A photovoltaic power station intelligent operation and maintenance system and method

Abstract

The application provides a photovoltaic power station intelligent operation and maintenance system and method, relates to the technical field of intelligent operation and maintenance, and comprises the following steps: collecting basic geographic and building data, equipment full life cycle data, operation monitoring data and environment data, and performing pretreatment based on a density clustering algorithm and Z-score standardization; performing data fusion on the pretreated basic geographic and building data and equipment full life cycle data based on multi-source heterogeneous data fusion, obtaining a multi-source data package, and then constructing a photovoltaic power station virtual model; performing dynamic simulation according to the photovoltaic power station virtual model and the pretreated operation monitoring data, constructing a photovoltaic power station digital twin model based on a digital twin model, and then obtaining twin equipment operation data; constructing a photovoltaic power station digital twin fault prediction model, and obtaining multi-dimensional fault precursor parameters; generating a fault heat map of the photovoltaic power station digital twin model according to the fault precursor parameters, and intelligently operating and maintaining the photovoltaic power station; and the operation and maintenance efficiency is greatly improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance technology, specifically to an intelligent operation and maintenance system and method for photovoltaic power plants. Background Technology

[0002] With the increasing global demand for clean energy, photovoltaic power plants, as an important renewable energy generation method, have been widely used and developed. However, the efficient and stable operation of photovoltaic power plants faces many challenges, placing enormous pressure on operation and maintenance management.

[0003] In terms of data processing, photovoltaic power plants generate massive amounts of multi-source heterogeneous data during operation, including basic geographic and building data, equipment lifecycle data, operational monitoring data, and environmental data. These data have different formats, sources, and dimensions, and their quality varies greatly, containing issues such as noise, outliers, and missing values. Traditional data processing methods struggle to effectively integrate and analyze this complex data, preventing it from fully realizing its value and failing to provide accurate and reliable support for photovoltaic power plant model building and operational decisions. For example, when building a photovoltaic power plant model, inaccurate or incomplete data may prevent the model from accurately reflecting the actual operating conditions of the power plant, thus affecting the evaluation and optimization of power plant performance.

[0004] In terms of fault prediction and operation and maintenance management, the current operation and maintenance of photovoltaic power plants mainly relies on regular inspections and reactive repairs after a fault occurs. This approach suffers from low efficiency, high costs, and untimely fault response. Because photovoltaic power plants have numerous and widely distributed devices, manual inspections are insufficient to comprehensively and promptly identify potential faults, often leading to the escalation of equipment failures and significant power generation losses. Furthermore, traditional fault diagnosis methods are primarily based on experience and simple data analysis, lacking real-time monitoring and in-depth analysis of equipment operating status. This makes it impossible to predict equipment failures in advance and hinders the achievement of intelligent and efficient operation and maintenance for photovoltaic power plants.

[0005] In recent years, with the continuous development of information technology, digital twin technology has been gradually applied to various fields, providing new ideas and methods for solving the operation and maintenance management problems of photovoltaic power plants. Digital twin technology achieves real-time monitoring, simulation, and optimization of the physical system by constructing a virtual model corresponding to the actual physical system and collecting and analyzing data from the actual system in real time. However, in the field of photovoltaic power plants, the application of digital twin technology still faces challenges in data processing and model building, requiring further research and improvement.

[0006] Therefore, in order to improve the operation and maintenance efficiency of photovoltaic power plants, reduce operation and maintenance costs, and ensure the stable operation and efficient power generation of the power plants, a smart operation and maintenance system and method for photovoltaic power plants is provided. Summary of the Invention

[0007] To address the aforementioned technical problems, the present invention aims to provide a smart operation and maintenance system and method for photovoltaic power plants.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a smart operation and maintenance method for photovoltaic power plants, the method comprising:

[0009] Collect basic geographic and building data, equipment life cycle data, operation monitoring data and environmental data of photovoltaic power plants, and preprocess them based on density clustering algorithm and Z-score standardization;

[0010] Based on multi-source heterogeneous data fusion technology, preprocessed basic geographic and building data as well as equipment life cycle data are fused to obtain multi-source data packets. Based on the multi-source data packets, a virtual model of a photovoltaic power station is constructed.

[0011] Dynamic simulation is performed based on the virtual model of the photovoltaic power station and pre-processed operation monitoring data, and a digital twin model of the photovoltaic power station is constructed based on the digital twin model; the operation data of the twin equipment is obtained based on the digital twin model of the photovoltaic power station.

[0012] Based on the operating data of the twin equipment and a neural network model, a digital twin fault prediction model for photovoltaic power plants is constructed to obtain multi-dimensional fault precursor parameters. Based on the multi-dimensional fault precursor parameters, a fault heat map of the digital twin model of the photovoltaic power plant is generated. Based on the fault heat map, intelligent operation and maintenance management of the photovoltaic power plant is carried out.

[0013] According to a preferred embodiment of the present invention, the process of collecting basic geographical and building data, equipment lifecycle data, operation monitoring data, and environmental data of a photovoltaic power station includes:

[0014] A data acquisition device is set up, which includes a geographic and building data acquisition unit, an equipment life cycle data acquisition unit, an operation monitoring data acquisition unit, and an environmental data acquisition unit; and a collection cycle is set, which includes several collection moments.

[0015] The geographic and building data acquisition unit collects geographic and building data of the photovoltaic power station based on GIS technology; the equipment life cycle data acquisition unit collects equipment life cycle data of the corresponding equipment in the photovoltaic power station; the operation monitoring data acquisition unit is used to collect operation monitoring data of the corresponding equipment in the photovoltaic power station; and the environmental data acquisition unit is used to collect environmental data of the photovoltaic power station.

[0016] According to a preferred embodiment of the present invention, the process of preprocessing basic geographic and building data, equipment lifecycle data, operation monitoring data, and environmental data includes:

[0017] Data cleaning is performed on basic geographic and building data, equipment lifecycle data, operation monitoring data, and environmental data, and logical noise judgment rules are set.

[0018] Based on the aforementioned logical noise judgment rules, logical noise reduction processing is performed on basic geographic and building data, equipment life cycle data, operation monitoring data, and environmental data to remove data that does not conform to the logical noise judgment rules for the corresponding data types.

[0019] Based on Z-score standardization, the basic geographic and building data, equipment life cycle data, operation monitoring data and environmental data after logical denoising are standardized.

[0020] Statistical noise reduction is performed on the standardized data based on density clustering algorithms, including:

[0021] Based on standardized basic geographic and building data, equipment life cycle data, operation monitoring data, and environmental data, set the neighborhood radius and minimum number of core points for the corresponding data.

[0022] Based on the neighborhood radius and the minimum number of core points, core points, boundary points, and noise points are generated;

[0023] Noise points corresponding to the standardized basic geographic and building data, equipment life cycle data, operation monitoring data, and environmental data are removed to obtain the corresponding statistically noise-processed basic geographic and building data, equipment life cycle data, operation monitoring data, and environmental data.

[0024] Non-numerical data of basic geographic and building data and equipment life cycle data, as well as basic geographic and building data and equipment life cycle data after statistical noise processing, are referred to as preprocessed basic geographic and building data and preprocessed equipment life cycle data.

[0025] The corresponding statistical noise-processed operation monitoring data and environmental data are denoted as preprocessed operation monitoring data and preprocessed environmental data.

[0026] According to a preferred embodiment of the present invention, the process of fusing preprocessed basic geographic and building data and equipment lifecycle data to obtain multi-source data packets includes:

[0027] The non-numerical and numerical data in the preprocessed basic geographic and building data and the preprocessed equipment lifecycle data are separated to obtain the corresponding non-numerical and numerical data.

[0028] For the non-numerical data in the preprocessed basic geographic and building data and the preprocessed equipment lifecycle data, based on natural language processing technology, the non-numerical data in the preprocessed basic geographic and building data and the preprocessed equipment lifecycle data are classified, including classification fields, text description fields, spatial direction fields and level status fields.

[0029] The non-numerical data in the preprocessed basic geographic and building data and the preprocessed equipment lifecycle data are used to generate non-numerical data packages based on the classification field, text description field, spatial direction field and level status field.

[0030] A numerical data fusion adversarial network is constructed, and numerical data packets are generated based on the numerical data fusion adversarial network; the non-numerical data packets and the numerical data packets are combined into multi-source data packets.

[0031] According to a preferred embodiment of the present invention, the process of constructing a numerical data fusion adversarial network includes:

[0032] The numerical data in the preprocessed basic geographic and building data and the preprocessing equipment lifecycle data are respectively denoted as: Obtain multi-source data samples ;

[0033] Configure a numerical data fusion generator The numerical data fusion generator The input is random noise Numerical data in preprocessed basic geographic and building data and preprocessed equipment lifecycle data are respectively denoted as That is, input ;

[0034] Based on neural network convolution, the input Perform a transformation to output a sample with the same dimensions as the real data. ;

[0035] Configure a numerical data fusion discriminant The numerical data fusion discriminator The input is a sample and multi-source data samples Numerical data fusion discriminator The output represents the input sample. It is the probability of real data, ranging from between;

[0036] The output numerical data fusion discriminator The output is: ; ;

[0037] Define a numerical data fusion generator Generator loss function The generator loss function for:

[0038] ;in, Expressing expectations, This indicates the input to the numerical data fusion generator. Distribution;

[0039] Define a numerical data fusion discriminator Discriminator loss function The discriminator loss function for: ;

[0040] in, Expressing expectations, Represents multi-source data samples The distribution, This indicates the input to the numerical data fusion generator. Distribution;

[0041] The process of training a numerical data fusion adversarial network includes:

[0042] Fixed numerical data fusion generator According to the numerical data fusion discriminator Discriminator loss function Calculate the discriminator gradient This leads to an update of the numerical data fusion discriminator. parameters The expression is:

[0043] ;in, Numerical data fusion discriminator The learning rate; Represents the discriminator loss function For parameters Partial derivatives; discriminator gradient ;

[0044] Fixed numerical data fusion discriminator According to the numerical data fusion generator Generator loss function Calculate generator gradient This leads to an update to the numerical data fusion generator. parameters The expression is:

[0045] ;in, Numerical data fusion generator The learning rate; Represents the discriminator loss function For parameters Partial derivatives; generator gradient .

[0046] According to a preferred embodiment of the present invention, the process of constructing a virtual model of a photovoltaic power station based on multi-source data packets includes:

[0047] Based on multi-source data packets, obtain the plan view, elevation view, section view of the photovoltaic power station at the corresponding geographical location, as well as the distribution map of the corresponding equipment in the photovoltaic power station;

[0048] The architectural drawings of the photovoltaic power station in the corresponding geographical location are imported into the BIM modeling software, and then a three-dimensional model of the photovoltaic power station building is constructed.

[0049] The distribution map of the corresponding equipment in the photovoltaic power station is imported into the three-dimensional model of the photovoltaic power station building, and a virtual model of the photovoltaic power station is constructed based on the actual size of the photovoltaic power station and the corresponding equipment.

[0050] According to a preferred embodiment of the present invention, the process of constructing a digital twin model of a photovoltaic power station and obtaining operating data of the twin equipment includes:

[0051] Acquire virtual models of photovoltaic power plants, preprocess operational monitoring data, and preprocess environmental data;

[0052] The virtual model of the photovoltaic power station is digitized to obtain the coordinate position data of the photovoltaic power station building and the coordinate position data of the photovoltaic power station equipment.

[0053] Based on the coordinate location data of the photovoltaic power station building and the coordinate location data of the photovoltaic power station equipment, and using digital twin model technology, a digital twin model of the photovoltaic power station is constructed.

[0054] The preprocessed operation monitoring data and preprocessed environmental data are linked with the digital twin model of the photovoltaic power station;

[0055] Based on digital twin simulation software and the aforementioned dynamic simulation parameters and conditions, a dynamic simulation of the photovoltaic power station digital twin model is performed. During the simulation process, the photovoltaic power station digital twin model will calculate and update the equipment's operating dataset in real time based on the input pre-processed operation monitoring data and pre-processed environmental data, and record the operating dataset as the twin equipment operating data.

[0056] According to a preferred embodiment of the present invention, the process of constructing a digital twin fault prediction model for a photovoltaic power station based on twin device operating data and obtaining multi-dimensional fault precursor parameters includes:

[0057] Based on the set collection period and collection time, acquire several sets of twin device operation data from different devices;

[0058] A digital twin fault prediction model for photovoltaic power plants is constructed based on a neural network model and the operating data of several sets of twin devices from different equipment.

[0059] The preprocessed operation monitoring data and preprocessed environmental data collected in the current acquisition cycle are input into the digital twin fault prediction model of the photovoltaic power station to obtain multi-dimensional fault precursor parameters of the equipment corresponding to the current acquisition cycle. The multidimensional fault precursor parameters for:

[0060] ;in, For the time of data collection Corresponding preprocessing operation monitoring data; This is the minimum value among the preprocessed monitoring data corresponding to the current acquisition cycle; This represents the maximum value in the preprocessed monitoring data corresponding to the current acquisition cycle. For the time of data collection Corresponding preprocessed environment data; This represents the maximum value in the preprocessed environment data corresponding to the current acquisition cycle. This represents the maximum value in the preprocessed environment data corresponding to the current acquisition cycle. This is an environmental correction factor, set by relevant technical personnel; Weights for preprocessing operational monitoring data; Weights for preprocessed environmental data.

[0061] According to a preferred embodiment of the present invention, the process of generating a fault heat map of a digital twin model of a photovoltaic power station based on multidimensional fault precursor parameters, and performing intelligent operation and maintenance management of the photovoltaic power station includes:

[0062] The digital twin model of the photovoltaic power station is divided into several areas according to different equipment. It should be further explained that each area represents one piece of equipment.

[0063] Based on the multidimensional fault precursor parameters of the corresponding equipment Set up a fault heatmap;

[0064] In the fault heatmap, low-risk areas are represented by green, medium-risk areas by yellow, and high-risk areas by red.

[0065] Preset standard fault threshold range ;

[0066] If multidimensional fault precursor parameters If so, the area corresponding to the device is marked as a low-risk area and indicated in green;

[0067] If multidimensional fault precursor parameters If this occurs, the area corresponding to the device will be marked as a medium-risk area and indicated in yellow, triggering a local yellow flashing alarm and remotely notifying maintenance personnel to inspect and maintain the device.

[0068] If multidimensional fault precursor parameters If the situation is critical, the area corresponding to the device will be marked as a high-risk area and indicated in red, triggering a local red flashing alarm, issuing a siren warning, and reminding maintenance personnel to take emergency measures to address the issue.

[0069] A second aspect of the present invention also provides a smart operation and maintenance system for a photovoltaic power plant, the system executing the above-described smart operation and maintenance method for a photovoltaic power plant, comprising:

[0070] The photovoltaic power station multi-source data acquisition module is used to collect basic geographic and building data, equipment life cycle data, operation monitoring data and environmental data of the photovoltaic power station and perform preprocessing.

[0071] The photovoltaic power plant model building module fuses the pre-processed basic geographic and building data as well as the equipment life cycle data of the photovoltaic power plant to obtain multi-source data packets, and builds a virtual model of the photovoltaic power plant based on the multi-source data packets;

[0072] The digital twin dynamic simulation module constructs a digital twin model of the photovoltaic power station based on the virtual model of the photovoltaic power station, and performs dynamic simulation on the digital twin model of the photovoltaic power station based on preprocessed operation monitoring data and environmental data; and obtains the operating data of the twin equipment based on the digital twin model of the photovoltaic power station.

[0073] The photovoltaic power plant fault prediction module is used to construct a digital twin fault prediction model for photovoltaic power plants based on the operating data of the twin equipment, and obtain multi-dimensional fault precursor parameters.

[0074] The photovoltaic power plant intelligent management module is used to generate a fault heat map of the digital twin model of the photovoltaic power plant based on multi-dimensional fault precursor parameters, and to perform intelligent operation and maintenance management of the photovoltaic power plant based on the fault heat map.

[0075] Compared with the prior art, the beneficial effects of the present invention are:

[0076] 1. Collecting multi-source data and preprocessing it using density clustering algorithms and Z-score normalization effectively removes noise and outliers, while normalizing data of different dimensions significantly improves data quality. Multi-source heterogeneous data fusion technology integrates basic geographic and building data, as well as equipment lifecycle data, creating a multi-source data package that provides rich and accurate data support for constructing virtual models of photovoltaic power plants. This accurately reflects the actual situation of the photovoltaic power plant, providing a reliable basis for subsequent analysis and decision-making, and reducing misjudgments and biases caused by inaccurate or incomplete data.

[0077] 2. Constructing a digital twin fault prediction model for photovoltaic power plants enables the extraction of multi-dimensional fault precursor parameters from massive amounts of data. Fault heatmaps generated based on these parameters visually represent the fault risk levels of different areas or equipment within the photovoltaic power plant. Maintenance personnel can then strategically schedule maintenance work based on the heatmaps, transforming traditional passive maintenance into proactive maintenance. This avoids unnecessary inspections and wasted maintenance resources, significantly improving maintenance efficiency and reducing power generation losses due to equipment failures, thus ensuring the stable operation and economic benefits of the photovoltaic power plant. Attached Figure Description

[0078] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0079] Figure 1 This is a schematic diagram illustrating the steps of a smart operation and maintenance method for a photovoltaic power plant.

[0080] Figure 2 This is a flowchart illustrating a smart operation and maintenance method for photovoltaic power plants.

[0081] Figure 3 This is a schematic diagram of a module of a smart operation and maintenance system for a photovoltaic power plant. Detailed Implementation

[0082] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0083] like Figure 1As shown, a smart operation and maintenance method for a photovoltaic power station includes:

[0084] Collect basic geographic and building data, equipment life cycle data, operation monitoring data and environmental data of photovoltaic power plants, and preprocess them based on density clustering algorithm and Z-score standardization;

[0085] Based on multi-source heterogeneous data fusion technology, preprocessed basic geographic and building data as well as equipment life cycle data are fused to obtain multi-source data packets. Based on the multi-source data packets, a virtual model of a photovoltaic power station is constructed.

[0086] Based on the virtual model of the photovoltaic power station and the digital twin model, a digital twin model of the photovoltaic power station is constructed. The digital twin model of the photovoltaic power station is dynamically simulated based on the pre-processed operation monitoring data and environmental data. Based on the digital twin model of the photovoltaic power station, the operation data of the twin equipment is obtained.

[0087] Based on the operating data of the twin equipment and a neural network model, a digital twin fault prediction model for photovoltaic power plants is constructed to obtain multi-dimensional fault precursor parameters. Based on the multi-dimensional fault precursor parameters, a fault heat map of the digital twin model of the photovoltaic power plant is generated. Based on the fault heat map, intelligent operation and maintenance management of the photovoltaic power plant is carried out.

[0088] It should be further explained that, in the specific implementation process, the collection of basic geographical and building data, equipment lifecycle data, operation monitoring data, and environmental data for photovoltaic power plants includes:

[0089] The data acquisition device is set up, which includes a geographic and building data acquisition unit, an equipment life cycle data acquisition unit, an operation monitoring data acquisition unit, and an environmental data acquisition unit. A collection cycle is set, which includes several collection times. It should be further noted that setting the same collection cycle is beneficial for data uniformity and facilitates data processing.

[0090] The geographic and architectural data acquisition unit is based on GIS technology. It acquires three-dimensional basic geographic data such as topography, altitude, slope, and aspect of the area where the photovoltaic power station is located through satellite remote sensing imagery and UAV mapping. It scans the buildings in the photovoltaic power station based on laser scanning technology to obtain architectural data such as the size, location, and orientation of the corresponding buildings. It should be further noted that the obtained architectural data such as the size, location, and orientation of the corresponding buildings are compared with the design drawings and construction archives to ensure the accuracy and completeness of the data.

[0091] The equipment lifecycle data acquisition unit is used to record data such as the brand, model, specifications, production date, and purchase date of the corresponding equipment; to collect data such as the installation location, installation time, and debugging parameters of the corresponding equipment during the installation process; to collect data such as the operating status, performance parameters, and fault records of the corresponding equipment during the operation phase; and to collect data such as the maintenance time, content, and replaced parts of the corresponding equipment during the maintenance and decommissioning phases, as well as the time and reason for the decommissioning of the equipment.

[0092] The operation monitoring data acquisition unit is used to install light sensors, temperature sensors, current sensors, and voltage sensors on photovoltaic modules to collect parameters such as light intensity, module temperature, current, and voltage in real time; corresponding sensors are installed on equipment such as inverters and combiner boxes to collect parameters such as power output and efficiency of the equipment; it should be further noted that the operation monitoring data acquisition unit summarizes the data collected by these sensors and transmits it to the data center through wired or wireless communication networks.

[0093] The environmental data acquisition unit monitors data such as ambient temperature, humidity, wind speed, wind direction, air pressure, rainfall, and pollutant content of the photovoltaic power station according to the acquisition cycle.

[0094] It should be further explained that, in the specific implementation process, the preprocessing of basic geographic and building data, equipment lifecycle data, operation monitoring data, and environmental data includes:

[0095] Acquire basic geographic and building data, equipment lifecycle data, operation monitoring data, and environmental data;

[0096] It should be further clarified that numerical data in basic geographic and building data, equipment life cycle data, operation monitoring data, and environmental data are preprocessed, while non-numerical data in basic geographic and building data and equipment life cycle data are not preprocessed.

[0097] Based on Z-score standardization, basic geographic and building data, equipment lifecycle data, operation monitoring data, and environmental data are standardized. The specific process includes:

[0098] Data cleaning is performed on basic geographic and building data, equipment lifecycle data, operation monitoring data, and environmental data, and logical noise judgment rules are set. It should be further noted that the logical noise judgment rules include logical noise judgment rules for the corresponding data types, and are set by professionals in the field.

[0099] Based on the aforementioned logical noise judgment rules, logical noise reduction processing is performed on basic geographic and building data, equipment life cycle data, operation monitoring data, and environmental data to remove data that does not conform to the logical noise judgment rules for the corresponding data types.

[0100] Based on Z-score standardization, the basic geographic and building data, equipment life cycle data, operation monitoring data and environmental data after logical denoising are standardized.

[0101] For example, standardizing the elevation data in basic geographic and building data involves grouping elevations from different collection times within the same collection period into a single set, denoted as . ; Calculate the mean altitude ; Calculate the standard deviation of altitude Based on the Z-score standardized formula After obtaining the standard altitude .

[0102] Statistical noise processing is performed on Z-score-normalized basic geographic and building data, equipment lifecycle data, operation monitoring data, and environmental data based on density clustering algorithms. The specific process includes:

[0103] Based on standardized basic geographic and building data, equipment life cycle data, operation monitoring data, and environmental data, set the neighborhood radius and minimum number of core points for the corresponding data.

[0104] Based on the neighborhood radius and the minimum number of core points, core points, boundary points, and noise points are generated. It should be further noted that points containing at least the minimum number of core points within the neighborhood of the stated neighborhood radius are designated as core points; these data points are located in high-density areas of the data and are considered the core portion of the corresponding data. Points within the neighborhood of the stated neighborhood radius but containing fewer data points than the minimum number of core points are designated as boundary points; these data points are located at the edge of the data but are not considered abnormal data points. Data points not within the neighborhood of the stated neighborhood radius are designated as noise points.

[0105] For example, in the monitoring data of light intensity, the corresponding neighborhood radius is set to 0.5 and the minimum number of core points is 5. If there are 6 data points of light intensity within the neighborhood radius during the collection period, then the corresponding data points of light intensity are core points; if there are 2 data points of light intensity within the neighborhood radius during the collection period, then the corresponding data points of light intensity are boundary points; if the light intensity during the collection period is not within the neighborhood radius, then the corresponding light intensity during the collection period is a noise point.

[0106] Noise points corresponding to the standardized basic geographic and building data, equipment life cycle data, operation monitoring data, and environmental data are removed to obtain the corresponding statistically noise-processed basic geographic and building data, equipment life cycle data, operation monitoring data, and environmental data.

[0107] Non-numerical data of basic geographic and building data and equipment life cycle data, as well as basic geographic and building data and equipment life cycle data after statistical noise processing, are referred to as preprocessed basic geographic and building data and preprocessed equipment life cycle data.

[0108] The corresponding statistical noise-processed operation monitoring data and environmental data are denoted as preprocessed operation monitoring data and preprocessed environmental data.

[0109] It should be further explained that, in the specific implementation process, the data fusion of preprocessed basic geographic and building data as well as equipment lifecycle data to obtain multi-source data packets includes:

[0110] Acquire preprocessed basic geographic and building data, as well as data throughout the entire lifecycle of preprocessing equipment;

[0111] The non-numerical and numerical data in the preprocessed basic geographic and building data and the preprocessing equipment lifecycle data are separated to obtain the corresponding non-numerical and numerical data.

[0112] For the non-numerical data in the preprocessed basic geographic and building data and the preprocessed equipment lifecycle data, based on natural language processing technology, the non-numerical data in the preprocessed basic geographic and building data and the preprocessed equipment lifecycle data are classified, including classification fields, text description fields, spatial direction fields and level status fields.

[0113] For example, the classification of preprocessed basic geographic and building data includes:

[0114] 1. Classification Fields: Geographical Region Classification: such as "Photovoltaic Power Station in Plains Areas," classifying photovoltaic power stations according to their geographical and geomorphological characteristics. Building Facility Classification: such as "Photovoltaic Module Array Area," classifying buildings and facilities with different functions within the photovoltaic power station. Land Use Classification: "Dedicated Construction Land for Photovoltaics," "Comprehensive Utilization Land for Agro-Photovoltaic Integration," "Land Used for Wasteland," clearly defining the land use type occupied by the photovoltaic power station.

[0115] 2. Text Description Fields: Geographical Environment Description: Describes the factors influencing power generation from the geographical environment of the photovoltaic power station. Building Structure Description: Provides a detailed description of the building's structural features and functional facilities. Surrounding Amenities Description: Describes the surrounding amenities.

[0116] 3. Spatial Orientation Fields: Overall Location of the Power Station: Describes the location and orientation of the photovoltaic power station within a large region. Module Array Orientation: Specifies the spatial orientation of the module array. Relative Location of Buildings and Facilities: Describes the spatial relationship between various buildings and facilities.

[0117] 4. Status and Rating Fields: Land Use Compliance Rating: "Fully Compliant," "Partially Compliant," "Non-Compliant," reflecting the degree of compliance with land use regulations. Building Safety Rating: "Level 1 Safety," "Level 2 Safety," "Level 3 Safety," assessing the safety status of building facilities. Geographical Adaptability Rating: "Highly Adaptable," "Moderately Adaptable," "Lowly Adaptable," measuring the suitability of the geographical environment for the photovoltaic power station.

[0118] The classification of data throughout the entire lifecycle of preprocessing equipment includes:

[0119] 1. Classification Fields: Equipment Type Classification: such as "Photovoltaic Modules," "Inverters," "Combiner Boxes," "Transformers," and "Monitoring Equipment," classifying different equipment in a photovoltaic power station. Equipment Function Classification: such as "Power Generation Equipment," "Power Conversion Equipment," "Step-Up Equipment," and "Monitoring and Control Equipment," categorizing equipment according to its function. Equipment Brand Classification: such as "Jinko Solar Photovoltaic Modules," "Huawei Inverters," and "Sungrow Power Inverters," classifying equipment according to its manufacturer's brand.

[0120] 2. Text Description Fields: Equipment Operation Manual: Records the equipment's operating procedures and precautions. Equipment Fault Report: Records detailed equipment fault conditions and handling measures. Equipment Maintenance Record: Describes the equipment's maintenance and upkeep.

[0121] 3. Spatial Direction Fields: Equipment Installation Location: Describes the specific installation location of the equipment within the power station. Equipment Connection Direction: Describes the direction of the connection lines between equipment. Equipment Movement Record: Records the spatial movement of the equipment.

[0122] 4. Equipment Status Levels: Equipment Operating Status Levels: "Good (All equipment performance indicators are normal, no faults occur)," "Average (Minor performance fluctuations exist, but do not affect normal power generation)," "Poor (Frequent equipment failures affect power generation efficiency)," assessing the equipment's operating condition. Equipment Aging Levels: "New Equipment (less than 1 year of use)," "Moderately Aging Equipment (3-5 years of use)," "Severely Aging Equipment (more than 10 years of use)," reflecting the degree of equipment aging. Equipment Importance Levels: "Critical Equipment (plays a decisive role in power generation, such as photovoltaic modules and inverters)," "Important Equipment (has a significant impact on power plant operation, such as transformers and combiner boxes)," "Auxiliary Equipment (plays an auxiliary role in power plant operation, such as monitoring equipment)," categorized according to the equipment's importance within the power plant.

[0123] The non-numerical data in the preprocessed basic geographic and building data and the preprocessed equipment lifecycle data are used to generate non-numerical data packages based on the classification field, text description field, spatial direction field and level status field.

[0124] The specific process of constructing a numerical data fusion adversarial network includes:

[0125] It should be further explained that the generative adversarial network is used to fuse numerical data from preprocessed basic geographic and building data and preprocessed equipment lifecycle data to generate numerical data packets.

[0126] The numerical data in the preprocessed basic geographic and building data and the preprocessing equipment lifecycle data are respectively denoted as: Obtain multi-source data samples ;

[0127] Configure a numerical data fusion generator It should be further explained that the numerical data fusion generator The input is random noise Numerical data in preprocessed basic geographic and building data and preprocessed equipment lifecycle data are respectively denoted as That is, input ;

[0128] Based on neural network convolution, the input Perform a transformation to output a sample with the same dimensions as the real data. .

[0129] Configure a numerical data fusion discriminant It should be further explained that the numerical data fusion discriminator... The input is a sample and multi-source data samples Numerical data fusion discriminator The output represents the input sample. It is the probability of real data, ranging from between;

[0130] The output numerical data fusion discriminator The output is: ; .

[0131] Define a numerical data fusion generator Generator loss function The generator loss function for:

[0132] ;in, Expressing expectations, This indicates the input to the numerical data fusion generator. The distribution of .

[0133] Define a numerical data fusion discriminator Discriminator loss function The discriminator loss function for: ;

[0134] in, Expressing expectations, Represents multi-source data samples The distribution, This indicates the input to the numerical data fusion generator. The distribution of .

[0135] The optimization of numerical data fusion adversarial networks includes the following specific processes:

[0136] It should be further explained that in training the numerical data fusion generator At that time, fixed numerical data fusion discriminator The parameters are updated in the numerical data fusion generator based on the backpropagation algorithm. parameters This makes the generator loss function Minimize; in training the numerical data fusion discriminator At that time, fixed numerical data fusion generator The parameters are updated based on the backpropagation algorithm to determine the numerical data fusion discriminant. parameters This makes the discriminator loss function minimize;

[0137] Fixed numerical data fusion generator According to the numerical data fusion discriminator Discriminator loss function Calculate the discriminator gradient This leads to an update of the numerical data fusion discriminator. parameters The expression is:

[0138] ;in, Numerical data fusion discriminator The learning rate; Represents the discriminator loss function For parameters The partial derivatives of .

[0139] It should be further explained that the discriminator gradient ;

[0140] Fixed numerical data fusion discriminator According to the numerical data fusion generator Generator loss function Calculate generator gradient This leads to an update to the numerical data fusion generator. parameters The expression is:

[0141] ;in, Numerical data fusion generator The learning rate; Represents the discriminator loss function For parameters The partial derivatives of .

[0142] It should be further explained that the generator gradient ;

[0143] Obtain the trained numerical data fusion adversarial network, and generate numerical data packets based on the numerical data fusion adversarial network;

[0144] The non-numerical data packets and numerical data packets are combined into multi-source data packets.

[0145] It should be further explained that, in the specific implementation process, the process of constructing a virtual model of a photovoltaic power station based on multi-source data packets includes:

[0146] Based on multi-source data packets, obtain the plan view, elevation view, section view of the photovoltaic power station at the corresponding geographical location, as well as the distribution map of the corresponding equipment in the photovoltaic power station;

[0147] The architectural drawings of the photovoltaic power station in the corresponding geographical location are imported into the BIM modeling software, and then a three-dimensional model of the photovoltaic power station building is constructed.

[0148] The distribution map of the corresponding equipment in the photovoltaic power station is imported into the three-dimensional model of the photovoltaic power station building, and a virtual model of the photovoltaic power station is constructed based on the actual size of the photovoltaic power station and the corresponding equipment.

[0149] It should be further explained that, in the specific implementation process, the process of constructing a digital twin model of a photovoltaic power station and thereby obtaining the operating data of the twin equipment includes:

[0150] Acquire virtual models of photovoltaic power plants, preprocess operational monitoring data, and preprocess environmental data;

[0151] The virtual model of the photovoltaic power station is digitized to obtain the coordinate position data of the photovoltaic power station building and the coordinate position data of the photovoltaic power station equipment.

[0152] Based on the coordinate location data of the photovoltaic power station building and the coordinate location data of the photovoltaic power station equipment, and using digital twin model technology, a digital twin model of the photovoltaic power station is constructed.

[0153] By linking preprocessed operation monitoring data and preprocessed environmental data with the digital twin model of the photovoltaic power station, and comparing the actual data with the expected output of the digital twin model, the accuracy and reliability of the digital twin model of the photovoltaic power station are verified.

[0154] For example, input the actual monitored light intensity data at a certain collection time into the virtual model and check whether the deviation between the photovoltaic module power generation calculated by the model and the actual measured value is within an acceptable range.

[0155] Based on the actual operation and needs of the photovoltaic power station, dynamic simulation parameters and conditions are set, including simulation time step, environmental conditions, and equipment operation mode.

[0156] For example, the simulation time step is set to 1 minute to simulate the change in light intensity from sunrise to sunset throughout the day, as well as the operating status of photovoltaic power station equipment under different light conditions.

[0157] Based on digital twin simulation software and the aforementioned dynamic simulation parameters and conditions, a dynamic simulation of the photovoltaic power station digital twin model is performed. During the simulation process, the photovoltaic power station digital twin model will calculate and update the equipment's operating dataset in real time based on the input pre-processed operation monitoring data and pre-processed environmental data, and record the operating dataset as the twin equipment operating data.

[0158] For example, digital twin simulation software calculates the output power and efficiency of photovoltaic modules based on current light intensity and temperature data, and transmits this data to the inverter equipment model to calculate the inverter's conversion efficiency and output power.

[0159] It should be further explained that, in the specific implementation process, the process of constructing a digital twin fault prediction model for photovoltaic power plants based on the operating data of the twin devices and obtaining multi-dimensional fault precursor parameters includes:

[0160] Based on the set collection period and collection time, acquire several sets of twin device operation data from different devices;

[0161] Several sets of twin device operation data are grouped and labeled, denoted as follows: It is a natural number;

[0162] Will The operating data of the twin devices were used as sample data, and Less than The natural numbers, and using the sample data, the mean of the sample data is obtained, denoted as the sample set;

[0163] The remaining sets of twin device operation data are used as the test set; a training sample set is formed based on the sample set and the test set; a standard fault prediction model is constructed based on the neural network model.

[0164] The training sample set is then input into the standard fault prediction model to train the standard fault prediction model. The trained standard fault prediction model is denoted as the photovoltaic power station digital twin fault prediction model.

[0165] It should be further noted that the relevant training sample data and test data include:

[0166] For example, for a photovoltaic power station, data is collected once per hour (for a total of 100 days, 24 hours a day, with a total of 2400 sets of data).

[0167] The data acquisition equipment includes:

[0168] Photovoltaic modules (number) The parameters collected are operating voltage (V), operating current (A), and temperature (°C).

[0169] Inverter (number) The parameters collected are input voltage (V), output power (kW), operating temperature (°C), and conversion efficiency (%).

[0170] Combiner box (number) The parameters collected are input current (A), output voltage (V), and contact resistance (mΩ).

[0171] Data format, such as: [Photovoltaic module: Voltage: 350V, Current: 8A, Temperature: 22℃; Inverter: Input voltage: 350V, Output power: 2.8kW, Temperature: 45℃, Efficiency: 97%; Combiner box input current: 80A, Output voltage: 350V, Contact resistance: 5mΩ];

[0172] Data grouping and labeling: Total data volume: 2400 groups (arranged in chronological order);

[0173] Group labeling: Numbered in order of collection time. (Natural number labels).

[0174] Sample set: The first 500 data sets (labeled as follows) ), used to calculate the mean and partial training.

[0175] The sample data for label 1 is as follows: [Photovoltaic module: voltage: 350V, current: 8A, temperature: 22℃; Inverter: input voltage: 350V, output power: 2.8kW, temperature: 45℃, efficiency: 97%; Combiner box input current: 80A, output voltage: 350V, contact resistance: 5mΩ];

[0176] ...;

[0177] The sample data for label 500 is as follows: [Photovoltaic module: voltage: 345V, current: 8.2A, temperature: 25℃; Inverter: input voltage: 345V, output power: 2.85kW, temperature: 48℃, efficiency: 96.5%; Combiner box: input current: 82A, output voltage: 345V, contact resistance: 4.8mΩ];

[0178] Test set: The last 1800 data sets (labeled as follows) Take the middle 100 sets of data as backups (labeled as follows). );

[0179] The sample data for label 601 is as follows: [Photovoltaic module: Voltage: 320V, Current: 6A, Temperature: 18℃; Inverter: Input voltage: 320V, Output power: 1.92kW, Temperature: 38℃, Efficiency: 97.5%; Combiner box: Input current: 60A (10 groups of modules in parallel, 6A per group), Output voltage: 320V, Contact resistance: 5.2mΩ];

[0180] ...;

[0181] The sample data for label 2400 is as follows: [Photovoltaic module: voltage: 335V, current: 9A, temperature: 35℃; Inverter: input voltage: 335V, output power: 3.015kW, temperature: 58℃, efficiency: 96.8%; Combiner box: input current: 90A, output voltage: 335V, contact resistance: 4.9mΩ];

[0182] Standard threshold values ​​are specified, for example, the normal range for the contact resistance of a combiner box is [value missing]. The normal temperature range for the inverter is: The normal range for inverter efficiency is greater than or equal to 1. The normal range of photovoltaic module voltage is: .

[0183] Abnormal samples, such as: poor contact in combiner box: [PV module: voltage: 350V, current: 8A, temperature: 22℃; inverter: input voltage: 350V, output power: 2.8kW, temperature: 45℃, efficiency: 97%; combiner box input current: 80A, output voltage: 340V, contact resistance: 15mΩ]; hot spot effect in PV module: [PV module: voltage: 280V, current: 6A, temperature: 60℃; inverter: input voltage: 280V, output power: 1.68kW, temperature: 45℃, efficiency: 97%; combiner box: input current: 60A, output voltage: 280V, contact resistance: 15mΩ] [5mΩ]; Inverter overheating: [Photovoltaic module: Voltage: 350V, Current: 8A, Temperature: 22℃; Inverter: Input voltage: 350V, Output power: 2.8kW, Temperature: 65℃, Efficiency: 97%; Combiner box: Input current: 80A, Output voltage: 350V, Contact resistance: 5mΩ]; Combiner box current imbalance: [Photovoltaic module: Voltage: 350V, Current: 8A, Temperature: 22℃; Inverter: Input voltage: 350V, Output power: 2.8kW, Temperature: 45℃, Efficiency: 97%; Combiner box: Input current: 70A, Output voltage: 350V, Contact resistance: 5mΩ].

[0184] The preprocessed operation monitoring data and preprocessed environmental data collected in the current acquisition cycle are input into the digital twin fault prediction model of the photovoltaic power station to obtain multi-dimensional fault precursor parameters of the equipment corresponding to the current acquisition cycle. The multidimensional fault precursor parameters for:

[0185] ;in, For the time of data collection Corresponding preprocessing operation monitoring data; This is the minimum value among the preprocessed monitoring data corresponding to the current acquisition cycle; This represents the maximum value in the preprocessed monitoring data corresponding to the current acquisition cycle. For the time of data collection Corresponding preprocessed environment data; This represents the maximum value in the preprocessed environment data corresponding to the current acquisition cycle. This represents the maximum value in the preprocessed environment data corresponding to the current acquisition cycle. This is an environmental correction factor, set by relevant technical personnel; Weights for preprocessing operational monitoring data; Weights for preprocessed environmental data.

[0186] It should be further explained that, in the specific implementation process, the process of generating a fault heatmap of the photovoltaic power station's digital twin model based on multi-dimensional fault precursor parameters, and carrying out intelligent operation and maintenance management of the photovoltaic power station includes:

[0187] The digital twin model of the photovoltaic power station is divided into several areas according to different equipment. It should be further explained that each area represents one piece of equipment.

[0188] Based on the multidimensional fault precursor parameters of the corresponding equipment Set up a fault heatmap;

[0189] In the fault heatmap, low-risk areas are represented by green, medium-risk areas by yellow, and high-risk areas by red.

[0190] Preset standard fault threshold range ;

[0191] If multidimensional fault precursor parameters If so, the area corresponding to the device is marked as a low-risk area and indicated in green;

[0192] If multidimensional fault precursor parameters If this occurs, the area corresponding to the device will be marked as a medium-risk area and indicated in yellow, triggering a local yellow flashing alarm and remotely notifying maintenance personnel to inspect and maintain the device.

[0193] If multidimensional fault precursor parameters If the situation is critical, the area corresponding to the device will be marked as a high-risk area and indicated in red, triggering a local red flashing alarm, issuing a siren warning, and reminding maintenance personnel to take emergency measures to address the issue.

[0194] like Figure 3 As shown, a smart operation and maintenance system for a photovoltaic power station includes:

[0195] The photovoltaic power station multi-source data acquisition module is used to collect basic geographic and building data, equipment life cycle data, operation monitoring data and environmental data of the photovoltaic power station and perform preprocessing.

[0196] The photovoltaic power plant model building module fuses the pre-processed basic geographic and building data as well as the equipment life cycle data of the photovoltaic power plant to obtain multi-source data packets, and builds a virtual model of the photovoltaic power plant based on the multi-source data packets;

[0197] The digital twin dynamic simulation module constructs a digital twin model of the photovoltaic power station based on the virtual model of the photovoltaic power station, and performs dynamic simulation on the digital twin model of the photovoltaic power station based on preprocessed operation monitoring data and environmental data; and obtains the operating data of the twin equipment based on the digital twin model of the photovoltaic power station.

[0198] The photovoltaic power plant fault prediction module is used to construct a digital twin fault prediction model for photovoltaic power plants based on the operating data of the twin equipment, and obtain multi-dimensional fault precursor parameters.

[0199] The photovoltaic power plant intelligent management module is used to generate a fault heat map of the digital twin model of the photovoltaic power plant based on multi-dimensional fault precursor parameters, and to perform intelligent operation and maintenance management of the photovoltaic power plant based on the fault heat map.

[0200] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for intelligent operation and maintenance of a photovoltaic power station, characterized in that, The method includes: Collect basic geographic and building data, equipment life cycle data, operation monitoring data and environmental data of photovoltaic power plants, and preprocess them based on density clustering algorithm and Z-score standardization; Based on multi-source heterogeneous data fusion technology, preprocessed basic geographic and building data as well as equipment life cycle data are fused to obtain multi-source data packets. Based on the multi-source data packets, a virtual model of a photovoltaic power station is constructed. Dynamic simulation is performed based on the virtual model of the photovoltaic power station and pre-processed operation monitoring data, and a digital twin model of the photovoltaic power station is constructed based on the digital twin model; the operation data of the twin equipment is obtained based on the digital twin model of the photovoltaic power station. Based on the operating data of the twin equipment and using a neural network model, the process of constructing a digital twin fault prediction model for photovoltaic power plants and obtaining multi-dimensional fault precursor parameters includes: Based on the set collection period and collection time, acquire several sets of twin device operation data from different devices; A digital twin fault prediction model for photovoltaic power plants is constructed based on a neural network model and the operating data of several sets of twin devices from different equipment. The preprocessed operation monitoring data and the preprocessed environment data collected in the current collection period are input into the digital twin fault prediction model of the photovoltaic power station to obtain a multi-dimensional fault precursor parameter of the equipment corresponding to the current collection period , the multi-dimensional fault precursor parameter is: ;in, For the time of data collection Corresponding preprocessing operation monitoring data; This is the minimum value among the preprocessed monitoring data corresponding to the current acquisition cycle; This represents the maximum value in the preprocessed monitoring data corresponding to the current acquisition cycle. For the time of data collection Corresponding preprocessed environment data; This represents the maximum value in the preprocessed environment data corresponding to the current acquisition cycle. This represents the maximum value in the preprocessed environment data corresponding to the current acquisition cycle. This is an environmental correction factor, set by relevant technical personnel; Weights for preprocessing operational monitoring data; Weights for preprocessed environmental data; Based on the parameters indicating potential faults, a fault heatmap of the digital twin model of the photovoltaic power station is generated. The process of intelligent operation and maintenance management of the photovoltaic power station based on this fault heatmap includes: The digital twin model of the photovoltaic power station is divided into several areas according to different equipment. It should be further explained that each area represents one piece of equipment. Based on the multidimensional fault precursor parameters of the corresponding equipment Set up a fault heatmap; In the fault heatmap, low-risk areas are represented by green, medium-risk areas by yellow, and high-risk areas by red. Pre-set standard failure threshold interval ; If multidimensional fault precursor parameters If so, the area corresponding to the device is marked as a low-risk area and indicated in green; If multidimensional fault precursor parameters If this occurs, the area corresponding to the device will be marked as a medium-risk area and indicated in yellow, triggering a local yellow flashing alarm and remotely notifying maintenance personnel to inspect and maintain the device. If the multi-dimensional fault precursor parameter If the multi-dimensional fault precursor parameter If the multi-dimensional fault precursor parameter 2.The photovoltaic power station intelligent operation and maintenance method of claim 1, wherein, The process of collecting basic geographic and building data, equipment lifecycle data, operation monitoring data, and environmental data for photovoltaic power plants includes: A data acquisition device is set up, which includes a geographic and building data acquisition unit, an equipment life cycle data acquisition unit, an operation monitoring data acquisition unit, and an environmental data acquisition unit; and a collection cycle is set, which includes several collection moments. The geographic and building data acquisition unit collects geographic and building data of the photovoltaic power station based on GIS technology; the equipment life cycle data acquisition unit collects equipment life cycle data of the corresponding equipment in the photovoltaic power station; the operation monitoring data acquisition unit is used to collect operation monitoring data of the corresponding equipment in the photovoltaic power station; and the environmental data acquisition unit is used to collect environmental data of the photovoltaic power station. 3.The photovoltaic power station intelligent operation and maintenance method of claim 2, characterized in that, The preprocessing process for basic geographic and building data, equipment lifecycle data, operational monitoring data, and environmental data includes: Data cleaning is performed on basic geographic and building data, equipment lifecycle data, operation monitoring data, and environmental data, and logical noise judgment rules are set. Based on the aforementioned logical noise judgment rules, logical noise reduction processing is performed on basic geographic and building data, equipment life cycle data, operation monitoring data, and environmental data to remove data that does not conform to the logical noise judgment rules for the corresponding data types. Based on Z-score standardization, the basic geographic and building data, equipment life cycle data, operation monitoring data and environmental data after logical denoising are standardized. Statistical noise reduction is performed on the standardized data based on density clustering algorithms, including: Based on standardized basic geographic and building data, equipment life cycle data, operation monitoring data, and environmental data, set the neighborhood radius and minimum number of core points for the corresponding data. Based on the neighborhood radius and the minimum number of core points, core points, boundary points, and noise points are generated; Noise points corresponding to the standardized basic geographic and building data, equipment life cycle data, operation monitoring data, and environmental data are removed to obtain the corresponding statistically noise-processed basic geographic and building data, equipment life cycle data, operation monitoring data, and environmental data, which are then recorded as preprocessed basic geographic and building data, preprocessed equipment life cycle data, preprocessed operation monitoring data, and preprocessed environmental data. 4.The photovoltaic power station intelligent operation and maintenance method of claim 3, characterized in that, The process of fusing preprocessed basic geographic and building data with equipment lifecycle data to obtain multi-source data packets includes: The non-numerical and numerical data in the preprocessed basic geographic and building data and the preprocessed equipment lifecycle data are separated to obtain the corresponding non-numerical and numerical data. For the non-numerical data in the preprocessed basic geographic and building data and the preprocessed equipment lifecycle data, based on natural language processing technology, the non-numerical data in the preprocessed basic geographic and building data and the preprocessed equipment lifecycle data are classified, including classification fields, text description fields, spatial direction fields and level status fields. The non-numerical data in the preprocessed basic geographic and building data and the preprocessed equipment lifecycle data are used to generate non-numerical data packages based on the classification field, text description field, spatial direction field and level status field. A numerical data fusion adversarial network is constructed, and numerical data packets are generated based on the numerical data fusion adversarial network; the non-numerical data packets and the numerical data packets are combined into multi-source data packets.

5. The intelligent operation and maintenance method for a photovoltaic power station according to claim 4, characterized in that, The process of constructing a numerical data fusion adversarial network includes: The pre-processed basic geographic and building data and the pre-processed equipment life cycle data are respectively denoted as ; and a plurality of source data samples are obtained. Configure a numerical data fusion generator The numerical data fusion generator The input is random noise Numerical data in preprocessed basic geographic and building data and preprocessed equipment lifecycle data are respectively denoted as That is, input ; Based on neural network convolution, the input Perform a transformation to output a sample with the same dimensions as the real data. ; Configure a numerical data fusion discriminant The numerical data fusion discriminator The input is a sample and multi-source data samples Numerical data fusion discriminator The output represents the input sample. It is the probability of real data, ranging from between; The output numerical data fusion discriminator is: ; ; Define a numerical data fusion generator Generator loss function The generator loss function for: ; wherein, represents a desired, represents an input to a numerical data fusion generator ; Define a numerical data fusion discriminator Discriminator loss function The discriminator loss function for: ;in, Expressing expectations, Represents multi-source data samples The distribution, This indicates the input to the numerical data fusion generator. Distribution; Optimizing numerical data fusion adversarial networks involves the following process: Fixed numerical data fusion generator According to the numerical data fusion discriminator Discriminator loss function Calculate the discriminator gradient This leads to an update of the numerical data fusion discriminator. parameters The expression is: ;in, Numerical data fusion discriminator The learning rate; Represents the discriminator loss function For parameters Partial derivatives; discriminator gradient ; Fixed numerical data fusion discriminator According to the numerical data fusion generator Generator loss function Calculate generator gradient This leads to an update to the numerical data fusion generator. parameters The expression is: ;in, Numerical data fusion generator The learning rate; Represents the discriminator loss function For parameters Partial derivatives; generator gradient .

6. The intelligent operation and maintenance method for a photovoltaic power station according to claim 5, characterized in that, The process of constructing a virtual model of a photovoltaic power plant based on multi-source data packets includes: Based on multi-source data packets, obtain the plan view, elevation view, section view of the photovoltaic power station at the corresponding geographical location, as well as the distribution map of the corresponding equipment in the photovoltaic power station; The architectural drawings of the photovoltaic power station in the corresponding geographical location are imported into the BIM modeling software, and then a three-dimensional model of the photovoltaic power station building is constructed. The distribution map of the corresponding equipment in the photovoltaic power station is imported into the three-dimensional model of the photovoltaic power station building, and a virtual model of the photovoltaic power station is constructed based on the actual size of the photovoltaic power station and the corresponding equipment.

7. The photovoltaic power plant intelligent operation and maintenance method according to claim 6, characterized in that, The process of constructing a digital twin model of a photovoltaic power station and obtaining the operating data of the twin equipment includes: Acquire virtual models of photovoltaic power plants, preprocess operational monitoring data, and preprocess environmental data; The virtual model of the photovoltaic power station is digitized to obtain the coordinate position data of the photovoltaic power station building and the coordinate position data of the photovoltaic power station equipment. Based on the coordinate location data of the photovoltaic power station building and the coordinate location data of the photovoltaic power station equipment, and using digital twin model technology, a digital twin model of the photovoltaic power station is constructed. The preprocessed operation monitoring data and preprocessed environmental data are linked with the digital twin model of the photovoltaic power station; Based on digital twin simulation software, and the aforementioned dynamic simulation parameters and conditions, a dynamic simulation of the photovoltaic power station digital twin model is performed. During the simulation process, the photovoltaic power station digital twin model will calculate and update the equipment's operating dataset in real time based on the input pre-processed operation monitoring data and pre-processed environmental data, and record the operating dataset as the twin equipment operating data.

8. A smart operation and maintenance system for a photovoltaic power plant, wherein the system executes the smart operation and maintenance method for a photovoltaic power plant as described in any one of claims 1 to 7, characterized in that, include: The photovoltaic power station multi-source data acquisition module is used to collect basic geographic and building data, equipment life cycle data, operation monitoring data and environmental data of the photovoltaic power station and perform preprocessing. The photovoltaic power plant model building module fuses the pre-processed basic geographic and building data as well as the equipment life cycle data of the photovoltaic power plant to obtain multi-source data packets, and builds a virtual model of the photovoltaic power plant based on the multi-source data packets; The digital twin dynamic simulation module constructs a digital twin model of the photovoltaic power station based on the virtual model of the photovoltaic power station, and performs dynamic simulation on the digital twin model of the photovoltaic power station based on pre-processed operation monitoring data and environmental data; Based on the aforementioned digital twin model of the photovoltaic power station, obtain the operating data of the twin equipment; The photovoltaic power plant fault prediction module is used to construct a digital twin fault prediction model for the photovoltaic power plant based on the operating data of the twin equipment. The process of obtaining multi-dimensional fault precursor parameters includes: Based on the set collection period and collection time, acquire several sets of twin device operation data from different devices; A digital twin fault prediction model for photovoltaic power plants is constructed based on a neural network model and the operating data of several sets of twin devices from different equipment. The preprocessed operation monitoring data and the preprocessed environment data collected in the current collection period are input into the digital twin fault prediction model of the photovoltaic power station to obtain a multi-dimensional fault precursor parameter of the equipment corresponding to the current collection period , the multi-dimensional fault precursor parameter is: ;in, For the time of data collection Corresponding preprocessing operation monitoring data; This is the minimum value among the preprocessed monitoring data corresponding to the current acquisition cycle; This represents the maximum value in the preprocessed monitoring data corresponding to the current acquisition cycle. For the time of data collection Corresponding preprocessed environment data; This represents the maximum value in the preprocessed environment data corresponding to the current acquisition cycle. This represents the maximum value in the preprocessed environment data corresponding to the current acquisition cycle. This is an environmental correction factor, set by relevant technical personnel; Weights for preprocessing operational monitoring data; Weights for preprocessed environmental data; The intelligent management module for photovoltaic power plants is used to generate a fault heat map of the digital twin model of the photovoltaic power plant based on multi-dimensional fault precursor parameters. The process of intelligent operation and maintenance management of the photovoltaic power plant based on the fault heat map includes: The digital twin model of the photovoltaic power station is divided into several areas according to different equipment. It should be further explained that each area represents one piece of equipment. Based on the multidimensional fault precursor parameters of the corresponding equipment Set up a fault heatmap; In the fault heatmap, low-risk areas are represented by green, medium-risk areas by yellow, and high-risk areas by red. Preset standard fault threshold range ; If multidimensional fault precursor parameters If so, the area corresponding to the device is marked as a low-risk area and indicated in green; If multidimensional fault precursor parameters If this occurs, the area corresponding to the device will be marked as a medium-risk area and indicated in yellow, triggering a local yellow flashing alarm and remotely notifying maintenance personnel to inspect and maintain the device. If multidimensional fault precursor parameters If the situation is critical, the area corresponding to the device will be marked as a high-risk area and indicated in red, triggering a local red flashing alarm, issuing a siren warning, and reminding maintenance personnel to take emergency measures to address the issue.

Images