Wind power transmission chain health state dynamic grey correlation analysis method and system

Abstract

The invention relates to the field of wind power plants, in particular to a dynamic grey correlation analysis method and system for the health state of a wind power transmission chain. The method comprises the following steps: constructing a multi-source heterogeneous monitoring data set by collecting SCADA operation data, CMS vibration spectrum data and gearbox oil temperature data of a wind turbine generator transmission chain in real time; and performing sliding window division and time domain alignment on the data to generate a standardized feature sequence with time consistency. Based on a grey correlation analysis algorithm, calculating a dynamic correlation degree between each monitoring parameter and a preset fault feature sequence, and generating a correlation degree matrix; and carrying out weighted fusion on the correlation degree matrix by adopting a self-adaptive weight adjustment strategy, and constructing a transmission chain health state evaluation model. And according to a model output result, generating a real-time health degree score and fault probability distribution of the transmission chain component. And when the health degree score is lower than a dynamic threshold value, triggering a multi-level early warning mechanism and generating a maintenance priority list. And carrying out sliding window iteration updating on the evaluation model based on a newly added fault sample, optimizing a grey relational degree calculation parameter, and realizing dynamic self-adaption of health state evaluation.

Description

Technical Field

[0001] The present invention relates to the field of wind farms, and in particular to a method and system for dynamic grey correlation analysis of the health status of a wind power transmission chain. Background Art

[0002] In the wind power industry, the drive train system of a wind turbine is a critical component, and its health directly impacts the operational efficiency and maintenance costs of a wind farm. Traditional drive train health monitoring methods primarily rely on single data sources, such as SCADA operational data or CMS vibration monitoring data. While these methods can provide some fault diagnosis information, they suffer from significant limitations in the complex wind power operating environment. First, a single data source cannot fully reflect the health of the drive train, easily overlooking potential fault hazards. Second, traditional methods lack effective fusion mechanisms when processing multi-source heterogeneous data, resulting in an inability to fully utilize the correlations between data. Furthermore, existing technologies have shortcomings in fault early warning, unable to accurately assess the health status and fault risk level of drive train components in real time, making preventive maintenance difficult. Finally, existing technologies experience lags in model updating and optimization, preventing timely adjustment of model parameters based on newly detected fault samples, leading to a gradual decline in model accuracy and reliability. Therefore, there is an urgent need for a wind farm data processing and fault early warning method and system that can address these issues. Summary of the Invention

[0003] The object of the present invention is to provide a method and system for dynamic grey correlation analysis of the health status of a wind power transmission chain, so as to solve the problems in the above-mentioned background technology.

[0004] To achieve the above objectives, the following technical solutions are adopted.

[0005] A dynamic grey correlation analysis method for the health status of a wind turbine transmission chain comprises the following steps: real-time collection of SCADA operation data, CMS vibration spectrum data, and gearbox oil temperature data of a wind turbine transmission chain, construction of a multi-source heterogeneous monitoring data set, sliding window partitioning and time domain alignment of the multi-source heterogeneous data, and generation of a standardized feature sequence with time consistency; calculation of the dynamic correlation between each monitoring parameter and a preset fault feature sequence based on a grey correlation analysis algorithm, and generation of a correlation matrix; weighted fusion of the correlation matrix using an adaptive weight adjustment strategy, and construction of a transmission chain health status assessment model; generation of real-time health scores and fault probability distributions of transmission chain components based on the output results of the model; triggering a multi-level early warning mechanism and generating a maintenance priority list when the health score is lower than a dynamic threshold; iteratively updating the assessment model with a sliding window based on newly added fault samples, optimizing grey correlation calculation parameters, and achieving dynamic adaptation of health status assessment.

[0006] Optionally, the preprocessing of the multi-source heterogeneous data includes: using wavelet packet decomposition to suppress noise on SCADA data, extracting frequency domain energy features on CMS vibration signals, and separating trend terms on gearbox oil temperature data; eliminating dimensional differences through Z-score standardization to generate standardized feature vectors with timestamps; and using a dynamic time warping algorithm to align data streams with different sampling frequencies to construct a multi-dimensional input sequence that is consistent in time and space.

[0007] Optionally, the grey correlation analysis algorithm includes: selecting typical fault features in historical fault data as a reference sequence; calculating the grey correlation coefficient between each monitoring parameter sequence and the reference sequence; generating a dynamic correlation matrix by statistically calculating the correlation mean and variance through a sliding window; using the entropy weight method to determine the initial weight of each parameter, and dynamically correcting the weight distribution in combination with the real-time health score.

[0008] Optionally, the adaptive weight adjustment strategy includes: dividing the health status level according to the stability index of the correlation matrix; when the health score continuously decreases, increasing the weight coefficient of the vibration spectrum parameter; when the oil temperature fluctuates abnormally, improving the contribution of oil temperature data in the correlation calculation; dynamically adjusting the weight attenuation factor based on the parameter change rate within the sliding window to achieve adaptive optimization of the evaluation model.

[0009] Optionally, the triggering of the multi-level warning mechanism includes: constructing an exponential dynamic threshold curve based on the health score, and triggering a third-level warning when the score is lower than the threshold and the duration exceeds a preset window; combining the equipment's historical maintenance records and remaining life prediction results, generating a priority list including fault location, maintenance plan and spare parts requirements through a fuzzy decision algorithm; pushing the warning results to the mobile terminal and synchronously updating the maintenance knowledge base.

[0010] Optionally, the iterative update of the model includes: when the confidence of the newly added fault sample exceeds a preset threshold, using a sliding window mechanism to screen the correlation features in the historical data; correcting the weight distribution of the reference sequence through the cumulative deviation of the grey correlation degree; and using an incremental learning algorithm to update the decision boundary of the health status assessment model to optimize the accuracy of fault probability prediction.

[0011] Optionally, it also includes: multi-dimensional visual mapping of health scores and SCADA operating parameters to generate a three-dimensional degradation trajectory diagram of the transmission chain components; based on user interaction feedback data, optimizing the grey correlation calculation rules and warning threshold parameters through reinforcement learning algorithms to form a human-machine collaborative decision-making closed loop.

[0012] Optionally, it also includes: building a correlation model between the health status of the transmission chain and the operation and maintenance costs, dynamically generating spare parts procurement plans and maintenance resource scheduling strategies based on real-time health scores; combining wind farm meteorological forecast data to optimize the execution time window of maintenance tasks and reduce unplanned downtime losses.

[0013] The dynamic grey correlation analysis system for the health status of wind power transmission chain includes:

[0014] Data acquisition module, used to obtain SCADA operation data, CMS vibration spectrum data and gearbox oil temperature data in real time;

[0015] The preprocessing module is used to perform noise suppression, feature extraction, and time domain alignment on multi-source heterogeneous data to generate a standardized input sequence;

[0016] Grey correlation analysis module uses dynamic weight adjustment strategy to calculate the correlation matrix between each parameter and fault characteristics;

[0017] Health assessment module, which outputs real-time health scores and fault probability distribution of transmission chain components based on the correlation matrix;

[0018] Early warning decision module, used to trigger multi-level early warnings and generate maintenance priority lists;

[0019] The model optimization module iteratively updates the grey relational degree calculation parameters through a sliding window to achieve adaptive optimization of the evaluation model;

[0020] Visual interaction module, used to display the health status degradation trajectory and receive user feedback.

[0021] Optionally, the grey correlation analysis module further includes: a reference sequence generation unit for extracting typical features of historical fault data; a dynamic weight adjustment unit for correcting parameter weights in real time according to changes in health scores; the early warning decision module integrates a remaining life prediction algorithm and combines it with a maintenance knowledge base to generate an optimized path for spare parts logistics; the model optimization module has a built-in incremental learning unit for online updating of correlation calculation rules.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] The present invention provides a dynamic grey correlation analysis method for the health status of a wind turbine transmission chain. This method constructs a multi-source heterogeneous monitoring dataset by real-time acquisition of multi-source heterogeneous data from the wind turbine transmission chain, including SCADA operational data, CMS vibration spectrum data, and gearbox oil temperature data. This method performs sliding window partitioning and time domain alignment on the multi-source heterogeneous data to generate a standardized feature sequence with temporal consistency, addressing the difficulties of data fusion and low processing efficiency in existing technologies. A grey correlation analysis algorithm is used to calculate the dynamic correlation between each monitoring parameter and a preset fault feature sequence, generating a correlation matrix. This correlation matrix is then weighted and fused using an adaptive weight adjustment strategy to construct a transmission chain health status assessment model. This method can output the health scores and fault probability distributions of transmission chain components in real time. When the health score falls below a dynamic threshold, a multi-level early warning mechanism is triggered and a maintenance priority list is generated. Furthermore, the method iteratively updates the assessment model using a sliding window based on newly added fault samples, optimizing grey correlation calculation parameters and achieving dynamic self-adaptation of health status assessment. This invention offers improvements and additions in various areas, including preprocessing multi-source heterogeneous data, implementing a grey correlation analysis algorithm, implementing an adaptive weight adjustment strategy, triggering a multi-level early warning mechanism, iterative model updates, visualizing health status, and dynamically optimizing operation and maintenance costs. These improvements not only enhance the accuracy and reliability of health status assessments but also strengthen the system's intelligence and practicality, enabling it to better adapt to the complex and ever-changing wind power operating environment and providing strong technical support for the efficient operation and maintenance of wind farms. DETAILED DESCRIPTION

[0024] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that, unless there is a conflict, the embodiments and features in the embodiments of the present application can be combined with each other.

[0025] The following detailed description is an exemplary description, which is intended to provide further detailed description of the present invention. Unless otherwise indicated, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art. The terms used in the present invention are only for describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present invention.

[0026] This invention provides a method and system for dynamic grey correlation analysis of the health status of a wind turbine transmission chain. This method aims to achieve real-time monitoring of the health status of wind turbine transmission chains and provide fault warnings through the integration of multi-source heterogeneous data and dynamic grey correlation analysis. The following describes specific implementations of this method and system.

[0027] Data collection and preprocessing

[0028] First, various operational data from the wind turbine drive chain is collected in real time, including SCADA operational data, CMS vibration spectrum data, and gearbox oil temperature data. This data comes from a wide range of sources, covers key operating parameters of the wind turbine drive chain, and can comprehensively reflect the health of the drive chain. SCADA operational data primarily includes key operating parameters such as wind turbine power, speed, and temperature. It is collected in real time through the wind turbine's SCADA system, and the sampling frequency can be set according to actual needs, for example, once per second. CMS vibration spectrum data is obtained through vibration sensors installed on key components of the drive chain. These sensors can monitor component vibration in real time, and the sampling frequency is typically high, such as hundreds of times per second, to capture subtle changes in the vibration signal. Gearbox oil temperature data is obtained through temperature sensors installed inside the gearbox. This data is important for analyzing the lubrication condition and wear level of the gearbox.

[0029] The collected multi-source heterogeneous data differ in time and format, so preprocessing is required. Specifically, wavelet packet decomposition is used to suppress noise in SCADA data. Wavelet packet decomposition is used to decompose the data into multiple frequency bands, remove high-frequency noise, and retain low-frequency valid signals. Frequency domain energy feature extraction is performed on the CMS vibration signal. The vibration signal is converted from the time domain to the frequency domain through Fourier transform, and the energy characteristics of each frequency band are extracted for subsequent fault diagnosis. Trend terms are separated from the gearbox oil temperature data. The long-term trend terms in the data are removed by the moving average method, and short-term fluctuation features are extracted. Z-score standardization is used to eliminate dimensional differences, and the characteristic values of different data sources are converted into dimensionless standardized values to generate standardized feature vectors with timestamps. A dynamic time warping algorithm is used to align data streams with different sampling frequencies, construct a multidimensional input sequence that is consistent in time and space, and ensure the temporal consistency of different data sources.

[0030] Grey relational analysis and health status assessment

[0031] Next, based on the grey correlation analysis algorithm, the dynamic correlation between each monitoring parameter and the preset fault feature sequence is calculated to generate a correlation matrix. Specifically, typical fault features in historical fault data are selected as reference sequences, and the dynamic correlation matrix is generated by calculating the grey correlation coefficient between each monitoring parameter sequence and the reference sequence. The dynamic correlation matrix is generated by statistically calculating the mean and variance of the correlation through a sliding window to reflect the dynamic correlation between each monitoring parameter and the fault feature. The entropy weight method is used to determine the initial weight of each parameter, and the weight distribution is dynamically corrected in combination with the real-time health score to ensure the rationality and dynamism of the weight distribution.

[0032] An adaptive weight adjustment strategy is employed to weightedly fuse the correlation matrix and construct a transmission chain health assessment model. Specifically, health status is categorized according to the stability index of the correlation matrix. When the health score continuously decreases, the weight coefficient of the vibration spectrum parameter is increased to improve sensitivity to vibration faults. When the oil temperature fluctuates abnormally, the contribution of oil temperature data in the correlation calculation is increased to enhance the detection capability of lubrication faults. The weight attenuation factor is dynamically adjusted based on the parameter change rate within a sliding window, achieving adaptive optimization of the assessment model and ensuring that the model can dynamically adjust the weight distribution based on real-time data.

[0033] Based on the output of the evaluation model, a real-time health score and failure probability distribution for the transmission chain components are generated. When the health score falls below the dynamic threshold, a multi-level warning mechanism is triggered and a maintenance priority list is generated. Specifically, an exponential dynamic threshold curve is constructed based on the health score. When the score falls below the threshold and the duration exceeds the preset window, a three-level warning is triggered. Combining historical equipment maintenance records and remaining life prediction results, a fuzzy decision-making algorithm is used to generate a priority list that includes fault location, maintenance plan, and spare parts requirements. The warning results are pushed to mobile terminals and the maintenance knowledge base is simultaneously updated, enabling real-time transmission of warning information and dynamic scheduling of maintenance tasks.

[0034] Model iteration and system architecture

[0035] Based on newly added fault samples, the assessment model is iteratively updated using a sliding window, optimizing the gray correlation calculation parameters to achieve dynamic self-adaptation of health status assessment. Specifically, when the confidence level of a newly added fault sample exceeds a preset threshold, a sliding window mechanism is used to screen for correlation features in historical data. The weight distribution of the reference sequence is then corrected using the accumulated deviation of the gray correlation. An incremental learning algorithm is used to update the decision boundary of the health status assessment model, optimizing the accuracy of fault probability prediction and ensuring that the model can dynamically adjust its parameters based on newly added data, thereby improving its accuracy and reliability.

[0036] The present invention also provides a dynamic grey correlation analysis system for the health status of a wind power transmission chain, which includes the following modules:

[0037] The data acquisition module is used to obtain real-time SCADA operation data, CMS vibration spectrum data, and gearbox oil temperature data. This module uses multiple sensors and data interfaces to ensure the real-time and integrity of the data.

[0038] Preprocessing Module: This module performs noise suppression, feature extraction, and time-domain alignment on multi-source heterogeneous data to generate a standardized input sequence. This module uses techniques such as wavelet packet decomposition, frequency-domain energy feature extraction, and trend term separation to improve data quality, eliminate dimensional differences, and construct a multidimensional input sequence that is consistent in both time and space.

[0039] Grey Correlation Analysis Module: This module uses a dynamic weight adjustment strategy to calculate the correlation matrix between each parameter and fault characteristics. This module uses the entropy weight method to determine the initial weights for each parameter and dynamically adjusts the weight distribution based on real-time health scores to ensure the rationality and dynamism of the weight distribution.

[0040] Health Assessment Module: This module outputs real-time health scores and failure probability distributions for drive train components based on a correlation matrix. This module dynamically optimizes the health assessment model through an adaptive weight adjustment strategy, ensuring that the model can dynamically adjust parameters based on real-time data, improving its accuracy and reliability.

[0041] Early Warning Decision Module: This module triggers multi-level early warnings and generates a maintenance priority list. Using an exponential dynamic threshold curve and fuzzy decision-making algorithm, this module generates a priority list that includes fault location, repair plans, and spare parts requirements. It then pushes the early warning results to mobile devices, enabling real-time delivery of early warning information and dynamic scheduling of maintenance tasks.

[0042] Model Optimization Module: This module iteratively updates the gray correlation calculation parameters through a sliding window to achieve adaptive optimization of the assessment model. This module uses an incremental learning algorithm to update the decision boundary of the health status assessment model, optimizing the accuracy of fault probability predictions and ensuring that the model can dynamically adjust parameters based on new data, thereby improving the model's accuracy and reliability.

[0043] Visualization Interaction Module: This module displays health degradation trajectories and receives user feedback. This module generates a three-dimensional degradation trajectory diagram of the transmission chain components through multi-dimensional visualization mapping. Based on user interaction feedback, it optimizes the grey correlation calculation rules and warning threshold parameters using a reinforcement learning algorithm, forming a closed loop for human-machine collaborative decision-making.

[0044] Through the collaborative operation of the above modules, the present invention enables real-time monitoring of the health status of wind turbine transmission chains and fault warnings, improving the operational efficiency and reliability of wind turbine transmission chains. This system not only monitors the health of the transmission chain in real time but also uses intelligent algorithms to predict potential faults, enabling proactive maintenance measures and reducing both the incidence of failures and repair costs. Furthermore, by optimizing data processing and model updating, the system reduces model lag, improves real-time performance and response speed, and provides strong technical support for the intelligent operation and maintenance of wind farms.

[0045] It is understood from common technical knowledge that the present invention may be implemented by other embodiments that do not depart from its spirit or essential features. Therefore, the embodiments disclosed above are, in all respects, merely illustrative and not exclusive. All modifications within the scope of the present invention or equivalent to the scope of the present invention are intended to be encompassed by the present invention.

Claims

1. A dynamic grey correlation analysis method for the health status of a wind power transmission chain, characterized by: The following steps are involved: Real-time collection of SCADA operation data, CMS vibration spectrum data, and gearbox oil temperature data from the wind turbine drive chain to construct a multi-source heterogeneous monitoring data set; sliding window partitioning and time domain alignment are performed on the multi-source heterogeneous data to generate a standardized feature sequence with time consistency; based on the grey correlation analysis algorithm, the dynamic correlation between each monitoring parameter and the preset fault feature sequence is calculated to generate a correlation matrix; an adaptive weight adjustment strategy is used to perform weighted fusion on the correlation matrix to construct a drive chain health status assessment model; based on the output results of the model, real-time health scores and fault probability distributions of the drive chain components are generated; When the health score falls below the dynamic threshold, a multi-level warning mechanism is triggered and a maintenance priority list is generated; The evaluation model is updated iteratively with a sliding window based on newly added fault samples, and the grey relational degree calculation parameters are optimized to achieve dynamic self-adaptation of health status evaluation.

2. The method for dynamic grey correlation analysis of wind power transmission chain health status according to claim 1, characterized in that: The preprocessing of multi-source heterogeneous data includes: using wavelet packet decomposition to suppress noise on SCADA data, extracting frequency domain energy features on CMS vibration signals, and separating trend terms on gearbox oil temperature data; eliminating dimensional differences through Z-score standardization to generate standardized feature vectors with timestamps; and using a dynamic time warping algorithm to align data streams with different sampling frequencies to construct a multidimensional input sequence that is consistent in time and space.

3. The method for dynamic grey correlation analysis of wind power transmission chain health status according to claim 1, characterized in that: The grey correlation analysis algorithm includes: selecting typical fault features in historical fault data as a reference sequence; calculating the grey correlation coefficient between each monitoring parameter sequence and the reference sequence; generating a dynamic correlation matrix by statistically calculating the correlation mean and variance through a sliding window; using the entropy weight method to determine the initial weight of each parameter, and dynamically correcting the weight distribution in combination with the real-time health score.

4. The method for dynamic grey correlation analysis of wind power transmission chain health status according to claim 1, characterized in that: The adaptive weight adjustment strategy includes: dividing the health status into levels according to the stability index of the correlation matrix; increasing the weight coefficient of the vibration spectrum parameter when the health score continuously decreases; improving the contribution of oil temperature data in the correlation calculation when the oil temperature fluctuates abnormally; and dynamically adjusting the weight attenuation factor based on the parameter change rate within the sliding window to achieve adaptive optimization of the evaluation model.

5. The method for dynamic grey correlation analysis of wind power transmission chain health status according to claim 1, characterized in that: The triggering of the multi-level warning mechanism includes: constructing an exponential dynamic threshold curve based on the health score, triggering a third-level warning when the score is lower than the threshold and the duration exceeds the preset window; combining the equipment's historical maintenance records and remaining life prediction results, generating a priority list including fault location, maintenance plan and spare parts requirements through a fuzzy decision algorithm; pushing the warning results to the mobile terminal and synchronously updating the maintenance knowledge base.

6. The method for dynamic grey correlation analysis of wind power transmission chain health status according to claim 1, characterized in that: The iterative update of the model includes: when the confidence of a new fault sample exceeds a preset threshold, a sliding window mechanism is used to screen the correlation features in the historical data; the weight distribution of the reference sequence is corrected by the cumulative deviation of the grey correlation degree; and the decision boundary of the health status assessment model is updated using an incremental learning algorithm to optimize the accuracy of fault probability prediction.

7. The method for dynamic grey correlation analysis of wind power transmission chain health status according to claim 1, characterized in that: Also includes: Perform multi-dimensional visualization mapping of health scores and SCADA operating parameters to generate a three-dimensional degradation trajectory diagram of the transmission chain components; Based on user interaction feedback data, the grey correlation calculation rules and warning threshold parameters are optimized through reinforcement learning algorithms to form a closed loop of human-machine collaborative decision-making.

8. The method for dynamic grey relational analysis of wind power transmission chain health status according to claim 1, characterized in that: Also includes: Build a correlation model between the health status of the transmission chain and operation and maintenance costs, and dynamically generate spare parts procurement plans and maintenance resource scheduling strategies based on real-time health scores; Combined with wind farm meteorological forecast data, the execution time window of maintenance tasks is optimized to reduce unplanned downtime losses.

9. A system for analyzing the health status of a wind turbine transmission chain using a dynamic grey correlation analysis method based on the health status of a wind turbine transmission chain according to any one of claims 1 to 8, characterized in that: include: Data acquisition module, used to obtain SCADA operation data, CMS vibration spectrum data and gearbox oil temperature data in real time; The preprocessing module is used to perform noise suppression, feature extraction, and time domain alignment on multi-source heterogeneous data to generate a standardized input sequence; Grey correlation analysis module uses dynamic weight adjustment strategy to calculate the correlation matrix between each parameter and fault characteristics; Health assessment module, which outputs real-time health scores and fault probability distribution of transmission chain components based on the correlation matrix; Early warning decision module, used to trigger multi-level early warnings and generate maintenance priority lists; The model optimization module iteratively updates the grey relational degree calculation parameters through a sliding window to achieve adaptive optimization of the evaluation model; Visual interaction module, used to display the health status degradation trajectory and receive user feedback.

10. The wind power transmission chain health status dynamic grey correlation analysis system according to claim 9, characterized in that: The grey correlation analysis module also includes: a reference sequence generation unit for extracting typical features of historical fault data; a dynamic weight adjustment unit for correcting parameter weights in real time based on changes in health scores; the early warning decision module integrates a remaining life prediction algorithm and combines it with a maintenance knowledge base to generate an optimized spare parts logistics path; the model optimization module has a built-in incremental learning unit for online updating of correlation calculation rules.