured power attenuation value and location information of the photovoltaic modules in the available data set; Retrieve the simulated data array generated by the digital twin model, the array containing simulated power attenuation values, simulated location information, and related feature data; Fault identification is completed by cross-validating the measured data array and the simulated data array; The upper-level control program is further configured to: locate and screen photovoltaic modules with severely worn anti-reflective films based on the output of the fault discrimination module, and generate a notification message to initiate the module replacement and maintenance process.

[0019] Compared with the prior art, the beneficial effects that the at least one technical solution adopted in the embodiments of the present invention can achieve include at least: This invention discloses a digital operation and maintenance system for photovoltaic power plants. Through cloud-edge-device collaboration and digital twin technology, it accurately monitors and identifies wear faults in the anti-reflective coating of photovoltaic modules caused by cleaning operations. This effectively decouples the wear from other faults such as cell aging and dust accumulation, avoiding misjudgments. Based on layered data acquisition and simulation analysis, it quantifies the degree of wear and predicts power attenuation, providing accurate data for operation and maintenance. Targeted operation and maintenance instructions reduce ineffective work, improve efficiency, and lower power generation losses and economic costs caused by anti-reflective coating wear, thereby enhancing the digital operation and maintenance monitoring capabilities of photovoltaic power plants. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a system service architecture diagram of a digital operation and maintenance system for a photovoltaic power plant according to an embodiment of the present invention; Figure 2 This is a system working mode of a digital operation and maintenance system for a photovoltaic power plant according to an embodiment of the present invention; Figure 3 This invention provides a digital detection process for a photovoltaic power plant digital operation and maintenance system according to an embodiment of the present invention. Figure 4 This invention relates to a digital twin model simulation and discrimination process for a digital operation and maintenance system for a photovoltaic power plant, as described in an embodiment of the present invention. Detailed Implementation

[0022] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0023] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. This application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] It should be noted that various aspects of embodiments within the scope of the appended claims are described below. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this application, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number and aspects set forth herein can be used to implement the device and / or practice the method. Additionally, this device and / or method can be implemented using structures and / or functionalities other than one or more of the aspects set forth herein.

[0025] It should also be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. The drawings only show the components related to this application and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0026] Additionally, specific details are provided in the following description to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that practice can be carried out without these specific details.

[0027] Based on the problems existing in the prior art, this invention proposes a method and a digital operation and maintenance system for simulating, monitoring and judging the power and power generation capacity attenuation caused by the thinning of the anti-reflective coating of photovoltaic module glass due to cleaning friction during the cleaning process throughout the entire life cycle of a photovoltaic power station. Furthermore, it can issue operation instructions based on the judgment results of the digital operation and maintenance system, thereby improving the operation and maintenance efficiency of the photovoltaic power station.

[0028] The technical solutions provided by the various embodiments of this application are described below with reference to the accompanying drawings.

[0029] like Figure 1-4As shown, this embodiment of the invention provides a digital operation and maintenance system for photovoltaic power plants. Based on a digital twin model deployed on a cloud-side server, the system completes cleaning and maintenance work for large ground-mounted power plants equipped with cleaning robots through real-time interaction between the digital twin and the photovoltaic power plant data acquisition system. It monitors, identifies, and simulates the cleaning tools on the cleaning robot, assesses the wear of the anti-reflective coating on the glass surface of the photovoltaic modules, and issues operation and maintenance instructions.

[0030] Specifically, it includes an integrated interactive platform and a data acquisition system. The integrated interactive platform and the data acquisition system work together through data links and command links to monitor and identify anti-reflective coating wear faults in photovoltaic modules caused by cleaning operations. The integrated interactive platform includes cloud-side service components, edge-side service components, and terminal-side service components. Among them, the cloud-side service components are equipped with a digital twin model of photovoltaic cleaning operations, upper-level control programs, a digital detection model of the power station, and a meteorological data reading program. The digital twin model is used to decouple anti-reflective coating wear faults in photovoltaic modules caused by cleaning operations from other types of faults, and to generate standard power generation performance benchmark data and fault simulation characteristic data for photovoltaic modules and strings. The edge-side service components are configured to acquire and process real-time power generation data of photovoltaic strings, photovoltaic subarrays, and photovoltaic power stations, and can receive commands from the cloud-side service components, forward them to the terminal-side service components, and upload stored data to the cloud. The terminal-side service components are configured to acquire real-time power generation data of individual photovoltaic modules.

[0031] The core function of the integrated interactive platform is to achieve accurate monitoring of wear faults in the antireflective coating of photovoltaic modules, decoupling of multiple faults, and closed-loop operation and maintenance commands. Its underlying collaborative principle is: the end side collects micro-module data, the edge side processes mid-level string / power station data, and the cloud side centrally performs simulation calculations and decision-making. Through the bidirectional flow of data uplink and command downlink, it solves the problems of difficult identification of wear faults and easy coupling of multiple faults.

[0032] Cloud-based service deployment: digital twin model of photovoltaic cleaning operation, photovoltaic power prediction model, upper-level control program, digital detection model of power station, and meteorological data reading program for reading real-time meteorological station data of large ground photovoltaic power station.

[0033] The edge service component acts as an intermediate hub, receiving and aggregating edge data and forwarding cloud commands, and processing string, subarray, and power plant-level macro-generation data locally to reduce cloud data transmission and computing pressure.

[0034] Edge-side service deployment: Real-time power generation data (PV string current I, voltage V, etc.) of PV strings, PV subarrays, and PV power plants are located on edge-side devices such as inverter controllers.

[0035] The role of the end-side service component is to act as a data acquisition terminal to obtain real-time microscopic power generation data of a single photovoltaic module, providing a precise data source for fault location to a specific module.

[0036] End-side service deployment: Real-time power generation data of photovoltaic modules (photovoltaic module current I, voltage V, etc.) are located on end-side equipment such as photovoltaic module optimizers and module-level power electronic (MLPE) devices.

[0037] Among them, the digital twin model of photovoltaic cleaning operation is used to decouple the wear failure of anti-reflective film from the failures of cell aging, dust accumulation and obstruction, and to generate standard benchmark data and failure simulation feature data, providing a basis for comparison and judgment.

[0038] The power plant's digital monitoring model is used for data evaluation and screening to ensure the validity of data input from the cloud side, providing a high-quality data source for model calculations. The meteorological data reading program is used to acquire environmental parameters, providing the necessary input for the digital twin model's simulation calculations.

[0039] The data acquisition system includes photovoltaic inverters, module-level power electronic MLPE devices, and meteorological data sources. It is used to acquire power generation performance data and meteorological data at the photovoltaic module level, string level, subarray level, and power plant level, and uploads the collected data to the cloud-side service component. The cloud-side service component performs simulation calculations and comparative analysis on the uploaded data through digital twin models and power plant digital detection models. Combining the location information of photovoltaic modules with simulated characteristic data, it completes fault decoupling and wear location. Finally, the upper-level control program issues operation and maintenance instructions to realize targeted operation and maintenance of photovoltaic power plants.

[0040] The data acquisition system utilizes photovoltaic inverters (string-level IV acquisition), MLPE devices (module-level IV acquisition), and meteorological data sources to achieve full coverage of power generation performance data (current I, voltage V, power P) and meteorological data (solar irradiance, ambient temperature, etc.) across the entire "module-string-subarray-power station" hierarchy, ensuring no data is missed from the micro-module level to the macro-power station level. Simultaneously, the acquired data is uploaded to cloud-side service components, providing complete and accurate raw data support for cloud-side model calculations and fault analysis, thus solving the problems of coarse granularity and limited dimensions in traditional operation and maintenance data acquisition.

[0041] Specifically, the data acquisition system is characterized by having both real-time and manual acquisition capabilities.

[0042] The data collection conditions are as follows: To avoid affecting the normal power generation of large ground-mounted power plants, cleaning operations are usually carried out at night or during periods or under conditions where solar irradiance is very low or zero. After the actual cleaning operation, it is necessary to collect the current and voltage data of the photovoltaic modules under conditions where irradiance has recovered well (as close as possible to the standard STC test conditions of 1000W / m2 and 25 degrees Celsius) to determine the power generation capacity of the photovoltaic power plant and the cleaning effect of the actual cleaning work.

[0043] The data acquisition system sources its data from meteorological data: when a large-scale ground-mounted photovoltaic power station is equipped with a real-time weather station, the meteorological data reading program on the "cloud" side server will read the solar irradiance data from the real-time weather station; when the "cloud" side server can access satellite meteorological data, then satellite meteorological data will be used. Power generation performance data includes real-time current (I), real-time voltage (V), and real-time power output (P) data for photovoltaic modules, photovoltaic strings, photovoltaic subarrays, and the photovoltaic power station.

[0044] The data acquisition equipment includes a photovoltaic inverter and a module-level power electronic MLPE device. The photovoltaic inverter is configured to collect current and voltage data at the photovoltaic string level in response to commands from the host control program; the module-level power electronic MLPE device is configured to work with the photovoltaic inverter to collect current and voltage data at the photovoltaic module level.

[0045] Among them, the photovoltaic inverter has string-level I / O acquisition function, and can independently acquire photovoltaic string current (I) and voltage (V) data after receiving instructions from the upper control program. The photovoltaic module-level power electronics (MLPE) equipment (optimizer or shutdown device) has module-level I / O acquisition function, and can work with the photovoltaic inverter to detect photovoltaic module-level electrical output data (photovoltaic module current I, voltage V, etc.).

[0046] The data acquisition conditions were as follows: On the day before IV data acquisition (i.e., the day of the cleaning operation), a solar irradiance prediction model was run to predict the solar irradiance on the data acquisition day. Data with solar irradiance not less than 800 W / m2 and as close as possible to 1000 W / m2 were selected as the irradiance selection threshold.

[0047] The core function of a photovoltaic (PV) inverter is to act as the central execution unit for string-level data acquisition, responding to instructions from the cloud-side upper-level control program to accurately collect current and voltage data at the PV string level. Its effectiveness lies in achieving targeted string-level data acquisition by receiving unified scheduling instructions from the cloud side, ensuring precise matching between data acquisition and system operation and maintenance needs. Simultaneously, string-level data, as intermediate-level data, provides the cloud side with macroscopic power generation performance references and serves as a preliminary judgment basis for subsequent component-level data acquisition (e.g., triggering component-level acquisition only after determining that string performance is normal), avoiding invalid acquisition, reducing data transmission and cloud-side computational pressure, and ensuring the efficiency and relevance of data acquisition.

[0048] The cloud-side service component completes fault simulation calculations through a digital twin model, generating standard power generation performance benchmark data and fault characteristic data. Combined with the power plant's digital detection model, it compares and analyzes the measured data. Its core function is to decouple "anti-reflective film wear faults caused by cleaning" from "cell aging, dust accumulation and obstruction" and other coupled faults, avoiding misjudgments of faults. At the same time, combined with the location information of photovoltaic modules, it can accurately locate specific modules with severe wear, providing a basis for decision-making for targeted operation and maintenance.

[0049] The upper-level control program issues operation and maintenance instructions based on the cloud-side analysis results, realizing closed-loop control of "data acquisition - analysis and judgment - instruction issuance - operation and maintenance execution". Its core function is to avoid blind operation and maintenance, ensure that operation and maintenance actions are focused on the anti-reflection film wear components, and solve the problem of "targeted" operation and maintenance in traditional operation and maintenance.

[0050] Preferably, the system is configured to run in model-triggered mode, wherein: The digital twin model for photovoltaic cleaning operations is configured to operate autonomously in real time. Based on real-time meteorological data, historical meteorological data, and forecast meteorological data collected by the data acquisition system, it simulates the power generation, output, and cleaning effect of the photovoltaic power station. The digital twin model is further configured to make autonomous judgments based on the simulation results and trigger the upper-level control program to issue instructions to control the data acquisition system to collect the power generation performance data of the photovoltaic power station in order to perform predictive maintenance on the wear of the anti-reflective coating on the glass surface of the photovoltaic modules.

[0051] Specifically, the digital twin model operates autonomously in real time, automatically simulating the power generation, output, and cleaning operations of large-scale ground-mounted photovoltaic power stations based on solar irradiance, historical meteorological data, and forecasted meteorological data collected on-site. The model autonomously judges the simulation results and then issues instructions from the higher level to collect power generation performance data of large-scale photovoltaic power stations and perform predictive maintenance on the wear of anti-reflective coatings on the glass surfaces of photovoltaic modules.

[0052] The core function of the model-triggered mode is to proactively predict and perform forward-looking maintenance on anti-reflective film wear failures. In this mode, the digital twin model of photovoltaic cleaning operations operates autonomously based on real-time, historical, and forecast meteorological data. It continuously simulates the impact of power plant power generation, electricity output, and cleaning operations on the anti-reflective film. Through autonomous judgment, it triggers data collection and maintenance commands without human intervention, thus identifying potential anti-reflective film wear risks in advance. This shifts maintenance from "post-event remediation" to "pre-event prevention," avoiding the cumulative power attenuation caused by continued wear, significantly reducing power generation losses due to wear, and reducing the blind spots of manual inspections, thereby improving the timeliness and accuracy of maintenance responses.

[0053] The system is configured to run in detection-triggered mode, where: The power station digital detection model is configured to operate manually or semi-automatically according to the preset photovoltaic power station cleaning work schedule and technical requirements, and control the data acquisition system to collect power generation performance data of photovoltaic modules, photovoltaic strings and photovoltaic subarrays in the photovoltaic power station. The power station digital detection model is further configured to transmit the collected data back to the photovoltaic cleaning operation digital twin model for interactive judgment, so as to perform planned maintenance on the wear of the anti-reflection film on the glass surface of the photovoltaic modules.

[0054] Specifically, the digital detection model operates manually or semi-automatically, collecting and evaluating power generation performance data of photovoltaic modules, photovoltaic strings, and photovoltaic subarrays in the power station according to the cleaning work schedule and technical requirements of large-scale ground-mounted photovoltaic power stations. The data is then transmitted back to the digital twin model for interactive judgment, and planned maintenance is performed to address the wear of the anti-reflective coating on the glass surface of the photovoltaic modules.

[0055] The core function of the detection trigger mode is to adapt to the cleaning operation plan, enabling targeted and precise data collection and planned maintenance. In this mode, the power plant's digital detection model operates manually or semi-automatically according to the preset cleaning schedule, collecting power generation performance data of components, strings, and subarrays and transmitting it back to the digital twin model for interactive judgment. This ensures that the data collection and cleaning operation rhythm are precisely matched, allowing for data verification at key points after cleaning. This effectively eliminates interference from non-cleaning factors, providing high-quality data support for anti-reflective film wear identification. Simultaneously, it aligns with the power plant's operation and maintenance plan, promoting maintenance work in an orderly manner, avoiding conflicts between operation and maintenance and power generation operations, ensuring the standardization and efficiency of the operation and maintenance process, and further enhancing the accuracy of anti-reflective film wear fault identification.

[0056] Preferably, the power plant digital inspection model is configured to perform the following inspection process: Based on the digital twin model, a standard clear-sky power generation performance benchmark dataset of photovoltaic modules and photovoltaic strings under fault-free conditions and a standard power location marker map containing the location information of photovoltaic modules are generated. The control data acquisition system collects measured power generation performance data of photovoltaic strings and photovoltaic modules after the cleaning operation; The measured power generation data of each photovoltaic module is compared with the corresponding standard clear-sky power generation in the benchmark dataset, and the photovoltaic modules are classified into usable datasets, verification datasets, or excluded datasets based on the comparison deviation value. Based on the available dataset, the power generation attenuation value and location information of each photovoltaic module are calculated and used as input parameters for interactive judgment with the digital twin model.

[0057] Specifically, using the electrical models of photovoltaic modules and photovoltaic strings in the digital twin model, benchmark datasets of photovoltaic module-level standard clear-sky current (I-module-standard), voltage (V-module-standard), and power generation (P-module-standard) are generated as benchmark values ​​for the power generation performance of photovoltaic modules and photovoltaic strings without the impact of faults.

[0058] The specific location of photovoltaic strings and photovoltaic modules in a large-scale ground-mounted photovoltaic power station is generated from the CAD drawings of the power station, which contains a standard power location map (P-module-standard, Location-standard) that includes the standard clear-sky power generation of the photovoltaic modules and their location.

[0059] The upper-level control program issues instructions, and the photovoltaic inverter collects the photovoltaic string current (I-string-measured), voltage (V-string-measured), and real-time output power (P-string-measured). After combining these data with the standard clear-sky power generation performance benchmark values ​​of the photovoltaic string to determine that the power generation performance of the photovoltaic string is normal, the photovoltaic inverter issues instructions to the photovoltaic module-level power electronic device (MLPE) to collect the corresponding photovoltaic module current (I-module-measured), voltage (V-module-measured), and real-time output power (P-module-measured).

[0060] The core function of this detection process is to construct a complete data support system of "benchmark-acquisition-screening-quantification" for the identification of antireflective coating wear faults. By generating a standard clear-sky benchmark dataset and location marker map of the fault-free state through a digital twin model, an objective comparison benchmark is established, solving the problem of lack of a unified reference for fault identification.

[0061] Specifically, the data acquisition system is used to collect measured data after cleaning to ensure a strong correlation between the data and the cleaning operation, and to eliminate interference from non-cleaning factors. By classifying deviations, three types of datasets are selected: usable, verification, and exclusion. Invalid interference data such as shadows and occlusions are removed to ensure the validity of the input data. Finally, the power attenuation value and location information of the usable dataset are calculated to provide accurate and quantitative core input parameters for subsequent interaction and judgment with the digital twin model.

[0062] Preferably, the classification based on the comparison deviation value includes: Photovoltaic modules with a comparison deviation less than a first predetermined threshold are classified as available datasets; Photovoltaic modules with a comparison deviation between the first predetermined threshold and the second predetermined threshold are classified into a verification dataset, and their data are verified to eliminate interference factors. Once the verification is passed, they are merged into the available dataset. Photovoltaic modules with a comparison deviation greater than the second predetermined threshold and which have not improved after data review are classified as excluded datasets.

[0063] After the cleaning operation, the power generation data in the IV data of each photovoltaic module is compared with the power generation data in the IV data of each photovoltaic module in the photovoltaic module-level standard clear sky benchmark dataset on the cleaning operation day, and the following classifications are made: Available dataset: For photovoltaic modules with a comparison deviation of less than 1%, the cleaning effect is considered to be good, and the IV data of this module can be used for fault diagnosis.

[0064] Dataset verification: For photovoltaic modules with a comparison deviation greater than 1% and less than 10%, location marker maps are used to analyze the specific location of each photovoltaic module and photovoltaic string, and data verification is performed; combined with the collection time, solar position, and irradiance, the data collection time node is changed to eliminate the influence of shadow shading on the collected data; then this set of data is merged into the available dataset.

[0065] Excluded datasets: For photovoltaic modules with a comparison deviation greater than 10%, if there is still no improvement after data verification to avoid the influence of shading, it is determined that the cleaning effect is poor, and the data of such photovoltaic modules does not need to be used for this fault judgment.

[0066] The power generation attenuation dP-module-measured and location information Location-measured of the corresponding photovoltaic modules in the available dataset are calculated as interaction parameters with the digital twin model.

[0067] Based on the above steps, standardized data filtering rules can be established to accurately distinguish the validity of collected data and eliminate interference for subsequent fault diagnosis. By setting first and second predetermined thresholds, the measured data of photovoltaic modules are divided into three categories: usable, verification, and exclusion datasets. This ensures that data that directly meets the accuracy requirements (deviation less than the first threshold) can be quickly put into use, provides verification and correction opportunities for data with slight deviations (between the two thresholds), and removes data with serious deviations that cannot be improved (greater than the second threshold). This avoids invalid data such as shading and collection errors interfering with the diagnosis of antireflective film wear faults, ensuring that the data input into the digital twin model is reliable and targeted.

[0068] Preferably, the digital twin model for photovoltaic cleaning operations includes: A 3D optical model of a photovoltaic module is configured to calculate the photon absorption power of photovoltaic cells and photovoltaic modules based on a 3D optical calculation algorithm. The photovoltaic module electrical model is configured to be based on the equivalent diode model to calculate the electrical power of photovoltaic cells and photovoltaic modules; A 3D optical model of the photovoltaic string is configured to calculate the optical loss between the photovoltaic string and the surrounding physical structure based on a 3D optical calculation algorithm. The electrical model of the photovoltaic string is configured to calculate the electrical power of the photovoltaic string based on Ohm's law; The photovoltaic power calculation model is configured to use the photovoltaic module electrical model and meteorological input data to calculate the power generation of photovoltaic modules, photovoltaic strings, photovoltaic subarrays and photovoltaic power plants; The photovoltaic power prediction model is configured to calculate the predicted power generation of photovoltaic power generation units at different levels by using the electrical model of photovoltaic modules and predicted meteorological data, or by adopting a data-driven method. The solar irradiance prediction model is configured to use a data-driven approach, combining ground-based sensor data and satellite inversion data, to predict the temporal solar irradiance at a target location.

[0069] The seven sub-models of the photovoltaic cleaning operation digital twin model form a collaborative system encompassing "optics-electricity-power-prediction," with its core function being to provide accurate, multi-level simulation data support for monitoring anti-reflective film wear faults and decoupling multiple faults. Specifically, the 3D optical model of the photovoltaic module / string overcomes the challenges of calculating photon absorption power of cells / modules and optical losses of strings and surrounding structures through 3D optical algorithms, accurately capturing the optical performance changes caused by anti-reflective film wear; the electrical model of the photovoltaic module / string, based on the equivalent diode model and Ohm's law, converts optical data into electrical power data, establishing a "light-to-electricity" conversion simulation link; the photovoltaic power calculation model integrates the electrical model and meteorological data to achieve full-level power generation simulation from module to power plant; and the photovoltaic power prediction model and solar irradiance prediction model provide time-series prediction data through meteorological data and data-driven methods, providing a preliminary basis for the triggering and execution of the two operating modes.

[0070] This model configuration, through the collaboration of multiple sub-models, significantly improves fault identification accuracy and operation and maintenance support capabilities. On the one hand, the detailed design of the optical and electrical models accurately reproduces the impact path of anti-reflective film wear on "photon absorption-electrical output," making the simulated data highly consistent with the actual power generation scenario and providing a reliable computational foundation for quantifying anti-reflective film wear attenuation. On the other hand, the full-level power calculation and time-series prediction functions not only cover the multi-dimensional simulation needs of "module-string-subarray-power station," but also provide data support for the autonomous prediction of model triggering modes and the accurate verification of detection triggering modes, facilitating the efficient implementation of both working modes. At the same time, the fault simulation feature data generated by each sub-model (such as photocurrent and power attenuation values) provides core evidence for decoupling anti-reflective film wear from cell aging and dust accumulation shading faults, significantly reducing the fault misjudgment rate and providing a solid guarantee for the issuance of targeted operation and maintenance instructions.

[0071] Preferably, the 3D optical computing algorithm is configured as follows: Energy integration is performed on incident light with wavelengths ranging from 300 nm to 1200 nm. Its calculation range and the area irradiated by incident photon flux can be customized, and it can be gradually expanded from the photovoltaic cell level to the photovoltaic module, photovoltaic string, photovoltaic sub-array and photovoltaic power station level; In the calculation, for the response of each beam of light on the photovoltaic material, at least the following factors are considered: the photon flux corresponding to different wavelengths, the absorption probability of the photovoltaic cell material for photons of different wavelengths, and the quantum efficiency of generating charge carriers under a specific absorption probability.

[0072] Specifically, the 3D optical calculation algorithm performs energy integration on the visible light portion from 300 to 1200 nanometers. Its calculation range starts from the smallest unit of a photovoltaic power plant—the photovoltaic cell—and gradually expands to photovoltaic modules, photovoltaic strings, photovoltaic subarrays, and ultimately, the entire photovoltaic power plant. Furthermore, the irradiation area of ​​the incident photon flux can be customized, typically starting from the smallest unit of the photovoltaic power plant—the photovoltaic cell—and gradually expanding to photovoltaic modules, photovoltaic strings, photovoltaic subarrays, and finally, the entire photovoltaic power plant. The following factors are mainly considered regarding the response of each beam of light on the photovoltaic material: Photon flux: Based on the wave-particle duality of light, it corresponds to the photon energy at different wavelengths.

[0073] Absorption probability: The probability of absorbing photons of different wavelengths, depending on the different absorption depths of different photovoltaic cell materials.

[0074] Quantum efficiency: The efficiency of generating charge carriers for photons of different wavelengths at a certain absorption probability.

[0075] This algorithm primarily provides high-fidelity optical foundation data for digital twin models. By limiting the wavelength range to 300-1200 nanometers (covering the visible and near-infrared bands effectively absorbed by photovoltaic modules), it achieves accurate integration of the effective energy of incident light, avoiding interference from ineffective bands. It supports hierarchical expansion from photovoltaic cells to power plants and allows for customization of the irradiation area, meeting the simulation needs of multiple granularities from "single module to string to power plant" and adapting to different levels of fault identification scenarios. It clearly considers three key factors: photon flux, absorption probability, and quantum efficiency, completely replicating the propagation and energy conversion mechanism of light in photovoltaic materials, ensuring that optical calculations can accurately reflect optical losses caused by factors such as anti-reflective film wear and dust accumulation.

[0076] Preferably, the simulation of the digital twin model has a configurable simulation granularity, including: simulation of the power generation performance of a single photovoltaic module, and / or simulation of the power generation performance of photovoltaic strings, photovoltaic subarrays, and photovoltaic power plants.

[0077] Preferably, the digital twin model is configured to perform time-series process simulations to decouple multiple coexisting faults in the photovoltaic module; specifically configured as follows: For the N cleaning operations planned during the entire life cycle of a photovoltaic power station, the time node (t≤N) corresponding to the t-th cleaning operation is simulated in the digital twin model to generate the corresponding simulation results; Based on the generated time-series simulation results, feature data for identifying the relationships between different faults are extracted.

[0078] Specifically, in model-triggered mode: digital twin simulation is performed in real time. In detection-triggered mode: digital twin simulation is performed on demand.

[0079] Since major faults in photovoltaic modules may coexist and evolve due to causal relationships, in order to decouple individual faults and improve discrimination accuracy, it is necessary to generate feature data based on time-series simulation to determine the relationships between various faults. When the planned number of cleaning rounds throughout the entire life cycle of a photovoltaic power station is N (e.g., daily cleaning over 30 years, which is approximately 10,000 cleaning rounds), a digital twin model is used to simulate each cleaning time point t (t is less than or equal to N), generating t simulation results.

[0080] The simulated granularity is as follows: Power generation performance of a single photovoltaic module.

[0081] Power generation performance of photovoltaic strings, photovoltaic subarrays, and photovoltaic power plants The digital twin model is configured to perform time-series simulations. Through time-series simulations of the entire lifecycle, it captures the dynamic evolution patterns of multiple coexisting faults, providing specific characteristic basis for fault decoupling. For N cleaning operations throughout the lifecycle of a photovoltaic power station, simulations are performed sequentially at time nodes t. This accurately reconstructs the evolution trajectory of three types of faults—anti-reflective film wear, cell aging, and dust accumulation—with the number of cleaning cycles. For example, wear faults are strongly correlated with cleaning frequency, aging faults exhibit continuous and gradual characteristics, and dust accumulation faults are intermittent. Time-series simulation transforms these differentiated patterns into identifiable characteristic data, overcoming the limitation of static analysis in distinguishing coexisting faults and laying the foundation for subsequent accurate decoupling.

[0082] Preferably, the digital twin model is configured to generate a power generation performance map of a single photovoltaic module, wherein: The graph is a time-series characteristic data representing the main faults of photovoltaic modules, generated through simulation using 3D optical and electrical models of photovoltaic modules. The graph should include at least the simulated photocurrent (I-module-simulated) and simulated power generation (P-module-simulated) indicators; The graph generates simulation results for the following fault types: uniform aging fault of photovoltaic cells themselves, residual dust blockage fault caused by poor cleaning effect, and thinning fault of anti-reflective film of photovoltaic module glass caused by cleaning process.

[0083] Specifically, the power generation performance map of a single photovoltaic module: using a 3D optical model module for photovoltaic modules, a digital twin model module is constructed for a single photovoltaic module (the smallest simulation unit is a photovoltaic cell), and the time-series characteristic data map of the main faults of the photovoltaic module is generated through simulation. For the following main faults, a photovoltaic module power generation performance map is generated for a single photovoltaic module.

[0084] The photovoltaic module power generation performance graph includes two performance indicators: photocurrent (I-module-simulated) and power generation (P-module-simulated).

[0085] This power generation performance map can be used to process typical or predictive simulation data.

[0086] Each simulation result represents the result at time point t (t ≤ N): Photovoltaic cell aging failure: Simulated based on the uniform degradation of each photovoltaic cell within the photovoltaic module.

[0087] Poor cleaning effect and residual dust obstructing the view: simulated according to the degradation of each cell in the surrounding photovoltaic module.

[0088] Thinning failure of glass antireflective film caused by cleaning process: simulation of optical loss according to the uniform thinning and grooved thinning of the reflective film.

[0089] After the digital twin model is configured to generate the power generation performance map of a single photovoltaic module, it generates time-series characteristic data maps representing three core faults—uniform aging of photovoltaic cells, residual dust accumulation and obstruction, and thinning of the anti-reflective film—through the collaborative simulation of the 3D optical and electrical models of the photovoltaic module. Using simulated photocurrent (I-module-simulated) and power generation (P-module-simulated) as key indicators, it accurately captures the differentiated impact patterns of different faults on the power generation performance of the module. This provides a specific characteristic basis for decoupling the three easily coupled faults and lays the foundation for extending the fault characteristics of single modules to the string and power plant levels. It ensures the accurate identification and quantification of anti-reflective film wear faults and supports the formulation of targeted operation and maintenance decisions for the system.

[0090] Preferably, the digital twin model is further configured to extend the power generation performance map of a single photovoltaic module to the power generation performance maps of photovoltaic strings, photovoltaic subarrays, and photovoltaic power plants. The extension method includes: Extract the characteristic value Ex of the uniformity of the photocurrent in the power generation performance spectrum of a single photovoltaic module. The characteristic value Ex is formed by the coordinates of the cell with the largest photocurrent in the module and the current uniformity (E1, E2, E3, E4) in its four quadrants. For the three fault types, the characteristic value arrays of the uniformity of photogenerated current at each simulation time node t are generated respectively: Ex1_t, Ex2_t, Ex3_t. Based on the location marking map and electrical topology of the photovoltaic power station, the characteristic value array of each photovoltaic module is classified and aggregated to generate a station-wide simulated uniformity characteristic distribution of photovoltaic current. According to the station-wide characteristic distribution, the simulated power attenuation value, the corresponding module location information and the characteristic value array are correlated to form a fault judgment basis that can be used for cross-level comparison.

[0091] Specifically, among the degradation caused by wear on the surface glass of photovoltaic modules, the most sensitive indicator of power generation performance is the reduction in photocurrent of photovoltaic cells after damage to the anti-reflective film. This reduction, after passing through the internal circuitry of the photovoltaic module, is reflected in the decrease (or degradation) of the output current (I) of the photovoltaic module under the same testing conditions.

[0092] In a photovoltaic module, a photovoltaic module is usually composed of several photovoltaic cells (132 cells, 144 cells, etc.) connected in series and in parallel.

[0093] In large-scale ground-mounted photovoltaic power plants, multiple cleaning robots are typically deployed, with each robot responsible for cleaning one photovoltaic string. Therefore, during the cleaning operations of these robots, the damage to the surface glass of the photovoltaic modules within different photovoltaic subarrays and strings is unevenly distributed.

[0094] Therefore, it is necessary to extend the power generation performance map of a single photovoltaic module to the power generation performance map of photovoltaic strings and photovoltaic subarrays, in order to generate fault comparison datasets at different locations and under different conditions, so as to make a judgment with the digital detection results.

[0095] Extension methods: For each different time point, the uniformity of photocurrent Ex in the power generation performance spectrum of a single photovoltaic module is used as the feature value to extract mode features: Select the photovoltaic cell with the highest photocurrent in this photovoltaic module.

[0096] Record the coordinates (a, b) of the photovoltaic cell's position within the photovoltaic module.

[0097] Based on the in-situ location, the photovoltaic module is divided into four quadrants, and the uniformity of the photocurrent in each quadrant is calculated and denoted as E1, E2, E3, and E4. Let Ex(a,b,E1,E2,E3,E4) be the characteristic array of the photocurrent value of a certain photovoltaic module.

[0098] For the three faults mentioned above, generate an array of eigenvalues ​​Ex representing the uniformity of the simulated photocurrent in a single photovoltaic module: For time node t, the three types of faults generate Ex1t, Ex2t, and Ex3t respectively.

[0099] In the N cleaning rounds throughout the entire life cycle of the photovoltaic power station, there are N Ex feature value arrays for each of the three types of faults (photovoltaic cell aging fault, poor cleaning effect, residual dust obstruction fault, and glass antireflective film thinning fault caused by the cleaning process).

[0100] Based on the location marker map, the electrical topology of the entire large-scale ground-mounted photovoltaic power station is simulated, that is, how many photovoltaic modules each photovoltaic string consists of, how many photovoltaic strings each photovoltaic subarray consists of, and the location distribution of each photovoltaic string and photovoltaic subarray.

[0101] For each photovoltaic module in a large-scale ground-mounted photovoltaic power station, based on location and electrical topology, the power generation performance spectrum Ex feature array of a single photovoltaic module is used to categorize and generate the uniformity characteristic distribution of the simulated photovoltaic current at the entire station level: Those that meet the characteristics of photovoltaic cell aging failure Ex are classified as this type of failure.

[0102] Faults that meet the characteristics of poor cleaning effect and residual dust obstruction (Fault Ex) are classified as this type of fault.

[0103] Faults that meet the characteristics of Ex caused by thinning of the antireflective coating on glass during the cleaning process are classified as this type of fault.

[0104] The uniformity distribution of photovoltaic current at the whole-station level is simulated. It includes the simulated power attenuation of photovoltaic modules after each cleaning operation, and the location information of photovoltaic modules used in the simulation. The fault identification criteria are mapped to each module of the large ground photovoltaic power station at different cleaning operation times and locations using the (dP-module-simulated, Location-simulated, Ex) array.

[0105] This extended method constructs a dedicated array by extracting the uniformity feature value Ex of the photocurrent of a single module. It generates time-series feature arrays Ex1_t, Ex2_t, and Ex3_t for three types of faults according to time nodes. Then, combined with the power station location marking map and electrical topology, the features of the single module are aggregated into a station-wide simulated uniformity feature distribution of photocurrent. Finally, the simulated power attenuation value, module location information, and feature array are correlated to form a fault discrimination basis for cross-level comparison. This not only solves the problem of uneven distribution of faults such as anti-reflective film wear caused by cleaning at the string, subarray, and power station levels, but also realizes the extension of fault features from single modules to the entire station. It provides core support for accurate decoupling of multiple faults and full-level wear fault location, ensuring the pertinence and comprehensiveness of operation and maintenance instructions.

[0106] Preferably, the power generation performance map is constructed based on any of the following meteorological datasets: the typical meteorological year TMY dataset, or the predicted meteorological year FMY dataset; and the photovoltaic module electrical model, the photovoltaic string electrical model, and the photovoltaic power calculation model contain algorithms for decomposing solar irradiance in the time-series meteorological data into direct irradiance and diffuse irradiance, and inputting them into the photovoltaic module 3D optical model to generate photocurrent results corresponding to different fault types.

[0107] Specifically, the power generation performance graphs of a single photovoltaic module, photovoltaic strings, photovoltaic subarrays, and photovoltaic power plants are constructed using the photovoltaic module electrical model, photovoltaic string electrical model, and photovoltaic power calculation model as follows: Typical meteorological dataset: The input meteorological data uses typical meteorological year (TMY) data of the location of the photovoltaic power station.

[0108] Forecast meteorological dataset: The input meteorological data uses the forecast meteorological year (FMY) of the location of the photovoltaic power station.

[0109] The above datasets are time-series datasets, which typically contain data such as solar irradiance, ambient temperature, wind speed, wind direction, and precipitation.

[0110] To reflect the intermittent variations in photovoltaic power generation caused by rapid changes in solar irradiance, the time intervals of the above time series datasets are in the minute or second range.

[0111] The photovoltaic module electrical model, photovoltaic string electrical model, and photovoltaic power calculation model include a decomposition and conversion algorithm for direct and diffused irradiance that varies with time in the time-series meteorological dataset. The decomposed direct and diffuse irradiance are input into the photovoltaic module 3D optical model module to analyze and generate photocurrent results for photovoltaic cell aging faults, residual dust blockage faults due to poor cleaning effect, and glass anti-reflective film thinning faults caused by the cleaning process.

[0112] The photocurrent results of each photovoltaic module are extended to the power generation performance spectrum of the entire life cycle of the photovoltaic power station, at various time points and locations, by using uniformity as an indicator, through photovoltaic strings, photovoltaic subarrays and photovoltaic power stations.

[0113] The power generation performance map serves as a benchmark for comparing the collected data and is used for fault classification.

[0114] The simulation results generated by digital twin models can be categorized based on the different input data: When the input data is a typical dataset, the generated result is a typical data result.

[0115] When the input data is a prediction dataset, the generated result is the prediction data result.

[0116] Among these, the project provides a meteorological data foundation and precise optical calculation input that fits the actual scenario for the construction of power generation performance maps. By using typical meteorological year (TMY) or predicted meteorological year (FMY) datasets, it ensures that the maps can cover the power generation simulation needs under the normal and predicted operating conditions of the power plant. At the same time, by using the irradiance decomposition algorithm, the solar irradiance in the time-series meteorological data is divided into direct and diffuse irradiance, accurately restoring the impact of different irradiance types on the photocurrent of photovoltaic modules. This provides refined input for the 3D optical model of photovoltaic modules, thereby generating exclusive photocurrent results that can distinguish between three types of faults: cell aging, dust accumulation and obstruction, and thinning of anti-reflective film. This lays a reliable simulation data foundation for subsequent fault decoupling and accurate identification.

[0117] Preferably, the system further includes a fault identification module, configured to: acquire the measured data array output by the digital detection process of the power plant, the array containing the measured power attenuation value and location information of the photovoltaic modules in the available data; retrieve the simulated data array generated by the digital twin model, the array containing the simulated power attenuation value, simulated location information and related feature data; and complete the fault identification by cross-validating the measured data array and the simulated data array.

[0118] The upper-level control program is further configured to: locate and screen photovoltaic modules with severely worn anti-reflective films based on the output of the fault discrimination module, and generate a notification message to initiate the module replacement and maintenance process.

[0119] Among them, the upper-level fault identification program in the "cloud" side server can collect the power generation attenuation dP-module-measured of the photovoltaic module in the data center and the location data of the photovoltaic module in the location marker map during the digital detection of the power station, and generate the measured data array (dP-module-measured, Location-measured).

[0120] The fault diagnosis program retrieves the simulated values ​​generated by the digital twin model, cross-validates the power attenuation and location markers (dP-module-simulated, Location-simulated) of the simulated values, and uses this correspondence to determine the fault.

[0121] Locate and screen photovoltaic modules with severely worn anti-reflective coatings, and set up pop-up messages to notify the maintenance team to replace the photovoltaic modules.

[0122] The fault diagnosis module obtains the measured power attenuation value and location information array from the digital detection output of the power station. It then retrieves the array generated by the digital twin model containing simulated attenuation values, location information, and feature data. Through cross-validation, it accurately decouples the wear of the anti-reflective film from faults such as cell aging and dust accumulation, thus clarifying the fault type and degree. Simultaneously, the upper-level control program locates and selects photovoltaic modules with severe anti-reflective film wear based on the diagnosis results, generates a notification message to initiate the module replacement maintenance process, and achieves seamless connection between accurate fault diagnosis and targeted maintenance execution. This avoids ineffective maintenance, reduces power generation loss due to wear, and ensures the long-term stable operation of the photovoltaic power station.

[0123] The beneficial effects of this invention include: real-time monitoring and determination of the degree of wear on the anti-reflective coating on the photovoltaic module surface due to the cleaning process; real-time calculation and prediction of the resulting power degradation of the photovoltaic power station; improved digital operation and maintenance monitoring capabilities of the photovoltaic power station; and, when the front of the photovoltaic module is severely worn, suggestions for replacing the photovoltaic module are made, thereby improving operation and maintenance efficiency.

[0124] In this specification, the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for the product embodiments described later, since they correspond to the methods, the descriptions are relatively simple, and the relevant parts can be referred to the descriptions in the system embodiments.

[0125] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A digital operation and maintenance system for photovoltaic power plants, characterized in that, It includes an integrated interactive platform and a data acquisition system. The integrated interactive platform and the data acquisition system work together through data links and command links to monitor and identify wear faults in the anti-reflective film of photovoltaic module glass caused by cleaning operations. The integrated interactive platform includes cloud-side service components, edge-side service components, and terminal-side service components, wherein: The cloud-side service component is equipped with a digital twin model of photovoltaic cleaning operations, a host control program, a digital detection model of the power station, and a meteorological data reading program. The digital twin model is used to decouple the anti-reflective film wear fault of photovoltaic modules caused by cleaning operations from other types of faults, and to generate standard power generation performance benchmark data and fault simulation characteristic data of photovoltaic modules and strings. The edge service component is configured to acquire and process real-time power generation data of photovoltaic strings, photovoltaic subarrays and photovoltaic power plants, and can receive instructions from the cloud service component, forward them to the edge service component and upload stored data to the cloud. The end-side service component is configured to acquire real-time power generation data of a single photovoltaic module; The data acquisition system includes a photovoltaic inverter, a module-level power electronic MLPE device, and a meteorological data source, used to acquire power generation performance data and meteorological data at the photovoltaic module level, string level, subarray level, and power station level, and upload the acquired data to the cloud-side service component. The cloud-side service component performs simulation calculations and comparative analysis on the uploaded data through the digital twin model and the power station digital detection model. Combining the location information of the photovoltaic modules with the simulated feature data, it completes fault decoupling and wear location. Finally, the upper-level control program issues operation and maintenance instructions to realize targeted operation and maintenance of the photovoltaic power station.

2. The digital operation and maintenance system for photovoltaic power plants according to claim 1, characterized in that, The system is configured to run in model-triggered mode, wherein: The photovoltaic cleaning operation digital twin model is configured to run autonomously in real time. Based on real-time meteorological data, historical meteorological data and predicted meteorological data collected by the data acquisition system, it simulates the power generation, power output and cleaning operation effect of the photovoltaic power station. The digital twin model is further configured to autonomously judge based on the simulation results and trigger the upper-level control program to issue instructions to control the data acquisition system to collect power generation performance data of the photovoltaic power station, so as to perform predictive maintenance on the wear of the anti-reflective coating on the glass surface of the photovoltaic modules; or The system is configured to operate in a detection-triggered mode, wherein: The digital detection model of the power station is configured to operate in a manual or semi-automatic manner according to the preset photovoltaic power station cleaning work schedule and technical requirements, and to control the data acquisition system to collect the power generation performance data of photovoltaic modules, photovoltaic strings and photovoltaic subarrays in the photovoltaic power station. The digital detection model of the power station is further configured to transmit the collected data back to the digital twin model of the photovoltaic cleaning operation for interactive judgment, so as to perform planned maintenance on the wear of the anti-reflective film on the glass surface of the photovoltaic module.

3. The digital operation and maintenance system for photovoltaic power plants according to claim 1, characterized in that, The data acquisition system includes: A photovoltaic inverter is configured to collect current and voltage data of the photovoltaic string in response to instructions from the host control program. A module-level power electronic MLPE device is configured to work with the photovoltaic inverter to collect current and voltage data at the photovoltaic module level.

4. The digital operation and maintenance system for photovoltaic power plants according to claim 3, characterized in that, The power plant digital detection model is configured to perform the following detection process: Based on the digital twin model, a standard clear-sky power generation performance benchmark dataset of photovoltaic modules and photovoltaic strings under fault-free conditions and a standard power location marker map containing the location information of photovoltaic modules are generated. The data acquisition system is controlled to collect measured power generation performance data of the photovoltaic strings and photovoltaic modules after the cleaning operation; The measured power generation data of each photovoltaic module is compared with the corresponding standard clear-sky power generation in the benchmark dataset, and the photovoltaic modules are classified into usable datasets, verification datasets, or excluded datasets based on the comparison deviation value. Based on the available dataset, the power generation attenuation value and location information of each photovoltaic module are calculated, which serve as input parameters for interactive judgment with the digital twin model.

5. The digital operation and maintenance system for photovoltaic power plants according to claim 4, characterized in that, The classification based on the comparison deviation value includes: Photovoltaic modules with a comparison deviation less than a first predetermined threshold are classified into the available dataset; Photovoltaic modules with a comparison deviation between the first predetermined threshold and the second predetermined threshold are classified into the verification dataset, and their data is verified to eliminate interference factors. After the verification is passed, they are incorporated into the available dataset. Photovoltaic modules whose comparison deviation is greater than the second predetermined threshold and which have not improved after data review are classified as the excluded dataset.

6. The digital operation and maintenance system for photovoltaic power plants according to claim 1, characterized in that, The digital twin model of the photovoltaic cleaning operation includes: A 3D optical model of a photovoltaic module is configured to calculate the photon absorption power of photovoltaic cells and photovoltaic modules based on a 3D optical calculation algorithm. The photovoltaic module electrical model is configured to be based on the equivalent diode model to calculate the electrical power of photovoltaic cells and photovoltaic modules; A 3D optical model of a photovoltaic string is configured to calculate the optical loss between the photovoltaic string and the surrounding physical structure based on the 3D optical calculation algorithm. The electrical model of the photovoltaic string is configured to calculate the electrical power of the photovoltaic string based on Ohm's law; The photovoltaic power calculation model is configured to use the photovoltaic module electrical model and meteorological input data to calculate the power generation of photovoltaic modules, photovoltaic strings, photovoltaic subarrays and photovoltaic power stations; The photovoltaic power prediction model is configured to calculate the predicted power generation of the photovoltaic power generation units at different levels using the photovoltaic module electrical model and predicted meteorological data, or by adopting a data-driven method. The solar irradiance prediction model is configured to use a data-driven approach, combining ground-based sensor data and satellite inversion data, to predict the temporal solar irradiance at a target location.

7. The digital operation and maintenance system for photovoltaic power plants according to claim 6, characterized in that, The 3D optical calculation algorithm is configured as follows: Energy integration is performed on incident light with wavelengths ranging from 300 nm to 1200 nm. Its calculation range and the area irradiated by incident photon flux can be customized, and it can be gradually expanded from the photovoltaic cell level to the photovoltaic module, photovoltaic string, photovoltaic sub-array and photovoltaic power station level; In the calculation, for the response of each beam of light on the photovoltaic material, at least the following factors are considered: the photon flux corresponding to different wavelengths, the absorption probability of the photovoltaic cell material for photons of different wavelengths, and the quantum efficiency of generating charge carriers under a specific absorption probability.

8. The digital operation and maintenance system for photovoltaic power plants according to claim 1, characterized in that, The digital twin model's simulation work has a configurable simulation granularity, which includes: simulation of the power generation performance of a single photovoltaic module, and / or simulation of the power generation performance of photovoltaic strings, photovoltaic subarrays, and photovoltaic power plants.

9. The digital operation and maintenance system for photovoltaic power plants according to claim 1, characterized in that, The digital twin model is configured to perform time-series process simulations to decouple multiple coexisting faults in photovoltaic modules; specifically, the configuration is as follows: For the N cleaning operations planned for a photovoltaic power station throughout its entire life cycle, the time node (t ≤ N) corresponding to the t-th cleaning operation is simulated in the digital twin model to generate the corresponding simulation results; Based on the generated time-series simulation results, feature data for identifying the relationships between different faults are extracted.

10. The digital operation and maintenance system for photovoltaic power plants according to claim 6, characterized in that, The digital twin model is configured to generate a power generation performance spectrum of a single photovoltaic module, wherein: the spectrum is time-series characteristic data characterizing the main faults of the photovoltaic module, generated by simulation using the 3D optical and electrical models of the photovoltaic module; the spectrum includes at least simulated photocurrent (I-module-simulated) and simulated power generation (P-module-simulated) indicators; the spectrum generates simulation results for the following fault types: uniform aging fault of the photovoltaic cell itself, residual dust shading fault caused by poor cleaning effect, and thinning fault of the anti-reflective film of the photovoltaic module glass caused by the cleaning process.

11. The digital operation and maintenance system for photovoltaic power plants according to claim 10, characterized in that, The digital twin model is further configured to extend the power generation performance map of the single photovoltaic module to the power generation performance maps of photovoltaic strings, photovoltaic subarrays, and photovoltaic power plants. The extension method includes: Extract the characteristic value Ex of the uniformity of the photocurrent in the power generation performance spectrum of the single photovoltaic module. The characteristic value Ex is formed into an array based on the coordinates of the cell with the largest photocurrent in the module and the current uniformity (E1, E2, E3, E4) in its four quadrants. For the three fault types, respectively generate photocurrent uniformity characteristic value arrays Ex1_t, Ex2_t, and Ex3_t at each simulation time node t; Based on the location marking map and electrical topology of the photovoltaic power station, the feature value array of each photovoltaic module is classified and aggregated to generate a station-wide simulated uniformity feature distribution of photovoltaic current. According to the station-wide feature distribution, the simulated power attenuation value, the corresponding module location information and the feature value array are associated to form a fault judgment basis that can be used for cross-level comparison.

12. The digital operation and maintenance system for photovoltaic power plants according to claim 10 or 11, characterized in that, The power generation performance map is constructed based on any of the following meteorological datasets: the typical meteorological year TMY dataset, or the predicted meteorological year FMY dataset; and the photovoltaic module electrical model, photovoltaic string electrical model and photovoltaic power calculation model contain algorithms for decomposing solar irradiance in time-series meteorological data into direct irradiance and diffuse irradiance, and inputting them into the photovoltaic module 3D optical model to generate photocurrent results corresponding to different fault types.

13. The digital operation and maintenance system for photovoltaic power plants according to claim 1, characterized in that, The system also includes a fault detection module, configured as follows: Obtain the measured data array output from the digital detection process of the power plant, wherein the array contains the measured power attenuation value and location information of the photovoltaic modules in the available data set; Retrieve the simulated data array generated by the digital twin model, the array containing simulated power attenuation values, simulated location information, and related feature data; Fault identification is completed by cross-validating the measured data array and the simulated data array; The upper-level control program is further configured to: locate and screen photovoltaic modules with severely worn anti-reflective films based on the output of the fault discrimination module, and generate a notification message to initiate the module replacement and maintenance process.

Images