A photovoltaic power station operation and maintenance management system and method based on big data

Abstract

The present invention discloses a photovoltaic power station operation and maintenance management system and method based on big data, which relates to the field of photovoltaic power generation, including collecting photovoltaic component operation data, pre-processing the data using edge computing equipment, and storing it on a cloud big data platform; based on the stored photovoltaic component operation data, using a self-supervised learning method to establish a basic model of photovoltaic power station power generation behavior; combining historical data, tracing the abnormal power generation behavior in the distributed ledger and triggering operation and maintenance tasks using smart contracts; after the operation and maintenance task is executed, the processed abnormal power generation behavior is fed back to the self-supervised learning model sample library, and the model is dynamically adjusted; based on the execution content of the operation and maintenance task, an operation and maintenance report is generated and uploaded to the blockchain; the present invention automatically triggers the operation and maintenance task through the smart contract, realizes the full process automated management from abnormality detection to task issuance, and improves the operation and maintenance efficiency and the operational stability of the photovoltaic power station.

Description

Technical Field

[0001] The present invention relates to the field of photovoltaic power generation, and in particular to a photovoltaic power station operation and maintenance management system and method based on big data. Background Art

[0002] With the global energy transition, photovoltaic power generation, as a clean and renewable energy source, has experienced rapid growth in recent years, gradually becoming a vital component of energy production. However, the efficient operation of a photovoltaic power plant depends not only on the performance of photovoltaic modules but also on proper operation and maintenance management to ensure its long-term stable operation. Traditional operation and maintenance management relies primarily on regular manual inspections and simple rule-based alarm systems. However, with the continuous expansion of photovoltaic power plant scale and the increasing complexity of operating environments, this approach is no longer able to meet the efficiency, intelligence, and automation requirements of modern photovoltaic power plants. In recent years, with the rise of technologies such as big data, blockchain, and machine learning, a growing number of research and applications have begun to explore how to incorporate these cutting-edge technologies into the operation and maintenance of photovoltaic power plants, aiming to improve data processing efficiency, anomaly detection accuracy, and intelligent operation and maintenance decision-making.

[0003] Therefore, the present invention provides a photovoltaic power station operation and maintenance management system and method based on big data to solve the problem of insufficient accuracy in operation and maintenance anomaly detection. Although the existing photovoltaic power station operation and maintenance technology has achieved status monitoring and fault alarms for photovoltaic components to a certain extent, it still has many shortcomings. Most of the existing fault detection systems are based on simple preset rules, which are difficult to cope with the complex and changeable environmental factors in actual operation, resulting in insufficient accuracy and real-time performance of anomaly detection. In contrast, self-supervised learning can autonomously learn the potential structure of data by generating pseudo-labels from unlabeled data, dynamically adapt to different operating environments and improve the accuracy of anomaly detection. In addition, existing operation and maintenance data management systems usually rely on centralized databases for storage, and the security, transparency and traceability of operation and maintenance data are relatively limited. Especially when multiple stakeholders are involved or operation and maintenance data needs to be displayed externally, traditional centralized data storage methods are prone to tampering problems and lack a reliable third-party audit mechanism. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a photovoltaic power station operation and maintenance management system and method based on big data to solve the problems of insufficient accuracy of anomaly detection and limited security and transparency of operation and maintenance data.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a photovoltaic power station operation and maintenance management system and method based on big data, which includes collecting photovoltaic module operation data, pre-processing the data using edge computing devices, and storing it on a cloud big data platform;

[0008] Based on the stored PV module operation data, a basic model of PV power station power generation behavior is established using self-supervised learning methods;

[0009] Compare PV module operating data with the basic power generation behavior model to identify abnormal power generation behavior, upload the abnormal power generation behavior to the blockchain, and record it in the distributed ledger;

[0010] Combined with historical data, abnormal power generation behavior in the distributed ledger is traced and analyzed, and smart contracts are used to trigger operation and maintenance tasks;

[0011] After the operation and maintenance task is completed, the abnormal power generation behavior is fed back to the self-supervised learning model sample library to dynamically adjust the model;

[0012] Based on the execution content of the operation and maintenance tasks, an operation and maintenance report is generated and uploaded to the blockchain.

[0013] In the second aspect, the present invention provides a photovoltaic power station operation and maintenance management system and method based on big data, including a data acquisition module, a model construction module, an anomaly detection module, an anomaly tracing module, an operation and maintenance feedback module, and a report generation module; the data acquisition module is used to collect photovoltaic component operation data, use edge computing equipment to pre-process the data, and store it on a cloud big data platform; the model construction module is used to establish a basic model of photovoltaic power station power generation behavior based on the stored photovoltaic component operation data and a self-supervised learning method; the anomaly detection module is used to compare the photovoltaic component operation data and the basic model of power generation behavior, identify abnormal power generation behavior, upload the abnormal power generation behavior to the blockchain, and record it in a distributed ledger; the anomaly tracing module is used to combine historical data to trace the abnormal power generation behavior in the distributed ledger and use smart contracts to trigger operation and maintenance tasks; the operation and maintenance feedback module is used to feed back the processed abnormal power generation behavior to the self-supervised learning model sample library after the operation and maintenance task is executed, and dynamically adjust the model; the report generation module is used to generate an operation and maintenance report based on the execution content of the operation and maintenance task and upload it to the blockchain.

[0014] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the photovoltaic power station operation and maintenance management system and method based on big data as described in the first aspect of the present invention is implemented.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the photovoltaic power station operation and maintenance management system and method based on big data as described in the first aspect of the present invention.

[0016] The beneficial effects of the present invention are as follows: the present invention introduces self-supervised learning and blockchain technology into the photovoltaic power station operation and maintenance management system, forming an efficient and intelligent operation and maintenance management system. Through the self-supervised learning model, the system can automatically extract features from the photovoltaic component operation data and establish a basic model of power generation behavior, reducing the dependence on manually labeled data, and improving the generalization ability of the model and the accuracy of anomaly detection. The introduction of blockchain technology solves the data tampering and trust problems that may exist in traditional centralized data management systems, enhances the system's traceability, and realizes data transparency and credibility assurance. By automatically triggering operation and maintenance tasks through smart contracts, the full process of automated management from anomaly detection to task issuance is realized, which improves the operation and maintenance efficiency and the operational stability of photovoltaic power stations. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 This is a flow chart of the photovoltaic power station operation and maintenance management system and method based on big data in Example 1.

[0019] Figure 2 This is a flow chart of the report generation module in Example 1. DETAILED DESCRIPTION

[0020] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0021] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0022] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0023] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a photovoltaic power station operation and maintenance management system and method based on big data, including:

[0024] S1: Collect PV module operation data, pre-process the data using edge computing devices, and store it on the cloud big data platform.

[0025] Specifically, the following steps are included:

[0026] S1.1: Deploy a sensor network in the PV power station to collect PV module operating data in real time.

[0027] Specifically, it is necessary to select sensors that support the RS485 communication protocol to collect the current, voltage, irradiance, temperature and humidity data of the photovoltaic modules.

[0028] Install one sensor for each set of PV panels. The irradiance sensors should be placed in an unobstructed area and aligned with the inclination of the PV panels. Each sensor should be individually numbered to facilitate subsequent data identification.

[0029] The communication connection is carried out through the RS485 protocol, and each sensor is connected in series to the edge computing device, and a 120Ω terminal resistor is connected at the terminal.

[0030] Finally, calibrate the sensors and connect the debugging software to verify the communication stability of each sensor.

[0031] S1.2: Use edge computing devices to remove outliers and noise interference using noise filtering algorithms, use wavelet transform to compress data, and output standardized data.

[0032] It should be noted that the data acquisition method is to install Modbus reading software in the edge computing device (such as Raspberry Pi or industrial PC), set the collection interval to 10 seconds, poll all sensor data regularly, and record current, voltage, irradiance, temperature and humidity.

[0033] The noise filtering method is to use the moving average method to smooth short-term fluctuations for 5 consecutive data points. Detect abnormal values ​​that exceed the physical reasonable range and remove them (for example, the current value is less than 0 or greater than the device rated value, or the irradiance detected at night is greater than 5W / m 2Wavelet transform can be called through the pywt library, and only the low-frequency signal part needs to be retained.

[0034] It should be understood that outputting standardized data means storing the processed data in the log backup in JSON format.

[0035] S1.3: Use the MQTT protocol for data transmission and upload the processed standardized data to the cloud.

[0036] Specifically, configure the MQTT client (edge ​​device) and server (cloud). The client needs to set the topic and QoS level, and regularly publish processed JSON data to the server. The server needs to set up Subscribe to receive the data stream and store the data in the database.

[0037] Optimally, efficient and stable data acquisition, precise data preprocessing, and appropriate data compression and transmission not only improve the real-time performance, accuracy, and reliability of the overall solution, but also reduce transmission costs and storage pressure. These technical measures provide solid data support for subsequent photovoltaic power plant operation and maintenance management, performance optimization, and predictive maintenance, helping to improve the operational efficiency and economic benefits of photovoltaic power plants.

[0038] S2: Based on the stored PV module operation data, a basic model of PV power station power generation behavior is established using self-supervised learning methods.

[0039] Specifically, the following steps are included:

[0040] S2.1: Perform multi-channel mapping and data augmentation on the normalized data.

[0041] It should be understood that multi-channel mapping is to ensure that each data channel can be used as an input feature independently.

[0042] Among them, data enhancement includes the use of interpolation, noise addition, random scaling and other methods.

[0043] S2.2: Use deep convolutional neural network to extract features from enhanced data.

[0044] Specifically, a convolution operation is applied to the enhanced data to extract the local features of each sensor. A pooling operation is then applied to reduce the feature dimension and extract the feature map h.

[0045] S2.3: Use a two-layer fully connected neural network to map the features to the contrastive learning space and obtain the feature representation z, which is expressed as:

[0046] z=W2·ReLU(W1·h+b1)+b2;

[0047] Among them, h is the output of the feature extraction network, W1 is the weight matrix of the first layer, W2 is the weight matrix of the second layer, b1 is the bias term of the first layer fully connected network, and b2 is the bias term of the second layer fully connected network.

[0048] S2.4: Using SimCLR to compare the learning architecture, define the contrast loss function, which is expressed as:

[0049]

[0050] Among them, z i To enhance the feature representation of the view, z j z i The corresponding positive sample pair, τ is the temperature scaling factor, N is the number of samples, 1 [k≠i] is the indicator function, i is the sample size index coefficient, j is z i The corresponding positive sample index coefficient, k is the index coefficient of the negative sample pair, sim(z i ,z j ) is the cosine similarity.

[0051] The periodic enhancement factor C(t) is introduced to modify the cosine similarity, and the expression is:

[0052]

[0053] The periodic enhancement factor C(t) is expressed as:

[0054]

[0055] Where t is the current time, T is the daily period, α is the periodic amplitude correction parameter, and β is the balance factor.

[0056] S2.5: Use stochastic gradient descent optimization and dynamically adjust the learning rate.

[0057] Specifically, the learning rate expression is:

[0058]

[0059] Among them, η t is the learning rate of the tth iteration, η0 is the initial learning rate, and T is the total number of training rounds.

[0060] S2.6: Use the KNN clustering method to visualize the feature representation z and verify the separability of the embedded representation by projecting it into a two-dimensional space.

[0061] Among them, the feature representation z is projected into two-dimensional space through t-SNE dimensionality reduction technology to visualize the similarity and separability between samples.

[0062] The optimal basic model of photovoltaic power plant power generation behavior, established through self-supervised learning methods, comprehensively improves the intelligence and accuracy of photovoltaic operation and maintenance management. Multi-channel mapping and data augmentation technologies ensure the multidimensional information integrity of input data. Combined with convolutional networks and fully connected networks, deep features are extracted to effectively describe the operating status of photovoltaic modules. The introduction of the SimCLR comparative learning architecture and periodic enhancement factors enhances the model's adaptability to periodic characteristics and complex scenarios. Dynamic learning rate optimization improves training efficiency, while KNN clustering and t-SNE dimensionality reduction techniques make feature embedding highly interpretable and visualizable. This model provides a high-quality baseline for subsequent anomaly detection, improving the reliability of photovoltaic power plant operations.

[0063] S3: Compare the PV module operating data with the basic power generation behavior model, identify abnormal power generation behavior, upload the abnormal power generation behavior to the blockchain, and record it in the distributed ledger.

[0064] Specifically, the following steps are included:

[0065] S3.1: Collect the real-time photovoltaic module operation data x t , through the feature extraction network, obtain the feature representation z t ;

[0066] S3.2: Calculate the anomaly index A(z t ,z), the expression is:

[0067]

[0068] Among them, A(z t ,z) is the distance between the model prediction and the actual observation, z t is the feature vector set at the current time t (extracted from the real-time collected photovoltaic module data), z t,i is the feature vector of the i-th sample generated at the current time point t, z is the basic feature vector set learned in the model, and z i is the feature vector of the i-th basic feature sample, ∥·∥2 is the Euclidean distance, which is used to measure the spatial difference between the two. is a similarity-based metric function used to supplement the distribution information that cannot be captured by the Euclidean distance, λ is a balance parameter, and N is the number of feature vector samples.

[0069] It should be noted that the similarity measurement function The expression is:

[0070]

[0071] This expression is used to measure the directional consistency between the current feature and the basic feature. The more the direction deviates (that is, the smaller the similarity), the more likely anomaly will occur.

[0072] The balance parameter λ is a hyperparameter that adjusts the weight of the Euclidean distance and similarity metrics. If λ>1, the model relies more on similarity to judge anomalies. If λ<1, the model tends to use distance to evaluate anomalies.

[0073] S3.3: Comparative abnormality index A(z t ,z) and set the abnormal judgment threshold θ to identify abnormal power generation behavior. Specifically, the abnormal judgment threshold θ is expressed as:

[0074] θ=μ+k·σ;

[0075] Where μ is the mean of the anomaly index A in the historical data, σ is the standard deviation of the anomaly index A in the historical data, and k is a constant ranging from 2 to 3. k = 2 is used in scenarios with higher fault tolerance to reduce false positives, and k = 3 is used in strict scenarios to enhance the ability to identify abnormal behaviors.

[0076] Contrast abnormality index A(z t ,z) and set the abnormal judgment threshold θ, when A(z t ,z)>θ, the current operating state is considered abnormal. t ,z)<θ, the current operating status is considered normal.

[0077] S3.4: Build a record of abnormal power generation behavior, create and call a smart contract, upload the abnormal power generation behavior to the blockchain, and generate a transaction hash value.

[0078] The smart contract created needs to include functions such as uploading exception records, storing the hash value of the record to the distributed ledger, and returning the transaction ID and block location of the successfully stored transaction.

[0079] It should be noted that the generated transaction hash value can be used to trace the uploaded abnormal records to ensure that the data cannot be tampered with.

[0080] The best approach is to build an efficient, intelligent, and secure photovoltaic power plant operation and maintenance system by integrating anomaly detection algorithms, dynamic threshold adjustment, SHA-256 hash encryption, and blockchain storage technology. This approach not only improves the accuracy of abnormal behavior identification but also ensures data immutability, convenient tracking, and transparent management through blockchain. This significantly reduces operation and maintenance costs, enhances system reliability, and provides strong support for the future large-scale development of the photovoltaic industry.

[0081] S4: Combine historical data to trace the abnormal power generation behavior in the distributed ledger and use smart contracts to trigger operation and maintenance tasks.

[0082] Specifically, the following steps are included:

[0083] S4.1: Extract the abnormal power generation behavior record R from the distributed ledger and collect the operating parameters of the PV power station during the period when the abnormal behavior occurred.

[0084] S4.2: Define a spatiotemporal correlation graph G = (V, E), where V is the set of abnormal event nodes, corresponding to each abnormal power generation behavior record R i , the edge set E is the spatiotemporal correlation between abnormal events, and the expression is:

[0085]

[0086] Among them, w ij is the association weight of the node, S ij is the similarity of feature vectors between nodes, θ ij is the angular difference of the node in the spatiotemporal distribution, γ is the similarity weight factor, κ is the similarity adjustment factor, δ is the similarity threshold, and φ is the geographical distribution correction factor.

[0087] S4.3: Construct a causal inference network N = (V, E, W) and calculate the causal weight matrix W.

[0088] Specifically, the causal weight matrix W is calculated based on historical anomaly records and operating data in combination with causal relationship modeling tools (such as Granger causal analysis).

[0089] S4.4: Using the causal weight matrix W and the traceability target node R t ,Determine the tracing path through the depth-first search algorithm.

[0090] S4.5: Design smart contract task allocation rules, expressed as:

[0091]

[0092] in, To assign priority, w ij is the association weight of the corresponding node, c ij is the causal weight of the corresponding node, β(T e ) is the time priority function, and its expression is:

[0093]

[0094] Among them, is the time impact factor, T c is the current time, T e is the event time.

[0095] Furthermore, smart contract rules are written while ensuring that task priorities are dynamically matched with association weights, causal weights, and time factors.

[0096] S4.6: After the contract is triggered, operation and maintenance instructions are issued through the IoT device and cloud platform interface.

[0097] Specifically, the operation and maintenance instructions issued include component inspection, cleaning, maintenance, sensor calibration, etc., and are targeted at operation and maintenance personnel or designated photovoltaic components.

[0098] Optimally, by combining historical data with distributed ledgers and using smart contracts to trigger operation and maintenance tasks, the system can accurately trace and analyze abnormal power generation behavior in photovoltaic power plants. The constructed spatiotemporal correlation graph and causal reasoning network help identify the correlations and causal relationships between abnormal events. Smart contracts dynamically assign operation and maintenance tasks based on correlation weights, causal weights, and time priorities, ensuring efficient and accurate task completion. Combining big data analysis with blockchain technology not only improves the accuracy of abnormal event tracing, but also enhances data security and system transparency. Ultimately, the system enables automated anomaly monitoring and operation and maintenance response, significantly improving the operation and maintenance efficiency of photovoltaic power plants.

[0099] S5: After the operation and maintenance task is executed, the processed abnormal power generation behavior is fed back to the self-supervised learning model sample library to dynamically adjust the model.

[0100] Specifically, the following steps are included:

[0101] S5.1: Extract the feedback data after the operation and maintenance task is processed and the abnormal data before processing, and format the data.

[0102] Specifically, the processed feedback data is recorded as x after , the abnormal data before processing is recorded as x before .

[0103] S5.2: Extract sample features through the feature extraction network, introduce the correction vector, map the data to the contrastive learning space, and obtain the corrected sample.

[0104] Specifically, the trained feature extraction network F(·) is used to extract features and obtain the feature vector z before , z after . Introduce the correction vector z fix , the expression is:

[0105] z fix =z after -z before ;

[0106] Perform data enhancement (such as noise perturbation, amplitude scaling, etc.) on the correction vector to generate multi-view features, denoted as z fix,aug .

[0107] The enhanced data is mapped to the contrastive learning space and mapped to the modified feature representation z through a two-layer fully connected network proj, the expression is:

[0108] z proj =W2·ReLU(W1·z fix,aug +b1)+b2;

[0109] S5.3: Store the corrected samples into the self-supervised learning model library.

[0110] S5.4: Based on the updated sample library, re-perform self-supervised learning on the basic model.

[0111] Among them, the method of re-self-supervised learning is to minimize the distance between the corrected feature representation and the original abnormal feature, and the expression is:

[0112]

[0113] S5.5: Add a modified sample weight factor to the original contrastive learning loss function and adjust the loss function.

[0114] Specifically, the modified sample weight factor is recorded as γ fix , the loss function expression is:

[0115]

[0116] S5.6: Dynamically adjust the learning rate based on the number of correction samples.

[0117] Specifically, the learning rate expression is:

[0118]

[0119] Among them, |D fix | is the corrected sample size, |D total | is the total sample size.

[0120] Optimally, through a closed-loop mechanism of operational feedback, the model can continuously adapt to new abnormal behaviors and improve its performance in different operational environments. The introduction of correction vectors and data augmentation methods enables the model to better learn and understand the anomaly repair process, thereby improving the accuracy of abnormal behavior repair. Multi-view learning and dynamic adjustment of the learning rate enhance the model's ability to handle new types of anomalies and reduce the risk of overfitting. Dynamic adjustment of the loss function weights and learning rate ensures effective learning progress and stability during training. These effects collectively promote continuous improvement in model performance, ensuring that the model maintains high accuracy and stability during actual operational processes.

[0121] S6: Generate an operation and maintenance report based on the execution content of the operation and maintenance task and upload it to the blockchain.

[0122] Specifically, the following steps are included:

[0123] S6.1: Collect operation and maintenance task execution data and generate operation and maintenance reports.

[0124] Among them, the operation and maintenance task execution data includes: power generation power and efficiency before and after processing, abnormality type, repair operation, repair result, task execution timestamp, location, equipment information and other related log data and diagnostic results.

[0125] The contents of the operation and maintenance report include: description and execution status of the operation and maintenance task; specific operation steps and results of abnormal repair; equipment and tools used in the operation and maintenance process; results and verification results after repair; metadata such as task execution time, personnel, and location.

[0126] S6.2: Digitally sign the operation and maintenance report using a private key.

[0127] S6.3: Call the blockchain interface, upload the operation and maintenance report to the distributed ledger, and obtain the transaction hash value.

[0128] S6.4: Store operation and maintenance reports and record the mapping relationship between report files and blockchain transaction hash values ​​in the database.

[0129] Ideally, through the above process, operations and maintenance reports not only provide high security, transparency, and immutability in storage and verification, but also significantly improve data management efficiency, system automation, and compliance. Leveraging blockchain technology, operations and maintenance reports can be stored and managed transparently, verifiably, and immutably, significantly impacting enterprises in improving operations and maintenance efficiency, ensuring data security, enhancing audit capabilities, and improving compliance.

[0130] This embodiment also provides a photovoltaic power station operation and maintenance management system and method based on big data, including:

[0131] The data acquisition module is used to collect PV module operating data, pre-process the data using edge computing devices, and store it on the cloud big data platform;

[0132] A model building module is used to build a basic model of the power generation behavior of a photovoltaic power station based on the stored photovoltaic module operation data using a self-supervised learning method;

[0133] Anomaly detection module, which compares PV module operating data with the basic power generation behavior model, identifies abnormal power generation behavior, uploads abnormal power generation behavior to the blockchain, and records it in the distributed ledger;

[0134] The abnormality tracing module is used to combine historical data to trace and analyze abnormal power generation behavior in the distributed ledger and use smart contracts to trigger operation and maintenance tasks;

[0135] The operation and maintenance feedback module is used to feed back abnormal power generation behavior to the self-supervised learning model sample library after the operation and maintenance tasks are completed, and dynamically adjust the model;

[0136] The report generation module is used to generate operation and maintenance reports based on the execution content of operation and maintenance tasks and upload them to the blockchain.

[0137] This embodiment also provides a computer device suitable for the photovoltaic power station operation and maintenance management system and method based on big data, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the photovoltaic power station operation and maintenance management system and method based on big data proposed in the above embodiment.

[0138] The computer device may be a terminal, comprising a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0139] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the photovoltaic power station operation and maintenance management system and method based on big data as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0140] In summary, the present invention achieves the following: The present invention introduces self-supervised learning and blockchain technology into the photovoltaic power station operation and maintenance management system to form an efficient and intelligent operation and maintenance management system. Through the self-supervised learning model, the system can automatically extract features from the photovoltaic component operation data and establish a basic model of power generation behavior, reducing the dependence on manually labeled data and improving the generalization ability of the model and the accuracy of anomaly detection. The introduction of blockchain technology solves the data tampering and trust problems that may exist in traditional centralized data management systems, enhances the system's traceability, and achieves data transparency and credibility assurance. The operation and maintenance tasks are automatically triggered by smart contracts, realizing the full process of automated management from anomaly detection to task issuance, improving the operation and maintenance efficiency and the operational stability of photovoltaic power stations.

[0141] Example 2, referring to Table 1, is the second embodiment of the present invention. In order to further verify the technical solution of the present invention, experimental simulation data of a photovoltaic power station operation and maintenance management system and method based on big data are provided.

[0142] The experiment selected 6 photovoltaic modules (numbered as module 1 to module 6) and collected their operating data in real time through the deployed sensor network. The raw data recorded by these sensors include voltage (V), current (A), temperature (℃), humidity (%), irradiance (W / m 2Data is first preprocessed by edge computing devices, using noise filtering algorithms to eliminate outliers and noise interference to ensure data quality. The data is then standardized to conform to a unified dimension. Finally, a wavelet transform is used to compress the data, reducing storage and transmission requirements while preserving key feature information. The standardized data is uploaded to a cloud-based big data platform and input into a self-supervised learning model for feature extraction. For each component, an anomaly index is calculated using a formula based on the current time point characteristics and features from the basic power generation behavior model.

[0143] The sensor collected data is shown in Table 1 below:

[0144] Table 1 Sensor data collection table

[0145] Component Number Current (A) Voltage (V) <![CDATA[Irradiance (W / m 2 )]]> Temperature (℃) humidity(%) Component 1 9.5 230 950 28 60 Component 2 4.2 210.5 850 32.5 70 Component 3 10.1 235 1050 27.5 58 Component 4 12.3 240 1000 35 72 Component 5 8.9 225.5 920 29 65 Component 6 7.5 215 800 30 75

[0146] The data is fed into the self-supervised learning model to calculate the feature vector z of each component.

[0147] The characteristic vectors of the components are shown in Table 2 below:

[0148] Table 2 Component feature vector table

[0149]

[0150]

[0151] Calculate the abnormal index A(z t ,z), the expression is:

[0152]

[0153] The specific process is:

[0154] Calculate the Euclidean distance, the expression is,

[0155]

[0156] Calculate the similarity measure, the expression is,

[0157]

[0158]

[0159] z t ·z i =z t,1 ·z i,1 +z t,2 ·z i,2 +z t,3 ·z i,3 ;

[0160]

[0161] Calculate the anomaly index, the expression is,

[0162]

[0163] Get λ=1, z t =[0.50,0.60,0.55], taking component 1 and component 2 as examples, calculate the abnormal index A(z t ,z).

[0164] Component 1:

[0165] From Table 2, we can see that z1 = [0.45, 0.60, 0.52].

[0166] Euclidean distance:

[0167]

[0168] Similarity measure:

[0169] z t ·z1=0.50·0.45+0.60·0.60+0.55·0.52=0.225+0.36+0.286= 0.871;

[0171]

[0172] Abnormal index:

[0173]

[0174] Component 2:

[0175] From Table 2, we can see that z2 = [1.30, 1.60, 1.50].

[0176] Euclidean distance:

[0177]

[0178] Similarity measure:

[0179] z t ·z2=0.50·1.30+0.60·1.60+0.55·1.50=0.65+0.96+0.825= 2.435;

[0181]

[0182]

[0183] Abnormal index:

[0184]

[0185] The other components are calculated according to the above example.

[0186] The abnormal index of the components is shown in Table 3 below:

[0187] Table 3 Component abnormality index table

[0188] Component Number Abnormal Index Component 1 0.00147 Component 2 0.817 Component 3 0.00103 Component 4 0.00117 Component 5 0.0021 Component 6 0.6745

[0189] The expression of abnormal judgment threshold θ is:

[0190] θ=μ+k·σ;

[0191] Take the historical anomaly index A data, calculate the mean μ and standard deviation σ, and in the strict scenario of k=3, the anomaly judgment threshold θ is 0.439.

[0192] Compare the abnormality judgment threshold θ with the abnormality index of the component,

[0193] Component 1:

[0194] 0.00147<0.439;

[0195] Component 2:

[0196] 0.817>0.439;

[0197] Component 3:

[0198] 0.00103<0.439;

[0199] Component 4:

[0200] 0.00117<0.439;

[0201] Component 5:

[0202] 0.0021<0.439;

[0203] Component 6:

[0204] 0.06745>0.439;

[0205] The anomaly indexes of components 2 and 6 are greater than the threshold θ, indicating that their feature vectors differ significantly from the model predictions, indicating that these components are faulty or have degraded performance. The anomaly indexes of the other components are less than the threshold θ, indicating that these components are functioning normally.

[0206] Through the above steps and data tables, the experiment can be used to evaluate the operating status of PV panels and identify potential faults or performance degradation. The specific fault content can then be determined through traceability analysis and operation and maintenance tasks can be assigned.

[0207] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A photovoltaic power station operation and maintenance management system and method based on big data, characterized by: include, Collect PV module operating data, pre-process the data using edge computing devices, and store it on a cloud-based big data platform; Based on the stored PV module operation data, a basic model of PV power station power generation behavior is established using self-supervised learning methods; Compare PV module operating data with the basic power generation behavior model to identify abnormal power generation behavior, upload the abnormal power generation behavior to the blockchain, and record it in the distributed ledger; Combined with historical data, abnormal power generation behavior in the distributed ledger is traced and analyzed, and smart contracts are used to trigger operation and maintenance tasks; After the operation and maintenance task is completed, the abnormal power generation behavior is fed back to the self-supervised learning model sample library to dynamically adjust the model; Generate an operation and maintenance report based on the execution content of the operation and maintenance task and upload it to the blockchain; The method of establishing a basic model of photovoltaic power station power generation behavior based on the stored photovoltaic module operation data and using a self-supervised learning method is as follows: Perform multi-channel mapping and data augmentation on standardized data; Use deep convolutional neural network to extract features from enhanced data; Use a two-layer fully connected neural network to map the features to the contrastive learning space and obtain the feature representation z, which is expressed as: ； in, is the output of the feature extraction network, is the weight matrix of the first layer, is the weight matrix of the second layer, is the bias term of the first layer of the fully connected network, is the bias term of the second layer fully connected network; Through the SimCLR contrast learning architecture, the contrast loss function is defined, and the expression is: ； in, To enhance the feature representation of the view, for The corresponding positive sample pairs, is the temperature scaling factor, is the sample size, is the indicator function, is the sample size index coefficient, for The corresponding positive sample index coefficient, is the index coefficient of the negative sample pair, is the cosine similarity; Introducing periodic enhancement factors Corrected cosine similarity, the expression is: ； Periodic enhancement factor , the expression is: ； in, is the current time, is the daily cycle, is the periodic amplitude correction parameter, is the balance factor; Use stochastic gradient descent optimization and dynamically adjust the learning rate; Use KNN clustering method to represent features Perform visualization and verify the separability of the embedded representation by projecting it into two-dimensional space.

2. The photovoltaic power station operation and maintenance management system and method based on big data according to claim 1, characterized in that: The specific steps of collecting photovoltaic module operation data, pre-processing the data using edge computing equipment, and storing it on the cloud big data platform are as follows: Deploy sensor networks in photovoltaic power plants to collect real-time operating data of photovoltaic modules; Through edge computing devices, noise filtering algorithms are used to remove outliers and noise interference, and wavelet transform is used for compression to output standardized operating data; The MQTT protocol is used for data transmission, and the processed standardized data is uploaded to the cloud.

3. The photovoltaic power station operation and maintenance management system and method based on big data according to claim 1, characterized in that: The specific steps of comparing the photovoltaic module operation data with the basic model of power generation behavior, identifying abnormal power generation behavior, uploading the abnormal power generation behavior to the blockchain, and recording it in the distributed ledger are as follows: The real-time collected photovoltaic module operation data , through the feature extraction network, obtain feature representation ; Calculate the anomaly index , the expression is: ； in is the distance between the model prediction and the actual observation, The current time point The resulting feature representation is is the feature learned in the basic model, is a similarity-based metric function, is the balance parameter, is the number of feature vector samples; Contrast Abnormality Index Setting abnormality judgment threshold , identify abnormal power generation behavior; Build a record of abnormal power generation behavior, create and call a smart contract, upload abnormal power generation behavior to the blockchain, and generate a transaction hash value.

4. The photovoltaic power station operation and maintenance management system and method based on big data according to claim 3, characterized in that: The above method combines historical data to trace the abnormal power generation behavior in the distributed ledger and uses smart contracts to trigger operation and maintenance tasks. The specific steps are as follows: Extracting abnormal power generation behavior records from distributed ledgers , and collect the operating parameters of the photovoltaic power station during the period when the abnormal behavior occurs; Defining a spatiotemporal correlation graph ,in is a set of abnormal event nodes, corresponding to each abnormal power generation behavior record , edge set is the spatiotemporal correlation between abnormal events, expressed as: ； in, is the association weight of the node, is the similarity of feature vectors between nodes, is the angle difference of the node in the space-time distribution, is the similarity weight factor, is the similarity adjustment factor, is the similarity threshold, is the geographical distribution correction factor; Building a causal inference network , calculate the causal weight matrix ; Using the causal weight matrix and the traceability target node ,determine the tracing path through the depth-first search algorithm; Design smart contract task allocation rules, the expression is: ； in, To assign priorities, is the association weight of the corresponding node, is the causal weight of the corresponding node, is the time priority function, and its expression is: ； in, is the time impact factor, is the current time, is the event time; After the contract is triggered, operation and maintenance instructions are issued through the IoT device and cloud platform interface.

5. The photovoltaic power station operation and maintenance management system and method based on big data according to claim 4, characterized in that: After the operation and maintenance task is executed, the abnormal power generation behavior is fed back to the self-supervised learning model sample library to dynamically adjust the model. The specific steps are as follows: Extract feedback data after operation and maintenance task processing and abnormal data before processing, and format the data; Extract sample features through feature extraction network, introduce correction vector, map data to contrastive learning space, and obtain corrected samples; Store the corrected samples in the self-supervised learning model library; Based on the updated sample library, the basic model is re-trained through self-supervision; Add a modified sample weight factor to the original contrastive learning loss function and adjust the loss function; The learning rate is dynamically adjusted based on the number of correction samples.

6. The photovoltaic power station operation and maintenance management system and method based on big data according to claim 5, characterized in that: The operation and maintenance report is generated based on the execution content of the operation and maintenance task and uploaded to the blockchain. The specific steps are: Collect operation and maintenance task execution data and generate operation and maintenance reports; Digitally sign the operation and maintenance report using a private key; Call the blockchain interface, upload the operation and maintenance report to the distributed ledger, and obtain the transaction hash value; Store operation and maintenance reports and record the mapping relationship between report files and blockchain transaction hash values ​​in the database.

7. A photovoltaic power station operation and maintenance management system based on big data, based on the photovoltaic power station operation and maintenance management system and method based on big data according to any one of claims 1 to 6, characterized in that: Including data collection module, model building module, anomaly detection module, anomaly tracing module, operation and maintenance feedback module, and report generation module; The data acquisition module is used to collect photovoltaic module operation data, pre-process the data using edge computing devices, and store it on the cloud big data platform; The model building module is used to establish a basic model of photovoltaic power station power generation behavior based on the stored photovoltaic module operation data using a self-supervised learning method; The anomaly detection module is used to compare the photovoltaic module operation data with the basic model of power generation behavior, identify abnormal power generation behavior, upload the abnormal power generation behavior to the blockchain, and record it in the distributed ledger; The abnormality tracing module is used to combine historical data to trace the abnormal power generation behavior in the distributed ledger and trigger operation and maintenance tasks using smart contracts; The operation and maintenance feedback module is used to feed back the processed abnormal power generation behavior to the self-supervised learning model sample library after the operation and maintenance task is executed, and dynamically adjust the model; The report generation module is used to generate an operation and maintenance report based on the execution content of the operation and maintenance task and upload it to the blockchain.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the photovoltaic power station operation and maintenance management system and method based on big data are implemented as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the photovoltaic power station operation and maintenance management system and method based on big data are implemented as described in any one of claims 1 to 6.

Images