A wind turbine data cleaning and enhancement method and system

Abstract

The application belongs to the field of wind power generation, and discloses a wind turbine data cleaning and enhancement method and system, which comprises the following steps: obtaining multi-source monitoring data of a target wind turbine under extreme environment; performing measurement deviation calculation and adaptive calibration on the multi-source monitoring data by using a pre-constructed physical mechanism model to obtain drift correction data; identifying missing data segments and abnormal data segments in the drift correction data, performing interpolation filling and sequence reconstruction on the missing data segments and the abnormal data segments, and obtaining continuous time series; and performing multi-frequency band decomposition and denoising processing on the continuous time series to obtain enhanced monitoring data. The application can solve the technical problems of sensor measurement drift, data fault missing and complex noise interference of wind turbine monitoring data under extreme environment.

Description

Technical Field

[0001] This application belongs to the field of wind power generation, specifically relating to a method and system for cleaning and enhancing wind turbine data. Background Technology

[0002] With the rapid development of wind power technology, large wind turbines are widely deployed in vast, complex terrains and extreme climate zones, such as high-altitude and freezing environments. To ensure the safe and stable operation and full lifecycle management of wind turbines, wind farms are generally equipped with multi-source online monitoring systems capable of collecting massive amounts of multi-dimensional time-series data, including meteorological parameters, mechanical vibration responses, and operating conditions, in real time. Under normal circumstances, maintenance personnel and intelligent diagnostic systems primarily rely on this field-collected monitoring data for long-term health assessments and fault warnings. Therefore, the basic acquisition, transmission, and routine signal smoothing of multi-source monitoring data constitute an important component of the current intelligent operation and maintenance and data analysis system for wind farms.

[0003] However, in extremely cold and severe weather environments, wind turbines face rigorous physical challenges, revealing significant limitations in existing conventional data processing and operation and maintenance monitoring methods. On one hand, changes in physical characteristics caused by low temperatures often lead to systematic measurement drift in sensors, and communication interruptions or freezing failures of wind measurement devices caused by severe weather frequently result in large-scale missing data or nonlinear distortions in monitoring sequences, severely disrupting the integrity and continuity of the time series. On the other hand, under complex operating conditions, high-frequency mechanical noise from the mechanical transmission chain intertwines with low-frequency trend interference caused by environmental changes, making it difficult for conventional signal filtering methods to effectively remove various noises and easily masking subtle fault-sensitive characteristics. These issues of data distortion, gaps, and low signal-to-noise ratio often result in significant errors in condition assessments and fault diagnosis based on on-site monitoring data, failing to meet the practical engineering requirements for high-reliability operation and maintenance of wind turbines under extreme conditions. Summary of the Invention

[0004] This application addresses the technical problems of sensor measurement drift, data gaps and missing data, and complex noise interference in wind turbine monitoring data under extreme environments by providing a method and system for cleaning and enhancing wind turbine data.

[0005] To achieve the above objectives, this application adopts the following technical solution: Firstly, a method for cleaning and enhancing wind turbine data includes: Acquire multi-source monitoring data of the target wind turbine under extreme environments; The measurement deviation is calculated and adaptively calibrated using a pre-built physical mechanism model to obtain drift correction data from the multi-source monitoring data; Identify missing and abnormal data segments in the drift correction data, and perform interpolation imputation and sequence reconstruction on the missing and abnormal data segments to obtain a continuous time series; The continuous time series is subjected to multi-band decomposition and denoising processing to obtain enhanced monitoring data.

[0006] In some implementations, the multi-source monitoring data includes ambient temperature sequences and communication status logs.

[0007] In some implementations, the step of using a pre-built physical mechanism model to calculate measurement deviations and adaptively calibrate the multi-source monitoring data to obtain drift correction data includes: A thermodynamic compensation model for the device is constructed based on the ambient temperature sequence. Based on the thermodynamic compensation model, the sensor thermal inertia compensation variable under the current temperature gradient is calculated. The sensor thermal inertia compensation variable is introduced as a control input into the Kalman filter algorithm to calculate the optimal state estimate at the current moment. The optimal state estimate is used to correct the baseline offset error in the multi-source monitoring data to generate the drift correction data.

[0008] In some implementations, identifying missing and abnormal data segments in the drift correction data includes: Based on the disconnection records in the communication status log, identify the missing interval of the drift correction data as the missing data segment; The drift correction data is scanned using a preset anomaly detection algorithm to identify nonlinear zero-value distortion intervals or constant-value dead zone intervals as the abnormal data segments.

[0009] In some implementations, the step of interpolating and imputing the missing data segments and the abnormal data segments to obtain a continuous time series includes: The spatial correlation features of neighboring wind turbines in the target wind farm and the historical time series features of the target wind turbines are extracted using the spatiotemporal kriging interpolation algorithm. Based on the spatial correlation features and the historical time series features, a basic trend sequence covering the missing data segment and the abnormal data segment is generated. The basic trend sequence is input into a pre-trained generative adversarial network to generate high-frequency nonlinear fluctuation components; The basic trend sequence is superimposed and fused with the high-frequency nonlinear fluctuation component to obtain the continuous time series.

[0010] In some embodiments, the step of performing multi-band decomposition and denoising on the continuous time series to obtain enhanced monitoring data includes: The continuous time series is decomposed into multiple intrinsic mode function components using the empirical mode decomposition algorithm; Calculate the Pearson correlation coefficient between each intrinsic mode function component and the continuous time series, and remove intrinsic mode function components with correlation coefficients lower than a preset threshold to obtain mid-to-high frequency intrinsic mode function components; The enhanced monitoring data is obtained by using wavelet packet denoising algorithm to perform noise threshold quantization and high-frequency filtering on the mid-to-high frequency intrinsic mode function components.

[0011] Secondly, a wind turbine data cleaning and enhancement system includes: The data acquisition module is used to acquire multi-source monitoring data of the target wind turbine under extreme environments; The drift correction module is used to calculate the measurement deviation and adaptively calibrate the multi-source monitoring data using a pre-built physical mechanism model to obtain drift correction data. The reconstruction module is used to identify missing data segments and abnormal data segments in the drift correction data, and to perform interpolation imputation and sequence reconstruction on the missing data segments and abnormal data segments to obtain a continuous time series; The denoising module is used to perform multi-band decomposition and denoising processing on the continuous time series to obtain enhanced monitoring data.

[0012] Thirdly, a computer device includes: a processor and a computer-readable storage medium; A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements the aforementioned method for cleaning and enhancing wind turbine data.

[0013] Fourthly, a computer-readable storage medium storing a computer program adapted to be loaded by a processor and executed as described in the wind turbine data cleaning and enhancement method.

[0014] Fifthly, a computer program product comprising a computer program that, when executed by a processor, implements the wind turbine data cleaning and enhancement method as described above.

[0015] Compared with the prior art, this application has the following beneficial effects: This application provides a method for cleaning and enhancing wind turbine data. By acquiring multi-source monitoring data of the target wind turbine under extreme environments, the method uses a pre-built physical mechanism model to calculate measurement deviations and adaptively calibrate the multi-source monitoring data to obtain drift correction data. The method identifies missing and abnormal data segments in the drift correction data and performs interpolation and sequence reconstruction to obtain a continuous time series. The method then performs multi-frequency band decomposition and noise reduction on the continuous time series to obtain enhanced monitoring data. This method overcomes the distortion problem caused by sensor drift, data fragmentation, and noise interference in wind turbine monitoring data under extreme environments.

[0016] Furthermore, by constructing a thermodynamic compensation model for the device based on the ambient temperature sequence, the sensor thermal inertia compensation variable under the current temperature gradient is calculated according to the model. This compensation variable is then used as a control input to the Kalman filter algorithm to calculate the optimal state estimate at the current moment. The optimal state estimate is then used to correct the benchmark offset error in the multi-source monitoring data to generate drift correction data. This approach can accurately eliminate the systematic measurement drift generated by the sensor under extreme low temperatures from the physical mechanism level, ensuring that the corrected benchmark data is true and reliable.

[0017] Furthermore, by identifying the missing intervals of drift correction data as missing data segments based on the disconnection records in the communication status log, and using a preset anomaly detection algorithm to scan the drift correction data and identify nonlinear zero-value distortion intervals or constant value dead zone intervals as anomalous data segments, hard missing data and specific distortions can be accurately isolated, avoiding data pollution caused by blind interpolation and ensuring the targeting of subsequent reconstructed data.

[0018] Furthermore, by employing a spatiotemporal kriging interpolation algorithm to extract the spatial correlation features of neighboring wind turbines within the target wind farm and the historical time series features of the target wind turbines, a basic trend sequence covering missing and anomalous segments is generated. This basic trend sequence is then input into a pre-trained generative adversarial network to generate high-frequency nonlinear fluctuation components. Finally, the basic trend sequence and the high-frequency nonlinear fluctuation components are superimposed and fused to obtain a continuous time series. This process not only restores the macroscopic basic trend but also reproduces the high-frequency nonlinear fluctuation details of the actual mechanical operating conditions, completely solving the reconstruction distortion problem caused by long-sequence data discontinuity.

[0019] Furthermore, by employing the empirical mode decomposition algorithm to decompose the continuous time series into multiple intrinsic mode function components, calculating the Pearson correlation coefficient between each intrinsic mode function component and the continuous time series, and removing intrinsic mode function components with correlation coefficients below a preset threshold to obtain mid-to-high frequency intrinsic mode function components, wavelet packet denoising algorithm is then used to perform noise threshold quantization and high-frequency filtering on the mid-to-high frequency intrinsic mode function components to obtain enhanced monitoring data. This method can progressively remove non-stationary noise, achieving deep denoising while preserving fault-sensitive characteristics, and significantly improving the signal-to-noise ratio of the data. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 A flowchart of a wind turbine data cleaning and enhancement method provided in this application embodiment; Figure 2 A structural diagram of a wind turbine data cleaning and enhancement system provided in this application embodiment; Figure 3 A computer device diagram provided for an embodiment of this application. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] In one embodiment of this application, such as Figure 1 As shown, a method for cleaning and enhancing wind turbine data is provided, including: S1, acquire multi-source monitoring data of the target wind turbine under extreme environments; Specifically, the target wind turbine operates in extreme climate zones such as high-altitude and freezing areas, and its multi-source monitoring data is typically collected in real time by the wind turbine's SCADA system or on-site sensor network. In this embodiment, the multi-source monitoring data not only includes the unit's operating power and mechanical vibration parameters, but also explicitly includes ambient temperature sequences (such as the external ambient temperature of the nacelle and the surface temperature of the anemometer) and communication status logs recorded by the underlying communication equipment. This data provides the basic data source for subsequent physical mechanism correction and communication anomaly identification.

[0024] S2, using a pre-built physical mechanism model to calculate the measurement deviation and perform adaptive calibration on the multi-source monitoring data to obtain drift correction data; Specifically, in extreme low-temperature environments, the electronic components and mechanical structures inside the sensor undergo thermal expansion and contraction, leading to systematic measurement drift. This drift cannot be eliminated using conventional extremum removal algorithms. In practice, a thermodynamic compensation model for the device is first constructed based on the acquired ambient temperature sequence to calculate the sensor's thermal inertia compensation variable under the current temperature gradient change. Due to the physical lag in temperature conduction, this thermal inertia compensation variable can objectively reflect the sensor's physical deviation state.

[0025] Subsequently, the sensor thermal inertia compensation variable is used as a control input and introduced into the prediction and measurement update models of the Kalman filter algorithm to calculate the optimal state estimate of the true physical state at the current moment. Finally, this optimal state estimate is used to dynamically correct the reference offset error in the multi-source monitoring data, generating drift correction data that reflects the true physical reference.

[0026] S3, identify missing data segments and abnormal data segments in the drift correction data, and perform interpolation imputation and sequence reconstruction on the missing data segments and abnormal data segments to obtain a continuous time series; Specifically, during the identification phase, on the one hand, based on the network disconnection records in the communication status log, the hard missing intervals caused by communication interruption in the drift correction data are accurately identified and marked as missing data segments; On the other hand, to address signal distortion caused by blade icing or anemometer freezing, a preset anomaly detection algorithm (such as the isolated forest algorithm) is used to scan the drift correction data, identify nonlinear zero-value distortion intervals or long-term constant value dead zones, and mark them as abnormal data segments.

[0027] During the reconstruction phase, to avoid the loss of dynamic signal features caused by traditional linear interpolation, this embodiment adopts a joint spatiotemporal generation strategy. First, a spatiotemporal kriging (ST-Kriging) interpolation algorithm is used to extract the spatial correlation features of neighboring wind turbines within the target wind farm that have not experienced faults. This is then combined with the target wind turbines' own historical time-series features without missing data segments to generate a basic trend sequence covering both missing and abnormal data segments. At this stage, the basic trend sequence only possesses macroscopic smoothing characteristics. Subsequently, the basic trend sequence is input into a pre-trained generative adversarial network (GAN), and the generator generates high-frequency nonlinear fluctuation components that match the actual mechanical operating conditions of the wind turbine (such as gearbox meshing frequency). Finally, the basic trend sequence is superimposed and fused with the high-frequency nonlinear fluctuation component to replace the original missing and abnormal intervals, thereby reconstructing a continuous time series that conforms to both macro-level wind conditions and micro-level mechanical dynamics.

[0028] S4, perform multi-band decomposition and denoising on the continuous time series to obtain enhanced monitoring data.

[0029] Specifically, the reconstructed continuous time series often includes low-frequency trend interference caused by gradual changes in wind speed, as well as high-frequency mechanical noise emitted by the transmission chain under extreme conditions.

[0030] To extract valuable fault features, the Empirical Mode Decomposition (EMD) algorithm is first used to adaptively decompose the continuous time series into multiple intrinsic mode function (IMF) components arranged in descending order of frequency.

[0031] Next, the Pearson correlation coefficient between each IMF component and the original continuous time series is calculated. Low-frequency IMF components with correlation coefficients below a preset threshold (e.g., 0.15) usually correspond to environmental trend interference and are therefore removed, while the remaining mid-to-high-frequency IMF components are retained.

[0032] Subsequently, for the retained mid-to-high frequency IMF components, wavelet packet denoising (WPD) algorithm was used for noise threshold quantization and high-frequency filtering to remove irregular high-frequency mechanical white noise. After the above bidirectional stripping of low and high frequencies, the fault-sensitive features hidden in the mid-frequency band were successfully extracted and reconstructed into enhanced monitoring data with high signal-to-noise ratio.

[0033] The method further includes: performing operational status assessment or fault diagnosis on the target wind turbine based on the enhanced monitoring data. Specifically, the cleaned and enhanced high-fidelity monitoring data is provided to the upper-level operation and maintenance monitoring system. Since the enhanced monitoring data has eliminated interference from sensor drift, gaps, and complex noise under extreme environments, the operation and maintenance system can accurately extract its time-frequency domain features and input them into the fault diagnosis classification model, outputting precise fault location results for the gearbox or bearing of the target wind turbine, and automatically triggering shutdown protection or power reduction operation commands accordingly to ensure the safe operation of the unit under extreme environments.

[0034] In one embodiment of this application, such as Figure 2 As shown, a wind turbine data cleaning and enhancement system is provided, including: The data acquisition module is used to acquire multi-source monitoring data of the target wind turbine under extreme environments; The drift correction module is used to calculate the measurement deviation and adaptively calibrate the multi-source monitoring data using a pre-built physical mechanism model to obtain drift correction data. The reconstruction module is used to identify missing data segments and abnormal data segments in the drift correction data, and to perform interpolation imputation and sequence reconstruction on the missing data segments and abnormal data segments to obtain a continuous time series; The denoising module is used to perform multi-band decomposition and denoising processing on the continuous time series to obtain enhanced monitoring data.

[0035] For specific limitations regarding the wind turbine data cleaning and enhancement system, please refer to the limitations of the wind turbine data cleaning and enhancement method described above. The corresponding technical effects are equivalent and will not be repeated here. Each module in the aforementioned wind turbine data cleaning and enhancement system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0036] Figure 3 An internal structural diagram of a computer device is shown in one embodiment. This computer device may specifically be a terminal or a server. Figure 3 As shown, the computer device includes a processor, memory, network interface, display, camera, and input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for cleaning and enhancing wind turbine data. The display screen can be an LCD screen or an e-ink display screen. The input device can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device casing, or an external keyboard, touchpad, or mouse.

[0037] As will be understood by those skilled in the art, computer equipment Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the computer device to which the present invention is applied. A specific computing device may include more or fewer components than those shown in the figure, or combine certain components, or have the same component arrangement.

[0038] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method described above.

[0039] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0040] In summary, the wind turbine data cleaning and enhancement method, system, computer equipment, and storage medium provided in this application overcome the shortcomings of existing technologies that easily produce distortion and bias when directly processing multi-source monitoring data under extreme operating conditions. This application combines physical mechanism compensation, spatiotemporal feature generation, and multi-band decomposition technology to sequentially achieve adaptive calibration for sensor low-temperature drift, high-fidelity reconstruction of missing and abnormal data segments, and targeted filtering of complex high- and low-frequency environmental and mechanical noise. Through the above processing, the temporal continuity and signal-to-noise ratio of monitoring data are effectively improved, providing high-quality data support for the operational status assessment and fault location of wind turbines under extreme environments.

[0041] The various embodiments in this specification are described in a progressive manner. For directly identical or similar parts of the embodiments, refer to each other. Each embodiment focuses on its differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. It should be noted that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0042] The above-described embodiments are merely preferred embodiments of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various improvements and substitutions without departing from the principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention. Therefore, the scope of protection of this invention should be determined by the scope of the claims.

Claims

1. A method for cleaning and enhancing wind turbine data, characterized in that, include: Acquire multi-source monitoring data of the target wind turbine under extreme environments; The measurement deviation is calculated and adaptively calibrated using a pre-built physical mechanism model to obtain drift correction data from the multi-source monitoring data; Identify missing and abnormal data segments in the drift correction data, and perform interpolation imputation and sequence reconstruction on the missing and abnormal data segments to obtain a continuous time series; The continuous time series is subjected to multi-band decomposition and denoising processing to obtain enhanced monitoring data.

2. The method for cleaning and enhancing wind turbine data according to claim 1, characterized in that, The multi-source monitoring data includes ambient temperature sequences and communication status logs.

3. The method for cleaning and enhancing wind turbine data according to claim 2, characterized in that, The step of using a pre-built physical mechanism model to calculate measurement deviations and perform adaptive calibration on the multi-source monitoring data to obtain drift correction data includes: A thermodynamic compensation model for the device is constructed based on the ambient temperature sequence. Based on the thermodynamic compensation model, the sensor thermal inertia compensation variable under the current temperature gradient is calculated. The sensor thermal inertia compensation variable is introduced as a control input into the Kalman filter algorithm to calculate the optimal state estimate at the current moment. The optimal state estimate is used to correct the baseline offset error in the multi-source monitoring data to generate the drift correction data.

4. The method for cleaning and enhancing wind turbine data according to claim 2, characterized in that, The process of identifying missing and abnormal data segments in the drift correction data includes: Based on the disconnection records in the communication status log, identify the missing interval of the drift correction data as the missing data segment; The drift correction data is scanned using a preset anomaly detection algorithm to identify nonlinear zero-value distortion intervals or constant-value dead zone intervals as the abnormal data segments.

5. The method for cleaning and enhancing wind turbine data according to claim 1, characterized in that, The step of interpolating and imputing the missing data segments and the abnormal data segments to obtain a continuous time series includes: The spatial correlation features of neighboring wind turbines in the target wind farm and the historical time series features of the target wind turbines are extracted using the spatiotemporal kriging interpolation algorithm. Based on the spatial correlation features and the historical time series features, a basic trend sequence covering the missing data segment and the abnormal data segment is generated. The basic trend sequence is input into a pre-trained generative adversarial network to generate high-frequency nonlinear fluctuation components; The basic trend sequence is superimposed and fused with the high-frequency nonlinear fluctuation component to obtain the continuous time series.

6. The method for cleaning and enhancing wind turbine data according to claim 1, characterized in that, The process of performing multi-band decomposition and denoising on the continuous time series to obtain enhanced monitoring data includes: The continuous time series is decomposed into multiple intrinsic mode function components using the empirical mode decomposition algorithm; Calculate the Pearson correlation coefficient between each intrinsic mode function component and the continuous time series, and remove intrinsic mode function components with correlation coefficients lower than a preset threshold to obtain mid-to-high frequency intrinsic mode function components; The enhanced monitoring data is obtained by using wavelet packet denoising algorithm to perform noise threshold quantization and high-frequency filtering on the mid-to-high frequency intrinsic mode function components.

7. A wind turbine data cleaning and enhancement system, characterized in that, include: The data acquisition module is used to acquire multi-source monitoring data of the target wind turbine under extreme environments; The drift correction module is used to calculate the measurement deviation and adaptively calibrate the multi-source monitoring data using a pre-built physical mechanism model to obtain drift correction data. The reconstruction module is used to identify missing data segments and abnormal data segments in the drift correction data, and to perform interpolation imputation and sequence reconstruction on the missing data segments and abnormal data segments to obtain a continuous time series; The denoising module is used to perform multi-band decomposition and denoising processing on the continuous time series to obtain enhanced monitoring data.

8. A computer device, characterized in that, include: Processor and computer-readable storage media; A processor, adapted to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, implements a wind turbine data cleaning and enhancement method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and executed as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements a wind turbine data cleaning and enhancement method as described in any one of claims 1 to 6.

Images