Photovoltaic operation and maintenance training method based on digital twinning

Abstract

The invention discloses a photovoltaic operation and maintenance training method based on digital twinning, and the method comprises the steps: extracting a standardized operation process, generating a standardized operation instruction set corresponding to a fault scene, and guaranteeing that an operation instruction is consistent with an actual working condition; embedding the standardized operation instruction set into a high-reality scene, generating an interactive training module, simulating a real operation environment, and outputting training scene data; analyzing operation data of trainees in the interactive training module, optimizing an operation process, and generating a standardized skill migration training scheme; extracting operation feedback data from the skill migration training scheme, adjusting virtual scene parameters in combination with a unified training standard, and generating an optimized training scene data set; and through the optimized training scene data set, updating the interactive training module, circularly executing fault point identification and operation process training, and outputting a training result. The skill level of operation and maintenance personnel and the operation reliability of the power station are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of power training, and particularly relates to a photovoltaic operation and maintenance training method based on digital twinning. BACKGROUND

[0002] As a core pillar of clean energy, the safe and stable operation and efficient power generation of photovoltaic power stations are crucial for promoting energy transformation and sustainable development. With the expansion of photovoltaic power station scale and diversification of types, the skill level of operation and maintenance personnel directly affects the power generation efficiency, equipment life and operating cost of the power station. However, the traditional operation and maintenance training method has significant limitations and cannot meet the needs of the rapid development of the industry.

[0003] The current training method highly depends on real equipment and sites, and is limited by geographical location, weather conditions and equipment availability, making it difficult to achieve large-scale and high-frequency training. In actual operation, students may cause safety accidents or equipment damage due to misoperation, resulting in high training cost and high risk. In addition, the simulation of complex fault scenarios is limited, and rare or high-risk faults are difficult to reproduce, making it difficult for students to fully master the diagnosis and handling capabilities in diversified scenarios. The experience of excellent engineers is also difficult to systematize, and the training effect varies with the level of the instructor, lacking a unified standard.

[0004] In photovoltaic operation and maintenance training, the core challenge is how to achieve high-fidelity scene simulation and standardized operation process integration. First, the operation state of a photovoltaic power station is affected by various dynamic factors such as weather, equipment aging, etc., and the training system needs to accurately reproduce the behavior of the equipment under these complex working conditions. For example, a certain power station has a power drop due to string mismatch, and students need to accurately identify the fault point and perform standard operations under different light and temperature conditions, but existing methods are difficult to dynamically simulate such scenarios. Second, the construction of high-fidelity scenes needs to be combined with standardized operation processes to ensure that the operation habits learned by students in the virtual environment can be seamlessly applied to actual work. Without this combination, training may result in non-standard operations or difficulty in transferring skills to real scenarios.

[0005] Therefore, how to build a training system that can accurately simulate the dynamic operation state of a photovoltaic power station while integrating standardized operation processes has become a key issue to improve the skill level of operation and maintenance personnel and the overall efficiency of the industry. SUMMARY

[0006] To solve the above technical problems, the present application proposes a photovoltaic operation and maintenance training method based on digital twinning, which realizes the efficiency of photovoltaic power station operation and maintenance and the standardization of training effect, and significantly improves the skill level of operation and maintenance personnel and the reliability of power station operation.

[0007] To achieve the above purpose, the present application provides a photovoltaic operation and maintenance training method based on digital twinning, comprising:

[0008] acquire historical operation data of the photovoltaic power station, extract dynamic operation state parameters, and construct a virtual scene dataset; a realistic scene is constructed based on the virtual scene dataset, operation behaviors of equipment of the photovoltaic power station are simulated, and a dynamic equipment behavior dataset is constructed;

[0009] equipment operation features are extracted according to the dynamic equipment behavior dataset, a diversified fault scene simulation environment is generated in combination with a pre-established fault scene library, a fault point identification model is trained according to the diversified fault scene simulation environment, and a fault diagnosis result is obtained;

[0010] standardized operation processes are extracted according to the fault diagnosis result, a standardized operation instruction set corresponding to the fault scene is generated, the standardized operation instruction set is embedded in the realistic scene, a real operation environment is simulated, a training scene dataset is obtained, operation data of a trainee in an interactive training module is analyzed, an operation process is optimized, and a skill transfer training scheme is generated;

[0011] operation feedback data is extracted according to the skill transfer training scheme, virtual scene parameters are adjusted, the training scene dataset is optimized, and an optimized training scene dataset is generated; the interactive training module is updated through the optimized training scene dataset, fault point identification and operation process training are cyclically executed, and a training result is obtained.

[0012] The technical effect of the present application is that the present application discloses a photovoltaic operation and maintenance training method based on digital twinning, which aims at the business problems that equipment faults are difficult to accurately identify and operation processes are difficult to standardize under complex working conditions such as light intensity, temperature change and string mismatch in the operation of a photovoltaic power station, constructs a virtual scene dataset, simulates a real operation environment, generates a high-realistic scene to extract dynamic behavior features of equipment, trains a machine learning model in combination with a fault scene library, accurately identifies the position and type of a fault point, generates a standardized operation instruction set, and embeds an interactive training module through virtual reality technology, optimizes an operation process in combination with reinforcement learning, and finally generates a standardized skill transfer training scheme. The present application cyclically optimizes a training scene dataset, continuously improves fault diagnosis accuracy and operation process standardization, realizes the efficiency of operation and maintenance of a photovoltaic power station and the standardization of training effects, and significantly improves the skill level of operation and maintenance personnel and the operation reliability of a power station. BRIEF DESCRIPTION OF DRAWINGS

[0013] The accompanying drawings, which form a part of the present application, are intended to provide further understanding of the present application, and the illustrative embodiments of the present application and their description serve the purpose of explaining the present application. The accompanying drawings do not constitute an inappropriate limitation on the present application. In the accompanying drawings:

[0014] Figure 1 FIG. 1 is a flowchart of a photovoltaic operation and maintenance training method based on digital twinning according to an embodiment of the present application. DETAILED DESCRIPTION

[0015] It should be noted that the embodiments and features in the present application can be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0016] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown herein.

[0017] As shown in the embodiment, a photovoltaic operation and maintenance training method based on digital twinning is provided, comprising: Figure 1

[0018] Obtaining historical operation data of a photovoltaic power station, extracting dynamic operation state parameters, and constructing a virtual scene data set; constructing a realistic scene based on the virtual scene data set, simulating the operation behavior of the photovoltaic power station equipment, and constructing a dynamic equipment behavior data set;

[0019] Extracting equipment operation features according to the dynamic equipment behavior data set, combining a pre-established fault scene library, and generating a diversified fault scene simulation environment; training a fault point identification model according to the diversified fault scene simulation environment to obtain a fault diagnosis result;

[0020] Extracting a standardized operation process according to the fault diagnosis result, generating a standardized operation instruction set corresponding to the fault scene; embedding the standardized operation instruction set into the realistic scene to simulate a real operation environment, and obtaining a training scene data set; analyzing operation data of a trainee in an interactive training module, optimizing the operation process, and generating a skill transfer training scheme;

[0021] Extracting operation feedback data according to the skill transfer training scheme, adjusting virtual scene parameters, optimizing the training scene data set, and generating an optimized training scene data set; updating the interactive training module through the optimized training scene data set, cyclically executing fault point identification and operation process training, and obtaining a training result.

[0022] Further, constructing a virtual scene data set comprises:

[0023] Based on the historical operation data of the photovoltaic power station, time series data is extracted, abnormal values and missing values are removed, and a cleaned time series data set is obtained;

[0024] According to the cleaned time series data set, the daily variation rate of the light intensity and the environmental temperature and the deviation value of the string mismatch are calculated to obtain the dynamic operation state parameters; ​

[0025] According to the dynamic operating state parameters, the light intensity, the ambient temperature and the string mismatch are classified to obtain an operating state set under multiple working conditions;

[0026] According to the operating state set, data synthesis is performed to obtain a virtual scene data set.

[0027] Specifically, the virtual scene data set is generated based on the operating state set, and a data synthesis method can be used. For high fluctuation working conditions, the light intensity sequence is synthesized from 800 W / m 2 to 1100 W / m 2 , the temperature is raised from 25℃ to 32℃, and the string mismatch deviation is kept at about 0.5A. The synthesized data can be generated by the statistical distribution of historical data to ensure that the virtual scene is close to the actual operating characteristics. This data set can be used to simulate the performance of the power station under complex working conditions, optimize the control strategy, and reduce the operating risk. The implementation of the above method can significantly improve the accuracy and prediction ability of the operation analysis of the photovoltaic power station. The cleaning data reduces the abnormal interference, the feature extraction quantifies the dynamic characteristics, the cluster analysis clearly divides the working conditions, and the virtual scene data set provides a reliable basis for optimization. These steps support each other to form a complete technical chain to help the efficient and stable operation of the power station.

[0028] Further, the dynamic device behavior data set includes:

[0029] According to the virtual scene data set, the dynamic changes of the light intensity and the ambient temperature are simulated to obtain a virtual scene parameter set containing a dynamic change rate;

[0030] The light intensity, the ambient temperature and the string mismatch of the virtual scene parameter set containing the dynamic change rate are classified to generate an operating state set under multiple working conditions, and an operating state data set after classification is obtained;

[0031] Based on the operating state data set after classification, the operating state set is subjected to three-dimensional visualization processing to construct a dynamic device behavior data set.

[0032] Specifically, a three-dimensional modeling tool is used to generate a virtual scene containing light intensity and ambient temperature. Based on time series data, a three-dimensional model of the photovoltaic power station is constructed to simulate the influence of light intensity change on the irradiation angle of the components, such as the low morning light angle, the shadow covering part of the components, and the noon light close to vertical with the highest coverage. The ambient temperature affects the surface temperature of the components, and then affects the power generation efficiency. The initial virtual scene data set can present the light from 150 W / m 2 to 950 W / m 2The gradual change of the irradiance and the change of the temperature from 18℃ to 33℃ form a dynamic power plant operation scenario. Based on the classified operation state dataset, a three-dimensional visualization is processed by virtual reality rendering tools such as Unity. The generated dynamic device behavior simulation of the photovoltaic power plant can show how the change of the irradiance affects the component light-receiving area, how the change of the temperature affects the component surface heat distribution, and how the mismatch of the module string leads to the fluctuation of the power output. For example, under the condition of high fluctuation, the component surface is red due to the temperature rise, and the power output decreases by 10%. The dynamic device behavior dataset intuitively presents the operation state of the power plant, which is convenient for analyzing the device response characteristics and improving the optimization capability.

[0033] Further, obtaining the fault diagnosis result comprises:

[0034] Obtaining time series data containing irradiance, ambient temperature and power output, cleaning and formatting the time series data to obtain a standardized operation data set;

[0035] Extracting operation features of irradiance and ambient temperature for the standardized operation data set to obtain a feature dataset containing operation features;

[0036] Extracting corresponding fault scenario data from a pre-established fault scenario library, associating the feature dataset with the fault scenario data to obtain an associated fault feature dataset;

[0037] Training a fault point recognition model for the associated fault feature dataset to judge the fault point position and fault type to obtain the fault diagnosis result.

[0038] Specifically, when obtaining time series data from the photovoltaic power plant historical operation database, key operation data can be extracted from the light sensor, temperature sensor and inverter through the data acquisition system. The irradiance data is recorded by the light power meter, for example, the irradiance is 300W / m 2 at 10:00 and 800W / m 2 at 16:00. The ambient temperature is collected by the temperature sensor, for example, 22℃ at 10:00 and 33℃ at 16:00. The power output is recorded by the inverter, for example, 420kW at a certain time. These data lay the foundation for subsequent analysis. Data cleaning can use pre-processing tools to process abnormal or missing values. For example, abnormal records of irradiance such as negative values or more than 2000W / m 2data; for the missing temperature data, it can be filled by linear interpolation method, such as 28℃ missing at a certain time can be calculated according to the data before and after. This cleaning method ensures the integrity and consistency of the data, which helps the accuracy of subsequent feature extraction. Based on the associated fault feature data set, support vector machine algorithm can be used to train the fault point recognition model. Support vector machine distinguishes normal and abnormal states by classification hyperplane. For example, the input feature data set contains light intensity 800W / m 2 , temperature 33℃, mismatch feature value 0.3, the model judges that the fault point is located in the junction box of a certain group of strings, and the fault type is poor contact. This classification method can quickly lock the fault location and type, provide clear guidance for maintenance, effectively shorten the troubleshooting time, and improve the stability of power station operation.

[0039] Further, the standardized operation instruction set corresponding to the fault scene is generated, including:

[0040] According to the fault diagnosis result, data containing fault point position and fault type are obtained, and the fault point position and fault type are classified and arranged to obtain a standardized fault diagnosis data set;

[0041] The key features of the fault point position and fault type are extracted from the standardized fault diagnosis data set to obtain an operation feature data set;

[0042] According to the corresponding operation process data obtained from the pre-established operation process library, the operation feature data set is associated with the operation process data to obtain an associated operation process data set;

[0043] The operation process data is formatted to generate a standardized operation instruction set corresponding to the fault scene.

[0044] Specifically, when obtaining the data of fault point position and fault type from the fault diagnosis result of the photovoltaic power station, the data can be classified and arranged through a data preprocessing tool. Assuming that the fault data at 10:00 on a certain day is extracted from the historical diagnosis record, it is shown that the fault point is located at the junction box of a certain group of strings, and the fault type is poor contact; the fault data at 16:00 points to the inverter, and the type is overheating. The data preprocessing tool will classify these data according to the fault point position such as group string, inverter and fault type such as poor contact, overheating, and generate a standardized fault diagnosis data set. This classification method facilitates subsequent feature extraction and ensures clear data structure. For different fault scenarios, multiple operation process matching methods can be extended. Assuming that a component aging caused power drop occurs, the operation weight feature is 0.65, and the process of "replacing photovoltaic components" is matched. Compared with the junction box inspection, this requires more complex operation steps, such as disassembling old components, installing new components, and testing output power. The combination of multiple matching methods makes the operation instruction set cover multiple fault scenarios and ensures the comprehensiveness of the processing scheme. This multi-angle process generation method supports flexible response to complex working conditions through data-driven support.

[0045] Further, obtaining the training scene data set comprises:

[0046] According to the environment parameter data obtained from the pre-established real environment database, the environment parameter data is processed by three-dimensional reconstruction to generate a virtual scene data set;

[0047] According to the operation instruction data obtained from the standardized operation instruction set, the operation instruction data is associated with the virtual scene data set for processing to obtain instruction embedding data set;

[0048] The instruction embedding data set is formatted and processed, and an interactive module data set is generated through user interaction parameter configuration to obtain an interactive training module;

[0049] The interactive training module is rendered to generate a training scene data set.

[0050] Specifically, the environment parameter data is obtained from the pre-established real environment database, which usually involves the actual operating environment of the photovoltaic power station. For example, the environment parameter data can include key indicators such as light intensity, environmental temperature, humidity, and wind speed. Specifically, assuming that the database of a certain photovoltaic power station records that the light intensity at 14:00 on a certain day is 800 W / m 2The environmental parameters are collected in real time by sensors and stored in a database, providing a basis for subsequent scenario modeling. It is worth noting that the accuracy of environmental parameter data directly affects the realism of the virtual scene, so data collection must ensure high precision. For example, using a scene modeling tool to perform three-dimensional reconstruction processing on the environmental parameter data can generate a virtual scene dataset. The scene modeling tool simulates the actual layout of the photovoltaic power station by mapping parameters such as light intensity and temperature to a three-dimensional space. Assuming that the photovoltaic power station contains 10 component groups and 2 inverters, the modeling tool generates a virtual scene containing components, inverters, and the surrounding environment based on the location data and environmental parameters in the database. The virtual scene dataset not only reflects the physical layout but also simulates the actual impact of light on the components, such as a group string being unevenly illuminated due to shadows. This virtual scene provides an intuitive operating environment for the embedding of subsequent operation instructions. In one possible implementation, after the virtual scene dataset is successfully generated, operation instruction data is obtained from the standardized operation instruction set.

[0051] Further, generating the skill transfer training scheme includes:

[0052] obtaining trainee operation data according to the training scene dataset, and performing structured processing on the trainee operation data to obtain a structured operation dataset;

[0053] extracting trainee behavior patterns according to the structured operation dataset, and performing classification processing on the behavior patterns to obtain a classified behavior dataset;

[0054] performing reconstruction processing on the classified behavior dataset to generate an optimized process dataset, and obtaining the optimized process dataset;

[0055] performing format processing on the optimized process dataset to generate the skill transfer training scheme.

[0056] Specifically, for the structured operation dataset, the data analysis tool extracts the behavior patterns of the trainees. Assuming that the preset standardized operation threshold requires the trainee to complete the shield check within 60 seconds, if a trainee takes 90 seconds, the system identifies that the behavior deviates from the threshold. The clustering algorithm further classifies the behavior patterns to generate a classified behavior dataset. For example, through the K-means algorithm, the trainees are divided into an "efficient operation group", a "delayed operation group", and an "error operation group". Specifically, the trainees in the delayed operation group may be slower in response due to unfamiliarity with the virtual interface, and the trainees in the error operation group may mistakenly select irrelevant components. These classifications provide the basis for personalized training. For example, based on the classified behavior dataset, the process optimization tool reconstructs the operation process to generate an optimized process dataset. For the delayed operation group, the tool may suggest shortening the interface guidance time, such as optimizing the prompt animation from 10 seconds to 5 seconds. For the error operation group, the tool may increase the error prompt function, such as popping up a "please reselect" guide when the component is mistakenly selected. These optimized processes improve training efficiency in a data-driven manner.

[0057] Further, the generation of the optimized training scene dataset includes:

[0058] According to the skill transfer training scheme, the trainee operation feedback data is obtained, and the structured feedback dataset is obtained by structuring the trainee operation feedback data;

[0059] The structured feedback dataset is grouped and processed to generate a classified feedback dataset, and the behavior patterns in the classified feedback dataset are determined;

[0060] The classified feedback dataset is compared with the training standard, and the behavior pattern deviates from the preset standardized operation threshold to obtain a deviation behavior dataset;

[0061] Based on the deviation behavior dataset, the virtual scene parameters are adjusted and processed to generate an optimized scene parameter dataset, and the optimized scene parameter dataset is formatted to generate an optimized training scene dataset.

[0062] Specifically, in the photovoltaic power station virtual reality training scene, the acquisition and processing of operation feedback data of trainees are important links for optimizing the training scheme. Operation feedback data refers to feedback information recorded by the system after the trainees perform tasks in the virtual scene, such as task completion quality, operation accuracy, and reaction speed. In the task of "checking the connection line of photovoltaic components", the system records whether the trainee correctly identifies the line fault, selects the appropriate repair tool, and the time length of completing the task. These data are structured by a data collection tool to generate a structured feedback data set. Based on the deviation behavior data set, a parameter optimization tool adjusts the virtual scene parameters to generate an optimized scene parameter data set. For the group that needs to be improved, the system may increase the frequency of interface prompts, such as popping up the guidance "suggest selecting voltage tester" when the trainee operates for more than 10 seconds without selecting the correct tool. The optimized scene parameter data set includes adjusted parameters such as "prompt interval: 10 seconds" and "animation duration: 3 seconds". Through a data set formatting tool, these parameters are formatted into an optimized training scene data set.

[0063] Further, obtaining the training result includes:

[0064] According to the optimized training scene data set, structured operation feedback data is obtained, the structured operation feedback data is classified and processed to generate a classified operation data set, operation behaviors are grouped to obtain a behavior pattern data set;

[0065] According to the behavior pattern data set, a standard threshold comparison tool is used to compare with a preset standard operation threshold. If the behavior pattern deviates from the preset threshold, a deviation behavior data set is generated, and the deviation behavior data set is filtered to obtain a fault point data set;

[0066] Combined with the fault point data set, the virtual scene parameters of the interactive training module are dynamically updated to generate an updated scene parameter data set, and the updated scene parameter data set is subjected to consistency verification to obtain a verified scene parameter data set;

[0067] According to the verified scene parameter data set, an operation flow training is performed in the interactive training module to generate a cycle training result data set, and the cycle training result data set is subjected to structured processing to obtain a training result.

[0068] Specifically, the K-means clustering algorithm is used to divide the students into "efficient group", "standard group" and "inefficient group" according to the time consumption and accuracy. The students in the efficient group may complete the configuration within 40 seconds without errors, the students in the standard group consume 50 seconds and have 1 parameter input deviation, and the students in the inefficient group consume 70 seconds and have multiple errors. The behavior pattern dataset reveals the operation habits of the students, such as the inefficient group may repeatedly adjust due to unfamiliarity with the parameter input interface. Based on the verified scene parameter dataset, the cycle training tool performs operation process training in the interactive training module to generate a cycle training result dataset. The students repeatedly practice in the updated scene, and the system records the time consumption and accuracy of each training, such as "student ID: B002, second training time consumption: 48 seconds, accuracy: 95%". Through the formatting tool, these data are structured and processed to generate a final training result dataset, including "task completion rate: 90%" and "average time consumption: 45 seconds". This data-driven optimization method ensures that the training scene is highly targeted and gradually improves the operation proficiency of the students.

[0069] The application discloses a photovoltaic operation and maintenance training method based on digital twinning, aiming at the business problems that equipment faults are difficult to accurately identify and operation processes are difficult to standardize under complex working conditions such as light intensity, temperature change and string mismatch in the operation of a photovoltaic power station, a virtual scene dataset is constructed to simulate a real operation environment, a high-realism scene is generated to extract dynamic behavior characteristics of equipment, a machine learning model is trained in combination with a fault scene library to accurately identify fault point positions and types, a standardized operation instruction set is generated, an interactive training module is embedded through virtual reality technology, an operation process is optimized in combination with reinforcement learning, and finally a standardized skill transfer training scheme is generated. The application continuously improves fault diagnosis accuracy and operation process standardization by cyclically optimizing a training scene dataset, realizes the efficiency of photovoltaic power station operation and maintenance and the standardization of training effects, and significantly improves the skill level of operation and maintenance personnel and the operation reliability of the power station.

[0070] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A photovoltaic operation and maintenance training method based on digital twinning, characterized in that, The method comprises the following steps: acquiring historical operation data of a photovoltaic power station, extracting dynamic operation state parameters, and constructing a virtual scene data set; constructing a realistic scene based on the virtual scene data set, simulating the operation behavior of the photovoltaic power station equipment, and constructing a dynamic equipment behavior data set; extracting equipment operation features from the dynamic equipment behavior data set, combining a pre-established fault scene library to generate a diversified fault scene simulation environment, training a fault point identification model according to the diversified fault scene simulation environment, and obtaining a fault diagnosis result; extracting a standardized operation process according to the fault diagnosis result, generating a standardized operation instruction set corresponding to the fault scene, embedding the standardized operation instruction set into the realistic scene, simulating a real operation environment, and obtaining a training scene data set; analyzing the operation data of the trainees in the interactive training module, optimizing the operation process, and generating a skill transfer training scheme; extracting operation feedback data according to the skill transfer training scheme, adjusting the virtual scene parameters, optimizing the training scene data set, generating an optimized training scene data set, updating the interactive training module through the optimized training scene data set, and cyclically executing fault point identification and operation process training to obtain a training result.

2. The photovoltaic operation and maintenance training method based on digital twinning according to claim 1, characterized in that, The construction of the virtual scene data set comprises: extracting time series data based on the historical operation data of the photovoltaic power station, removing abnormal values and missing values, and obtaining a cleaned time series data set; calculating the daily variation rates of the light intensity and the environmental temperature and the deviation values of the string mismatch according to the cleaned time series data set, and obtaining dynamic operation state parameters; classifying the light intensity, the environmental temperature and the string mismatch according to the dynamic operation state parameters, and obtaining a running state set under multiple working conditions; synthesizing data according to the running state set, and obtaining a virtual scene data set.

3. The photovoltaic operation and maintenance training method based on digital twinning according to claim 1, characterized in that, The construction of the dynamic equipment behavior data set comprises: simulating the dynamic changes of the light intensity and the environmental temperature according to the virtual scene data set, and obtaining a virtual scene parameter set containing dynamic variation rates; classifying the light intensity, the environmental temperature and the string mismatch of the virtual scene parameter set containing dynamic variation rates, generating a running state set under multiple working conditions, and obtaining a classified running state data set; based on the classified running state data set, performing three-dimensional visualization processing on the running state set, and constructing a dynamic equipment behavior data set.

4. The photovoltaic operation and maintenance training method based on digital twinning of claim 1, wherein, The obtaining of the fault diagnosis result comprises: acquiring time series data containing light intensity, environmental temperature and power output, cleaning and formatting the time series data, and obtaining a standardized running data set; extracting the running features of the light intensity and the environmental temperature for the standardized running data set, and obtaining a feature data set containing running features; extracting corresponding fault scene data from a pre-established fault scene library, associating the feature data set with the fault scene data, and obtaining an associated fault feature data set; training a fault point identification model for the associated fault feature data set, judging the fault point position and the fault type, and obtaining a fault diagnosis result.

5. The photovoltaic operation and maintenance training method based on digital twinning according to claim 1, characterized in that, The generation of the standardized operation instruction set corresponding to the fault scene comprises: According to the fault diagnosis result, data containing fault point position and fault type are obtained, the fault point position and fault type are classified and arranged, and a standardized fault diagnosis data set is obtained; According to the standardized fault diagnosis data set, key features of the fault point position and fault type are extracted, and an operation feature data set is obtained; According to the corresponding operation process data obtained from the pre-established operation process library, the operation feature data set is associated with the operation process data, and an associated operation process data set is obtained; The operation process data is formatted, and a standardized operation instruction set corresponding to the fault scene is generated.

6. The photovoltaic operation and maintenance training method based on digital twinning according to claim 1, characterized in that, Obtaining a training scene data set includes: According to the environment parameter data obtained from the pre-established real environment database, the environment parameter data is processed by three-dimensional reconstruction, and a virtual scene data set is generated; According to the operation instruction data in the standardized operation instruction set, the operation instruction data is associated with the virtual scene data set, and an instruction embedding data set is obtained; The instruction embedding data set is formatted, an interactive module data set is generated through user interaction parameter configuration, and an interactive training module is obtained; The interactive training module is rendered to generate a training scene data set.

7. The photovoltaic operation and maintenance training method based on digital twinning according to claim 1, characterized in that, Generating a skill transfer training scheme includes: According to the training scene data set, student operation data is obtained, and the student operation data is processed by structuring to obtain a structured operation data set; According to the structured operation data set, student behavior patterns are extracted, and the behavior patterns are classified to obtain a classified behavior data set; The classified behavior data set is reconstructed to generate an optimized process data set, and an optimized process data set is obtained; The optimized process data set is formatted to generate a skill transfer training scheme.

8. The photovoltaic operation and maintenance training method based on digital twinning according to claim 1, characterized in that, Generating an optimized training scene data set includes: According to the skill transfer training scheme, student operation feedback data is obtained, and the student operation feedback data is processed by structuring to obtain a structured feedback data set; The structured feedback data set is grouped to generate a classified feedback data set, and the behavior patterns in the classified feedback data set are determined; The classified feedback data set is compared with the training standard, and the behavior patterns deviate from the pre-set standardized operation threshold value to obtain a deviation behavior data set; Based on the deviation behavior data set, the virtual scene parameters are adjusted to generate an optimized scene parameter data set, and the optimized scene parameter data set is formatted to generate an optimized training scene data set. 9.The digital-twin-based photovoltaic operation and maintenance training method of claim 1, wherein, Obtaining a training result includes: According to the optimized training scene data set, structured operation feedback data is obtained, and the structured operation feedback data is classified to generate a classified operation data set, and the operation behavior is grouped to obtain a behavior pattern data set; According to the behavior pattern data set, a standard threshold comparison tool is used to compare with a pre-set standard operation threshold value, and if the behavior pattern deviates from the pre-set threshold value, a deviation behavior data set is generated, and the deviation behavior data set is filtered to obtain a fault point data set; The virtual scene parameters of the interactive training module are dynamically updated in combination with the fault point data set, an updated scene parameter data set is generated, consistency check is performed on the updated scene parameter data set, and a checked scene parameter data set is obtained; According to the checked scene parameter data set, operation process training is performed in the interactive training module, a cycle training result data set is generated, structured processing is performed on the cycle training result data set, and a training result is obtained.

Images