CN-122014523-A - Dynamic early warning method and related device for specific faults of wind turbine generator

CN122014523ACN 122014523 ACN122014523 ACN 122014523ACN-122014523-A

Abstract

The invention discloses a dynamic early warning method for specific faults of a wind turbine generator, which relates to the field of state monitoring of wind power generation equipment and comprises the following steps of collecting real-time vibration data and real-time operation condition data of the wind turbine generator; the method comprises the steps of utilizing a pre-trained depth feature extraction model to extract vibration depth features from real-time vibration data, inputting real-time operation condition data into a pre-trained health state prediction model to obtain a predicted value of an enhanced specificity index under the current operation condition, calculating a confidence interval according to the predicted value, taking the confidence interval as a dynamic threshold range, triggering an early warning signal when the actual value of the enhanced specificity index exceeds the dynamic threshold range, and enabling the enhanced specificity index to comprise the vibration depth features and a preset traditional specificity index. The invention can solve the problems of single characteristic representation and fixed early warning threshold in the prior art.

Inventors

LIU ZHIWEN
HUANG YONGQUAN
ZHANG YANMIN
MIAO TAO
Wei Lincheng
XUE FENG

Assignees

特变电工新疆新能源股份有限公司

Dates

Publication Date: 20260512
Application Date: 20260203

Claims (10)

1. A dynamic early warning method for specific faults of a wind turbine generator is characterized by comprising the following steps: Collecting real-time vibration data and real-time operation condition data of the wind turbine generator; extracting vibration depth features from the real-time vibration data by using a pre-trained depth feature extraction model; Inputting the real-time operation condition data into a pre-trained health state prediction model to obtain a predicted value of the enhanced specificity index under the current operation condition; Triggering an early warning signal when the actual value of the enhanced specificity index exceeds the dynamic threshold range, wherein the enhanced specificity index comprises the vibration depth characteristic and a preset traditional specificity index.
2. The method for dynamically pre-warning specific faults of a wind turbine generator according to claim 1, further comprising the following steps after triggering the pre-warning signal: Inputting the real-time vibration data into a convolutional neural network for feature extraction to obtain a vibration feature vector, inputting the real-time operation condition data into a long-and-short-term memory network for feature extraction to obtain an SCADA feature vector; And inputting the time sequence of the enhanced specificity index into a time convolution network, predicting to obtain the residual service life of the wind turbine generator, and completing the service life diagnosis.
3. The wind turbine generator system specific fault dynamic early warning method according to claim 1, wherein the pre-trained depth feature extraction model is trained by the following steps: and acquiring historical vibration data of the wind turbine generator, taking the historical vibration data as an original input signal, performing unsupervised training on the one-dimensional convolution self-encoder model by minimizing reconstruction loss, obtaining a pre-trained one-dimensional convolution self-encoder model after the unsupervised training is completed, and taking an encoder in the pre-trained one-dimensional convolution self-encoder model as a pre-trained depth feature extraction model.
4. The wind turbine specific fault dynamic early warning method according to claim 1, wherein the pre-trained health state prediction model is trained by the following steps: And acquiring historical operation condition data of the wind turbine generator, taking the historical operation condition data as input, taking the actual value of the enhanced specificity index as output, performing supervised training on the long-short-period memory network, inputting the historical operation condition data into the long-short-period memory network after the supervised training is finished to obtain a predicted value of the enhanced specificity index, calculating a residual error between the predicted value and the actual value of the enhanced specificity index, determining a confidence interval offset according to the statistical distribution of the residual error, and completing the pre-training of the health state prediction model.
5. The method for dynamically pre-warning the specific faults of the wind turbine generator according to claim 4, wherein the confidence interval is calculated according to the predicted value, specifically comprising: and adding and subtracting the predicted value and the confidence interval offset to obtain a confidence interval.
6. The utility model provides a wind turbine generator system specific fault dynamic early warning system which characterized in that includes: The multi-source heterogeneous data acquisition module is used for acquiring real-time vibration data and real-time operation condition data of the wind turbine generator; The vibration depth feature extraction module is used for extracting vibration depth features from the real-time vibration data by utilizing a pre-trained depth feature extraction model; The self-adaptive threshold learning module is used for inputting the real-time operation condition data into a pre-trained health state prediction model to obtain a predicted value of the enhanced specificity index under the current operation condition; And the dynamic early warning module is used for triggering an early warning signal when the actual value of the enhanced specificity index exceeds the dynamic threshold range, wherein the enhanced specificity index comprises the vibration depth characteristic and a preset traditional specificity index.
7. The wind turbine-specific fault dynamic early warning system of claim 6, further comprising: The fault diagnosis module is used for inputting the real-time vibration data into a convolutional neural network to perform feature extraction to obtain a vibration feature vector, inputting the real-time operation condition data into a long-and-short-term memory network to perform feature extraction to obtain an SCADA feature vector; and the residual service life prediction module is used for inputting the time sequence of the enhanced specificity index into a time convolution network, predicting the residual service life of the wind turbine generator, and completing the service life diagnosis.
8. An electronic device, comprising a memory, a processor and a computer program stored in the memory and operable in the processor, wherein the processor implements a wind turbine specific fault dynamic early warning method according to any one of claims 1 to 6 when executing the computer program.
9. A computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program when executed by a processor implements a wind turbine generator specific fault dynamic early warning method according to any one of claims 1 to 6.
10. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements a wind turbine specific fault dynamic pre-warning method according to any one of claims 1-6.

Description

Dynamic early warning method and related device for specific faults of wind turbine generator Technical Field The invention relates to the field of wind power generation equipment state monitoring, in particular to a wind turbine generator specific fault dynamic early warning method and a related device. Background Currently, in the field of wind power generation equipment state monitoring and fault prediction, operational reliability monitoring of wind turbines mainly depends on a vibration-based state monitoring system. The system generally extracts traditional characteristic indexes such as time domain and frequency domain by collecting unit vibration signals, and sets a fixed threshold value according to industry standards or experiences to perform abnormal early warning so as to realize preliminary judgment of the health state of equipment and fault prevention. However, the wind turbine generator has obvious individual differences in dynamic characteristics due to different types, installation positions and operation conditions, so that the fixed threshold is difficult to be universally used, the false alarm or missing alarm of the early warning system is often caused, and the reliability is insufficient. And secondly, the characteristic characterization capability of early weak faults and composite faults is weak by relying on limited traditional indexes such as effective values, peaks and frequency spectrums, so that the accurate positioning and type identification of the faults are difficult. In addition, vibration data, SCADA operation parameters, acoustic signals and other multi-source monitoring information are often stored and analyzed independently in practical application, are mutually fractured, cannot realize effective cross verification and information complementation, form an information island, and limit the depth and the integrity of fault diagnosis. Disclosure of Invention The invention aims to provide a dynamic early warning method and a related device for specific faults of a wind turbine generator, which are used for solving the problems that the characteristic representation of the prior art is single and the early warning threshold is fixed. In order to achieve the above purpose, the invention adopts the following technical scheme: in a first aspect, a wind turbine generator specific fault dynamic early warning method includes the following steps: Collecting real-time vibration data and real-time operation condition data of the wind turbine generator; Extracting vibration depth features from real-time vibration data by using a pre-trained depth feature extraction model; inputting the real-time operation condition data into a pre-trained health state prediction model to obtain a predicted value of the enhanced specificity index under the current operation condition; triggering an early warning signal when the actual value of the enhanced specificity index exceeds the dynamic threshold range, wherein the enhanced specificity index comprises the vibration depth characteristic and a preset traditional specificity index. In some embodiments, after triggering the pre-warning signal, the method further comprises the steps of: Inputting real-time vibration data into a convolutional neural network for feature extraction to obtain vibration feature vectors, inputting real-time operation condition data into a long-and-short-term memory network for feature extraction to obtain SCADA feature vectors; And inputting the time sequence of the enhanced specificity index into a time convolution network, predicting to obtain the residual service life of the wind turbine generator, and completing the service life diagnosis. In some implementations, the pre-trained depth feature extraction model is trained by: And acquiring historical vibration data of the wind turbine generator, taking the historical vibration data as an original input signal, performing unsupervised training on the one-dimensional convolution self-encoder model by minimizing reconstruction loss, obtaining a pre-trained one-dimensional convolution self-encoder model after the unsupervised training is completed, and taking an encoder in the pre-trained one-dimensional convolution self-encoder model as a pre-trained depth feature extraction model. In some embodiments, the pre-trained state of health prediction model is trained by: And acquiring historical operation condition data of the wind turbine generator, taking the historical operation condition data as input and the actual value of the enhanced specificity index as output, performing supervised training on the long-short-period memory network, inputting the historical operation condition data into the long-short-period memory network after the supervised training is finished to obtain a predicted value of the enhanced specificity index, calculating a residual error between the predicted value and the actual value of the enhanced specificity index, determining a confidence interval