New energy station equipment state online checking system based on longitudinal data chain

Abstract

The invention discloses a new energy station equipment state online checking system based on a longitudinal data chain, and belongs to the technical field of new energy stations. According to the method, data acquisition is performed on equipment in a new energy station to obtain corresponding equipment data, data precision demand analysis is performed after data preprocessing to determine whether a data precision retention measure is taken or not, and then the collected equipment data is input into a digital twin model to form a longitudinal data chain; and finally, completing consistency check and integrity check of cross-level and multi-source data based on the longitudinal data chain, outputting an equipment state through the state evaluation model, and constructing an equipment state feature vector through the checked longitudinal data chain to obtain the equipment state and output a corresponding check result. The online checking precision of the new energy station equipment is ensured, and the problem that the checking precision is reduced due to the fact that high-frequency data needed by checking are filtered in the longitudinal link in the prior art is solved.

Description

Technical Field

[0001] This invention relates to the field of new energy power station technology, and in particular to an online verification system for the status of new energy power station equipment based on a vertical data link. Background Technology

[0002] With the development of computer and internet technologies, the operation methods in the new energy industry have gradually shifted from manual semi-automation, such as paper records and manual meter reading, to information-based operations with remote monitoring and control and automatic data storage. The widespread application of digital technologies such as cloud computing, big data, IoT, and mobile internet has propelled the information-based operations of new energy enterprises into the digital age, and they are beginning to explore the intelligent stage of machine learning, artificial intelligence, virtualization, and robotics.

[0003] To adapt to technological development, respond to policy requirements, implement the group's strategy, and meet its own high-quality development needs, it is necessary to build a big data-driven smart new energy centralized production and operation center that is comprehensive, multi-faceted, effective, advanced, and reliable to address various factors affecting the development of new energy operations.

[0004] For example, Chinese invention patent application CN121012210A discloses a panoramic parameter monitoring system for grid-connected equipment in new energy power stations, which includes: a panoramic parameter database; a multi-source data acquisition module; a data quality assessment module; an equipment performance monitoring module; a panoramic visualization module; a safety closed-loop management module; and an early warning and diagnosis module. Through panoramic monitoring of grid-connected equipment in new energy power stations, power station status assessment, and grid-connected equipment performance monitoring, combined with new energy grid-connected technology supervision, the system achieves panoramic real-time monitoring of new energy power station operation data and dynamic assessment of operation status.

[0005] Currently, the implementation of such technologies in new energy power plants (such as wind farms and photovoltaic farms) typically involves equipment status monitoring, data acquisition, transmission, and processing to ensure the reliability, efficiency, and safety of equipment operation. Specifically, the equipment in new energy power plants (such as wind turbines, photovoltaic panels, and inverters) needs to be equipped with sensors to collect various operating data in real time (such as voltage, current, power, temperature, humidity, etc.). There are multiple data acquisition methods, such as: PLCs (Programmable Logic Controllers) for equipment control and status monitoring, which can acquire data through interface modules with sensors; SCADA (Supervisory Control and Data Acquisition) systems for centralized monitoring and data acquisition, often combined with PLC systems for real-time data acquisition; and smart sensors.

[0006] A vertical data chain typically refers to the entire data transmission and processing chain from the device end to the data center or cloud platform. It is implemented hierarchically to complete tasks such as data acquisition, transmission, storage, processing, and analysis. On-site devices collect data through sensors and transmit it to edge computing devices or remote data hubs via local data acquisition systems (such as PLCs). Edge computing nodes, located at the device site or regional control center, perform preliminary data processing, filtering out invalid data or compressing the data to reduce the amount of data transmitted to the upper layers. Simultaneously, edge nodes can also perform preliminary equipment fault diagnosis, such as calculating equipment health indicators. The cloud platform or data center centrally stores, processes, and analyzes data from various sites, using big data analytics, machine learning, and artificial intelligence technologies to monitor equipment operating status in real time, predict and diagnose faults, and generate equipment health reports. The core purpose of online equipment status verification technology is to assess whether the equipment is in normal operating condition in real time and to immediately alarm when potential faults are detected. Online verification typically includes the following technical aspects: real-time data monitoring and analysis, health management and fault diagnosis, online verification, and fault diagnosis and early warning.

[0007] The above-mentioned technology has at least the following technical problems:

[0008] In the current technology, new energy power generation still faces many challenges in the trend of digital transformation. For example, updating and developing existing information systems and integrating existing systems at different levels for deep business integration. However, in the process of online verification of actual equipment, many station control or centralized control nodes will average, downsample, and threshold filter the data (only upload data with changes greater than a certain value) in order to save bandwidth. However, the online verification model may need high-frequency vibration data, high-frequency current harmonics, and rapid voltage fluctuations. The link characteristics limit the model's capabilities, resulting in inaccurate verification. There is a problem that the high-frequency data required for verification is filtered in the vertical link, which reduces the verification accuracy. Summary of the Invention

[0009] To address the technical problem in existing technologies where high-frequency data required for verification is filtered in the vertical data link, leading to a decrease in verification accuracy, this invention provides an online verification system for the status of new energy power station equipment based on a vertical data link. The technical solution is as follows:

[0010] This invention provides an online verification system for the status of new energy power station equipment based on a vertical data chain. The system includes: a new energy power station equipment data acquisition module, an online equipment status verification module, and a verification result output module. The equipment data acquisition module collects data from the equipment within the new energy power station, and after preprocessing the data, analyzes data accuracy requirements to determine whether to implement data accuracy preservation measures. The online equipment status verification module inputs the collected equipment data into a preset digital twin model for automatically analyzing equipment status and identifying anomalies, forming a vertical data chain. The verification result output module performs consistency and integrity verification of cross-level, multi-source data based on the vertical data chain, outputs the equipment status through a status assessment model, and constructs an equipment status feature vector through the verified vertical data chain to obtain the equipment status and output the corresponding verification result.

[0011] Beneficial effects

[0012] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0013] 1. The online verification system for the status of new energy power station equipment based on a vertical data chain provided by this invention collects data from the equipment within the new energy power station to obtain corresponding equipment data. It then preprocesses the equipment data and performs data accuracy requirement analysis to determine whether to adopt data accuracy preservation measures, thereby improving the accuracy of input data to subsequent models. This solves the problem of limited bandwidth in new energy power stations, which prevents the indiscriminate uploading of all raw data, and the lack of decision-making basis for high-precision data collection strategies. This not only improves data quality and consistency, making subsequent model input cleaner and more accurate, but also helps avoid model misjudgments and omissions caused by low-precision data, improving modeling accuracy. Next, the collected equipment data is input into a digital twin model to form a vertical data chain, and the system automatically analyzes the status of the corresponding equipment, identifies corresponding abnormal situations, and issues early warnings. The system assesses the accuracy requirements of the verification data to determine whether to implement a data accuracy improvement plan. This avoids the problem that traditional monitoring can only see measured values ​​and cannot see the internal status of the equipment or the occurrence of potential risks. It enables proactive anomaly identification and intelligent early warning, and improves the model's fault identification and prediction capabilities. Finally, based on the vertical data chain, it completes the consistency and integrity verification of cross-level and multi-source data. Then, it calculates the equipment status through the status assessment model and constructs the equipment status feature vector through the verified vertical data chain to obtain the equipment status and output the corresponding verification results. This makes the vertical data chain a more reliable input source, thereby improving the data credibility and operational security of the overall system, ensuring the accuracy of online equipment verification, and effectively solving the problem in existing technologies where the high-frequency data required for verification is filtered in the vertical link, resulting in a decrease in verification accuracy.

[0014] 2. This invention calculates the total input current of each string and compares it with the total output current to obtain the corresponding current deviation. If the current deviation is greater than the set current threshold, it indicates that the DC combiner box is abnormal. This solves the problem that traditional monitoring only records the current values ​​of each channel but does not summarize and compare them, making it difficult to identify faults such as blown fuses, poor contact, and open circuits in a timely manner. It helps to quickly identify typical DC side faults such as string current imbalance, blown fuses, and open circuits in branches.

[0015] 3. After comparing the total input current of each string, the total input power of each string is calculated and compared with the total output power to obtain the corresponding power deviation. If the power deviation is greater than the set power threshold, it indicates that the DC combiner box is abnormal. This not only helps to improve the accuracy and stability of the early warning, but also fills the gap in the existing technology of lacking a complete consistency verification method for the DC side energy link. Then, it is judged whether the total output current and total output power have constant or zero values, and the duration of the occurrence is judged. If a constant or zero value occurs and the duration exceeds the set duration, it indicates that the DC combiner box is abnormal. This solves the technical problem that it is difficult to distinguish between data acquisition abnormalities and actual equipment abnormalities, which can easily lead to misjudgment. In addition, it realizes the automatic identification of data abnormalities such as communication failure of acquisition module, sensor lag, and measurement board crash. Furthermore, by judging the duration, it not only avoids false alarms caused by brief data jitter, but also improves the stability of the early warning.

[0016] 4. By using network communication indicators for quantitative analysis to obtain an assessment of bandwidth savings, and comparing it with preset abnormal event alarm thresholds, this method overcomes the shortcomings of existing technologies where insufficient network bandwidth cannot be quantified and the system cannot accurately determine whether a transmission strategy needs to be switched. This helps improve the reliability and timeliness of data upload strategy switching and ensures that the system does not experience data loss or delay due to bandwidth overload.

[0017] 5. If the bandwidth saving assessment value exceeds the abnormal event alarm threshold, the system will switch to bandwidth control mode. Bandwidth control mode means automatically exiting bandwidth saving mode and uploading device data at the set highest data upload frequency. This ensures that critical device data can still be uploaded quickly and preferentially when the network is under heavy pressure, avoiding data loss. This solves the problem of traditional systems being slow to respond to network anomalies and unable to automatically improve data transmission priority.

[0018] 6. If the bandwidth saving assessment is not greater than the abnormal event alarm threshold, then the bandwidth saving is divided and corresponding data accuracy preservation measures are taken. This fills the gap in the existing system, which cannot automatically adjust data accuracy and sampling frequency according to bandwidth conditions and lacks a refined bandwidth management mechanism. In this way, it realizes graded response according to bandwidth pressure and adopts different data accuracy preservation strategies under different network conditions. Under the premise of ensuring data quality, it reduces the overall link load, extends the stable operation time of the system, and improves the stability, anti-congestion capability and communication quality of the entire new energy power station monitoring system. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A schematic diagram of the structure of the online status verification system for new energy power station equipment based on vertical data link provided in this application embodiment;

[0021] Figure 2 A logic diagram of data acquisition and processing provided for embodiments of this application;

[0022] Figure 3 This is a schematic diagram illustrating the bidirectional data connection provided in an embodiment of this application.

[0023] Figure 4 This is a schematic diagram of the early warning overview interface provided in an embodiment of this application;

[0024] Figure 5 A flowchart illustrating the data accuracy requirement analysis provided for embodiments of this application;

[0025] Figure 6 This is a flowchart illustrating the process of assessing the accuracy requirements of verification data in an embodiment of this application. Detailed Implementation

[0026] The technical solution provided in this application will now be described with reference to the accompanying drawings.

[0027] To facilitate understanding of the embodiments of this application, the following points will be explained first:

[0028] First, in this application, "at least one" means one or more, and "more than one" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates an "or" relationship between the preceding and following related objects, but it does not exclude the possibility of indicating an "and" relationship; the specific meaning can be understood in context. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can mean: a, b, c; a and b; a and c; b and c; or a and b and c. Here, a, b, and c can be single or multiple.

[0029] Second, the use of prefixes such as "first" and "second" in this application is solely for the purpose of distinguishing and describing different things belonging to the same category, and does not constrain the order, size, or quantity of things. For example, "first message" and "second message" are simply different messages, and there is no chronological, size, or priority relationship between them.

[0030] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0031] Example 1: As Figure 1 The diagram shown is a structural schematic of the online verification system for the status of new energy power station equipment based on a vertical data chain provided in this application embodiment, including: a new energy power station equipment data acquisition module, an online verification module for equipment status, and a verification result output module.

[0032] Among them, the new energy power station equipment data acquisition module is used to collect data from the equipment in the new energy power station to obtain the corresponding equipment data. After preprocessing the equipment data, the data accuracy requirement analysis determines whether to take data accuracy retention measures to improve the accuracy of the input data to the subsequent model.

[0033] It should be added that data acquisition refers to real-time monitoring of the operating status of new energy power plants, including key indicators such as power generation and equipment performance, and collecting multi-source data, such as meteorological data and equipment operation data. This data forms the basis for building digital twin models. The collected data is then cleaned, transformed, and analyzed for subsequent modeling and simulation. Figure 2 The diagram shown is a logic diagram of data acquisition and processing provided in the embodiments of this application, which is used to intuitively demonstrate the process of acquiring, processing and modeling simulation of equipment data from new energy power plants.

[0034] The online equipment status verification module is used to input the collected data from each device into a preset digital twin model for automatically analyzing equipment status and identifying abnormal situations, forming a vertical data chain.

[0035] The vertical data chain refers to a full lifecycle data association system that structurally connects the operation data of new energy power plants along two dimensions: time axis and system level, according to the multi-level information flow of "grid dispatch, regional centralized control, and power plant equipment". It mainly includes three hierarchical dimensions, such as vertical dimension, time dimension, and data association dimension. Taking the hierarchical dimension as an example, the vertical data chain consists of the grid dispatch side (upper-level instructions, market data), the regional centralized control side (regional coordination, multi-power plant aggregation), the power plant monitoring side (power plant-level operation status), the equipment level / system level (inverters, wind turbines, transformer substations, etc.), and the component / parameter level (components, blades, temperature sensors, etc.). The vertical data chain is formed by associating multi-level data from the power plant side, regional side, and grid side according to time sequence. The steps include multi-source data access and standardization, timestamp alignment and time sequence reconstruction, cross-level association mapping, and data chain storage and indexing. The association relationships include equipment topology association, control command association, fault propagation association, and meteorological impact association.

[0036] It needs to be explained that digital twin technology is a technology that integrates multi-disciplinary and multi-scale simulation processes to map the entire lifecycle of physical equipment in virtual space, thereby reflecting the corresponding physical equipment's entire lifecycle. It consists of three parts: the physical product in the physical space, the virtual product in the virtual space, and the data and information interaction interface between the physical and virtual spaces. Specifically, as shown below... Figure 3 The diagram illustrates the bidirectional data connection between the physical and digital spaces. Figure 3 This is a schematic diagram illustrating the bidirectional data connection provided in an embodiment of this application.

[0037] Specifically, based on the collected data, a digital twin model of the new energy power plant is constructed. This model is virtual, but it can realistically reflect the structure and behavior of the physical power plant. During the model construction process, various methods such as physical simulation models, mathematical models, or statistical models are used. Through IoT technology and sensors, the state information of the physical entity is collected in real time and transmitted to the virtual model. The virtual model updates in real time based on the received information to maintain a high degree of synchronization with the physical entity. Furthermore, the main functions of this system are as follows:

[0038] (1) Real-time data monitoring: Digital twin technology collects various operating data of new energy power plants in real time through sensors and Internet of Things devices, such as power generation, equipment status, environmental parameters, etc.

[0039] (2) Anomaly Detection and Diagnosis: Based on real-time data, the system can automatically analyze equipment status, identify abnormal situations, and issue early warnings to help maintenance personnel take timely measures to prevent the fault from escalating, such as... Figure 4 The image shown is a schematic diagram of the early warning overview interface provided in an embodiment of this application.

[0040] (3) Data analysis and visualization: Digital twin technology provides powerful data analysis tools to transform complex data into intuitive charts and reports, helping operation and maintenance personnel to quickly understand the power plant's operating status.

[0041] (4) Data prediction: By establishing a digital twin model of a photovoltaic power station or wind power plant, the impact of natural conditions such as solar radiation and wind speed on power generation can be simulated, thereby predicting power generation in advance. This prediction helps the power grid dispatching department to make preparations for dispatching in advance and optimize power grid operation.

[0042] (5) Trend Viewing Page: In the operation of new energy power plants, digital twin technology can monitor the dynamic changes of key parameters in real time and record historical data. As the system operates for a longer period of time, the analysis, processing, and use of this data become particularly important for studying changes in equipment performance and the relationships between equipment performance. Digital twin technology provides a powerful tool that can display data arbitrarily and support viewing real-time and historical trends in order to analyze the changing trends and interrelationships between different performance parameters.

[0043] The verification result output module is used to complete the consistency verification and integrity verification of cross-level and multi-source data based on the vertical data chain. Then, it outputs the equipment status through the status assessment model, and constructs the equipment status feature vector through the verified vertical data chain to obtain the equipment status and output the corresponding verification results.

[0044] The input to each state assessment model is the collected operating data of the corresponding device, and the output is the corresponding device state. The main steps to construct the device state feature vector include basic feature extraction (from each level of the vertical data chain), feature dimension design (four-dimensional feature system), and dynamic weighting based on the verification results. The output verification results generally include a multi-dimensional structure of the verification results, such as data quality verification results and device state verification results, as well as a complete verification result output structure.

[0045] It should be noted that before designing an online verification system for the status of new energy power plant equipment based on a vertical data chain, technical personnel typically first construct a pre-set database to support the operation of various control strategies. This database integrates multiple key process parameters, including preset first proportional, preset second proportional, preset nacelle ambient temperature proportional, preset proportional, set wind speed value, set ambient temperature, set humidity, abnormal event alarm threshold, bandwidth saving range, and maximum model accuracy value. These parameters are all pre-set by personnel with professional technical backgrounds based on the analysis methods used and the on-site hardware configuration conditions. Furthermore, this pre-set database provides the core data foundation for subsequent automated processes such as data uploading, storage optimization, and filtering.

[0046] It should be explained that the condition assessment models include, but are not limited to, gearbox oil temperature abnormality early warning models, DC combiner box failure early warning models, photovoltaic module dust accumulation and inefficiency early warning models, photovoltaic string failure early warning models, power curve abnormality early warning models, generator bearing temperature abnormality early warning models, main control cabinet temperature abnormality early warning models, blade icing and power reduction operation early warning models, pitch motor heat dissipation system abnormality early warning models, hydraulic system operation abnormality early warning models, turbulent power reduction operation early warning models, photovoltaic string inefficiency early warning models, inverter failure early warning models, photovoltaic inverter power generation inefficiency early warning models, inverter communication abnormality models, DC combiner box communication abnormality early warning models, and string current data acquisition early warning models.

[0047] Example 1: The design concept of the gearbox oil temperature anomaly early warning model includes horizontal comparison and vertical comparison. Horizontal comparison means: based on different operating conditions represented by intervals defined by humidity, power, and gearbox speed, an alarm is triggered when the difference between the gearbox oil temperature data and the average oil temperature under non-full load exceeds a preset first proportion. That is, the average difference between the gearbox oil temperature data and the average oil temperature of each unit under non-full load is the difference value. An alarm is triggered when the difference between the gearbox oil temperature data and the average oil temperature of each unit under full load exceeds a preset second proportion. The average oil temperature is the average gearbox oil temperature of the new energy power station at the same time. The preset first proportion is higher than the preset second proportion, indicating that the alarm urgency under non-full load conditions is lower than under full load conditions. Typically, the preset first proportion is 15%, and the preset second proportion is 10%. Non-full load refers to all operating states where the equipment output is lower than its maximum design capacity, while full load refers to the maximum power or maximum working capacity that the equipment can output under continuous and stable operation under design conditions.

[0048] Specifically, taking power range division as an example, the possible division methods include the rated percentage method, the efficiency curve inflection point method, and the operating time distribution method. If the rated percentage method is used, when the power is below 5%, the equipment does not produce effective power, and the operating condition is shutdown / standby. When the power is between 5% and 30%, it belongs to the low efficiency range and the operating condition is classified as low load. When the power is between 30% and 70%, it belongs to the economic operating range and the operating condition is classified as medium load. When the power is between 70% and 95%, it is close to the maximum capacity and the operating condition is classified as high load. When the power is above 95%, it reaches the limit of the equipment and the operating condition is classified as full load / overload.

[0049] The longitudinal comparison compares the temperature rise rate with the unit power. The temperature rise coefficient represents the gear temperature rise caused by each unit increase in power. If the temperature rise coefficient exceeds the temperature rise threshold (generally set to 30), a warning is issued. The comparison also involves comparing the oil temperature with the gearbox inlet temperature and calculating their deviation rate. This involves calculating the difference between the oil temperature and the gearbox inlet temperature and then the ratio to the gearbox inlet temperature. If this deviation rate exceeds the oil temperature deviation ratio (generally 15%), a warning is issued. Furthermore, the comparison involves comparing the oil temperature with the engine room ambient temperature and calculating the deviation rate between the oil temperature and the engine room cabinet temperature. If this deviation rate exceeds the preset engine room ambient temperature ratio (generally 10%), a warning is issued. The gearbox oil temperature curve is derived by analyzing historical data, for example, through multiple regression fitting or neural network modeling. This curve comprehensively describes the ambient temperature, humidity, power, gearbox speed, gearbox inlet temperature, and engine room ambient temperature. If the currently operating gearbox oil temperature exceeds the set ratio of the fitted curve, a warning is issued. All of the above temperatures are obtained through temperature sensors deployed in the corresponding hardware.

[0050] Example 2: The specific content of the DC combiner box failure early warning model is as follows:

[0051] The first step is to calculate the total input current of each string and compare it with the total output current to obtain the corresponding current deviation. That is, to perform a ratio calculation between the total input current and the total output current. If the current deviation is greater than the set current threshold, it indicates that the DC combiner box is abnormal. Generally, a current sensor (such as a Hall sensor or shunt) is installed in each input branch of the DC combiner box to directly measure the real-time current of each string. Then, the real-time currents of each string are added together to obtain the total current.

[0052] The second step is to calculate the total input power of each string and compare it with the total output power to obtain the corresponding power deviation. If the power deviation is greater than the set power threshold, it indicates that the DC combiner box is abnormal. Generally, a DC power sensor is installed on each string input branch of the combiner box to directly measure the real-time power of the corresponding string, and the total power is obtained by adding them together.

[0053] The third step is to determine whether the total output current and total power are constant or zero, and to determine the duration of the occurrence. If a constant or zero value is found and the duration exceeds the set time (e.g., 1 hour), it indicates that the DC combiner box is abnormal.

[0054] Example 3: The specific content of the early warning model for inefficient photovoltaic module dust accumulation is as follows:

[0055] Direct radiation is divided into different intervals to represent different operating conditions. Based on the same operating conditions, under the same direct radiation conditions, the wind speed is greater than the set wind speed value (e.g., 0), the ambient temperature is greater than the set ambient temperature (e.g., 5 degrees), and the humidity is greater than the set humidity (e.g., 5). The deviation between the decrease in current of a single branch and the average current of all branches is compared. This means that the difference between the decrease in current of a single branch and the average current of all branches is calculated and then compared with the average current. If the current of a single branch is lower than the average value by a preset percentage (usually 15%), an early warning is issued; otherwise, no additional processing is performed. The wind speed, ambient temperature, and humidity are all obtained from the on-site self-built meteorological station.

[0056] Example 4: The design concept of the early warning model for turbulent power reduction operation is as follows:

[0057] First, when wind direction or wind speed values ​​change frequently, if the power output decreases compared to previous levels, i.e., under the same operating conditions with frequent changes in wind direction or wind speed values, if the decrease in output power exceeds a set threshold, an early warning will be issued. Frequent changes are specifically quantified using the coefficient of variation (COP). The COP is a statistic that measures the degree of data dispersion; it is the ratio of the standard deviation to the mean and is usually used to compare the dispersion of different datasets. The larger the COP, the greater the data volatility; the smaller the COP, the smaller the data volatility.

[0058] Next, the relationship between turbulence intensity and power is fitted. Turbulence intensity reflects the characteristics of wind speed fluctuation. When the average wind speed is constant, the greater the turbulence intensity, the greater the wind speed fluctuation and the smaller the wind power. When the turbulence intensity increases, the power decreases. If the power decrease exceeds the set threshold, it is considered abnormal.

[0059] It's important to explain that studying the relationship between turbulence and power has profound significance in the field of wind power generation. It directly relates to the effective capture and conversion efficiency of wind energy. Turbulence, as a common atmospheric phenomenon, presents a significant challenge to the performance and stability of wind turbine generators due to its rapid and unpredictable wind speed variations. In-depth research into turbulence characteristics helps to better understand the impact of wind speed fluctuations on the dynamic response of wind turbine generators, thereby optimizing the operating state of wind turbines and improving their power output efficiency under varying wind speeds. Furthermore, turbulence research can promote innovation in wind power technology, driving the design of new wind turbine generators, the application of materials, and the development of intelligent control systems, ultimately improving the overall performance and reliability of wind power generation. Through in-depth analysis of the effects of turbulence, more efficient maximum power point tracking algorithms can be developed, reducing energy losses caused by wind speed fluctuations, increasing the power generation of wind turbine generators, reducing power generation costs, and enhancing the market competitiveness of wind power. Simultaneously, turbulence research also contributes to the scientific planning and site selection of wind farms, choosing areas with optimal wind resources and turbulence conditions to achieve both economic efficiency and environmental friendliness in wind power projects. At the power grid level, turbulence research helps to better predict and control the volatility of wind power generation, improve the stability and economy of the power grid, reduce dependence on traditional energy sources, and promote the transformation and upgrading of the energy structure.

[0060] Specifically, the data sources required for the model include equipment ledger data, equipment operation data, and external environment data. First, the locations required for this model need to be sorted out. After the location information is sorted out, the valid data needs to be verified, and the model is developed based on the valid data. Among them, equipment ledger data refers to a structured data set that records the static and dynamic information of all equipment in the site throughout its entire life cycle. It is the digital ID card and resume file of equipment management, including but not limited to equipment identification and basic information, technical parameters and performance data, as well as manufacturer and supplier information.

[0061] Data acquisition: Obtain wind speed and active power measurement data, and calculate the standard wind speed of the site; in particular, the standard wind speed of the site is obtained by averaging the wind speeds of each wind turbine.

[0062] Data aggregation: Converting data from different sources into a unified format and standard to ensure comparability and consistency between data.

[0063] Data cleaning involves using statistical methods or machine learning algorithms to detect outliers in the data and processing them appropriately, such as deleting, replacing, or interpolating. For missing values, appropriate methods are used to impute them based on the data's distribution and correlation, such as mean imputation, interpolation imputation, or model-based prediction imputation. The cleaned data is then validated to ensure accuracy and consistency. Any errors or inconsistencies found are corrected promptly.

[0064] Turbulence intensity calculation:

[0065] ;

[0066] Turbulence intensity is the ratio of the standard deviation of wind speed over a 10-minute period to the average wind speed over the same period, indicating the severity of turbulence. Among these, This represents the standard deviation of wind speed over a 10-minute period. This represents the average wind speed. Turbulence intensity reflects the characteristics of wind speed fluctuations.

[0067] Finally, based on the above judgment criteria, when the set value is met, the model will issue a warning about turbulent power reduction. The warning will be issued if one of the following conditions is met:

[0068] (1) When the wind direction or wind speed value changes frequently, if the power decreases compared to the past, that is, under the same working condition where the wind direction or wind speed value changes frequently, if the output power decreases by more than the set threshold, an early warning will be issued.

[0069] (2) When the turbulence intensity increases, does the power decrease? If the power decrease exceeds the set threshold, it is abnormal.

[0070] In this embodiment, the collaborative action of the data acquisition module, the online equipment status verification module, and the verification result output module forms a fully intelligent closed loop for monitoring the status of new energy power station equipment, resulting in significant comprehensive technical effects.

[0071] First, after completing the acquisition of raw data from the device, the data acquisition module preprocesses the data and dynamically retains accuracy based on data accuracy requirements analysis. It can automatically adjust the data accuracy and upload frequency according to the device's operating status, communication bandwidth, and model input requirements. This mechanism not only avoids low-quality and noisy data from entering the model, but also helps improve the effectiveness, consistency, and usability of the data, reduces communication pressure, and achieves an optimal balance between data quality and bandwidth resources, providing a higher-quality input foundation for subsequent digital twin models.

[0072] Secondly, the online equipment status verification module utilizes a digital twin model to automatically analyze and identify anomalies in equipment operation status. By integrating multi-source and cross-level data, a vertical data chain is constructed, enabling data flow from the field level, control level, and management level, thus achieving visualization, calculability, and traceability of equipment operation status. Simultaneously, the digital twin model can capture the correlation and consistency between equipment operating parameters in real time, effectively identifying potential faults, abnormal trends, and performance degradation issues, significantly improving the accuracy and timeliness of equipment status diagnosis.

[0073] Finally, the verification result output module performs consistency and integrity verification based on the vertical data chain, ensuring the reliability of data during transmission, acquisition, and fusion, thereby constructing a highly reliable equipment status feature vector. Through the status assessment model, it can output accurate equipment health status, risk level, and anomaly type, providing a reliable basis for operation and maintenance decisions. Furthermore, this module achieves deep integration of data credibility evaluation and equipment status output, further improving the reliability and interpretability of intelligent diagnosis.

[0074] This invention achieves end-to-end optimization of data quality improvement, model diagnosis enhancement, and result credibility assurance, forming a new energy power station equipment status monitoring and verification system with higher intelligence, stronger reliability, and higher scalability, which greatly improves operation and maintenance efficiency and equipment operation safety.

[0075] like Figure 5 The diagram shown illustrates the data accuracy requirement analysis process provided in this application embodiment. The specific logic is as follows: A bandwidth saving assessment is obtained through quantitative analysis of network communication indicators, and this assessment is compared with a preset abnormal event alarm threshold. If the bandwidth saving assessment is greater than the abnormal event alarm threshold, the bandwidth saving mode is automatically exited, and device data is uploaded at the set highest data upload frequency. If the bandwidth saving assessment is not greater than the abnormal event alarm threshold, bandwidth saving is divided, and corresponding data accuracy preservation measures are taken. Through the above process, a dynamic balance between data accuracy and bandwidth utilization is achieved, which helps improve the accuracy of device status verification.

[0076] Example 2: Conducting data accuracy requirement analysis to determine whether to take data accuracy preservation measures, the specific process is as follows:

[0077] The bandwidth saving assessment is obtained by quantitative analysis of network communication indicators and compared with the preset abnormal event alarm threshold. The network communication indicators include bandwidth utilization, retransmission rate, CPU utilization, and transmission queue backlog.

[0078] Specifically, bandwidth utilization is collected directly through SNMP integration in the industrial control SCADA system, retransmission rate is calculated through Performance Counter, CPU utilization can be read through PLC communication protocols (Modbus / TCP, S7 protocol, etc.), and the backlog of the transmission queue is obtained through OS network queue statistics.

[0079] It should be added that after the network communication indicators are normalized, they are multiplied by preset weights and then summed to obtain the bandwidth saving assessment. The preset weights include bandwidth utilization weight, retransmission rate weight, CPU utilization weight, and transmission queue backlog weight, and the sum of the four is 1. They are generally preset by professionals based on historical data and experience rules within a historical time period and stored in a preset database.

[0080] If the bandwidth saving assessment value is greater than the abnormal event alarm threshold, the system will switch to bandwidth control mode. Bandwidth control mode means that the system will automatically exit bandwidth saving mode and upload device data at the highest set data upload frequency.

[0081] If the bandwidth saving assessment is not greater than the abnormal event alarm threshold, then the bandwidth saving is classified and corresponding data accuracy preservation measures are taken.

[0082] Among them, the alarm threshold for abnormal events and the maximum data upload frequency are preset by professional personnel based on historical data and experience rules, and are stored in a preset database in advance.

[0083] It should be noted that the specific process for classifying bandwidth savings and taking corresponding data precision preservation measures is as follows:

[0084] The bandwidth saving assessment is compared with the preset bandwidth saving range, which includes a first bandwidth saving range, a second bandwidth saving range, and a third bandwidth saving range, with the bandwidth saving represented by each range increasing progressively. The bandwidth saving range is a range that is preset by professionals based on historical data and experience rules, and is stored in a preset database for later retrieval.

[0085] Based on the bandwidth saving assessment, if the data falls within the bandwidth saving range, corresponding data accuracy preservation measures are implemented. These measures include adjusting the average window size, adjusting the downsampling rate, and adjusting the reporting frequency.

[0086] By classifying bandwidth saving levels and implementing corresponding data accuracy preservation measures, the system achieves refined and intelligent management of network resources. Specifically, by comparing the bandwidth saving assessment with preset multi-level bandwidth saving intervals, the system can more accurately identify the current network pressure level, continuously characterizing bandwidth insufficiency from mild to severe. Based on different bandwidth insufficiency levels, the system invokes corresponding data accuracy preservation strategies, including average window size adjustment, downsampling rate adjustment, and reporting frequency adjustment, enabling the data upload strategy to adapt to changes in network status. This tiered control mechanism effectively reduces network load while ensuring that key data features are not lost, thereby maintaining the stability of the upload link. Simultaneously, it avoids data quality degradation due to excessive downsampling during mild bandwidth insufficiency and data backlog and delays caused by excessive upload pressure during severe insufficiency. This scheme achieves a dynamic balance between data accuracy and bandwidth utilization, improving the system's reliability, flexibility, and communication efficiency in complex network environments.

[0087] Specifically, the average window size adjustment means that the average window size is obtained by mapping the bandwidth saving assessment value to the bandwidth saving mapping table, and the device data is collected based on the average window size.

[0088] It should be noted that the bandwidth saving assessment values ​​are input into a pre-trained bandwidth saving mapping table to output the corresponding average window size, representing the impact of each bandwidth saving assessment value on the accuracy of device data acquisition. This mapping table is constructed using a logistic regression algorithm and trained using the scikit-learn framework based on the cross-entropy loss criterion. The training data includes bandwidth saving assessment values ​​obtained from historical periods and a preset average window size set by staff according to empirical rules, used to fit the mapping relationship between the bandwidth saving assessment values ​​and the average window size.

[0089] Downsampling rate adjustment means projecting the bandwidth saving assessment value onto the corresponding bandwidth saving range to obtain the corresponding downsampling rate adjustment value. After adjusting the downsampling rate using the downsampling rate adjustment value, device data is collected. This means multiplying the downsampling rate adjustment value by the downsampling rate.

[0090] Specifically, the bandwidth saving assessment is calculated by subtracting the minimum value of the corresponding bandwidth saving interval, and then the ratio is calculated by subtracting the difference between the maximum and minimum values ​​of the bandwidth saving interval to obtain the corresponding proportion, denoted as the bandwidth proportion. The bandwidth proportion is input into the downsampling rate adjustment sequence, and the corresponding downsampling rate adjustment value is output. The downsampling rate adjustment sequence is used to fit the mapping relationship between the bandwidth proportion and the downsampling rate adjustment value. The downsampling rate adjustment sequence is trained using an initial data sequence constructed by a linear regression algorithm and based on the least squares criterion and the statsmodels framework. The training data used are the bandwidth proportions obtained in historical time periods and the downsampling rate adjustment values ​​set according to empirical rules.

[0091] The reporting frequency adjustment means calculating the deviation rate between the bandwidth saving assessment value and the bandwidth saving range to obtain the reporting frequency deviation value. By comparing the reporting frequency deviation value, the corresponding reporting frequency optimization value is obtained. The reporting frequency is then optimized based on the reporting frequency optimization value to obtain the optimized reporting frequency. This involves multiplying the optimized reporting frequency value with the reported frequency and collecting device data based on the optimized reporting frequency.

[0092] It should be added that, after calculating the difference between the bandwidth saving assessment value and the minimum value of the bandwidth saving range, the ratio is calculated with the maximum value and the difference between the maximum values ​​of the bandwidth saving range to obtain the reporting frequency deviation value. The reporting frequency deviation value is then input into the trained reporting frequency mapping set, and the corresponding reporting frequency optimization value is obtained by comparison. The reporting frequency mapping set is used to reflect the correlation between the reporting frequency deviation value and the reporting frequency optimization value. Furthermore, the reporting frequency deviation values ​​within the historical time period, as well as the preset reporting frequency optimization value set by professionals based on experience rules, are input into the initial dataset constructed by the logistic regression algorithm. The dataset is then trained using the least squares criterion and the statsmodels framework to obtain the corresponding reporting frequency mapping set.

[0093] In this embodiment, by quantitatively analyzing network communication indicators, the system intelligently assesses the degree of bandwidth saving and dynamically adjusts it according to preset abnormal event alarm thresholds, thereby optimizing the device data upload strategy and improving system stability and efficiency. Real-time monitoring and quantitative analysis of network communication indicators allow for a more accurate assessment of the bandwidth saving effect. When the bandwidth saving exceeds the set threshold, the system automatically switches to bandwidth control mode, exits bandwidth saving mode, and restores the highest upload frequency, ensuring that critical device data can be uploaded more promptly and completely, avoiding data loss or delay due to insufficient bandwidth. This mechanism not only improves the reliability of data transmission but also ensures the continuity of device status monitoring. Secondly, when the bandwidth saving does not reach the abnormal threshold, the system performs refined management based on bandwidth conditions, adjusting data upload accuracy through precision preservation measures. This optimizes bandwidth resource utilization while ensuring data quality, reducing bandwidth pressure and lowering the overall system load. This solution, through automated and intelligent bandwidth control and data accuracy management, improves the efficiency and accuracy of device data upload while ensuring reasonable bandwidth utilization, contributing to enhanced system scalability and anti-interference capabilities.

[0094] like Figure 6 The diagram shows a flowchart of the verification data accuracy requirement assessment provided in this application embodiment. The specific logic is as follows: If both the AUC value and the F1 value are greater than the corresponding model accuracy judgment value, it indicates that the performance accuracy of the online verification model is qualified, and no data accuracy improvement scheme is implemented; if neither the AUC value nor the F1 value is greater than the corresponding model accuracy judgment value, it indicates that the performance accuracy of the online verification model is unqualified, and a data accuracy improvement scheme is implemented; if either the AUC value or the F1 value is greater than the corresponding model accuracy judgment value, the model accuracy value that is not greater than the model accuracy judgment value is differentially quantified to obtain the corresponding data accuracy adjustment amount, and a bandwidth saving control ratio is obtained by mapping based on the data accuracy adjustment amount, and the degree of data accuracy retention measures is adjusted based on the bandwidth saving control ratio; through the above process, it is beneficial to improve the reliability, stability and accuracy of the digital twin model, thereby ensuring the accuracy of online verification of the device status.

[0095] Example 3: Identifying abnormal situations, followed by assessing the required data accuracy to determine whether to implement a data accuracy improvement plan. The specific process is as follows:

[0096] After the digital twin model performs online verification of the device status, the model accuracy value corresponding to the model is read for judgment. The model accuracy value includes the AUC value and the F1 value.

[0097] It should be explained that the AUC value is calculated from the ROC curve and obtained using roc_auc_score or the framework's built-in AUC function, while the F1 value is calculated using the f1_score function.

[0098] If both the AUC and F1 values ​​are greater than the corresponding model accuracy judgment values, it means that the performance accuracy of the online verification model is qualified, and no data accuracy improvement scheme will be implemented. The model accuracy judgment values ​​include the AUC value judgment value and the F1 value judgment value, which are preset values ​​set by professionals in advance.

[0099] If the AUC and F1 values ​​are not greater than the corresponding model accuracy judgment values, it means that the performance accuracy of the online verification model is unqualified, and a data accuracy improvement plan should be implemented.

[0100] If the AUC or F1 value is greater than the corresponding model accuracy judgment value, then the model accuracy values ​​that are not greater than the model accuracy judgment value are differentially quantified to obtain the corresponding data accuracy adjustment amount. Based on the data accuracy adjustment amount, the bandwidth saving control ratio is obtained by mapping. Based on the bandwidth saving control ratio, the degree of data accuracy preservation measures is adjusted.

[0101] Specifically, the absolute value of the difference between the model accuracy value (not greater than the model accuracy judgment value) and the corresponding model accuracy judgment value is taken to obtain the data accuracy adjustment amount. The data accuracy adjustment amount is input into the data accuracy adjustment queue after training is completed, and the corresponding bandwidth saving control ratio is obtained by comparison. The data accuracy adjustment queue is used to reflect the correlation between the data accuracy adjustment amount and the bandwidth saving control ratio. The data accuracy adjustment amount in the historical time period and the bandwidth saving control ratio set by professionals based on experience rules are input into the initial data queue constructed by the logistic regression algorithm. The data is trained based on the least squares criterion and the statsmodels framework to obtain the corresponding data accuracy adjustment queue.

[0102] In this embodiment, after anomaly identification, a model accuracy evaluation and data accuracy improvement mechanism is further introduced, so that the online verification process of equipment status forms a closed-loop system of "anomaly early warning - model self-check - accuracy adaptive adjustment", which is conducive to improving the reliability, stability and long-term sustainable operation capability of the digital twin model.

[0103] By synchronously reading the model accuracy value after the digital twin model completes online verification, real-time self-checking of model performance is achieved. AUC and F1, as core indicators for evaluating classification models, can simultaneously reflect the model's overall discriminative ability and the balance between positive and negative class recognition. The system compares these two indicators with the set model accuracy judgment value, which can more accurately determine whether the current model is in a high-confidence state. If both indicators meet the requirements, there is no need to implement a data accuracy improvement scheme, ensuring that the system maintains efficient and stable operation when performance is good, and avoiding unnecessary resource consumption.

[0104] Secondly, if neither AUC nor F1 reaches the corresponding judgment value, it indicates that the model accuracy is showing a general trend of degradation. At this time, the system automatically executes the data accuracy improvement scheme. By improving the data acquisition accuracy and upload quality, such as increasing the sampling rate, reducing the average window, and increasing the reporting frequency, it is beneficial to enhance the quality of the model input data, enabling the model to regain high-quality training samples and promote its performance recovery. This mechanism ensures that the model can quickly obtain external data support when it degrades, thus achieving proactive repair of the verification accuracy.

[0105] Furthermore, when only AUC or F1 meets the standard, the system does not simply judge it as qualified or unqualified. Instead, it performs differential quantification on the indicators that exceed the judgment value to obtain a precise data precision adjustment amount. This adjustment amount is then mapped to the corresponding bandwidth saving control ratio, thereby enabling fine-grained dynamic control of data precision preservation measures. This process achieves deep linkage between model performance and network resource scheduling: the worse the model performance, the more the system tends to improve data quality; the better the model performance, the more it allows for an appropriate reduction in upload precision to save bandwidth. Through differential quantification and mapping calculation, the control process is both reasonable and continuous, avoiding abrupt "black and white" strategies and better conforming to the dynamic characteristics of the actual operating environment.

[0106] In summary, this solution achieves intelligent integration of model performance monitoring, data accuracy control, and bandwidth resource management, enabling the system to automatically optimize data acquisition strategies based on real-time changes in model accuracy. This ensures the long-term high-precision operation of the digital twin model. This process not only improves the accuracy and reliability of equipment status verification but also achieves efficient utilization of bandwidth resources, providing a smart operation and maintenance technology system for new energy power plants with self-diagnosis, self-repair, and autonomous control capabilities.

[0107] The specific plan for improving data accuracy is as follows:

[0108] The AUC value and F1 value were processed to identify the differences between the AUC value and the corresponding model accuracy judgment value. Then, the data was normalized and the mean was calculated to obtain the representative value of the data accuracy improvement.

[0109] It should be explained that data difference processing means that the AUC value and F1 value are respectively calculated with the corresponding model accuracy judgment value, the absolute value is taken, and then the data is normalized and the mean is calculated to obtain the representative value of the data accuracy improvement.

[0110] The compensation calculation is performed based on the representative value of the data accuracy improvement and the preset maximum model accuracy value. That is, the representative value of the data accuracy improvement and the maximum model accuracy value are multiplied to obtain the corresponding data accuracy improvement range. The maximum model accuracy value is preset and set in advance by professionals based on historical data and experience rules.

[0111] The degree of adjustment of data accuracy preservation measures is based on the magnitude of the improvement in data accuracy. The degree of adjustment means multiplying the average window size, downsampling rate adjustment value, and reporting frequency optimization value in the data accuracy preservation measures based on the magnitude of the improvement in data accuracy.

[0112] By processing, normalizing, and calculating the mean of the differences between AUC and F1 scores and the model accuracy judgment values, this approach achieves fine-grained control over model accuracy. Through difference processing and data normalization, the scheme quantifies the shortcomings of model performance, thereby generating a representative value for data accuracy improvement. Then, based on this representative value and the preset maximum model accuracy value, a compensation calculation is performed to calculate a more accurate data accuracy improvement range. According to this improvement range, the data accuracy retention measures are dynamically adjusted, allowing data acquisition accuracy to be optimized and adjusted while ensuring model recovery performance. This mechanism not only ensures that data quality can be improved in a timely manner when model accuracy is insufficient, avoiding excessive improvement and unnecessary resource consumption through fine-grained control, but also avoids unnecessary data redundancy when model accuracy is too high. Overall, through precise quantification and control, this scheme improves data acquisition efficiency and the adaptive capability of model performance on the one hand, and enhances the system's intelligence level and resource utilization on the other.

[0113] The various features and processes described above can be used independently of each other or can be combined in various ways. All possible combinations and sub-combinations are intended to fall within the scope of this disclosure. Furthermore, certain method or process blocks may be omitted in some embodiments. The methods and processes described herein are not limited to any particular order, and the blocks or states associated with them may be performed in other suitable orders. For example, the described blocks or states may be performed in an order different from the order specifically disclosed, or multiple blocks or states may be combined in a single block or state. Example blocks or states may be performed serially, in parallel, or in some other manner. Blocks or states may be added to or removed from the disclosed example embodiments. The exemplary systems and components described herein may be configured differently from those described. For example, elements may be added to, removed from, or rearranged compared to the disclosed example embodiments.

[0114] The various operations of the example methods described herein can be performed at least in part by an algorithm. This algorithm can be contained in program code or instructions stored in memory (e.g., the aforementioned non-transitory computer-readable storage medium). Such an algorithm may include a machine learning algorithm. In some embodiments, the machine learning algorithm may not be explicitly programmed into the computer to perform the function, but can learn from training data to create a predictive model that performs the function.

[0115] The various operations of the example methods described herein can be performed, at least in part, by one or more processors that are temporarily configured (e.g., by software) or permanently configured to perform the relevant operations. Whether temporarily or permanently configured, such processors can constitute the engine of a processor implementation that operates to perform one or more of the operations or functions described herein.

[0116] Similarly, the methods described herein can be implemented at least in part by a processor, where one or more specific processors are examples of hardware. For example, at least some operations of a method can be performed by one or more processors or an engine implemented by a processor. Furthermore, one or more processors can also be operated to support the performance of related operations in a “cloud computing” environment or as “Software as a Service” (SaaS). For example, at least some operations can be performed by a set of computers (as an example of a machine including processors), where these operations are accessible via a network (e.g., the Internet) and via one or more suitable interfaces (e.g., application programming interfaces (APIs)).

[0117] The performance of certain operations can be distributed across processors, residing not only within a single machine but also deployed across multiple machines. In some example embodiments, the processor or processor-implemented engine may reside in a single geographic location (e.g., within a home environment, office environment, or server cluster). In other example embodiments, the processor or processor-implemented engine may be distributed across multiple geographic locations.

[0118] In this specification, multiple instances may implement components, operations, or structures described as single instances. Although individual operations of one or more methods are shown and described as separate operations, one or more of the separate operations may be performed simultaneously and do not need to be performed in the order shown. Structures and functions presented as separate components in the example configuration may be implemented as composite structures or components. Similarly, structures and functions presented as single components may be implemented as separate components. These and other variations, modifications, additions, and improvements fall within the scope of this document.

[0119] While an overview of the subject matter has been described with reference to specific example embodiments, various modifications and changes can be made to these embodiments without departing from the broader scope of embodiments of this disclosure. Such embodiments of the subject matter are referred to herein, individually or collectively, by the term "invention," and are used for convenience only and are not intended to limit the scope of this application to any single disclosure or concept, should more than one disclosure or concept be disclosed in fact.

[0120] The embodiments described herein have been described in sufficient detail to enable those skilled in the art to practice the disclosed teachings. Other embodiments may be used and derived therefrom, such that structural and logical substitutions and changes may be made without departing from the scope of this disclosure. Therefore, the detailed description should not be construed as limiting, and the scope of the various embodiments is defined only by the appended claims and the full scope of their equivalents.

Claims

1. A new energy station equipment state online checking system based on longitudinal data chain, characterized in that, The method comprises the following steps: The new energy station equipment data acquisition module is used for collecting data of equipment in the new energy station to obtain corresponding equipment data, and after data preprocessing of the equipment data, data precision requirement analysis is performed to determine whether to take data precision retention measures; The equipment state online checking module is used for inputting the collected equipment data into a preset digital twin model for automatically analyzing the equipment state and identifying abnormal conditions, forming a longitudinal data chain; The checking result output module is used for completing consistency checking and integrity checking of cross-level and multi-source data based on the longitudinal data chain, outputting the equipment state through a state evaluation model, and constructing an equipment state feature vector through the checked longitudinal data chain to obtain the equipment state and output the corresponding checking result. The state evaluation model includes but is not limited to a gear box oil temperature abnormality early warning model, a direct current combiner box failure early warning model, and a photovoltaic module dust covering low efficiency early warning model; 2. The new energy station equipment state online checking system based on longitudinal data chain of claim 1, wherein: The design idea of the gear box oil temperature abnormality early warning model includes horizontal comparison and longitudinal comparison; The horizontal comparison means that different working conditions are represented as a reference by dividing intervals according to humidity, power and gear box speed, and if the difference between the unit gear box oil temperature data and the average oil temperature under non-full load exceeds a preset first proportion, an alarm is given, and if the difference between the unit gear box oil temperature data and the average oil temperature under full load exceeds a preset second proportion, an alarm is given, the average oil temperature of the gear box at the same time in the new energy station, the preset first proportion is higher than the preset second proportion, which means that the alarm emergency degree under non-full load condition is lower than that under full load condition. The longitudinal comparison means comparing the temperature rise rate with the unit power, and if the temperature rise coefficient exceeds the temperature rise threshold, an early warning is given, the temperature rise coefficient represents the temperature rise of the gear temperature caused by each unit power increase; 3. The new energy station equipment state online checking system based on longitudinal data chain of claim 2, wherein: The oil temperature is compared with the gear box inlet temperature and the deviation rate is calculated, and if the deviation rate exceeds the oil temperature deviation proportion, an early warning is given; The oil temperature is compared with the cabin environment temperature and the deviation rate of the oil temperature and the cabin cabinet temperature is calculated, and if the deviation rate exceeds the preset cabin environment temperature proportion, an early warning is given; The gear box oil temperature curve is obtained by analyzing the historical data, which is a curve that comprehensively describes the corresponding curves of environmental temperature, humidity, power, gear box speed, gear box inlet temperature and cabin environment temperature; If the current running gear box oil temperature exceeds the set proportion of the fitted curve, an early warning is given. The specific content of the direct current combiner box failure early warning model is as follows:

4. The new energy station equipment state online checking system based on longitudinal data chain of claim 2, wherein: First, calculate the total current of each group string and compare it with the total output current to obtain the corresponding current deviation, if the current deviation is greater than the set current threshold, it means that the direct current combiner box is abnormal; Second, calculate the total power of each group string and compare it with the total output power to obtain the corresponding power deviation, if the power deviation is greater than the set power threshold, it means that the direct current combiner box is abnormal; ​ In the third step, it is judged whether the output total current and the output total power appear constant value or zero value, and the duration is judged, if the constant value or zero value appears and the duration exceeds the set duration, it indicates that the DC junction box is abnormal.

5. The new energy station equipment state online checking system based on longitudinal data chain of claim 2, wherein: The specific content of the dust covering low efficiency early warning model of the photovoltaic module is as follows: The direct radiation is divided into different intervals to represent different working conditions, and the same working condition is taken as the benchmark, under the condition of the same direct radiation, the wind speed is greater than the set wind speed value, the environment temperature is greater than the set environment temperature, and the humidity is greater than the set humidity, and the deviation value between the single branch current reduction and the average of all branch currents is compared, if the single branch current is lower than the preset proportion of the average value, the early warning is carried out, otherwise no additional processing is carried out.

6. The new energy station equipment state online checking system based on longitudinal data chain of claim 1, wherein: The specific process of determining whether to take data precision retention measures by analyzing data precision requirements is as follows: The bandwidth saving degree evaluation quantity is obtained by quantitative analysis of network communication indicators, and is compared with the preset abnormal event alarm threshold, the network communication indicators include bandwidth utilization rate, retransmission rate, CPU occupancy rate and sending queue backlog quantity; If the bandwidth saving degree evaluation quantity is greater than the abnormal event alarm threshold, switch to the bandwidth regulation mode, the bandwidth regulation mode means automatically exiting the bandwidth saving mode, and uploading device data with the set highest data upload frequency; If the bandwidth saving degree evaluation quantity is not greater than the abnormal event alarm threshold, the bandwidth saving degree is divided and the corresponding data precision retention measures are taken.

7. The new energy station equipment state online checking system based on longitudinal data chain of claim 6, wherein: The specific process of dividing the bandwidth saving degree and taking the corresponding data precision retention measures is as follows: The bandwidth saving degree evaluation quantity is compared with the preset bandwidth saving degree interval, the bandwidth saving degree interval includes bandwidth first saving interval, bandwidth second saving interval and bandwidth third saving interval, and the bandwidth saving degree represented by them increases gradually; Based on the bandwidth saving degree evaluation quantity belonging to the bandwidth saving degree interval, the corresponding data precision retention measures are taken, the data precision retention measures include average window quantity adjustment, down sampling rate adjustment and reporting frequency adjustment.

8. The new energy station equipment state online checking system based on longitudinal data chain of claim 7, wherein: The average window quantity adjustment means that the corresponding average window quantity is obtained by mapping the bandwidth saving degree evaluation quantity in the bandwidth saving mapping table, and device data is collected according to the average window quantity; The down sampling rate adjustment means that the corresponding down sampling rate adjustment degree value is obtained by projecting the proportion of the bandwidth saving degree evaluation quantity in the corresponding bandwidth saving degree interval, and device data is collected after the down sampling rate is adjusted by the down sampling rate adjustment degree value; The reporting frequency adjustment means that the reporting frequency deviation value is obtained by calculating the deviation rate of the bandwidth saving degree evaluation quantity and the bandwidth saving degree interval, the corresponding reporting frequency optimization quantity is obtained by comparing the reporting frequency deviation value, the optimized reporting frequency is obtained by optimizing the reporting frequency according to the reporting frequency optimization quantity, and the device data is collected based on the optimized reporting frequency.

9. The new energy station equipment state online checking system based on longitudinal data chain of claim 1, wherein: The specific process of determining whether to execute the data precision improvement scheme by evaluating the data precision requirement after identifying the abnormal condition is as follows: After the device state online checking is performed on the digital twin model, a model precision value corresponding to the model is read for judgment, the model precision value including an AUC value and an F1 value; If the AUC value and the F1 value are both greater than corresponding model precision judgment values, it indicates that the performance precision of the online checking model is qualified, and no data precision improvement scheme is performed; If the AUC value and the F1 value are both not greater than the corresponding model precision judgment values, it indicates that the performance precision of the online checking model is unqualified, and the data precision improvement scheme is performed; If the AUC value or the F1 value is greater than the corresponding model precision judgment value, a data precision adjustment amount is obtained by difference quantization of the model precision value not greater than the model precision judgment value, a bandwidth saving control proportion is obtained by mapping based on the data precision adjustment amount, and the data precision retention measure is degree-regulated based on the bandwidth saving control proportion.

10. The new energy station equipment state online checking system based on longitudinal data chain of claim 9, wherein: The data precision improvement scheme is specifically as follows: Data difference processing is performed on the AUC value and the F1 value and the corresponding model precision judgment values, and then data normalization and mean value operation are sequentially performed to obtain data precision improvement representative values; Compensation operation is performed on the data precision improvement representative values and a preset maximum model precision value to obtain a corresponding data precision improvement amplitude; The data precision retention measure is degree-regulated based on the data precision improvement amplitude.

Images