Method, device, storage medium and electronic device for processing photovoltaic module data

Abstract

The application discloses a data processing method and device of a photovoltaic assembly, a storage medium and an electronic device. A federal learning coordination platform is deployed on a central server, and a globally shared base network parameter is initialized. The central server distributes the initialized globally shared base network to authenticated photovoltaic power station nodes, each photovoltaic power station node completes the construction of a local computing environment, and carries out model training in parallel based on local historical data. During the training process, each photovoltaic power station node regularly calculates the update amount of the model parameter, and uploads the encrypted and differentially private processed update amount to the central server. The central server integrates the parameters of each node by using a secure aggregation algorithm, and generates an updated global model. Thus, the technical problem of realizing accurate corrosion grading under a small sample condition on the premise of protecting the data privacy of each power station is solved.

Description

Technical Field

[0001] This invention relates to photovoltaic power plants, and more specifically, to a method, apparatus, storage medium, and electronic device for processing photovoltaic module data. Background Technology

[0002] Coastal photovoltaic power plants are exposed to corrosive environments characterized by high salt spray and high humidity. The metal frames of the modules are susceptible to salt spray corrosion, leading to decreased structural strength and deteriorated grounding performance, ultimately affecting module power generation efficiency and operational safety. Currently, power plants primarily rely on daytime inspections to collect images of the module frames, with maintenance personnel judging the degree of corrosion based on experience. However, significant differences in marine environments result in varying feature distributions in corrosion images collected from different power plants, making it difficult for a single power plant to accumulate sufficient labeled samples, thus limiting the performance of deep learning-based corrosion recognition models. Furthermore, existing methods largely focus on corrosion morphology recognition, failing to establish a quantitative correlation between corrosion severity and changes in module power generation, resulting in a lack of data support for maintenance decisions. Traditional centralized training requires aggregating data from various power plants, posing a risk of data privacy breaches, and the scarcity of cross-domain samples leads to insufficient model generalization ability. Therefore, achieving accurate corrosion classification under small sample conditions while protecting the data privacy of each power plant, and establishing a correlation model between corrosion severity and power generation, has become a key challenge for the intelligent operation and maintenance of coastal photovoltaic power plants. Summary of the Invention

[0003] This invention provides a method, apparatus, storage medium, and electronic device for processing photovoltaic module data, to at least solve the problem in the prior art of achieving accurate corrosion classification under small sample conditions while protecting the data privacy of each power station.

[0004] According to one aspect of the present invention, a data processing method for photovoltaic modules is provided, comprising: deploying a federated learning coordination platform on a central server and initializing global shared base network parameters; the central server distributing the initialization of the global shared base network to certified photovoltaic power station nodes, each photovoltaic power station node completing the construction of its local computing environment and conducting parallel model training based on local historical data; during the training process, each photovoltaic power station node periodically calculating the update amount of the model parameters, encrypting and performing differential privacy processing on them before uploading them to the central server; the central server using a secure aggregation algorithm to integrate the parameters of each node and generate an updated global model.

[0005] Optionally, after generating the updated global model, the method further includes: each photovoltaic power station node using the optimized local model to identify the corrosion level of the real-time acquired component border images; for identification results identified by the local model with a confidence level lower than a set threshold, automatic manual review is triggered, and the reviewed and confirmed samples are included in the local training set.

[0006] Optionally, after generating the updated global model, the method further includes: precisely aligning the time series to match the corrosion detection results with the power generation data within the corresponding time period; correcting the impact on power generation through environmental factors; and establishing a quantitative relationship model between corrosion state and power generation performance using statistical analysis methods.

[0007] Optionally, after establishing the quantitative relationship model between corrosion state and power generation performance, the method further includes: continuously monitoring the operating status, and automatically triggering an incremental learning process when new samples that meet predetermined conditions are accumulated or when performance degradation is detected.

[0008] Optionally, each photovoltaic power station node completes the setup of its local computing environment, including: each photovoltaic power station node synchronously completes the setup of its local computing environment, including the calibration of the image acquisition system, the establishment of the data storage structure, and the testing of a secure connection with the central server; wherein, each photovoltaic power station node submits a description of its local environment characteristics to the central server during the registration process.

[0009] Optionally, each photovoltaic power station node completes the construction of its local computing environment and conducts model training in parallel based on local historical data, including: collecting visible light images and infrared thermal imaging data of the component frame, and extracting complementary visual information through feature-level fusion technology; in the data preprocessing stage, adopting a virtual sample generation method based on generative adversarial networks; and in the feature extraction stage, using a cross-modal attention mechanism to automatically learn the importance weights of different modal features.

[0010] According to another aspect of the present invention, a data processing apparatus for photovoltaic modules is also provided, comprising: an initialization module for deploying a federated learning coordination platform on a central server and initializing global shared base network parameters; a distribution module for the central server to distribute the initialized global shared base network to certified photovoltaic power station nodes, wherein each photovoltaic power station node completes the construction of its local computing environment and performs model training in parallel based on local historical data; an upload module for each photovoltaic power station node to periodically calculate the update amount of model parameters during the training process, and upload the updated parameters to the central server after encryption and differential privacy processing; and an update module for the central server to integrate the parameters of each node using a secure aggregation algorithm to generate an updated global model.

[0011] Optionally, the device further includes: an identification module, used for each photovoltaic power station node to identify the corrosion level of the real-time acquired component border image using the optimized local model after the updated global model is generated; and a triggering module, used to automatically trigger manual review for identification results identified by the local model with a confidence level lower than a set threshold, and to include the reviewed and confirmed samples into the local training set.

[0012] In this embodiment of the invention, a federated learning coordination platform is deployed on a central server to initialize the global shared base network parameters. The central server distributes the initialized global shared base network to the certified photovoltaic power station nodes. Each photovoltaic power station node completes the construction of its local computing environment and conducts parallel model training based on local historical data. During training, each photovoltaic power station node periodically calculates the update amount of the model parameters, encrypts and performs differential privacy processing on them, and then uploads them to the central server. The central server uses a secure aggregation algorithm to integrate the parameters of each node and generate an updated global model. That is, the server uses a secure aggregation algorithm to fuse update information from different nodes, ensuring that the local training data of each node cannot be reconstructed at any processing stage. This solves the technical problem in the prior art of achieving accurate corrosion classification under small sample conditions while protecting the data privacy of each power station. Attached Figure Description

[0013] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a structural diagram of a photovoltaic module data processing method according to an embodiment of the present invention; Figure 2 This is a flowchart of a photovoltaic module data processing method according to an embodiment of the present invention; Figure 3 This is a flowchart of a method for classifying and identifying photovoltaic module frame salt spray corrosion and associating it with power generation based on federated small sample data, according to an embodiment of the present invention. Figure 4 This is a structural diagram of a photovoltaic module data processing device according to an embodiment of the present invention. Detailed Implementation

[0014] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0015] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a sequence of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0016] The photovoltaic module data processing method provided in this application can be executed on a mobile terminal, computer terminal, or similar computing device. Taking running on a central server as an example... Figure 1 As shown, the mobile terminal 10 may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 (which may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and a memory 104 for storing data are also shown. Optionally, the mobile terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the mobile terminal described above. For example, the mobile terminal 10 may also include components that are more... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0017] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the multimodal fusion-based autonomous driving behavior prediction method in this embodiment of the invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, thereby implementing the above-described method. The memory 104 may include high-speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the mobile terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0018] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the mobile terminal 10. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0019] This embodiment also provides a method for processing photovoltaic module data. Figure 2 This is a flowchart of a photovoltaic module data processing method according to an embodiment of the present invention, such as... Figure 2 As shown, the process flow for processing photovoltaic module data includes the following steps: Step S202: Deploy the federated learning coordination platform on the central server and initialize the global shared base network parameters.

[0020] In step S204, the central server distributes the initialization of the global shared base network to the certified photovoltaic power station nodes. Each photovoltaic power station node completes the construction of its local computing environment and conducts model training in parallel based on local historical data.

[0021] In step S206, during the training process, each photovoltaic power station node periodically calculates the update amount of the model parameters, and uploads them to the central server after encryption and differential privacy processing.

[0022] In step S208, the central server uses a secure aggregation algorithm to integrate the parameters of each node and generate an updated global model.

[0023] In this embodiment, a privacy-preserving distributed collaborative learning architecture is constructed, comprising a central coordination server and multiple edge computing nodes forming a star topology. The central server does not directly access the raw data from each node; instead, it receives processed model parameter updates via an encrypted channel. Each photovoltaic power station node completes data preprocessing and model training locally, transmitting only differentially privacy-processed gradient information to the server. The server employs a secure aggregation algorithm to fuse update information from different nodes, ensuring that the local training data of each node cannot be reconstructed at any processing stage, technically eliminating the possibility of data privacy leakage.

[0024] To address the technical challenge of scarce corrosion samples from coastal photovoltaic power plants, a few-shot learning method based on multimodal data augmentation is proposed. The system simultaneously acquires visible light images and infrared thermal imaging data of the component outlines, extracting complementary visual information through feature-level fusion technology. In the data preprocessing stage, a virtual sample generation method based on generative adversarial networks is employed to effectively expand the training sample set while maintaining the authenticity of the data distribution. In the feature extraction stage, a cross-modal attention mechanism is designed to automatically learn the importance weights of features from different modalities, significantly improving the feature representation capability under few-shot conditions.

[0025] A hierarchical model architecture balancing generality and specificity is proposed, consisting of a shared base network and a personalized adaptation network. The shared base network employs a deep convolutional neural network structure, extracting common corrosion features with strong generalization ability from multi-power plant data through federated learning. The personalized adaptation network adopts a lightweight fully connected layer structure, learning specific environmental adaptation parameters for each power plant using a small number of local samples. The two networks are trained end-to-end by jointly optimizing the loss function, maintaining the model's generality across different power plants while also considering the adaptation requirements of specific environments.

[0026] This example deploys a federated learning coordination platform on a central server to initialize the globally shared base network parameters. The central server distributes the initialized globally shared base network to certified photovoltaic (PV) power plant nodes. Each PV power plant node completes the setup of its local computing environment and conducts parallel model training based on local historical data. During training, each PV power plant node periodically calculates the update amount of model parameters, encrypts and performs differential privacy processing before uploading it to the central server. The central server uses a secure aggregation algorithm to integrate the parameters from each node, generating an updated global model. In other words, the server uses a secure aggregation algorithm to fuse update information from different nodes, ensuring that the local training data of each node cannot be reconstructed at any processing stage. This solves the technical problem in existing technologies of achieving accurate corrosion classification under small sample conditions while protecting the data privacy of each power plant.

[0027] Optionally, after generating the updated global model, the above may also include: each photovoltaic power station node using the optimized local model to identify the corrosion level of the real-time acquired component border images; for identification results identified by the local model with a confidence level lower than a set threshold, automatic manual review is triggered, and the reviewed and confirmed samples are included in the local training set.

[0028] Optionally, after generating the updated global model, the above method may further include: precisely aligning the time series to match the corrosion detection results with the power generation data within the corresponding time period; correcting the impact on power generation through environmental factors; and establishing a quantitative relationship model between corrosion state and power generation performance using statistical analysis methods.

[0029] In this embodiment, a deep correlation analysis model between corrosion state and power generation performance is constructed. First, a data association mechanism based on timestamp alignment is established to ensure that corrosion detection results and power generation data are accurately matched in the time dimension. Then, multi-scale feature extraction technology is used to quantify corrosion features from multiple perspectives such as texture, shape, and color of corrosion images. Finally, a deep learning model is used to establish a mapping relationship between corrosion state evolution and power generation performance degradation, achieving accurate modeling from visual features to performance prediction.

[0030] Optionally, after establishing a quantitative relationship model between corrosion state and power generation performance, the above may also include: continuously monitoring the operating status, and automatically triggering an incremental learning process when new samples that meet predetermined conditions are accumulated or when performance degradation is detected.

[0031] In this embodiment, a system adaptive evolution module based on continuous learning is designed. A real-time monitoring module continuously tracks the model's performance on new data, automatically identifying performance degradation or concept drift. When a decline in model performance is detected, the system automatically initiates a gradual update process, incorporating new learning content while preserving existing knowledge. Simultaneously, a quality assessment module is introduced to verify the credibility of new samples, ensuring the quality of training data and achieving continuous system optimization and long-term stable operation.

[0032] Optionally, each photovoltaic power station node completes the setup of its local computing environment, which may include: each photovoltaic power station node synchronously completing the setup of its local computing environment, including the calibration of the image acquisition system, the establishment of the data storage structure, and the testing of the secure connection with the central server; wherein, each photovoltaic power station node submits a description of its local environment characteristics to the central server during the registration process.

[0033] Optionally, each photovoltaic power station node completes the construction of its local computing environment and conducts parallel model training based on local historical data. This may include: collecting visible light images and infrared thermal imaging data of the component frame and extracting complementary visual information through feature-level fusion technology; in the data preprocessing stage, adopting a virtual sample generation method based on generative adversarial networks; and in the feature extraction stage, using a cross-modal attention mechanism to automatically learn the importance weights of different modal features.

[0034] As an optional embodiment, this application also provides a method for graded identification and power generation correlation of photovoltaic module frame salt spray corrosion based on federated small samples. A federated learning network composed of multiple coastal photovoltaic power plants is constructed. Each power plant trains a graded model using locally collected frame corrosion images, uploading only the model parameter gradients to the consortium server to avoid leakage of raw data. Simultaneously, a lightweight model architecture combining a general corrosion feature base and a power plant adaptation layer is adopted. The general base learns shared corrosion features across multiple power plants through federated aggregation, while the adaptation layer fine-tunes using a small number of locally labeled samples to achieve a three-level classification of corrosion severity (mild, moderate, and severe). Furthermore, the identification results are correlated with the historical power generation data of the corresponding modules over time to establish a correlation between corrosion level and power generation attenuation rate, providing a quantitative basis for power plant maintenance decisions.

[0035] The specific content of this embodiment includes the following: 1. A privacy-preserving distributed collaborative learning architecture was constructed, comprising a central coordination server and multiple edge computing nodes forming a star topology. The central server does not directly access the raw data of each node, but instead receives processed model parameter updates through an encrypted channel. Each photovoltaic power station node completes data preprocessing and model training locally, only transmitting differentially privacy-processed gradient information to the server. The server employs a secure aggregation algorithm to fuse update information from different nodes, ensuring that the local training data of each node cannot be reconstructed at any processing stage, technically eliminating the possibility of data privacy leakage.

[0036] 2. To address the technical challenge of scarce corrosion samples from coastal photovoltaic power plants, a few-shot learning method based on multimodal data augmentation is proposed. The system simultaneously acquires visible light images and infrared thermal imaging data of the component outlines, extracting complementary visual information through feature-level fusion technology. In the data preprocessing stage, a virtual sample generation method based on generative adversarial networks is adopted to effectively expand the training sample set while maintaining the authenticity of the data distribution. In the feature extraction stage, a cross-modal attention mechanism is designed to automatically learn the importance weights of features from different modalities, significantly improving the feature representation capability under few-shot conditions.

[0037] 3. A hierarchical model architecture balancing generality and specificity is proposed, consisting of a shared base network and a personalized adaptation network. The shared base network employs a deep convolutional neural network structure, extracting common corrosion features with strong generalization ability from multi-power station data through federated learning; the personalized adaptation network adopts a lightweight fully connected layer structure, learning specific environmental adaptation parameters of the power station using a small number of local samples. The two networks are trained end-to-end by jointly optimizing the loss function, maintaining the model's generality across different power stations while also considering the adaptation requirements of specific environments.

[0038] 4. A deep correlation analysis model between corrosion state and power generation performance was constructed. First, a data association mechanism based on timestamp alignment was established to ensure that corrosion detection results and power generation data are accurately matched in the time dimension. Then, multi-scale feature extraction technology was used to quantify corrosion features from multiple perspectives such as texture, shape, and color of corrosion images. Finally, through a deep learning model, a mapping relationship between corrosion state evolution and power generation performance degradation was established, realizing accurate modeling from visual features to performance prediction.

[0039] 5. A system adaptive evolution module based on continuous learning was designed. The system continuously tracks the model's performance on new data through a real-time monitoring module, automatically identifying performance degradation or concept drift. When a decline in model performance is detected, the system automatically initiates a gradual update process, incorporating new learning content while preserving existing knowledge. Simultaneously, a quality assessment module is introduced to verify the credibility of new samples, ensuring the quality of training data and achieving continuous optimization and long-term stable operation of the system.

[0040] The implementation process in this embodiment is as follows: Figure 3 As shown, the specific steps are as follows: 1. First, a federated learning coordination platform is deployed on the central server to initialize the globally shared base network parameters and configure secure communication protocols and differential privacy parameters. Simultaneously, each photovoltaic power station node completes the setup of its local computing environment, including the calibration of the image acquisition system, the establishment of data storage structures, and secure connection testing with the server. During the registration process, nodes submit a description of their local environment characteristics to the server, including geographical location, equipment model, and environmental conditions, providing foundational data for subsequent personalized adaptation.

[0041] 2. After system initialization, the distributed model training phase begins. The server distributes the initialized shared base network to all authenticated nodes, and each node trains the model in parallel based on its local historical data. During training, nodes periodically calculate the update amount of model parameters, encrypt and perform differential privacy processing before uploading them to the server. The server uses a secure aggregation algorithm to integrate the contributions of each node and generate an updated global model. This process continuously optimizes model performance through multiple iterations until the preset convergence conditions are met.

[0042] 3. After federated training is complete, the system transitions to the online application phase. Each node uses its optimized local model to identify the erosion level of the component bounding box images acquired in real time. The identification process not only outputs the specific erosion level classification but also provides a corresponding confidence score. For identification results with a confidence score below a set threshold, the system automatically triggers a manual review process and incorporates the reviewed and confirmed samples into the local training set, forming a closed-loop mechanism for quality assurance.

[0043] 4. After obtaining the corrosion identification results, the system initiates a multimodal data correlation analysis process. This process first precisely aligns the time series, matching the corrosion detection results with the power generation data within the corresponding time period; then, through environmental factor correction, it eliminates the influence of external factors such as weather and seasons on power generation; finally, it uses statistical analysis methods to establish a quantitative relationship model between corrosion status and power generation performance, providing a scientific basis for operation and maintenance decisions.

[0044] 5. Based on the aforementioned analysis results, the system generates targeted operation and maintenance suggestions, including maintenance priority ranking and recommended maintenance plans. Simultaneously, the system continuously monitors its operational status, automatically triggering an incremental learning process when sufficient new samples are accumulated or performance degradation is detected. The incremental learning process carefully incorporates new learning content while preserving existing knowledge, ensuring system stability and continuous improvement capabilities.

[0045] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0046] This embodiment also provides a photovoltaic module data processing device for implementing the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0047] Figure 4 This is a structural block diagram of a photovoltaic module data processing device according to an embodiment of the present invention, such as... Figure 4 As shown, the photovoltaic module data processing device includes: Initial module 401 is used to deploy the federated learning coordination platform on the central server and initialize the global shared base network parameters; The distribution module 403 is used by the central server to distribute the initialization of the global shared base network to the certified photovoltaic power station nodes. Each photovoltaic power station node completes the construction of its local computing environment and conducts model training in parallel based on local historical data. Upload module 405 is used during the training process to periodically calculate the update amount of model parameters of each photovoltaic power station node, and then encrypt and perform differential privacy processing before uploading them to the central server; Update module 407 is used by the central server to integrate the parameters of each node using a secure aggregation algorithm to generate an updated global model.

[0048] Through the embodiments provided in this application, the initialization module 401 deploys a federated learning coordination platform on a central server and initializes the global shared base network parameters; the distribution module 403, the central server, distributes the initialized global shared base network to the certified photovoltaic power station nodes, and each photovoltaic power station node completes the construction of its local computing environment and conducts model training in parallel based on local historical data; during the training process, the upload module 405, the photovoltaic power station nodes periodically calculate the update amount of the model parameters, encrypt and perform differential privacy processing, and then upload them to the central server; the update module 407, the central server, uses a secure aggregation algorithm to integrate the parameters of each node and generate an updated global model. That is, the server uses a secure aggregation algorithm to fuse update information from different nodes, ensuring that the local training data of each node cannot be reconstructed at any processing stage. This solves the technical problem in the prior art of achieving accurate corrosion classification under small sample conditions while protecting the data privacy of each power station.

[0049] Optionally, the above-mentioned device may further include: an identification module, used for each photovoltaic power station node to identify the corrosion level of the real-time acquired component border image using the optimized local model after the updated global model is generated; and a triggering module, used to automatically trigger manual review for identification results identified by the local model with a confidence level lower than a set threshold, and to include the reviewed and confirmed samples into the local training set.

[0050] It should be noted that the above modules can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above modules are located in the same processor; or, the above modules are located in different processors in any combination.

[0051] Embodiments of the present invention also provide a storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above method embodiments when running.

[0052] Optionally, in this embodiment, the storage medium may be configured to store a computer program for performing the following steps: S1, deploy the federated learning coordination platform on the central server and initialize the global shared base network parameters; S2, the central server distributes the initialization of the global shared base network to the certified photovoltaic power station nodes. Each photovoltaic power station node completes the construction of its local computing environment and conducts model training in parallel based on local historical data. S3, during the training process, each photovoltaic power station node periodically calculates the update amount of the model parameters, and uploads them to the central server after encryption and differential privacy processing. S4, the central server uses a secure aggregation algorithm to integrate the parameters of each node and generate an updated global model.

[0053] It should be noted that the above modules can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above modules are located in the same processor; or, the above modules are located in different processors in any combination.

[0054] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0055] Embodiments of the present invention also provide an electronic device including a memory and a processor, the memory storing a computer program and the processor being configured to run the computer program to perform the steps in any of the above method embodiments.

[0056] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.

[0057] Optionally, in this embodiment, the processor can be configured to perform the following steps via a computer program: S1, deploy the federated learning coordination platform on the central server and initialize the global shared base network parameters; S2, the central server distributes the initialization of the global shared base network to the certified photovoltaic power station nodes. Each photovoltaic power station node completes the construction of its local computing environment and conducts model training in parallel based on local historical data. S23. During the training process, each photovoltaic power station node periodically calculates the update amount of the model parameters, and uploads them to the central server after encryption and differential privacy processing. S4, the central server uses a secure aggregation algorithm to integrate the parameters of each node and generate an updated global model.

[0058] Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, and will not be repeated here.

[0059] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those presented herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0060] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A data processing method for photovoltaic modules, characterized in that, include: Deploy the federated learning coordination platform on the central server and initialize the globally shared base network parameters; The central server distributes the initialization of the global shared base network to the certified photovoltaic power station nodes. Each photovoltaic power station node completes the construction of its local computing environment and conducts model training in parallel based on local historical data. During the training process, each photovoltaic power station node periodically calculates the update amount of the model parameters, and uploads them to the central server after encryption and differential privacy processing. The central server uses a secure aggregation algorithm to integrate the parameters of each node and generate an updated global model.

2. The method according to claim 1, characterized in that, After generating the updated global model, the method further includes: Each photovoltaic power station node uses an optimized local model to identify the corrosion level of the real-time acquired component border images; If the confidence level of the identification result identified by the local model is lower than the set threshold, manual review will be automatically triggered, and the reviewed and confirmed samples will be included in the local training set.

3. The method according to claim 2, characterized in that, After generating the updated global model, the method further includes: Accurate alignment of the time series data is performed to match corrosion detection results with power generation data within the corresponding time period; The impact of environmental factor correction on power generation; A quantitative relationship model between corrosion state and power generation performance was established using statistical analysis methods.

4. The method according to claim 3, characterized in that, After establishing the quantitative relationship model between corrosion state and power generation performance, the method further includes: The system continuously monitors the operational status and automatically triggers the incremental learning process when new samples that meet predetermined conditions are accumulated or when performance degradation is detected.

5. The method according to claim 1, characterized in that, Each photovoltaic power station node completes the setup of its local computing environment, including: Each photovoltaic power station node synchronously completes the construction of its local computing environment, including the calibration of the image acquisition system, the establishment of the data storage structure, and the testing of a secure connection with the central server. During the registration process, each photovoltaic power station node submits a description of its local environmental characteristics to the central server.

6. The method according to claim 1, characterized in that, Each photovoltaic power station node completes the setup of its local computing environment and conducts parallel model training based on local historical data, including: Visible light images and infrared thermal imaging data of the component frame are collected, and complementary visual information is extracted through feature-level fusion technology. In the data preprocessing stage, a virtual sample generation method based on generative adversarial networks is adopted; During the feature extraction stage, a cross-modal attention mechanism automatically learns the importance weights of features from different modalities.

7. A data processing device for a photovoltaic module, characterized in that, include: The initial module is used to deploy the federated learning coordination platform on the central server and initialize the globally shared base network parameters; The distribution module is used by the central server to distribute the initialization of the global shared base network to the certified photovoltaic power station nodes. Each photovoltaic power station node completes the construction of its local computing environment and conducts model training in parallel based on local historical data. The upload module is used during the training process to periodically calculate the update amount of model parameters of each photovoltaic power station node, and then encrypt and perform differential privacy processing before uploading them to the central server. The update module is used by the central server to integrate the parameters of each node using a secure aggregation algorithm to generate an updated global model.

8. The apparatus according to claim 7, characterized in that, The device further includes: The identification module is used to identify the corrosion level of the component border images collected in real time after the updated global model is generated. The triggering module is used to automatically trigger manual review when the recognition results identified by the local model have a confidence level lower than a set threshold, and to include the reviewed and confirmed samples into the local training set.

9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, wherein the computer program is configured to execute the method described in any one of claims 1 to 6 when it is run.

10. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to perform the method as described in any one of claims 1 to 6.

Images