Sensor-based test system and method for grid performance of a wind turbine generator

Abstract

This invention discloses a sensor-based test system and method for assessing the grid connection performance of wind turbine generators, relating to the field of wind power generation testing technology. The system includes: a sensor module for real-time acquisition of grid connection performance-related parameters of the wind turbine generator through preset sensors; preprocessing the acquired parameters to remove redundant and abnormal interference data; and integrating the preprocessed parameters into a standardized dataset. A simulation module is used to call the standardized dataset output by the sensor module and construct a predictive model of the wind turbine generator's operating conditions and the grid's operational status based on the parameter information within the dataset. This invention addresses the pain point of insufficient coverage in traditional test scenarios, achieving quantitative and accurate evaluation of grid connection performance and outputting standardized indicators; reducing test costs and risks, shortening the cycle, and providing comprehensive technical support for wind turbine R&D, grid connection testing, wind farm operation and maintenance, and grid planning.

Description

Technical Field

[0001] This invention relates to the field of wind power generation testing technology, specifically to a sensor-based test system and method for the grid connection performance of wind turbine generator sets. Background Technology

[0002] With the rapid development of the new energy industry, wind turbines, as core equipment for clean energy power generation, are seeing their grid-connected operation expand continuously, and their proportion in the power system is constantly increasing. They have become an important component in ensuring the safe, stable, and efficient operation of the power system. The grid connection performance of wind turbines directly determines their compatibility with the power grid, encompassing multiple key dimensions such as grid connection stability, fault ride-through capability, power regulation accuracy, and harmonic mitigation effectiveness. Their performance not only affects the operational reliability and power generation efficiency of the wind turbine itself, but can also significantly impact the voltage quality and frequency stability of the entire power grid, and may even trigger grid fluctuations, grid disconnection, and other safety incidents. Therefore, conducting precise and comprehensive testing of the grid connection performance of wind turbines has significant engineering importance and practical value.

[0003] Currently, existing testing methods for the grid connection performance of wind turbine generators mainly fall into two categories: on-site real-machine testing and laboratory simulation testing. On-site real-machine testing requires conducting tests in actual grid-connected operation scenarios of wind turbine generators. It involves deploying a small number of sensors to collect some grid connection parameters to complete performance evaluation. However, this method has several limitations: Firstly, on-site operating conditions are greatly affected by the natural environment (such as wind speed, wind direction, and air pressure), exhibiting strong randomness and making it difficult to accurately capture extreme conditions (such as typhoons, strong gusts, and sudden voltage spikes and drops in the grid). This results in missing grid connection performance data under extreme conditions, failing to fully reflect the generator's grid connection adaptability under complex operating conditions. Secondly, the grid connection parameters collected on-site are easily affected by environmental interference, line losses, and other factors, resulting in high data redundancy, numerous outliers, and a lack of standardized preprocessing procedures. This leads to insufficient accuracy and effectiveness of the collected data, resulting in significant deviations in subsequent performance evaluation results. To address these issues, we propose a sensor-based testing system and method for the grid connection performance of wind turbine generators. Summary of the Invention

[0004] To address the aforementioned technical problems, this technical solution provides a sensor-based test system and method for assessing the grid connection performance of wind turbine generator sets.

[0005] To achieve the above objectives, the technical solution adopted by this invention is: a sensor-based test system for the grid connection performance of wind turbine generator sets, comprising: The sensor module is used to collect grid-connection performance-related parameters of wind turbine generators in real time through preset sensors, preprocess the collected grid-connection performance-related parameters, remove redundant data and abnormal interference data, and integrate the preprocessed qualified parameters to form a standardized dataset. The simulation module is used to call the standardized dataset output by the sensor module, construct a prediction model of wind turbine generator operating conditions and grid operation status based on the parameter information in the dataset, introduce a virtual operating condition interpolation algorithm to supplement the extreme operating condition samples in the dataset, fill the technical gap of scarce extreme operating condition samples in the actual collection process, and then simulate the actual operating conditions of wind turbine generators and grid operation status through the prediction model to obtain the normal grid-connected operation status, fault operation status of wind turbine generators, and the corresponding grid-connected parameters of wind turbine generators under the two different operation statuses. The evaluation module is used to receive the grid connection parameters of the wind turbine generator set under different operating conditions output by the simulation module, perform quantitative calculation and systematic analysis on the grid connection parameters, adopt a hierarchical weighted dynamic evaluation system to comprehensively calculate the grid connection performance value of the wind turbine generator set, and output standardized grid connection performance evaluation indicators. The visualization module is used to receive the grid connection performance evaluation indicators output by the evaluation module, and to intuitively display the grid connection performance evaluation indicators through a preset visualization presentation method, so as to realize the visualization output of the grid connection performance evaluation results of the wind turbine generator set.

[0006] Preferably, the sensor module has pre-set sensors that are placed at key locations on the wind turbine generator under test to collect the mechanical, electrical and environmental parameters required for the grid connection performance test. The sensors include intelligent speed and torque sensors, intelligent vibration sensors, current sensors, and temperature sensors. The intelligent speed and torque sensors are installed at the output end of the wind turbine gearbox and on the main shaft to collect real-time speed and torque parameters of the main shaft. Intelligent vibration sensors are installed in the gearbox, servo motor and nacelle of the wind turbine to collect vibration acceleration and vibration frequency parameters during wind turbine operation. Current sensors are installed at the output and grid-connected input terminals of the wind turbine generator set to collect three-phase voltage, three-phase current, and power parameters. Temperature sensors are installed on the gearbox, servo motor, and electronic rotor of the wind turbine generator set to collect temperature parameters during operation.

[0007] Preferably, the preprocessing steps within the sensor module are as follows: By integrating parameters from multiple sources to construct the original dataset, the source can be traced, and the parameter units, data types, and dimension fields can be standardized to achieve format standardization. Using the acquisition time, acquisition node, and parameter type as the sole criterion, completely duplicate records are identified and deleted; for near-duplicate data with acquisition time deviations within the acquisition period and identical parameter values, one valid record is retained, and the remaining duplicates are deleted. Delete data with parameter values ​​that are empty, all zeros, or have no actual business meaning; For highly relevant network performance parameters, retain the core analysis parameters; The preprocessed parameters are integrated to obtain a standardized dataset.

[0008] Preferably, the prediction model within the simulation module is built based on a standardized dataset, extracting core information on the wind turbine generator operating parameters and power grid operating parameters from the dataset, using the wind turbine generator operating conditions and power grid operating status as prediction targets, and constructing a prediction model for the wind turbine generator operating conditions and power grid operating status using deep learning algorithms; A virtual working condition interpolation algorithm is introduced to interpolate existing conventional working condition samples in the standardized dataset, generate virtual extreme working condition samples that conform to actual physical laws, and supplement them into the dataset to fill the technical gap of scarce extreme working condition samples. The complete dataset, supplemented with extreme working condition samples, is re-input into the prediction model for training and optimization. The model parameters are then adjusted to obtain the prediction model.

[0009] Preferably, the prediction model is initially built from a standardized dataset. Extracting core parameters ,by For the purpose of prediction; in , is a set of operating parameters for wind turbine generator sets, including generator speed and nacelle azimuth angle, where m is the number of generator parameters; This is a set of power grid operating parameters, including grid voltage, frequency, power factor, and line losses. This represents the total number of core parameters. ; Wind turbine generator set operating conditions; The power grid operating status is uniformly denoted as: ; The initial prediction model is constructed using deep learning algorithms, and the formula is expressed as follows: in A deep learning model consists of an input layer, hidden layers, and an output layer. The initial parameters of the model, including weights and biases, are used by the model to fit... and The mapping relationship is used to complete the initial construction; A virtual operating condition interpolation algorithm is introduced, based on a normal operating condition sample set. Interpolation calculations are performed to generate virtual extreme working condition samples that conform to actual physical laws. The interpolation calculation formula is expressed as: in For the first Core parameter values ​​of a virtual extreme working condition sample; For the sample set of normal working conditions Two adjacent normal operating condition parameter values, ; These are the interpolation coefficients. This ensures that the generated samples are extreme values ​​and conform to the physical operating limits of the wind turbine generator and the power grid; This is a virtual operating condition interpolation algorithm; The generated virtual extreme condition samples are added to the original standardized dataset to obtain the complete dataset, expressed as: ;in This is the supplementary standardized dataset; Complete dataset Core parameters in and their corresponding prediction targets Re-input to the initial prediction model In this process, the model parameters are adjusted by minimizing the prediction error. The optimized model parameters are obtained. With the final prediction model The final prediction model expression is: .

[0010] Preferably, the steps for obtaining grid connection parameters of wind turbine generators within the simulation module are as follows: The optimized prediction model is launched to simulate the actual operating scenarios of wind turbine generators and simultaneously simulate the grid operation status under the corresponding scenarios. During the simulation, the model will combine various parameter features in the dataset to restore the operating logic of the unit and the response state of the power grid under different operating conditions, and to simulate the operating state of the unit and the power grid in all aspects. After the simulation is completed, various types of data generated during the simulation are automatically collected. The data of the two core scenarios of normal grid connection and fault operation of the wind turbine generator are distinguished and extracted. The grid connection parameters of the wind turbine generator under the two types of operation are organized accordingly. The parameters include the grid-connected power of the generator, reactive power compensation, voltage deviation, and parameter mutation values ​​when the fault is triggered. The collected parameters are classified and summarized to output complete simulation result data.

[0011] Preferably, the specific steps for outputting standardized network performance evaluation indicators within the evaluation module are as follows: Receive grid connection parameters of wind turbine generator sets, form an initial evaluation parameter set, and perform verification; Standardize the network-related parameters that have passed the verification process, and convert the network-related parameters into quantitative values ​​of a uniform magnitude; For each grid-related parameter, targeted quantitative calculations are performed to obtain the quantitative index of each individual parameter, clarifying the basic impact of each parameter on the grid-related performance of the unit; Based on individual parameter indicators, a multi-dimensional systematic analysis of grid-related parameters is conducted. A hierarchical weighted dynamic evaluation system is adopted to comprehensively calculate the grid-related performance value of the unit. By weighted summation, the quantitative indicators of each individual parameter are integrated into the comprehensive grid-related performance value of the unit, thus completing the comprehensive calculation. The comprehensive network performance values ​​and quantitative indicators of each individual parameter obtained from the comprehensive calculation are standardized, and the quantitative data and analysis conclusions in the evaluation process are organized to form a complete evaluation report.

[0012] Preferably, the weighted summation method integrates the quantitative indicators of each individual network-related parameter into a comprehensive network-related performance value F of the unit, and its calculation formula is as follows: in It is a single quantitative indicator, obtained through targeted quantitative calculation; The dynamic weighting reflects the core of "hierarchical weighted dynamic evaluation," and is set based on the results of multi-dimensional systematic analysis. The dynamic adjustment logic is as follows: Under normal operating conditions, the focus is on increasing the grid-connected power of the generating units. Reactive power compensation The weights; In fault operation mode, the focus is on improving voltage deviation. Sudden changes in fault parameters The weights; As a comprehensive performance value, the calculation results can be directly used for subsequent standardization processing and transformed into standardized network performance evaluation indicators; , which is the serial number of the grid-related single parameter, corresponding to 4 core grid-related parameters, including the unit's grid-connected power, reactive power compensation, voltage deviation and fault parameter sudden change value; For the first Quantitative indicators of network-related parameters; For the first The weights of the quantitative indicators of network-related parameters; The value is 4, corresponding to 4 network-related parameters.

[0013] Preferably, the visualization module is based on preset visualization rules and diverse display templates, and uses a layered and categorized visualization presentation method; the display methods include bar charts, heat maps and line charts.

[0014] The test method for the grid connection performance of sensor-based wind turbine generators includes the following steps: S1. The sensor module collects relevant raw parameters of the network through preset sensors, and after preprocessing to remove redundant and abnormal data, it forms a standardized dataset and outputs it. S2. The simulation module calls the dataset to build a prediction model of unit operating conditions and grid status, introduces a virtual operating condition interpolation algorithm to supplement extreme operating condition samples, and obtains relevant parameters of the unit's grid connection under normal and fault states through model simulation. S3. The evaluation module receives the simulation parameters, performs quantitative calculations and systematic analysis, and then uses a hierarchical weighted dynamic evaluation system to calculate the comprehensive performance value and output standardized evaluation indicators. S4. The visualization module receives evaluation metrics and presents them intuitively through preset visualization methods. Compared with the prior art, the beneficial effects of the present invention are as follows: This invention significantly improves the scientific rigor and practicality of wind turbine grid connection performance testing through a closed-loop design throughout the entire process. The sensor module ensures accurate and standardized data collection, laying a solid foundation for subsequent testing. The simulation module fills the gap in extreme operating condition samples, comprehensively reproducing normal and fault operating conditions of the unit, solving the pain point of insufficient coverage of traditional test scenarios, achieving quantitative and accurate evaluation of grid connection performance, and outputting standardized indicators. It reduces test costs and risks, shortens the cycle, and provides comprehensive technical support for wind turbine R&D, grid connection testing, wind farm operation and maintenance, and grid planning, combining theoretical innovation with engineering practicality. Attached Figure Description

[0015] Figure 1 This is a system framework diagram of the present invention; Figure 2 This is a flowchart of the steps of the present invention. Detailed Implementation

[0016] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0017] Reference Figure 1As shown in this embodiment, the sensor-based wind turbine grid connection performance test system mainly includes a sensor module, a simulation module, an evaluation module, and a visualization module. The modules establish communication connections via industrial Ethernet and use the TCP / IP protocol to achieve real-time data transmission and interaction, ensuring stable and coordinated operation of all modules. Specific implementation details are as follows: Sensor module As the core of data acquisition for the entire test system, the sensor module is equipped with high-precision industrial sensors adapted to the grid connection performance testing of wind turbine generator sets. These sensors include: voltage sensor (measurement range 0-35kV, accuracy ±0.2%FS), current sensor (measurement range 0-2000A, accuracy ±0.2%FS), power sensor (measurement range 0-5MW, accuracy ±0.1%FS), wind speed sensor (measurement range 0-70m / s, accuracy ±0.3m / s), speed sensor (measurement range 0-3000r / min, accuracy ±1r / min), and fault detection sensor. Each sensor is installed corresponding to the grid connection interface, nacelle, hub, and grid connection point of the wind turbine generator set to achieve comprehensive acquisition of grid connection performance-related parameters. During parameter acquisition, the sensor module collects the aforementioned grid performance-related parameters in real time at a sampling frequency of 100Hz. The collected parameters include, but are not limited to: grid voltage, grid current, active power, reactive power, wind turbine speed, wind speed, converter output parameters, and grid frequency. After acquisition, the parameters are preprocessed. The preprocessing process is as follows: first, a moving average filtering algorithm is used to remove random interference noise from the parameters; then, the 3σ criterion is used to identify and remove outlier data (i.e., data exceeding the mean ± 3 times the standard deviation); finally, data normalization is performed (mapping the parameters to the [0,1] interval), redundant and duplicate data are removed, and all preprocessed parameters are integrated according to a preset format (XML format) to form a standardized dataset, which is stored in the module's built-in cache unit for use by the simulation module.

[0018] Simulation module The simulation module uses an industrial-grade server (CPU: Intel Xeon E5-2690, memory: 32GB, hard disk: 1TB SSD) and has built-in preset model building and operating condition simulation software. Its specific working process is as follows: First, it calls the standardized dataset in the sensor module's cache unit through the communication interface, divides the dataset into a training set and a test set in a 7:3 ratio; based on the parameter information in the training set, it uses the LSTM neural network algorithm to build a prediction model of the wind turbine generator's operating condition and the power grid's operating status. The model input is the collected grid-related parameters, and the output is the wind turbine generator's operating condition and the power grid's operating status. To address the scarcity of extreme operating condition samples during actual data collection, a virtual operating condition interpolation algorithm (using cubic spline interpolation) is introduced to supplement the extreme operating condition samples in the dataset. The extreme operating conditions specifically include: sudden voltage drop (dropping to 60% of the rated voltage), sudden voltage rise (rising to 120% of the rated voltage), extreme wind speed (≥50m / s), and the superposition of wind turbine full-load and grid fluctuations. During the interpolation process, the parameter differences between adjacent operating condition samples are calculated based on the existing operating condition samples. The interpolation algorithm generates extreme operating condition sample data that conforms to actual physical laws, filling the gaps in the extreme operating condition sample. The interpolation process is deeply integrated with the physical laws of wind turbine operation, the characteristics of power grid operation, and the correlation with actual operating conditions. Using cubic spline interpolation algorithm as a carrier, consistency between virtual samples and real samples is achieved. The specific guarantee methods are as follows: Following the physical constraints of equipment operation, the interpolation boundary is defined. Combining the rated parameters and withstand limits of the wind turbine and the power grid, the reasonable parameter range for each extreme operating condition is clarified. For example, the voltage drop should not be lower than the minimum withstand voltage of the equipment (not lower than 60% of the rated voltage), the voltage rise should not be higher than the upper limit of the insulation withstand voltage of the equipment (not higher than 120% of the rated voltage), and the extreme wind speed should not exceed the design limit of the wind turbine. The superimposed operating condition parameters conform to the power balance principle, avoid interpolation to generate invalid samples that violate physical laws, and ensure the feasibility of virtual samples. Based on the correlation of real samples and anchoring the interpolation benchmark law, this study analyzes the changing trends and quantitative relationships of core parameters such as voltage, wind speed, and power between adjacent samples, using existing normal and near-extreme operating condition samples as a basis. It extracts the parameter change rate and coupling law (such as the nonlinear correlation between wind speed and power, and the linkage between voltage fluctuation and wind turbine output). The interpolation strictly follows this law to ensure that the parameter changes of the generated extreme operating condition samples are continuous and smooth, and consistent with the parameter evolution logic of real operating conditions. By adapting to the characteristics of cubic spline interpolation and strengthening the verification of sample authenticity, the algorithm's smoothness and second-order differentiability are utilized to fit the parameter curves of adjacent real samples. This ensures that the extreme operating condition samples generated by interpolation not only fill the gaps but also avoid parameter abrupt changes. At the same time, based on actual operating experience, the rationality of the interpolated samples is verified, and samples that do not conform to the operating characteristics of the equipment and the fluctuation patterns of the power grid are removed. This ensures that the supplementary extreme operating condition samples not only enrich the dataset but also truly reflect the operating status of the equipment under extreme conditions, providing reliable data support for subsequent model training.

[0019] After the model training is completed, the model is validated using a test set to ensure that the model prediction accuracy is ≥95%. After the validation is qualified, the actual operating conditions of the wind turbine generator and the grid operation status are simulated through the prediction model. The normal grid operation status (stable grid parameters and normal wind turbine output) and the fault operation status (voltage drop fault and converter fault) are simulated respectively. The corresponding grid parameters under the two operation statuses are collected in real time. After the collected parameters are processed, they are transmitted to the evaluation module.

[0020] Evaluation module The evaluation module and the simulation module use the same industrial server with built-in quantitative calculation and evaluation analysis software. The specific working process is as follows: the evaluation module receives network parameters under different operating conditions transmitted by the simulation module. First, the parameters are quantitatively calculated. The quantitative indicators include: parameter deviation rate, parameter fluctuation amplitude, response time and stability coefficient. The quantitative calculation adopts a preset standard formula to ensure the accuracy and standardization of the calculation results. After the quantitative calculation is completed, a hierarchical weighted dynamic evaluation system is adopted to comprehensively measure the grid-connected performance value of the wind turbine generator set. The system is divided into three layers: grid-side evaluation layer, wind turbine-side evaluation layer, and comprehensive evaluation layer. The weight of the grid-side evaluation layer is set to 0.4, the weight of the wind turbine-side evaluation layer is set to 0.4, and the weight of the comprehensive evaluation layer is set to 0.2. The weight of the indicators in each evaluation layer can be dynamically adjusted according to the actual grid connection scenario (grid connection commissioning, daily operation and maintenance), with an adjustment range of ±0.1. The comprehensive grid connection performance value is calculated by weighted summation. The formula is: Grid connection performance value = Grid side evaluation score × 0.4 + Wind turbine side evaluation score × 0.4 + Comprehensive evaluation score × 0.2. The final output is a standardized grid connection performance evaluation index, including the comprehensive performance score (out of 100), the evaluation scores and performance levels of each level (Excellent ≥ 90 points, Good 80-89 points, Pass 60-79 points, Unqualified < 60 points). The evaluation index is then transmitted to the visualization display module.

[0021] Visualization module The visualization module includes an industrial monitor (27 inches in size, 1920×1080 resolution) and visualization software. The software supports a variety of preset visualization presentation methods, including: line chart (showing the trend of network parameters over time), bar chart (comparing parameter differences under different operating conditions), radar chart (showing the scores of each evaluation indicator), and digital dashboard (real-time display of comprehensive performance score and performance level). After receiving the grid connection performance evaluation indicators output by the evaluation module, the visualization module displays them intuitively through the above visualization presentation method. It supports parameter filtering (which can be filtered by operating status and time range), data export (export format is Excel and PDF), and abnormal alarm (when the performance level is unqualified, an audible and visual alarm is triggered). This realizes the visualization and convenient output of the grid connection performance evaluation results of wind turbine generator sets, making it easier for test personnel and operation and maintenance personnel to quickly grasp the grid connection performance status of the units.

[0022] The preprocessing steps within the sensor module are as follows: Data preprocessing within the sensor module is crucial for ensuring accurate and efficient subsequent data analysis. The core is to transform multi-source heterogeneous raw data into standardized, high-quality datasets. First, various core monitoring, environmental, and equipment status parameters are integrated to construct the raw dataset. Source identification is added to ensure traceability. At the same time, parameter units, data types, and dimension fields are standardized to complete format standardization and solve the problem of data heterogeneity. Using collection time, collection node, and parameter type as the sole dimensions, completely duplicate records are deleted; for near-duplicate data with collection time deviations within the collection period and consistent parameter values, only one valid record is retained, and invalid data, including data with empty parameter values, zero values ​​without reasonable justification, and fixed value data without actual business significance, are deleted to improve the purity of the dataset. For highly relevant network performance parameters, core analytical parameters are selected and retained through correlation analysis, while redundant parameters are deleted to improve analysis efficiency. All preprocessed valid parameters are integrated, and the data standardization and validity are verified again to form a standardized dataset that can be directly used for subsequent feature extraction and model analysis, providing data support for the efficient operation of sensor modules.

[0023] The prediction model within the simulation module is constructed based on a standardized dataset for core information extraction and initial model building. Core information on wind turbine operating parameters and grid operating parameters is accurately extracted from the standardized dataset. For wind turbine operating parameters, key indicators such as speed, pitch angle, output power, nacelle azimuth angle, and yaw speed are prioritized. For grid operating parameters, core parameters such as voltage, frequency, line load, and power factor are focused on. Redundant information irrelevant to the prediction objective is eliminated to reduce the computational load for model training. The prediction objectives are clearly defined as wind turbine operating conditions (including normal operation, minor anomalies, severe anomalies, and shutdown scenarios) and grid operating states (including stable operation, voltage fluctuations, frequency shifts, and load overload types). Combining the temporal characteristics and nonlinear correlations of these two types of prediction objectives, the LSTM deep learning algorithm is selected to construct the prediction model. By building a network structure of input, hidden, and output layers, the mapping relationship between core parameters and prediction objectives is modeled, and the basic parameters of the model are initialized to lay the foundation for subsequent training and optimization. A virtual operating condition interpolation algorithm is introduced to fill the technical gap of scarce extreme operating condition samples. Since the probability of extreme wind speeds and grid short circuits occurring in actual operation of wind turbines is extremely low, the number of extreme operating condition samples in the standardized dataset is very small. If directly used for model training, the model's prediction accuracy for extreme operating conditions would be insufficient, resulting in prediction blind spots. To address this, a physical constraint-based virtual operating condition interpolation algorithm is introduced. Based on existing conventional operating condition samples (such as rated wind speed and operating condition data under normal load) in the standardized dataset, and combined with the actual operating physical laws of wind turbines and the power grid, reasonable interpolation boundaries are set (such as upper limits for wind speed and voltage thresholds to avoid generating unrealistic virtual samples). Through a combination of linear interpolation and cubic spline interpolation, the core parameters of the conventional operating condition samples are interpolated to generate a series of virtual extreme operating condition samples that conform to actual operating logic and cover various extreme scenarios. These virtual extreme operating condition samples are then added to the dataset in a reasonable proportion, achieving a balanced distribution of conventional and extreme operating condition samples, completely solving the problem of scarce extreme operating condition samples, and improving the model's generalization ability and robustness. The final prediction model is obtained through training and optimization using a complete dataset. The complete dataset, supplemented with virtual extreme operating condition samples, is divided into training, validation, and test sets in a 7:2:1 ratio and re-inputted into the built LSTM prediction model for full-process training and optimization. During training, prediction accuracy and mean squared error are used as the core evaluation indicators, and model parameters, including the number of hidden layer neurons, learning rate, number of iterations, and regularization coefficient, are dynamically adjusted. The training effect of the model is monitored in real time using the validation set to avoid overfitting and underfitting. At the same time, the performance of the optimized model is validated using the test set to ensure that the model can not only accurately predict the wind turbine and grid status under normal operating conditions, but also effectively identify various extreme operating conditions, meeting the prediction requirements of the simulation module. Finally, a stable and accurate prediction model for wind turbine operating conditions and grid operation status is obtained.

[0024] The initial construction of the prediction model is achieved by using a standardized dataset. Extracting core parameters ,by For the purpose of prediction; in , is a set of operating parameters for wind turbine generator sets, including generator speed and nacelle azimuth angle, where m is the number of generator parameters; This is a set of power grid operating parameters, including grid voltage, frequency, power factor, and line losses. This represents the total number of core parameters. ; Wind turbine generator set operating conditions; The power grid operating status is uniformly denoted as: ; The initial prediction model is constructed using deep learning algorithms, and the formula is expressed as follows: in A deep learning model consists of an input layer, hidden layers, and an output layer. The initial parameters of the model, including weights and biases, are used by the model to fit... and The mapping relationship is used to complete the initial construction; A virtual operating condition interpolation algorithm is introduced, based on a normal operating condition sample set. Interpolation calculations are performed to generate virtual extreme working condition samples that conform to actual physical laws. The interpolation calculation formula is expressed as: in For the first Core parameter values ​​of a virtual extreme working condition sample; For the sample set of normal working conditions Two adjacent normal operating condition parameter values, ; These are the interpolation coefficients. This ensures that the generated samples are extreme values ​​and conform to the physical operating limits of the wind turbine generator and the power grid; This is a virtual operating condition interpolation algorithm; The generated virtual extreme condition samples are added to the original standardized dataset to obtain the complete dataset, expressed as: ;in This is the supplementary standardized dataset; Complete dataset Core parameters in and their corresponding prediction targets Re-input to the initial prediction model In this process, the model parameters are adjusted by minimizing the prediction error. The optimized model parameters are obtained. With the final prediction model The final prediction model expression is: .

[0025] In the initial construction phase, this application accurately delineates the core parameter X and the prediction target Y, eliminates redundant parameters, reduces training costs, and uses formulaic modeling to make the model structure interpretable and the parameters traceable, laying the foundation for subsequent optimization. Through a virtual operating condition interpolation algorithm, extreme samples conforming to physical laws are generated according to a predetermined formula, solving the problem of scarce extreme operating condition samples without additional collection costs, filling prediction blind spots, and improving the model's generalization ability. After integrating the complete dataset, the model is retrained, and the prediction error is minimized by adjusting parameters, avoiding overfitting and underfitting problems, significantly improving prediction accuracy and stability. The overall process closely aligns with engineering practice, and the parameter selection and algorithm design are consistent with equipment operating patterns. The constructed model can adapt to full-condition prediction needs, providing accurate support for wind turbine and power grid operation monitoring, decision-making, and scheduling, demonstrating strong practicality and scalability.

[0026] The steps for obtaining grid connection parameters of wind turbine generators within the simulation module are as follows: The optimized prediction model is launched to simulate the actual operating scenarios of wind turbine generators and simultaneously simulate the grid operation status under the corresponding scenarios. During the simulation, the model will combine various parameter features in the dataset to restore the operating logic of the unit and the response state of the power grid under different operating conditions, and to simulate the operating state of the unit and the power grid in all aspects. After the simulation is completed, various types of data generated during the simulation are automatically collected. The data of the two core scenarios of normal grid connection and fault operation of the wind turbine generator are distinguished and extracted. The grid connection parameters of the wind turbine generator under the two types of operation are organized accordingly. The parameters include the grid-connected power of the generator, reactive power compensation, voltage deviation, and parameter mutation values ​​when the fault is triggered. The collected parameters are classified and summarized to output complete simulation result data.

[0027] This application utilizes an optimized predictive model to simulate actual operating scenarios of generating units and the power grid. Leveraging the advantages of model optimization, it can accurately reproduce the operating logic of generating units and the response status of the power grid under different operating conditions, avoiding the limitations of actual data collection, reducing on-site testing costs and equipment wear. The simulation process comprehensively covers the operating status of generating units and the power grid, ensuring that no grid-related parameters are missed and closely matches actual operating needs. After the simulation is completed, the data of two core scenarios, normal and fault, are extracted. Key grid-related parameters such as the generating unit's on-grid power and reactive power compensation are classified and organized. At the same time, the sudden change values ​​of parameters during faults are captured, ensuring the relevance of the parameters and providing core basis for fault analysis.

[0028] The specific steps for outputting standardized network performance evaluation indicators within the evaluation module are as follows: Receive grid connection parameters of wind turbine generator sets, form an initial evaluation parameter set, and perform verification; Standardize the network-related parameters that have passed the verification process, and convert the network-related parameters into quantitative values ​​of a uniform magnitude; For each grid-related parameter, targeted quantitative calculations are performed to obtain the quantitative index of each individual parameter, clarifying the basic impact of each parameter on the grid-related performance of the unit; Based on individual parameter indicators, a multi-dimensional systematic analysis of grid-related parameters is conducted. A hierarchical weighted dynamic evaluation system is adopted to comprehensively calculate the grid-related performance value of the unit. By weighted summation, the quantitative indicators of each individual parameter are integrated into the comprehensive grid-related performance value of the unit, thus completing the comprehensive calculation. The comprehensive network performance values ​​and quantitative indicators of each individual parameter obtained from the comprehensive calculation are standardized, and the quantitative data and analysis conclusions in the evaluation process are organized to form a complete evaluation report.

[0029] This application receives grid-related parameters to form an initial set, verifies them in terms of completeness, rationality, and consistency, removes abnormal data and provides feedback for correction to ensure the validity of the parameters. Next, the verified parameters are normalized to a uniform magnitude, eliminating differences in units and magnitudes. Subsequently, for each parameter, targeted quantitative calculations are performed based on evaluation benchmarks to clarify the fundamental impact of each parameter on grid-related performance. A hierarchical weighted dynamic evaluation system is adopted, with weighted summation according to the parameter's impact weight to calculate the unit's comprehensive grid-related performance value. The comprehensive performance value and individual indicators are standardized and categorized, evaluation levels are assigned, and the calculated data and conclusions are compiled to form a complete evaluation report, providing support for unit optimization and grid dispatch.

[0030] The weighted summation method integrates the quantitative indicators of each individual grid-related parameter into a comprehensive grid-related performance value F of the unit, and its calculation formula is as follows: in It is a single quantitative indicator, obtained through targeted quantitative calculation; The dynamic weighting reflects the core of "hierarchical weighted dynamic evaluation," and is set based on the results of multi-dimensional systematic analysis. The dynamic adjustment logic is as follows: Under normal operating conditions, the focus is on increasing the grid-connected power of the generating units. Reactive power compensation The weights; In fault operation mode, the focus is on improving voltage deviation. Sudden changes in fault parameters The weights; As a comprehensive performance value, the calculation results can be directly used for subsequent standardization processing and transformed into standardized network performance evaluation indicators; , which is the serial number of the grid-related single parameter, corresponding to 4 core grid-related parameters, including the unit's grid-connected power, reactive power compensation, voltage deviation and fault parameter sudden change value; For the first Quantitative indicators of network-related parameters; For the first The weights of the quantitative indicators of network-related parameters; The value is 4, corresponding to 4 network-related parameters.

[0031] This application integrates individual indicators through weighted summation, which not only retains the basic impact of each parameter on grid connection performance, but also comprehensively measures the overall grid connection level of the unit, taking into account both individual performance and overall performance. The calculated F can be directly used for subsequent standardization processing, seamlessly connecting with the subsequent processes of the evaluation module, improving evaluation efficiency, and providing accurate data support for the output of standardized grid connection performance evaluation indicators and the optimization and adjustment of the unit.

[0032] The visualization module uses a layered and categorized visualization approach based on preset visualization rules and diverse display templates. The display methods include bar charts, heatmaps, and line charts.

[0033] The test method for the grid connection performance of sensor-based wind turbine generators includes the following steps: S1. The sensor module collects relevant raw parameters of the network through preset sensors, and after preprocessing to remove redundant and abnormal data, it forms a standardized dataset and outputs it. S2. The simulation module calls the dataset to build a prediction model of unit operating conditions and grid status, introduces a virtual operating condition interpolation algorithm to supplement extreme operating condition samples, and obtains relevant parameters of the unit's grid connection under normal and fault states through model simulation. S3. The evaluation module receives the simulation parameters, performs quantitative calculations and systematic analysis, and then uses a hierarchical weighted dynamic evaluation system to calculate the comprehensive performance value and output standardized evaluation indicators. S4, the visualization module receives evaluation indicators and presents them intuitively through preset visualization methods.

[0034] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A sensor-based test system for the grid connection performance of wind turbine generator sets, characterized in that, include: The sensor module is used to collect grid-connection performance-related parameters of wind turbine generators in real time through preset sensors, preprocess the collected grid-connection performance-related parameters, remove redundant data and abnormal interference data, and integrate the preprocessed qualified parameters to form a standardized dataset. The simulation module is used to call the standardized dataset output by the sensor module, construct a prediction model of wind turbine generator operating conditions and grid operation status based on the parameter information in the dataset, introduce a virtual operating condition interpolation algorithm to supplement the extreme operating condition samples in the dataset, fill the technical gap of scarce extreme operating condition samples in the actual collection process, and then simulate the actual operating conditions of wind turbine generators and grid operation status through the prediction model to obtain the normal grid-connected operation status, fault operation status of wind turbine generators, and the corresponding grid-connected parameters of wind turbine generators under the two different operation statuses. The evaluation module is used to receive the grid connection parameters of the wind turbine generator set under different operating conditions output by the simulation module, perform quantitative calculation and systematic analysis on the grid connection parameters, adopt a hierarchical weighted dynamic evaluation system to comprehensively calculate the grid connection performance value of the wind turbine generator set, and output standardized grid connection performance evaluation indicators. The visualization module is used to receive the grid connection performance evaluation indicators output by the evaluation module, and to intuitively display the grid connection performance evaluation indicators through a preset visualization presentation method, so as to realize the visualization output of the grid connection performance evaluation results of the wind turbine generator set.

2. The sensor-based test system for grid connection performance of wind turbine generator sets according to claim 1, characterized in that: The sensor module contains preset sensors that are placed at key locations on the wind turbine generator under test to collect mechanical, electrical and environmental parameters required for grid connection performance testing. The sensors include intelligent speed and torque sensors, intelligent vibration sensors, current sensors, and temperature sensors. The intelligent speed and torque sensors are installed at the output end of the wind turbine gearbox and on the main shaft to collect real-time speed and torque parameters of the main shaft. Intelligent vibration sensors are installed in the gearbox, servo motor and nacelle of the wind turbine to collect vibration acceleration and vibration frequency parameters during wind turbine operation. Current sensors are installed at the output and grid-connected input terminals of the wind turbine generator set to collect three-phase voltage, three-phase current, and power parameters. Temperature sensors are installed on the gearbox, servo motor, and electronic rotor of the wind turbine generator set to collect temperature parameters during operation.

3. The sensor-based test system for grid connection performance of wind turbine generator sets according to claim 1, characterized in that, The preprocessing steps within the sensor module are as follows: By integrating parameters from multiple sources to construct the original dataset, the source can be traced, and the parameter units, data types, and dimension fields can be standardized to achieve format standardization. Using the collection time, collection node, and parameter type as the sole criteria, completely duplicate records are identified and deleted. For nearly duplicate data with the same parameter values ​​and a time deviation within the acquisition period, retain one valid record and delete the rest of the duplicates. Delete data with parameter values ​​that are empty, all zeros, or have no actual business meaning; For highly relevant network performance parameters, retain the core analysis parameters; The preprocessed parameters are integrated to obtain a standardized dataset.

4. The sensor-based test system for grid connection performance of wind turbine generator sets according to claim 1, characterized in that: The prediction model within the simulation module is built on a standardized dataset. Core information on the wind turbine generator operating parameters and power grid operating parameters in the dataset is extracted. The wind turbine generator operating conditions and power grid operating status are used as prediction targets. A prediction model for the wind turbine generator operating conditions and power grid operating status is built using a deep learning algorithm. A virtual working condition interpolation algorithm is introduced to interpolate existing conventional working condition samples in the standardized dataset, generate virtual extreme working condition samples that conform to actual physical laws, and supplement them into the dataset to fill the technical gap of scarce extreme working condition samples. The complete dataset, supplemented with extreme working condition samples, is re-input into the prediction model for training and optimization. The model parameters are then adjusted to obtain the prediction model.

5. The sensor-based test system for grid connection performance of wind turbine generator sets according to claim 4, characterized in that: The initial construction of the prediction model is achieved by using a standardized dataset. Extracting core parameters ,by For the purpose of prediction; in , is a set of operating parameters for wind turbine generator sets, including generator speed and nacelle azimuth angle, where m is the number of generator parameters; This is a set of power grid operating parameters, including grid voltage, frequency, power factor, and line losses. This represents the total number of core parameters. ; Wind turbine generator set operating conditions; The power grid operating status is uniformly denoted as: ; The initial prediction model is constructed using deep learning algorithms, and the formula is expressed as follows: in A deep learning model consists of an input layer, hidden layers, and an output layer. The initial parameters of the model, including weights and biases, are used by the model to fit... and The mapping relationship is used to complete the initial construction; A virtual operating condition interpolation algorithm is introduced, based on a normal operating condition sample set. Interpolation calculations are performed to generate virtual extreme working condition samples that conform to actual physical laws. The interpolation calculation formula is expressed as: in For the first Core parameter values ​​of a virtual extreme working condition sample; For the sample set of normal working conditions Two adjacent normal operating condition parameter values, ; These are the interpolation coefficients. This ensures that the generated samples are extreme values ​​and conform to the physical operating limits of the wind turbine generator and the power grid; This is a virtual operating condition interpolation algorithm; The generated virtual extreme condition samples are added to the original standardized dataset to obtain the complete dataset, expressed as: ;in This is the supplementary standardized dataset; Complete dataset Core parameters in and their corresponding prediction targets Re-input to the initial prediction model In this process, the model parameters are adjusted by minimizing the prediction error. The optimized model parameters are obtained. With the final prediction model The final prediction model expression is: .

6. The sensor-based test system for grid connection performance of wind turbine generator sets according to claim 1, characterized in that, The steps for obtaining grid connection parameters of wind turbine generators within the simulation module are as follows: The optimized prediction model is launched to simulate the actual operating scenarios of wind turbine generators and simultaneously simulate the grid operation status under the corresponding scenarios. During the simulation, the model will combine various parameter features in the dataset to restore the operating logic of the unit and the response state of the power grid under different operating conditions, and to simulate the operating state of the unit and the power grid in all aspects. After the simulation is completed, various types of data generated during the simulation will be automatically collected, and data from two core scenarios will be distinguished and extracted: the normal operation status of the wind turbine generator and the fault operation status. The grid-connected parameters of wind turbine generators under two different operating conditions are compiled and organized. The parameters include the generator's grid-connected power, reactive power compensation, voltage deviation, and parameter mutation values ​​when a fault is triggered. The collected parameters are classified and summarized to output complete simulation results data.

7. The sensor-based test system for grid connection performance of wind turbine generator sets according to claim 1, characterized in that, The specific steps for outputting standardized network performance evaluation indicators within the evaluation module are as follows: Receive grid connection parameters of wind turbine generator sets, form an initial evaluation parameter set, and perform verification; Standardize the network-related parameters that have passed the verification process, and convert the network-related parameters into quantitative values ​​of a uniform magnitude; For each grid-related parameter, targeted quantitative calculations are performed to obtain the quantitative index of each individual parameter, clarifying the basic impact of each parameter on the grid-related performance of the unit; Based on individual parameter indicators, a multi-dimensional systematic analysis of grid-related parameters is conducted. A hierarchical weighted dynamic evaluation system is adopted to comprehensively calculate the grid-related performance value of the unit. By weighted summation, the quantitative indicators of each individual parameter are integrated into the comprehensive grid-related performance value of the unit, thus completing the comprehensive calculation. The comprehensive network performance values ​​and quantitative indicators of each individual parameter obtained from the comprehensive calculation are standardized, and the quantitative data and analysis conclusions in the evaluation process are organized to form a complete evaluation report.

8. The sensor-based test system for grid connection performance of wind turbine generator sets according to claim 7, characterized in that: The weighted summation method integrates the quantitative indicators of each individual grid-related parameter into a comprehensive grid-related performance value F of the unit, and its calculation formula is as follows: in It is a single quantitative indicator, obtained through targeted quantitative calculation; The dynamic weighting reflects the core of "hierarchical weighted dynamic evaluation," and is set based on the results of multi-dimensional systematic analysis. The dynamic adjustment logic is as follows: Under normal operating conditions, the focus is on increasing the grid-connected power of the generating units. Reactive power compensation The weights; In fault operation mode, the focus is on improving voltage deviation. Sudden changes in fault parameters The weights; As a comprehensive performance value, the calculation results can be directly used for subsequent standardization processing and transformed into standardized network performance evaluation indicators; , which is the serial number of the grid-related single parameter, corresponding to 4 core grid-related parameters, including the unit's grid-connected power, reactive power compensation, voltage deviation and fault parameter sudden change value; For the first Quantitative indicators of network-related parameters; For the first The weights of the quantitative indicators of network-related parameters; The value is 4, corresponding to 4 network-related parameters.

9. The sensor-based test system for grid connection performance of wind turbine generator sets according to claim 1, characterized in that: The visualization module uses a layered and categorized visualization approach based on preset visualization rules and diverse display templates. The display methods include bar charts, heatmaps, and line charts.

10. A test method for the grid connection performance of a sensor-based wind turbine generator set, applied to the test system for the grid connection performance of a sensor-based wind turbine generator set as described in any one of claims 1 to 9, characterized in that, The experimental steps are as follows: S1. The sensor module collects relevant raw parameters of the network through preset sensors, and after preprocessing to remove redundant and abnormal data, it forms a standardized dataset and outputs it. S2. The simulation module calls the dataset to build a prediction model of unit operating conditions and grid status, introduces a virtual operating condition interpolation algorithm to supplement extreme operating condition samples, and obtains relevant parameters of the unit's grid connection under normal and fault states through model simulation. S3. The evaluation module receives the simulation parameters, performs quantitative calculations and systematic analysis, and then uses a hierarchical weighted dynamic evaluation system to calculate the comprehensive performance value and output standardized evaluation indicators. S4, the visualization module receives evaluation indicators and presents them intuitively through preset visualization methods.

Images