CN117454314A - Wind turbine component running state prediction method, device, equipment and storage medium - Google Patents
Wind turbine component running state prediction method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN117454314A CN117454314A CN202311745314.XA CN202311745314A CN117454314A CN 117454314 A CN117454314 A CN 117454314A CN 202311745314 A CN202311745314 A CN 202311745314A CN 117454314 A CN117454314 A CN 117454314A
- Authority
- CN
- China
- Prior art keywords
- sensor data
- wind turbine
- feature subset
- target
- turbine component
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 58
- 230000004927 fusion Effects 0.000 claims abstract description 64
- 238000012423 maintenance Methods 0.000 claims description 30
- 238000012549 training Methods 0.000 claims description 30
- 230000015654 memory Effects 0.000 claims description 17
- 230000008859 change Effects 0.000 claims description 6
- 238000007621 cluster analysis Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 description 13
- 238000004891 communication Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000007781 pre-processing Methods 0.000 description 3
- 230000000306 recurrent effect Effects 0.000 description 3
- 230000002195 synergetic effect Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000004590 computer program Methods 0.000 description 2
- 238000007499 fusion processing Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/211—Selection of the most significant subset of features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Wind Motors (AREA)
Abstract
The application discloses a wind turbine component running state prediction method, a device, equipment and a storage medium, which belong to the technical field of wind turbine component running state prediction, and the method comprises the following steps: acquiring various sensor data; clustering a plurality of sensor data to obtain at least one feature subset; determining a target feature subset, in which the correlation degree between different sensor data meets a preset condition, from at least one feature subset; obtaining the data fusion weight of each sensor data by utilizing the target feature subset; fusing a plurality of sensor data by utilizing the data fusion weight to obtain a unit comprehensive state feature vector; and inputting the unit comprehensive state feature vector into a state prediction model to obtain the predicted state of the unit. The method and the device can improve the accuracy of the fan assembly running state prediction result.
Description
Technical Field
The present disclosure relates to the field of wind turbine component status prediction, and in particular, to a method, an apparatus, a device, and a storage medium for predicting an operation status of a wind turbine component.
Background
At present, in order to monitor the running state of a wind driven generator component and predict the failure of the wind driven generator component, the running state of the wind driven generator component is monitored and the failure is predicted mainly by utilizing a data fusion analysis mode of various sensors.
However, by means of data fusion analysis of multiple sensors, the operation state of the wind turbine assembly is monitored and faults are predicted, and the inherent relation and the differences among different sensor data are not considered, but the sensor data are subjected to simple superposition or splicing processing, so that potential relevant information among the data of how far can not be fully mined in the characteristic selection and model training process, and the operation state of the wind turbine assembly can not be accurately predicted.
Disclosure of Invention
The main purpose of the application is to provide a wind turbine component running state prediction method, a device, equipment and a storage medium, and aims to solve the technical problem that the running state of a wind turbine component cannot be accurately predicted.
To achieve the above object, the present application provides a method for predicting an operation state of a wind turbine component, including:
acquiring various sensor data; the plurality of sensor data includes at least two of vibration sensor data, sound sensor data, electrical signal sensor data, temperature sensor data, and humidity sensor data;
Clustering a plurality of sensor data to obtain at least one feature subset;
determining a target feature subset, in which the correlation degree between different sensor data meets a preset condition, from at least one feature subset;
obtaining the data fusion weight of each sensor data by utilizing the target feature subset;
fusing a plurality of sensor data by utilizing the data fusion weight to obtain a unit comprehensive state feature vector;
and inputting the unit comprehensive state feature vector into a state prediction model to obtain the predicted state of the unit.
Optionally, after the fusing the plurality of sensor data by using the data fusion weights to obtain the unit comprehensive state feature vector, the method further includes:
if any sensor data in the plurality of sensor data is monitored to change, acquiring a plurality of changed first sensor data;
clustering the plurality of first sensor data to obtain at least one first feature subset;
determining a first target feature subset, in which the correlation degree between different first sensor data meets a preset condition, from at least one first feature subset;
Obtaining first data fusion weights of the first sensor data by using the first target feature subset;
updating the data fusion weight by using the first data fusion weight to obtain an updated data fusion weight;
and fusing the first sensor data by using the updated data fusion weight to obtain a new unit comprehensive state feature vector.
Optionally, the obtaining the data fusion weight of each sensor data by using the target feature subset includes:
acquiring a training feature subset;
training supervised learning models corresponding to the sensor data by utilizing the training feature subsets to obtain at least two target supervised learning models;
inputting the target feature subset into all the target supervised learning models to obtain at least two prediction results;
determining an actual result corresponding to each predicted result from each sensor data based on each predicted result and the target feature subset;
for each prediction result, obtaining the fitting degree of a corresponding target supervised learning model based on the prediction result and the corresponding actual result;
And respectively taking each fitting degree as the data fusion weight of the corresponding sensor data.
Optionally, after the unit comprehensive state feature vector is input into a state prediction model to obtain a predicted state of the unit, the method further includes:
inputting the prediction state into an optimized maintenance algorithm model to obtain a maintenance scheme;
outputting the maintenance scheme to enable a worker to maintain the wind turbine assembly based on the maintenance scheme.
Optionally, the outputting the maintenance schedule, so that after the staff performs maintenance on the wind turbine assembly based on the maintenance schedule, the method further comprises:
acquiring a plurality of second sensor data of the maintained wind turbine assembly;
performing cluster analysis on a plurality of second sensor data to obtain at least one second feature subset;
determining a second target feature subset, in which the correlation degree between different second sensor data meets a preset condition, from at least one second feature subset;
and inputting the second target feature subset into all the target supervised learning models for training to obtain the target supervised learning models after fine adjustment.
Optionally, inputting the second target feature subset into all the target supervised learning models for training to obtain a fine-tuned target supervised learning model, including:
Acquiring the actual output power of a wind turbine component;
inputting the second target feature subset into a wind turbine component output power prediction model to obtain predicted output power of the wind turbine component;
obtaining a root mean square error between the predicted output power and the actual output power;
and if the root mean square error is greater than a preset threshold value, inputting the second target feature subset into all the target supervised learning models for training to obtain the target supervised learning models after fine adjustment.
Optionally, the supervised learning model is a transducer model.
In a second aspect, the present application provides a wind turbine component operational state prediction apparatus, the wind turbine component operational state prediction apparatus comprising:
the acquisition module is used for acquiring various sensor data; the plurality of sensor data includes at least two of vibration sensor data, sound sensor data, electrical signal sensor data, temperature sensor data, and humidity sensor data;
the clustering module is used for clustering a plurality of sensor data to obtain at least one feature subset;
a determining module, configured to determine, from at least one of the feature subsets, a target feature subset for which a correlation between different sensor data satisfies a preset condition;
The obtaining module is used for obtaining the data fusion weight of each sensor data by utilizing the target feature subset;
the data fusion module is used for fusing the plurality of sensor data by utilizing the data fusion weight to obtain a unit comprehensive state feature vector;
and the input module is used for inputting the unit comprehensive state characteristic vector into a state prediction model to obtain the predicted state of the unit.
In a third aspect, the present application provides a wind turbine component operational state prediction apparatus comprising: a processor, a memory and a wind turbine component operational state prediction program stored in the memory, which when executed by the processor, performs the steps of the wind turbine component operational state prediction method according to any one of the preceding claims.
In a fourth aspect, the present application provides a computer readable storage medium having stored thereon a wind turbine component operational state prediction program which when executed by a processor implements a wind turbine component operational state prediction method as defined in any one of the preceding claims.
According to the wind turbine component running state prediction method, various sensor data are acquired; the plurality of sensor data includes at least two of vibration sensor data, sound sensor data, electrical signal sensor data, temperature sensor data, and humidity sensor data; clustering a plurality of sensor data to obtain at least one feature subset; determining a target feature subset, in which the correlation degree between different sensor data meets a preset condition, from at least one feature subset; obtaining the data fusion weight of each sensor data by utilizing the target feature subset; fusing a plurality of sensor data by utilizing the data fusion weight to obtain a unit comprehensive state feature vector; and inputting the unit comprehensive state feature vector into a state prediction model to obtain a predicted state of the unit, so that the accuracy of a fan assembly running state prediction result can be improved.
Compared with the prior art, the method has the advantages that the integrated state feature vector of the unit is obtained by directly carrying out simple superposition or splicing processing on the sensor data, after at least one feature subset of each sensor is determined, the target feature subset which can most represent the running state of the wind turbine component and can generate a synergistic effect in the state prediction model is determined from the feature subset, further, based on the target feature subset, the internal coupling structure and the association mode between the sensor data are deeply excavated, different weights are given to different sensor data, finer and accurate data fusion is realized, data redundancy is reduced, and therefore the simple superposition or splicing processing on the sensor data is avoided, the accuracy of the prediction result is improved, and the predicted running state of the wind turbine component is more consistent with the real running state of the wind turbine component.
Drawings
FIG. 1 is a schematic diagram of a wind turbine component operational state prediction apparatus of the present application;
FIG. 2 is a flow chart of a first embodiment of a method for predicting an operational state of a wind turbine component according to the present application;
FIG. 3 is a functional block diagram of a first embodiment of a wind turbine component operating state prediction apparatus according to the present application.
The realization, functional characteristics and advantages of the present application will be further described with reference to the embodiments, referring to the attached drawings.
Detailed Description
It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present application.
At present, in order to monitor the running state of a wind driven generator component and predict the failure of the wind driven generator component, the running state of the wind driven generator component is monitored and the failure is predicted mainly by utilizing a data fusion analysis mode of various sensors.
However, by means of data fusion analysis of multiple sensors, the operation states of the wind turbine components are monitored and faults are predicted, internal relation and differences among different sensor data are not considered, and the sensor data are subjected to simple superposition or splicing processing, so that potential relevant information among multi-source data cannot be fully mined in the characteristic selection and model training process, and the prediction result of the wind turbine components is inaccurate.
Compared with the prior art, the method has the advantages that the simple superposition or splicing processing is directly carried out on the sensor data to obtain the unit comprehensive state feature vector, after at least one feature subset of each sensor is determined, the target feature subset which can most represent the running state of the wind turbine component and can generate a synergistic effect in the state prediction model is determined from the feature subset, further, based on the target feature subset, the intrinsic coupling structure and the association mode between the sensor data are deeply excavated, different weights are given to different sensor data, finer and accurate data fusion is realized, and data redundancy is reduced, so that the simple superposition or splicing processing is avoided on the sensor data directly, the accuracy of a prediction result is improved, and the predicted running state of the wind turbine component is more consistent with the real running state of the wind turbine component.
Referring to fig. 1, fig. 1 is a schematic structural diagram of a wind turbine component operation state prediction apparatus of a hardware operation environment according to an embodiment of the present application.
As shown in fig. 1, the wind turbine component operation state prediction apparatus may include: a processor 1001, such as a central processing unit (Central Processing Unit, CPU), a communication bus 1002, a user interface 1003, a network interface 1004, a memory 1005. Wherein the communication bus 1002 is used to enable connected communication between these components. The user interface 1003 may include a Display, an input unit such as a Keyboard (Keyboard), and the optional user interface 1003 may further include a standard wired interface, a wireless interface. The network interface 1004 may optionally include a standard wired interface, a WIreless interface (e.g., a WIreless-FIdelity (WI-FI) interface). The Memory 1005 may be a high-speed random access Memory (Random Access Memory, RAM) Memory or a stable nonvolatile Memory (NVM), such as a disk Memory. The memory 1005 may also optionally be a storage device separate from the processor 1001 described above.
It will be appreciated by those skilled in the art that the configuration shown in FIG. 1 is not limiting of the wind turbine assembly operational state prediction apparatus and may include more or fewer components than shown, or certain components may be combined, or a different arrangement of components.
As shown in fig. 1, an operating system, a data storage module, a network communication module, a user interface module, and a wind turbine component operation state prediction program may be included in the memory 1005 as one type of storage medium.
In the wind turbine component operational state prediction apparatus shown in FIG. 1, the network interface 1004 is mainly used for data communication with a network server; the user interface 1003 is mainly used for data interaction with a user; the processor 1001 and the memory 1005 in the wind turbine component running state prediction device can be arranged in the wind turbine component running state prediction device, and the wind turbine component running state prediction device calls the wind turbine component running state prediction program stored in the memory 1005 through the processor 1001 and executes the wind turbine component running state prediction method provided by the embodiment of the application.
Based on the hardware structure of the wind turbine component operation state prediction device, but not limited to the hardware structure, the application provides a first embodiment of a wind turbine component operation state prediction method. Referring to FIG. 2, FIG. 2 is a flow chart illustrating a first embodiment of a method for predicting an operational state of a wind turbine assembly according to the present application.
It should be noted that although a logical order is depicted in the flowchart, in some cases the steps depicted or described may be performed in a different order than presented herein.
In this embodiment, the method for predicting the running state of the wind turbine component includes:
and step S10, acquiring various sensor data.
The plurality of sensor data includes at least two of vibration sensor data, sound sensor data, electrical signal sensor data, temperature sensor data, and humidity sensor data.
The execution main body of the wind turbine component running state prediction method is wind turbine component running state prediction equipment, a wind turbine component running state prediction program is stored on the wind turbine component running state prediction equipment, and when the wind turbine component running state prediction equipment executes the wind turbine component running state prediction program, the wind turbine component running state prediction method of the embodiment is realized.
In particular, a plurality of different types of sensors may be mounted on the wind turbine assembly, such as temperature sensors, vibration sensors, humidity sensors, etc. After the sensor data such as the temperature sensor data, the vibration sensor data, the humidity sensor data and the like are collected, each sensor may perform data transmission with the wind turbine component operation state prediction device through a serial communication protocol, for example, UART (Universal Asynchronous Receiver/transceiver), SPI (Serial Peripheral Interface ) or I2C (Inter-Integrated Circuit, serial two-wire interface), so as to transmit the collected various sensor data to the wind turbine component operation state prediction device, or may also transmit various sensor data to the wind turbine component operation state prediction device through wireless technologies such as Wi-Fi, bluetooth, zigbee (ad hoc mesh network), lorewan (Long Range Wide Area Network, low power consumption wide area network) and the like, which is not limited in this embodiment.
And step S20, clustering a plurality of sensor data to obtain at least one feature subset.
And step S30, determining a target feature subset, of which the correlation degree between different sensor data meets a preset condition, from at least one feature subset.
In this embodiment, the non-supervised learning method may be used to cluster a plurality of sensor data, and compared with the method of using the supervised learning method to cluster sensor data based on the labels or target outputs of the sensor data, the method of using the non-supervised learning method to cluster a plurality of sensor data may better discover the available structures, patterns, and rules among the sensor data.
In this embodiment, the feature subset may be a feature that is related to the sensor data or has a prediction capability after removing redundant or irrelevant features from the sensor data, and in order to obtain a feature with a better prediction capability, further, a feature subset that is most related to the sensor data or has a prediction capability, that is, a target feature subset with a degree of correlation that satisfies a preset condition, may be determined from at least one feature subset.
Specifically, after the wind turbine component running state prediction device acquires various sensor data, preprocessing such as data cleaning and abnormal value detection can be performed on the various sensor data so as to ensure the accuracy of the sensor data. Further, after preprocessing the sensor data, a conventional method such as variance and correlation, or a machine learning method such as a feature selection algorithm and information gain can be utilized to select appropriate features from the sensor data according to specific requirements and problems. It will be appreciated that different sensors have different units of measurement and ranges between them, so that after preprocessing and feature selection of each sensor data, it is also possible to normalize each sensor data in order to eliminate deviations due to different scales.
After the standardized processing is performed on each sensor data, different clustering clusters can be obtained based on the similarity or distance between each sensor data by using clustering algorithms such as K-means, hierarchical clustering, DBSCAN (Density-Based Spatial Clustering of Applications with Noise, density-based clustering algorithm and application of noise points) and the like. Further, after the clustering result is obtained, association rule mining can be performed based on methods such as distance between clusters, similarity measurement and the like, so as to determine clusters with high association. After determining the cluster with high relevance, feature subsets can be extracted from each cluster by using different methods to obtain at least one feature subset, such as calculating the average value and variance of features in each cluster or selecting features with larger variance in the cluster. Finally, after obtaining at least one feature subset, a feature selection method, a feature importance evaluation method or domain expert knowledge may be used to determine a target feature subset from at least one of the feature subsets, where the correlation between different sensor data meets a preset condition.
And S40, obtaining the data fusion weight of each sensor data by utilizing the target feature subset.
It can be understood that, for determining the problem of fusion of multiple sensor data, the weighted average idea is mainly utilized to perform data fusion processing on the multiple sensor data so as to obtain the unit comprehensive state feature vector. However, the weighted average concept is utilized to perform fusion processing on the multiple sensor data to obtain a unit comprehensive state feature vector, which is essentially that the simple superposition or stitching processing is performed on the multiple sensor data, where the unit comprehensive state feature vector obtained by the simple superposition or stitching processing may result in a large amount of redundant data, and when the operation state of the wind turbine component is predicted, the large amount of redundant data may interfere with the prediction result, so that the accuracy of the prediction result is low, so in order to accurately predict the operation state of the wind turbine component, further, as an optional implementation manner, step S40 specifically includes:
step S401, obtaining a training feature subset.
And step S402, training the supervised learning models corresponding to the sensor data by utilizing the training feature subsets to obtain at least two target supervised learning models.
And S403, inputting the target feature subset into all the target supervised learning models to obtain at least two prediction results.
Step S404, determining an actual result corresponding to each predicted result from each sensor data based on each predicted result and the target feature subset.
Step S405, for each prediction result, obtaining a fitting degree of a corresponding target supervised learning model based on the prediction result and the corresponding actual result.
Step S406, each fitting degree is used as the data fusion weight of the corresponding sensor data.
In this embodiment, the training feature subset may be a plurality of historical target feature subsets, that is, historical target feature subsets corresponding to historical sensor data are used as the training feature subsets to train the supervised learning model corresponding to each sensor data, so that the target supervised learning model better adapts to the target feature subsets of each sensor data.
Specifically, the historical target feature subset can be used as a training sample set, the target feature subset is used as a test sample set, the training sample set is utilized to train the supervised learning model corresponding to each sensor data, and after each target supervised learning model is obtained, the test sample set is input into the target supervised learning model to obtain a prediction result.
Further, after the prediction results of each target supervised learning model are obtained, based on each prediction result and the target feature subset, determining an actual result corresponding to each prediction result from each sensor data, that is, determining an actual target value or a label corresponding to each test result, if the prediction result is the predicted vibration frequency of the wind turbine component blade, and at the moment, the actual result corresponding to the test result is the actual vibration frequency of the wind turbine component blade.
After obtaining the actual and predicted results, a fitness (R-squared) of the corresponding target supervised learning model may be calculated, i.e., R-squared=1- (SSR/SST), where SSR is the sum of squares of residuals (Sum of Squares of Residuals), SST is the sum of squares of total (Total Sum of Squares), ssr= Σ (y_pred-y_true) 2 Where y_pred is the predicted outcome of each target supervised learning model and y_true is the actual outcome corresponding to each predicted outcome. Sst= Σ (y_true-mean (y_true)) 2 Where mean (y_true) is the average of the actual results. After the fitting degrees are obtained, each fitting degree can be used as the data fusion weight of the corresponding sensor data.
In this embodiment, the supervised learning model may be a transducer model, or may be another model, which is not limited in this embodiment.
And S50, fusing a plurality of sensor data by utilizing the data fusion weights to obtain a unit comprehensive state feature vector.
After the data fusion weights of the sensor data are obtained, each sensor data can be linearly or nonlinearly combined according to the data fusion weights to obtain a unit comprehensive state feature vector, namely a unit comprehensive state feature vector=data fusion weight 1×sensor data 1+data fusion weight 2×sensor data 2+ … … +data fusion weight n×sensor data n.
It should be noted that, the sensor data collected by each sensor is not fixed, but fluctuates with changes of some external factors, for example, for summer and winter, the sensor data collected by the summer temperature sensor and the humidity sensor are higher than the sensor data collected by the winter temperature sensor and the humidity sensor. In windless weather, breeze weather and windy weather, the vibration sensor data collected by the vibration sensor are different, and the arrangement sequence of the vibration sensor data from big to small is as follows: strong wind weather > gentle wind weather > no wind weather.
Thus, further, as an alternative embodiment, to more accurately predict the operation state of the wind turbine component, after step S50, it further includes:
step S51, if any sensor data of the plurality of sensor data is monitored to change, acquiring a plurality of changed first sensor data.
Step S52, clustering the plurality of first sensor data to obtain at least one first feature subset.
Step S53, determining a first target feature subset, in which the correlation degree between different first sensor data satisfies a preset condition, from at least one first feature subset.
And step S54, obtaining first data fusion weights of the first sensor data by using the first target feature subset.
And step S55, updating the data fusion weight by using the first data fusion weight to obtain the updated data fusion weight.
And step S56, fusing the first sensor data by utilizing the updated data fusion weight to obtain a new unit comprehensive state feature vector.
It can be understood that, based on the foregoing, the sensor data collected by the sensors will change with the change of the external factors, but in practical application, the sensor data collected by each sensor will fluctuate in the normal fluctuation range, so when any one of the sensor data fluctuates in the corresponding normal fluctuation range, the data fusion weight of each sensor data may not be adjusted at this time. However, when the sensor data is not within the normal fluctuation range corresponding to the sensor data, the changed various sensor data, namely, the various first sensor data, needs to be acquired, and the various first sensor data are clustered to obtain at least one first feature subset. After at least one first feature subset is obtained, a first target feature subset, in which the correlation degree between different sensor data meets the preset condition, is determined from the at least one first feature subset, the first data fusion weight of each first sensor data is obtained by utilizing the first target feature subset, the data fusion weight is updated by utilizing the first data fusion weight, the updated data fusion weight is obtained, and finally the updated data fusion weight is utilized to fuse the first sensor data, so that a new unit comprehensive state feature vector is obtained.
In the embodiment, through real-time monitoring of the sensor data, the relevance change of the sensor data can be found in time, so that the data fusion weight of data fusion is adjusted in real time, the unit comprehensive state feature vector is always matched with the current unit state, and the accuracy of a prediction result is greatly improved.
And step S60, inputting the unit comprehensive state feature vector into a state prediction model to obtain a predicted state of the unit.
In this embodiment, the state prediction model may be a recurrent neural network (RNN, recurrent Neural Network) model, such as an LSTM (Long Short-Term Memory) model or a GRU (Gated Recurrent Unit, gated loop unit) model.
Specifically, the historical unit comprehensive state feature vector of the fan assembly and the corresponding target state can be used as a training data set, a proper deep learning network architecture such as LSTM or GRU is selected, a proper layer number and a proper neuron number are set, the historical unit comprehensive state feature vector is used as input, and an output layer of a model is defined to predict the state of the unit so as to obtain an initial model. Further, the initial model is trained by using the training data set, and model parameters can be optimized through a minimum loss function in the training process, so that the prediction performance is improved.
After the initial model is trained, the accuracy of the trained initial model can be verified by using a verification data set, namely, the accuracy of the trained initial model can be estimated based on error indexes such as root mean square error or average absolute error between a predicted state and an actual state. When the accuracy of the initial model after training meets the requirement, the model is used as a state prediction model, and the unit comprehensive state feature vector is input into the state prediction model to obtain the prediction state of the unit.
Further, after the predicted state of the unit is obtained, if the predicted state is normal operation, no measures can be taken; if the predicted state is a fault operation, a maintenance scheme corresponding to the fault needs to be determined, so as an optional implementation manner, after step S60, the method further includes:
and step S61, inputting the prediction state into an optimized maintenance algorithm model to obtain a maintenance scheme.
And step S62, outputting the maintenance scheme so that a worker maintains the wind turbine assembly based on the maintenance scheme.
In this embodiment, when the predicted state is normal, no corresponding maintenance measures are needed at this time, and when the predicted state is faulty, in order to quickly determine a corresponding maintenance scheme, the predicted state with the faulty predicted state may be input into an optimized maintenance algorithm model, so as to determine an optimal shutdown maintenance opportunity and maintenance scheme, so as to achieve the purposes of prolonging the service life of the unit and reducing the operation and maintenance costs. Meanwhile, the maintenance scheme is determined by using the optimized maintenance algorithm model, so that the workload of operation and maintenance personnel can be reduced, the operation and maintenance personnel can concentrate on more complex fault diagnosis and maintenance work, and the repetitive work is reduced.
It will be appreciated that, after maintenance of the wind turbine component, the sensor data collected by the sensor in the normal operation state of the wind turbine component is necessarily different from the collected sensor data in the failure operation state of the wind turbine component, and the target feature subset of the sensor data in the normal operation state is also necessarily different, so, in order to make the supervised learning model better adapt to the target feature subset of the new sensor data, further, as an optional embodiment, after step S62, the method further includes:
and step S63, acquiring various second sensor data of the maintained wind turbine assembly.
And step S64, performing cluster analysis on a plurality of second sensor data to obtain at least one second feature subset.
Step S65, determining a second target feature subset, in which the correlation degree between the different second sensor data satisfies a preset condition, from at least one second feature subset.
And step S66, inputting the second target feature subset into all the target supervised learning models for training to obtain the target supervised learning models after fine adjustment.
It should be noted that, when training all the target supervised learning models by using the second target feature subset to obtain the target supervised learning model after fine tuning, the determination needs to be performed according to the actual situation, so further, as an optional implementation manner, step S66 specifically includes:
Step S661, obtaining the actual output power of the wind turbine component.
Step S662, inputting the second target feature subset into a wind turbine component output power prediction model to obtain the predicted output power of the wind turbine component.
Step S663, obtaining the root mean square error between the predicted output power and the actual output power.
And step S664, inputting the second target feature subset into all the target supervised learning models for training if the root mean square error is greater than a preset threshold value, and obtaining the target supervised learning model after fine adjustment.
In this embodiment, the fan assembly power prediction model may be constructed based on wind turbine history data and a machine learning algorithm, and the specific construction method refers to the foregoing operations and is not described herein.
In this embodiment, a power meter may be installed at the output end of the wind turbine component or on the connection grid, and after the power meter collects the actual output power of the wind turbine component, the power meter may transmit the actual output power to the wind turbine component operation state prediction device in the foregoing manner.
Specifically, after the actual output power of the wind turbine component is obtained, the second target feature subset may be input into the wind turbine component output power prediction model, and the wind turbine component output power prediction model may calculate the corresponding output power, that is, the predicted output power, according to the input second target feature subset.
Further, after obtaining the actual output power and the predicted output power of the wind turbine component, the formula may be based on:root mean square error (Root Mean Square Error, RMSE) is calculated, where n represents the number of samples, y_i represents the actual output power of the ith sample, and yhat_i represents the predicted output power of the ith sample. When the root mean square error is greater than a preset threshold, it may be determined that the performance of the current supervised learning model may not meet the requirements, and a fine-tuning (fine-tuning) method may be employed to retrain the model in order to improve the performance of the model.
Specifically, the second target feature subset is used as input data, all target supervised learning models are connected together, the connected target supervised learning models are trained, the weights of all target supervised learning models are used as initial weights, the weights of the target supervised learning models are updated through a back propagation algorithm until the supervised models reach a convergence state, whether the performance of the target supervised learning models after fine tuning reaches an expected threshold value is checked, when the performance of the target supervised learning models reaches the expected threshold value, training is stopped, the target supervised learning models after fine tuning are obtained, and if the performance of the target supervised learning models does not reach the expected threshold value, training is continued until the target supervised learning models reach the expected threshold value.
In the present embodiment, by acquiring a plurality of sensor data; the plurality of sensor data includes at least two of vibration sensor data, sound sensor data, electrical signal sensor data, temperature sensor data, and humidity sensor data; clustering a plurality of sensor data to obtain at least one feature subset; determining a target feature subset, in which the correlation degree between different sensor data meets a preset condition, from at least one feature subset; obtaining the data fusion weight of each sensor data by utilizing the target feature subset; fusing a plurality of sensor data by utilizing the data fusion weight to obtain a unit comprehensive state feature vector; and inputting the unit comprehensive state feature vector into a state prediction model to obtain a predicted state of the unit, so that the accuracy of a fan assembly running state prediction result can be improved.
Compared with the prior art, the method has the advantages that the integrated state feature vector of the unit is obtained by directly carrying out simple superposition or splicing processing on the sensor data, after at least one feature subset of each sensor is determined, the target feature subset which can most represent the running state of the wind turbine component and can generate a synergistic effect in the state prediction model is determined from the feature subset, further, based on the target feature subset, the internal coupling structure and the association mode between the sensor data are deeply excavated, different weights are given to different sensor data, finer and accurate data fusion is realized, data redundancy is reduced, and therefore the simple superposition or splicing processing on the sensor data is avoided, the accuracy of the prediction result is improved, and the predicted running state of the wind turbine component is more in accordance with the real running state of the wind turbine component.
Based on the same application conception, the wind turbine component running state prediction device is provided, and referring to fig. 3, fig. 3 is a schematic diagram of functional modules of the wind turbine component running state prediction device.
The wind turbine component running state prediction device comprises:
an acquisition module 10 for acquiring various sensor data; the plurality of sensor data includes at least two of vibration sensor data, sound sensor data, electrical signal sensor data, temperature sensor data, and humidity sensor data;
a clustering module 20, configured to cluster a plurality of sensor data to obtain at least one feature subset;
a determining module 30, configured to determine, from at least one of the feature subsets, a target feature subset in which a correlation between different sensor data satisfies a preset condition;
an obtaining module 40, configured to obtain a data fusion weight of each sensor data by using the target feature subset;
the data fusion module 50 is configured to fuse a plurality of sensor data by using the data fusion weights, so as to obtain a unit comprehensive state feature vector;
the input module 60 is configured to input the unit comprehensive state feature vector into a state prediction model to obtain a predicted state of the unit.
It should be noted that, more modules may be further provided in the wind turbine component operation state prediction device, and the technical effects achieved by the embodiments of the wind turbine component operation state prediction device in this embodiment may refer to various implementations of the wind turbine component operation state prediction method in the foregoing embodiments, which are not described herein again.
In addition, the embodiment of the application also provides a computer storage medium, wherein the storage medium stores a wind turbine component running state prediction program, and the wind turbine component running state prediction program realizes the steps of the wind turbine component running state prediction method when being executed by a processor. Therefore, a detailed description will not be given here. In addition, the description of the beneficial effects of the same method is omitted. For technical details not disclosed in the embodiments of the computer-readable storage medium according to the present application, please refer to the description of the method embodiments of the present application. As an example, the program instructions may be deployed to be executed on one computing device or on multiple computing devices at one site or distributed across multiple sites and interconnected by a communication network.
Those skilled in the art will appreciate that implementing all or part of the above-described methods may be accomplished by way of computer programs, which may be stored on a computer-readable storage medium, and which, when executed, may comprise the steps of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random access Memory (Random AccessMemory, RAM), or the like.
It should be further noted that the above-described apparatus embodiments are merely illustrative, where elements described as separate elements may or may not be physically separate, and elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the drawings of the embodiment of the device provided by the application, the connection relation between the modules represents that the modules have communication connection therebetween, and can be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above description of the embodiments, it will be apparent to those skilled in the art that the present application may be implemented by means of software plus necessary general purpose hardware, or of course may be implemented by dedicated hardware including application specific integrated circuits, dedicated CPUs, dedicated memories, dedicated components and the like. Generally, functions performed by computer programs can be easily implemented by corresponding hardware, and specific hardware structures for implementing the same functions can be varied, such as analog circuits, digital circuits, or dedicated circuits. However, a software program implementation is a preferred embodiment in many cases for the present application. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a readable storage medium, such as a floppy disk, a usb disk, a removable hard disk, a Read-only memory (ROM), a random-access memory (RAM, randomAccessMemory), a magnetic disk or an optical disk of a computer, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method of the embodiments of the present application.
The foregoing description is only of the preferred embodiments of the present application, and is not intended to limit the scope of the claims, and all equivalent structures or equivalent processes using the descriptions and drawings of the present application, or direct or indirect application in other related technical fields are included in the scope of the claims of the present application.
Claims (10)
1. A method of predicting an operational state of a wind turbine component, the method comprising:
acquiring various sensor data; the plurality of sensor data includes at least two of vibration sensor data, sound sensor data, electrical signal sensor data, temperature sensor data, and humidity sensor data;
clustering a plurality of sensor data to obtain at least one feature subset;
determining a target feature subset, in which the correlation degree between different sensor data meets a preset condition, from at least one feature subset;
obtaining the data fusion weight of each sensor data by utilizing the target feature subset;
fusing a plurality of sensor data by utilizing the data fusion weight to obtain a unit comprehensive state feature vector;
and inputting the unit comprehensive state feature vector into a state prediction model to obtain the predicted state of the unit.
2. The method for predicting an operational state of a wind turbine component according to claim 1, wherein after the merging of the plurality of sensor data by using the data merging weights to obtain the set integrated state feature vector, the method further comprises:
if any sensor data in the plurality of sensor data is monitored to change, acquiring a plurality of changed first sensor data;
clustering the plurality of first sensor data to obtain at least one first feature subset;
determining a first target feature subset, in which the correlation degree between different first sensor data meets a preset condition, from at least one first feature subset;
obtaining first data fusion weights of the first sensor data by using the first target feature subset;
updating the data fusion weight by using the first data fusion weight to obtain an updated data fusion weight;
and fusing the first sensor data by using the updated data fusion weight to obtain a new unit comprehensive state feature vector.
3. The method of claim 1, wherein said obtaining data fusion weights for each of said sensor data using said target feature subset comprises:
Acquiring a training feature subset;
training supervised learning models corresponding to the sensor data by utilizing the training feature subsets to obtain at least two target supervised learning models;
inputting the target feature subset into all the target supervised learning models to obtain at least two prediction results;
determining an actual result corresponding to each predicted result from each sensor data based on each predicted result and the target feature subset;
for each prediction result, obtaining the fitting degree of a corresponding target supervised learning model based on the prediction result and the corresponding actual result;
and respectively taking each fitting degree as the data fusion weight of the corresponding sensor data.
4. A method for predicting an operational state of a wind turbine assembly according to claim 3, wherein after said inputting said aggregate state feature vector into a state prediction model to obtain a predicted state of the aggregate, said method further comprises:
inputting the prediction state into an optimized maintenance algorithm model to obtain a maintenance scheme;
outputting the maintenance scheme to enable a worker to maintain the wind turbine assembly based on the maintenance scheme.
5. The method of claim 4, wherein the outputting the maintenance schedule causes a worker to maintain the wind turbine assembly based on the maintenance schedule, the method further comprising:
acquiring a plurality of second sensor data of the maintained wind turbine assembly;
performing cluster analysis on a plurality of second sensor data to obtain at least one second feature subset;
determining a second target feature subset, in which the correlation degree between different second sensor data meets a preset condition, from at least one second feature subset;
and inputting the second target feature subset into all the target supervised learning models for training to obtain the target supervised learning models after fine adjustment.
6. The method for predicting an operational state of a wind turbine assembly according to claim 5, wherein inputting the second subset of target features into all of the target supervised learning models for training to obtain the trimmed target supervised learning models comprises:
acquiring the actual output power of a wind turbine component;
inputting the second target feature subset into a wind turbine component output power prediction model to obtain predicted output power of the wind turbine component;
Obtaining a root mean square error between the predicted output power and the actual output power;
and if the root mean square error is greater than a preset threshold value, inputting the second target feature subset into all the target supervised learning models for training to obtain the target supervised learning models after fine adjustment.
7. A method of predicting the operational state of a wind turbine component according to claim 3, wherein the supervised learning model is a transducer model.
8. A wind turbine component operational state prediction apparatus, characterized in that the wind turbine component operational state prediction apparatus comprises:
the acquisition module is used for acquiring various sensor data; the plurality of sensor data includes at least two of vibration sensor data, sound sensor data, electrical signal sensor data, temperature sensor data, and humidity sensor data;
the clustering module is used for clustering a plurality of sensor data to obtain at least one feature subset;
a determining module, configured to determine, from at least one of the feature subsets, a target feature subset for which a correlation between different sensor data satisfies a preset condition;
the obtaining module is used for obtaining the data fusion weight of each sensor data by utilizing the target feature subset;
The data fusion module is used for fusing the plurality of sensor data by utilizing the data fusion weight to obtain a unit comprehensive state feature vector;
and the input module is used for inputting the unit comprehensive state characteristic vector into a state prediction model to obtain the predicted state of the unit.
9. A wind turbine component operational state prediction apparatus, comprising: a processor, a memory and a wind turbine component operational state prediction program stored in the memory, which when executed by the processor, implements the steps of the wind turbine component operational state prediction method according to any one of claims 1 to 7.
10. A computer-readable storage medium, wherein a wind turbine component operation state prediction program is stored on the computer-readable storage medium, which when executed by a processor, implements the wind turbine component operation state prediction method according to any one of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311745314.XA CN117454314B (en) | 2023-12-19 | 2023-12-19 | Wind turbine component running state prediction method, device, equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311745314.XA CN117454314B (en) | 2023-12-19 | 2023-12-19 | Wind turbine component running state prediction method, device, equipment and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN117454314A true CN117454314A (en) | 2024-01-26 |
CN117454314B CN117454314B (en) | 2024-03-05 |
Family
ID=89589442
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311745314.XA Active CN117454314B (en) | 2023-12-19 | 2023-12-19 | Wind turbine component running state prediction method, device, equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117454314B (en) |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106503731A (en) * | 2016-10-11 | 2017-03-15 | 南京信息工程大学 | A kind of based on conditional mutual information and the unsupervised feature selection approach of K means |
CN111679654A (en) * | 2020-06-17 | 2020-09-18 | 衢州职业技术学院 | Electric drive system vibration signal diagnosis method and device based on neural network |
US20210270244A1 (en) * | 2020-03-02 | 2021-09-02 | Wuhan University | Method and system for fault diagnosis of gearbox of wind turbine generator |
CN113359682A (en) * | 2021-06-30 | 2021-09-07 | 西安力传智能技术有限公司 | Equipment fault prediction method, device, equipment fault prediction platform and medium |
CN114239885A (en) * | 2022-01-11 | 2022-03-25 | 中国科学院深圳先进技术研究院 | Operation fault prediction method and device |
CN116845868A (en) * | 2023-06-30 | 2023-10-03 | 云南电网有限责任公司 | Distributed photovoltaic power prediction method and device, electronic equipment and storage medium |
CN117034001A (en) * | 2023-06-26 | 2023-11-10 | 中国石油大学(北京) | Wind turbine generator system fault prediction method and device and electronic equipment |
-
2023
- 2023-12-19 CN CN202311745314.XA patent/CN117454314B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106503731A (en) * | 2016-10-11 | 2017-03-15 | 南京信息工程大学 | A kind of based on conditional mutual information and the unsupervised feature selection approach of K means |
US20210270244A1 (en) * | 2020-03-02 | 2021-09-02 | Wuhan University | Method and system for fault diagnosis of gearbox of wind turbine generator |
CN111679654A (en) * | 2020-06-17 | 2020-09-18 | 衢州职业技术学院 | Electric drive system vibration signal diagnosis method and device based on neural network |
CN113359682A (en) * | 2021-06-30 | 2021-09-07 | 西安力传智能技术有限公司 | Equipment fault prediction method, device, equipment fault prediction platform and medium |
CN114239885A (en) * | 2022-01-11 | 2022-03-25 | 中国科学院深圳先进技术研究院 | Operation fault prediction method and device |
CN117034001A (en) * | 2023-06-26 | 2023-11-10 | 中国石油大学(北京) | Wind turbine generator system fault prediction method and device and electronic equipment |
CN116845868A (en) * | 2023-06-30 | 2023-10-03 | 云南电网有限责任公司 | Distributed photovoltaic power prediction method and device, electronic equipment and storage medium |
Non-Patent Citations (2)
Title |
---|
董晓峰 等: "基于聚类分析与加权模糊逻辑的汽轮机组振动故障诊断方法研究", 《热能动力工程》, vol. 26, no. 3, 31 May 2011 (2011-05-31), pages 275 - 279 * |
赵仕策;赵洪山;寿佩瑶;: "智能电力设备关键技术及运维探讨", 《电力系统自动化》, vol. 44, no. 20, 25 October 2020 (2020-10-25), pages 1 - 10 * |
Also Published As
Publication number | Publication date |
---|---|
CN117454314B (en) | 2024-03-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US8606418B1 (en) | Wind prediction for wind farms through the use of weather radar | |
CN107707431A (en) | The data safety monitoring method and system of a kind of facing cloud platform | |
CN111143173A (en) | Server fault monitoring method and system based on neural network | |
US11016515B2 (en) | System operation decision-making assistance device and method | |
US20210151989A1 (en) | Supply and demand adjustment monitoring device of power grid and supply and demand adjustment monitoring method for power grid | |
KR20180039205A (en) | Method and device for intelligent fault diagnosis using improved rtc(real-time contrasts) method | |
US20190048740A1 (en) | Turbine diagnostic feature selection system | |
US20210232104A1 (en) | Method and system for identifying and forecasting the development of faults in equipment | |
US11619674B2 (en) | State estimation method and state estimation apparatus | |
CN110659672B (en) | Method and device for predicting step-by-step uncertainty of output of wind turbine generator | |
CN111949429A (en) | Server fault monitoring method and system based on density clustering algorithm | |
CN114239734A (en) | Distributed vehicle-mounted health management system | |
CN116125204A (en) | Fault prediction system based on power grid digitization | |
CN116307641A (en) | Digital power plant-oriented resource collaborative scheduling management method and system | |
CN116428124A (en) | Fault diagnosis method based on large number of equipment of same type | |
CN117390403B (en) | Power grid fault detection method and system for new energy lighthouse power station | |
Hassani et al. | Intelligent decision support and fusion models for fault detection and location in power grids | |
CN117454314B (en) | Wind turbine component running state prediction method, device, equipment and storage medium | |
CN117763449A (en) | Multi-source fault autonomous positioning and classifying method | |
Alghuried et al. | Anomaly detection in large-scale networks: A state-space decision process | |
CN107590008A (en) | A kind of method and system that distributed type assemblies reliability is judged by weighted entropy | |
CN116714469A (en) | Charging pile health monitoring method, device, terminal and storage medium | |
CN110928741B (en) | System state monitoring method, device and storage medium | |
JP2002169611A (en) | Fault diagnosis system and automated design system therefor | |
CN111382946B (en) | Autonomous evaluation method and system for health state of equipment and industrial internet equipment |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |