Risk prediction method and system based on fan lubricating oil

Abstract

The invention provides a risk prediction method and system based on fan lubricating oil, and the method comprises the steps: obtaining the operation parameters of a wind turbine generator and the physical and chemical parameters and wear particles of the lubricating oil in real time, and carrying out the preprocessing of the operation parameters and the physical and chemical parameters and wear particles of the lubricating oil; inputting the preprocessed operation data of the wind turbine generator, the physical and chemical parameters of the lubricating oil and the wear particles into a pre-trained risk trend prediction model to obtain a risk prediction result output by the risk trend prediction model; wherein the risk trend prediction model comprises an aging prediction sub-model, a fault prediction sub-model and a full connection layer; the risk prediction result comprises a lubricating oil aging prediction result and a fault probability prediction result of the wind turbine generator. According to the technical scheme, risk prediction is carried out based on the fan lubricating oil, the equipment safety is guaranteed, the fan operation efficiency is improved, and the technical effects of real-time follow-up and prediction of maintenance are achieved.

Description

Technical Field

[0001] This application relates to the field of fault diagnosis, and in particular to a risk prediction method and system based on wind turbine lubricating oil. Background Technology

[0002] As the core equipment of wind power generation, the stability and safety of wind turbine operation are crucial. The lubrication system is a vital component of wind turbine operation, responsible for providing necessary lubrication to various components to ensure normal and efficient operation. Lubricating oil plays a crucial role in the wind turbine lubrication system, not only reducing friction and wear, extending the service life of wind turbine components, but also lowering operating temperature and improving wind turbine efficiency. However, with the increase in wind turbine operating time, lubricating oil gradually ages and becomes contaminated, leading to performance degradation and potentially even wind turbine failure. Current technologies suffer from limitations in real-time monitoring and fault prediction based on lubricating oil issues, causing grid disturbances and compromising equipment safety and economic benefits. Summary of the Invention

[0003] This application provides a risk prediction method and system based on wind turbine lubricating oil, which at least solves the technical problem that the inability to perform real-time monitoring and fault prediction of wind turbines based on the existence of lubricating oil causes power grid disturbances, and the equipment safety and economic benefits cannot be guaranteed.

[0004] The first aspect of this application proposes a risk prediction method based on wind turbine lubricating oil, the method comprising:

[0005] The system acquires the operating parameters of the wind turbine and the physical and chemical parameters of the lubricating oil, as well as the wear particles, in real time, and preprocesses the operating parameters, the physical and chemical parameters of the lubricating oil, and the wear particles.

[0006] The preprocessed operating data of the wind turbine, the physicochemical parameters of the lubricating oil, and the wear particles are input into a pre-trained risk trend prediction model to obtain the risk prediction results output by the risk trend prediction model.

[0007] The risk trend prediction model includes: an aging prediction sub-model, a failure prediction sub-model, and a fully connected layer.

[0008] The risk prediction results include: lubricating oil aging prediction results and wind turbine failure probability prediction results.

[0009] Preferably, the training process of the risk trend prediction model includes:

[0010] Acquire lubricating oil monitoring sample data and test data of wind turbine generators. The lubricating oil monitoring sample data includes: lubricating oil physicochemical parameters, wear particles and wind turbine generator operating parameters, and fault information. The test data includes: lubricating oil composition information, test parameters and corresponding lubricating oil physicochemical parameters, and aging information.

[0011] A training sample dataset is constructed based on the lubricating oil monitoring sample data and test data of the wind turbine. The training sample dataset includes: lubricating oil physicochemical parameters, aging information, wind turbine operating parameters, wear particles, and fault information.

[0012] Extract a first training sample from the training sample dataset. The first training sample includes: lubricating oil physicochemical parameters, wind turbine operating parameters and aging information.

[0013] Using the lubricating oil physicochemical parameters and wind turbine operating parameters in the first training sample as inputs and aging information as outputs, the initial first network sub-model is iteratively trained to obtain a trained aging prediction sub-model.

[0014] Extract a second training sample from the training sample dataset. The second training sample includes: lubricating oil physicochemical parameters, wind turbine operating parameters, wear particles, aging information and fault information.

[0015] Using the lubricating oil physicochemical parameters, wind turbine operating parameters, wear particles, and aging information from the second training sample as inputs and fault information as outputs, the initial second network sub-model is iteratively trained to obtain a trained fault prediction sub-model.

[0016] The aging prediction sub-model and the fault prediction sub-model are connected, and the outputs of the aging prediction sub-model and the fault prediction sub-model are merged and output through the fully connected layer to obtain the risk trend prediction model.

[0017] Furthermore, the construction of the training sample dataset based on the lubricating oil monitoring sample data and test data of the wind turbine includes:

[0018] Align the lubricating oil physicochemical parameters and wind turbine operating parameters in the lubricating oil monitoring sample data with the lubricating oil physicochemical parameters and test parameters in the test data to establish the correlation between the test data and the wind turbine lubricating oil monitoring sample data. Then, merge the test data and the wind turbine lubricating oil monitoring sample data to construct a training sample dataset.

[0019] Furthermore, the process of acquiring the experimental data includes:

[0020] Collect wind turbine fault information, decompose the wind turbine fault information into wind turbine operating parameters, and determine the wind turbine operating environment parameters, including speed, temperature, and friction.

[0021] Using the aforementioned rotational speed, temperature, and friction as independent variables, and fault information as the dependent variable, a relationship fitting was performed to determine the influence relationship of the wind turbine's operating environment parameters.

[0022] The test parameters were configured based on the influence relationship of the wind turbine operating environment parameters.

[0023] Test samples were obtained based on the lubricating oil composition information, and wind turbine operation simulation tests were conducted on the test samples based on the test parameters. The physicochemical parameters and aging information of the lubricating oil were detected and recorded to construct test data.

[0024] Furthermore, the process of aligning the lubricating oil physicochemical parameters and wind turbine operating parameters in the lubricating oil monitoring sample data with the lubricating oil physicochemical parameters and test parameters in the test data includes:

[0025] The physical and chemical parameters of the lubricating oil, the test parameters, and the operating parameters of the wind turbine were sequentially cleaned and formatted.

[0026] After data cleaning and formatting, the characteristics of each parameter in the lubricating oil physicochemical parameters, test parameters, and wind turbine operating parameters are obtained, and parameter range analysis is performed.

[0027] Based on the parameter range, the lubricating oil physicochemical parameters, test parameters, and wind turbine operating parameters are decentered to obtain the data difference distribution.

[0028] Alignment rules are determined based on the data difference distribution, and the lubricating oil physicochemical parameters, test parameters, and wind turbine operating parameters are aligned according to the alignment rules.

[0029] Furthermore, determining the alignment rule based on the data difference distribution includes:

[0030] When the range of the data difference distribution is within a preset range, it is aligned according to the actual value.

[0031] When the range of the data difference distribution is outside the preset range, fuzzy alignment is performed according to the preset alignment interval.

[0032] Furthermore, the step of inputting the preprocessed operating data of the wind turbine and the physicochemical parameters of the lubricating oil and wear particles into a pre-trained risk trend prediction model to obtain the risk prediction result output by the risk trend prediction model includes:

[0033] The pre-processed operating data of the wind turbine and the physicochemical parameters of the lubricating oil are input into the pre-trained aging prediction sub-model to obtain the lubricating oil aging prediction results.

[0034] The aging prediction results, the pre-processed physicochemical parameters of the lubricating oil, and the wear particles are input into the pre-trained fault prediction sub-model to obtain the fault probability prediction results of the wind turbine.

[0035] The risk prediction result is obtained by outputting the lubricating oil aging prediction result and the wind turbine failure probability prediction result as the output result of the risk trend prediction model through the fully connected layer.

[0036] A second aspect of this application provides a risk prediction system based on wind turbine lubricating oil, comprising:

[0037] The acquisition module is used to acquire the operating parameters of the wind turbine and the physical and chemical parameters of the lubricating oil and wear particles in real time, and to preprocess the operating parameters, the physical and chemical parameters of the lubricating oil and wear particles.

[0038] The prediction module is used to input the preprocessed operating data of the wind turbine and the physicochemical parameters of the lubricating oil and wear particles into the pre-trained risk trend prediction model to obtain the risk prediction result output by the risk trend prediction model.

[0039] The risk trend prediction model includes: an aging prediction sub-model, a failure prediction sub-model, and a fully connected layer.

[0040] The risk prediction results include: lubricating oil aging prediction results and wind turbine failure probability prediction results.

[0041] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements the method described in the first aspect embodiment.

[0042] A fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method described in the first aspect.

[0043] The technical solutions provided by the embodiments of this application bring at least the following beneficial effects:

[0044] This application proposes a risk prediction method and system based on wind turbine lubricating oil. The method includes: acquiring the operating parameters of the wind turbine and the physicochemical parameters and wear particles of the lubricating oil in real time, and preprocessing the operating parameters, physicochemical parameters, and wear particles; inputting the preprocessed operating data of the wind turbine and the physicochemical parameters and wear particles of the lubricating oil into a pre-trained risk trend prediction model to obtain the risk prediction result output by the risk trend prediction model; wherein, the risk trend prediction model includes: an aging prediction sub-model, a fault prediction sub-model, and a fully connected layer; the risk prediction result includes: lubricating oil aging prediction result and wind turbine fault probability prediction result. The technical solution proposed in this application, based on wind turbine lubricating oil for risk prediction, ensures equipment safety, improves wind turbine operating efficiency, and achieves the technical effects of real-time tracking and predictive maintenance.

[0045] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0046] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0047] Figure 1 This is a flowchart illustrating a risk prediction method based on wind turbine lubricating oil according to an embodiment of this application;

[0048] Figure 2 This is a first structural diagram of a risk prediction system based on wind turbine lubricating oil according to an embodiment of this application;

[0049] Figure 3 This is a second structural diagram of a risk prediction system based on wind turbine lubricating oil according to an embodiment of this application. Detailed Implementation

[0050] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0051] This application proposes a risk prediction method and system based on wind turbine lubricating oil. The method includes: acquiring the operating parameters of the wind turbine and the physicochemical parameters and wear particles of the lubricating oil in real time, and preprocessing the operating parameters, physicochemical parameters, and wear particles; inputting the preprocessed operating data of the wind turbine and the physicochemical parameters and wear particles of the lubricating oil into a pre-trained risk trend prediction model to obtain the risk prediction result output by the risk trend prediction model; wherein, the risk trend prediction model includes: an aging prediction sub-model, a fault prediction sub-model, and a fully connected layer; the risk prediction result includes: lubricating oil aging prediction result and wind turbine fault probability prediction result. The technical solution proposed in this application, based on wind turbine lubricating oil for risk prediction, ensures equipment safety, improves wind turbine operating efficiency, and achieves the technical effects of real-time tracking and predictive maintenance.

[0052] The following description, with reference to the accompanying drawings, illustrates a risk prediction method and system based on fan lubricating oil according to an embodiment of this application.

[0053] Example 1

[0054] Figure 1 Here is a flowchart of a risk prediction method based on wind turbine lubricating oil according to an embodiment of this application, as shown below. Figure 1 As shown, the method includes:

[0055] Step 1: Real-time acquisition of the operating parameters of the wind turbine and the physical and chemical parameters of the lubricating oil, as well as the wear particles, and preprocessing of the operating parameters, the physical and chemical parameters of the lubricating oil, and the wear particles;

[0056] Step 2: Input the pre-processed operating data of the wind turbine, the physicochemical parameters of the lubricating oil, and the wear particles into the pre-trained risk trend prediction model to obtain the risk prediction results output by the risk trend prediction model;

[0057] The risk trend prediction model includes: an aging prediction sub-model, a failure prediction sub-model, and a fully connected layer.

[0058] The risk prediction results include: lubricating oil aging prediction results and wind turbine failure probability prediction results.

[0059] In this embodiment of the disclosure, step 2 specifically includes:

[0060] 2-1: Input the pre-processed operating data of the wind turbine and the physicochemical parameters of the lubricating oil into the pre-trained aging prediction sub-model to obtain the lubricating oil aging prediction results;

[0061] 2-2: Input the aging prediction results, the pre-processed physicochemical parameters of the lubricating oil, and the wear particles into the pre-trained fault prediction sub-model to obtain the fault probability prediction results of the wind turbine.

[0062] 2-3: The lubricating oil aging prediction result and the wind turbine failure probability prediction result are output as the risk trend prediction model output through the fully connected layer to obtain the risk prediction result.

[0063] For example, online monitoring data of wind turbines is imported into a risk trend prediction model. This data should contain the same or similar features as those used during model training, such as lubricating oil physicochemical parameters and wind turbine operating parameters. The imported online monitoring data is matched and filtered to identify and extract relevant information such as lubricating oil physicochemical parameters and wind turbine operating parameters. This data is then input into the aging prediction sub-model for aging prediction, and the aging prediction result is output. The output aging prediction result is further input into the fault prediction sub-model to combine lubricating oil physicochemical parameters and wear particles for fault prediction, and the fault prediction result is output. Finally, the aging prediction result and the fault prediction result are merged through a fully connected layer. That is, the fully connected layer uses the outputs of these two sub-models as input, and through certain weights and calculation methods, derives the final risk prediction result. This risk prediction result is a comprehensive assessment of the current and future state of the wind turbine, helping users better understand the operating status and risk situation of the wind turbine.

[0064] In this embodiment of the disclosure, the training process of the risk trend prediction model includes:

[0065] Step F1: Obtain lubricating oil monitoring sample data and test data of the wind turbine. The lubricating oil monitoring sample data includes: lubricating oil physicochemical parameters, wear particles and wind turbine operating parameters, and fault information. The test data includes: lubricating oil composition information, test parameters and corresponding lubricating oil physicochemical parameters, and aging information.

[0066] Furthermore, the process of acquiring the experimental data includes:

[0067] 1) Collect wind turbine fault information, decompose the wind turbine operating parameters into the fault information, and determine the wind turbine operating environment parameters, including speed, temperature, and friction.

[0068] 2) Using the aforementioned rotational speed, temperature, and friction as independent variables, and fault information as the dependent variable, a relationship fitting experiment is performed to determine the influence relationship of the wind turbine's operating environment parameters;

[0069] 3) Configure test parameters based on the influence relationship of the wind turbine operating environment parameters;

[0070] 4) Obtain test samples based on the lubricating oil composition information, and conduct wind turbine operation simulation tests on the test samples based on the test parameters. Detect and record the physicochemical parameters and aging information of the lubricating oil to construct test data.

[0071] Specifically, historical wind turbine fault information is collected, and the collected wind turbine data undergoes preprocessing such as cleaning, filtering, and standardization. Features related to wind turbine operation and faults, such as peak current, vibration frequency, and temperature change rate, are extracted from the preprocessed data. Mathematical methods or machine learning algorithms are used to decompose the extracted features and identify components related to the wind turbine's operating environment parameters, including speed, temperature, and friction. For example, by analyzing the wind turbine's current and voltage signals, combined with the wind turbine's speed sensor data, the real-time speed of the wind turbine can be determined.

[0072] Data on rotational speed, temperature, and friction were compiled, and wind turbine fault information was tagged by converting different fault types or severity into numerical or categorical labels to facilitate subsequent analysis and data fitting. Using rotational speed, temperature, and friction as independent variables and fault information as the dependent variable, a suitable fitting model was selected for relationship fitting, such as linear regression, logistic regression, or decision trees. Based on the model's output and parameters, the influence of independent variables such as rotational speed, temperature, and friction on fault information was analyzed. The weight or coefficient of each independent variable was calculated to quantify their contribution to the fault. Experimental parameters were configured and experiments were conducted based on the results of the influence relationship analysis.

[0073] During the experiment, representative test samples were selected and collected based on known lubricant composition information. Test parameters were configured based on the previously analyzed influence of wind turbine operating environment parameters, and a wind turbine operation simulation test platform was built to ensure its stable and accurate simulation of wind turbine operation. Furthermore, wind turbine operation simulation tests were conducted according to the set test parameters based on the test samples, and test data were recorded, including lubricant composition information, test parameters and corresponding lubricant physicochemical parameters, and aging information detection data.

[0074] By conducting wind turbine operation simulation tests based on experimental parameters, a data foundation is provided for subsequent model building and fault analysis. This is of great significance for evaluating lubricant performance, optimizing the wind turbine operating environment, and preventing faults.

[0075] It should be noted that the lubricating oil composition information corresponding to the current wind turbine is obtained, and test data is extracted based on the lubricating oil composition information. The test data includes lubricating oil physicochemical parameters, test parameters, and aging information.

[0076] The condition of the fan lubricating oil is closely related to the operating condition of the fan. The lubricating oil plays a key role in the operation of the fan. It can not only reduce friction and wear, but also help cool and clean mechanical parts. By analyzing the condition of the lubricating oil, we can indirectly understand the health status of the fan and the potential risks, discover potential problems in time, and take corresponding measures for prevention and treatment.

[0077] In the process of lubricating oil risk prediction, the first step is to acquire information on the composition of the lubricating oil to provide a data foundation for subsequent risk analysis. The composition of lubricating oil mainly consists of two parts: base oil and additives. Base oil is the main component of lubricating oil, determining its basic properties; it can be mineral base oil, synthetic base oil, or bio-based base oil. Additives are used to compensate for and improve the performance deficiencies of the base oil, while also imparting certain new properties to the lubricating oil. For example, antioxidants can reduce the oxidation rate of the oil, and alkaline additives can neutralize harmful acidic compounds, mitigating the damage caused by corrosion and wear. Furthermore, certain types of lubricating oil may contain other components to meet specific operational requirements. Therefore, acquiring information on the composition of lubricating oil provides an analytical basis for subsequent failure analysis, ensuring the customization and accuracy of the analysis.

[0078] To extract experimental data for subsequent analysis, a series of experiments and tests were conducted based on known lubricant composition information. This experimental data includes the lubricant's physicochemical parameters, test parameters, and aging information. The physicochemical parameters include viscosity, density, flash point, acid value, and moisture content. These parameters allow for in-depth analysis and evaluation of the target lubricant, facilitating subsequent experimental analysis. Test parameters include the temperature conditions used during lubricant performance testing, the duration of testing under specific conditions, and the load on the lubricated parts during simulated actual fan operation. This parameter information was recorded and further analyzed for aging to obtain aging information. For example, changes in lubricant performance were observed through prolonged heating or exposure to oxygen to assess its anti-aging properties. The degree of oil aging was assessed based on the quantity and nature of sediments that may form after a period of use. Furthermore, the aging status of the oil could be evaluated by comparing the initial viscosity with the viscosity after aging.

[0079] After the experiment, the data were organized and analyzed to draw accurate conclusions about the performance of the lubricating oil. This data is of great significance for evaluating the quality of the lubricating oil, predicting its service life, and guiding the selection and maintenance of the lubricating oil.

[0080] It should be noted that, in order to conduct correlation analysis between wind turbine failures and lubricating oil, monitoring sample data of the wind turbine lubricating oil were acquired. Using appropriate sampling methods, such as sampling valves, pressure testing connectors, or sampling pumps, accurate and representative lubricating oil samples were obtained from the wind turbine lubrication system. The obtained lubricating oil samples were then subjected to physicochemical parameter testing using professional lubricating oil testing and analysis methods, including viscosity, acid value, water content, and mechanical impurities. Furthermore, the lubricating oil samples were examined under a microscope to identify and count wear particles, which may include metallic and non-metallic particles. Their presence and quantity can reflect the wear state of the wind turbine. For further in-depth and detailed analysis, methods such as spectral analysis and ferrography can also be used to observe the sample lubricating oil to ensure the accuracy of the correlation analysis.

[0081] Operating parameter data for sample wind turbines are acquired through monitoring systems or sensors, including parameters such as speed, temperature, current, and voltage. These parameters reflect the operating status and performance of the wind turbines. Simultaneously, fault information is collected, such as abnormal sounds, vibrations, and overheating, which may be related to lubricant performance or the wear condition of the wind turbine. Furthermore, the acquired lubricant physicochemical parameters, wear particle data, and wind turbine operating parameters and fault information are organized and analyzed. Statistical methods and trend analysis are used to identify patterns and anomalies in the data, providing decision support for subsequent wind turbine analysis.

[0082] Step F2: Construct a training sample dataset based on the lubricating oil monitoring sample data and test data of the wind turbine, wherein the training sample dataset includes: lubricating oil physicochemical parameters, aging information, wind turbine operating parameters, wear particles, and fault information;

[0083] It should be noted that step F2 specifically includes:

[0084] Align the lubricating oil physicochemical parameters and wind turbine operating parameters in the lubricating oil monitoring sample data with the lubricating oil physicochemical parameters and test parameters in the test data to establish the correlation between the test data and the wind turbine lubricating oil monitoring sample data. Then, merge the test data and the wind turbine lubricating oil monitoring sample data to construct a training sample dataset.

[0085] Furthermore, the process of aligning the lubricating oil physicochemical parameters and wind turbine operating parameters in the lubricating oil monitoring sample data with the lubricating oil physicochemical parameters and test parameters in the test data includes:

[0086] The physical and chemical parameters of the lubricating oil, the test parameters, and the operating parameters of the wind turbine were sequentially cleaned and formatted.

[0087] After data cleaning and formatting, the characteristics of each parameter in the lubricating oil physicochemical parameters, test parameters, and wind turbine operating parameters are obtained, and parameter range analysis is performed.

[0088] Based on the parameter range, the lubricating oil physicochemical parameters, test parameters, and wind turbine operating parameters are decentered to obtain the data difference distribution.

[0089] Alignment rules are determined based on the data difference distribution, and the lubricating oil physicochemical parameters, test parameters, and wind turbine operating parameters are aligned according to the alignment rules.

[0090] For example, the test data and the wind turbine lubricating oil monitoring sample data are cleaned to remove duplicates, missing data, or excessive outliers, ensuring the integrity and accuracy of both. The data is then formatted to ensure consistency in timestamps, parameter units, and other formats, facilitating subsequent data alignment and fusion. Next, key features, such as lubricating oil physicochemical parameters (viscosity, acid value, etc.) and test parameters (test temperature, time, etc.), are extracted from the test data. Simultaneously, wind turbine operating parameters (speed, current, etc.) and fault information are extracted from the wind turbine lubricating oil monitoring sample data. The test data and wind turbine operating parameters are then time-aligned according to timestamps to ensure data matching within the same time period. Furthermore, the consistency of lubricating oil samples collected during the test and wind turbine operation must be ensured to guarantee the correctness of parameter correlation.

[0091] Furthermore, the characteristics of experimental data and wind turbine operating parameters can be combined. For example, the viscosity and acid value of lubricating oil can be combined with the speed and current of wind turbines to form a more comprehensive feature set through data fusion. This feature set can then be used to construct the training sample dataset.

[0092] In the process of constructing the training sample dataset, the fault information in the wind turbine lubricating oil monitoring sample data is first used to label the fused data. Then, the fused dataset is divided into training set, validation set and test set for training, adjustment and performance testing of the model.

[0093] The construction of training sample datasets through data fusion can provide strong support for in-depth analysis and prediction of wind turbine lubricant performance and wind turbine operating status, thereby ensuring the completeness, accuracy and reliability of subsequent analysis results.

[0094] For example, the characteristics of each parameter in the corresponding lubricating oil physicochemical parameters, test parameters, and wind turbine operating parameters are obtained. Statistical analysis is performed on each parameter to determine its maximum, minimum, average, and standard deviation, etc., and the distribution characteristics of each parameter are analyzed to understand the range and pattern of parameter variation. The mean of each parameter dataset is calculated and decentralized. The difference distribution of the decentralized data is analyzed to understand the dispersion and distribution of data points. One or more parameters are selected as alignment benchmarks for data alignment, such as using time as the alignment benchmark to set alignment rules for data alignment, including: aligning according to the time points of the test steps, interpolating missing data points, or pruning data that exceeds the alignment range. The lubricating oil physicochemical parameters, test parameters, and wind turbine operating parameters are synchronized to the same alignment benchmark, and the alignment results are verified to ensure that the alignment relationship between each parameter is correct.

[0095] The aligned data are combined into a complete sample, with each sample containing lubricating oil physicochemical parameters, experimental parameters, aging information, wear particles, wind turbine operating parameters, and fault information. This ensures that each sample is complete and consistent, facilitating subsequent data analysis and model training. Furthermore, based on the alignment relationship between experimental parameters and wind turbine operating parameters, the structure of the sample dataset is further adjusted and optimized to ensure that these parameters are aligned temporally or logically, reflecting the actual situation of the wind turbine during operation, thus laying the foundation for subsequent model construction.

[0096] The step of determining the alignment rule based on the data difference distribution includes:

[0097] When the range of the data difference distribution is within a preset range, it is aligned according to the actual value.

[0098] When the range of the data difference distribution is outside the preset range, fuzzy alignment is performed according to the preset alignment interval.

[0099] It should be noted that when determining the alignment rules, a preset threshold is set in advance. Different alignment methods are selected by analyzing the range of the difference distribution to ensure the accuracy and rationality of the alignment operation and improve the analysis efficiency. Those skilled in the art can set the preset threshold according to the actual situation.

[0100] Specifically, for each set of parameters that need to be aligned, the difference is calculated and the range of the difference distribution is determined. If the range of the data difference distribution is within the preset threshold, it means that the difference between the parameters is small, and alignment can be performed directly according to the actual values. In this case, the aligned data can accurately reflect the actual situation. If the range of the data difference distribution exceeds the preset threshold, it means that the difference between the parameters is large, and direct alignment may lead to a large error. In this case, an corresponding alignment interval should be set, and fuzzy alignment should be performed according to the alignment interval. Fuzzy alignment can use interpolation, averaging, or other suitable mathematical methods to distribute the difference within the alignment interval in order to reduce the alignment error. It should be understood that the result of fuzzy alignment should be as close as possible to the actual value.

[0101] Step F3: Extract the first training sample from the training sample dataset. The first training sample includes: lubricating oil physicochemical parameters, wind turbine operating parameters, and aging information.

[0102] Step F4: Using the lubricating oil physicochemical parameters and wind turbine operating parameters in the first training sample as inputs and aging information as outputs, iteratively train the initial first network sub-model to obtain the trained aging prediction sub-model.

[0103] Step F5: Extract the second training sample from the training sample dataset. The second training sample includes: lubricating oil physicochemical parameters, wind turbine operating parameters, wear particles, aging information, and fault information.

[0104] Step F6: Using the lubricating oil physicochemical parameters, wind turbine operating parameters, wear particles, and aging information from the second training sample as inputs and the fault information as output, iteratively train the initial second network sub-model to obtain the trained fault prediction sub-model.

[0105] Step F7: Connect the aging prediction sub-model and the fault prediction sub-model, and merge the output results of the aging prediction sub-model and the fault prediction sub-model through the fully connected layer to obtain the risk trend prediction model.

[0106] It should be noted that feature scaling, feature selection, or feature dimensionality reduction are performed on the feature data in the training sample dataset to standardize or reduce the dimensionality of the feature data, thereby improving the learning effect and efficiency of the model.

[0107] Choose a suitable neural network model, such as a fully connected neural network, convolutional neural network (CNN), recurrent neural network (RNN), or long short-term memory network (LSTM), and design a reasonable model structure, including: determining the number of layers in the neural network, the number of neurons in each layer, activation function, loss function, and optimizer, etc. Then initialize the model parameters, input the feature data from the training set of the divided training sample dataset into the initialized model for training, calculate the propagation loss, update the model parameters on the validation set, iteratively optimize the training, and evaluate and adjust the performance of the trained model on the test set to obtain the final risk trend prediction model.

[0108] By obtaining a risk trend prediction model, it is possible to predict the risk trend of wind turbine lubricating oil based on input characteristics (physicochemical parameters of lubricating oil, operating parameters of wind turbine, etc.), including the probability of failure and risk level, so as to promptly identify potential risks and take corresponding measures for prevention and maintenance.

[0109] Specifically, the risk trend prediction model framework includes a first network sub-model, a second network sub-model, and a fully connected layer. The first network sub-model is used to predict lubricating oil aging. Training samples, including lubricating oil physicochemical parameters, wind turbine operating parameters, and aging information, are extracted from the training sample dataset. The aging information is then labeled to obtain a first training sample set. The first network sub-model is then iteratively trained using this first training sample set. Here, the first network sub-model can be a multilayer perceptron (MLP) or other neural network structures suitable for processing continuous and categorical data. A convergence condition is set, which can be that the difference between the output aging prediction result and the labeled aging information (such as cross-entropy loss or mean squared error) reaches a target probability or falls below a certain threshold. When the model meets the convergence condition, training stops, and the aging prediction sub-model is obtained.

[0110] The second network sub-model is a sub-model for fault prediction. Similarly, a second training sample set, including identification information such as lubricating oil physicochemical parameters, wind turbine operating parameters, wear particles, aging information, and fault information, is extracted from the training sample dataset. The second network sub-model is trained and converged using the second training sample set to obtain the fault prediction sub-model. Furthermore, the outputs of the aging prediction sub-model and the fault prediction sub-model are used as inputs to a fully connected layer to connect the sub-models, and the merged output is used to predict risk trends.

[0111] By merging the outputs of the aging prediction sub-model and the failure prediction sub-model and processing them through a fully connected layer, a risk trend prediction model is finally obtained. This model can comprehensively consider factors such as lubricating oil physicochemical parameters, wind turbine operating parameters, and wear particles to predict the aging and failure risks of wind turbines.

[0112] In summary, the risk prediction method based on wind turbine lubricating oil proposed in this embodiment solves the technical problem that the existing technology cannot perform real-time tracking and fault prediction of wind turbines based on the lubrication system, i.e., the lubricating oil, which causes power grid disturbances and compromises equipment safety and economic benefits. It achieves the technical effect of improving wind turbine operating efficiency, enabling real-time tracking and predictive maintenance, and ensuring safe, economical and efficient operation of equipment.

[0113] Example 2

[0114] Figure 2 This is a structural diagram of a risk prediction system based on wind turbine lubricating oil according to an embodiment of this application, as shown below. Figure 2 As shown, the system includes:

[0115] The acquisition module 100 is used to acquire the operating parameters of the wind turbine and the physical and chemical parameters of the lubricating oil and wear particles in real time, and to preprocess the operating parameters, the physical and chemical parameters of the lubricating oil and wear particles.

[0116] The prediction module 200 is used to input the preprocessed operating data of the wind turbine and the physicochemical parameters of the lubricating oil and wear particles into a pre-trained risk trend prediction model to obtain the risk prediction result output by the risk trend prediction model.

[0117] The risk trend prediction model includes: an aging prediction sub-model, a failure prediction sub-model, and a fully connected layer.

[0118] The risk prediction results include: lubricating oil aging prediction results and wind turbine failure probability prediction results.

[0119] In the embodiments disclosed herein, such as Figure 3 As shown, the system further includes: a training module 300, the training module 300 comprising:

[0120] The acquisition unit 301 is used to acquire lubricating oil monitoring sample data and test data of the wind turbine. The lubricating oil monitoring sample data includes: lubricating oil physicochemical parameters, wear particles and wind turbine operating parameters, and fault information. The test data includes: lubricating oil composition information, test parameters and corresponding lubricating oil physicochemical parameters, and aging information.

[0121] The construction unit 302 is used to construct a training sample dataset based on the lubricating oil monitoring sample data and test data of the wind turbine, wherein the training sample dataset includes: lubricating oil physicochemical parameters, aging information, wind turbine operating parameters, wear particles, and fault information;

[0122] The first extraction unit 303 is used to extract a first training sample from the training sample dataset. The first training sample includes: lubricating oil physicochemical parameters, wind turbine operating parameters and aging information.

[0123] The first training unit 304 is used to iteratively train the initial first network sub-model with the lubricating oil physicochemical parameters and wind turbine operating parameters in the first training sample as inputs and aging information as outputs, so as to obtain a trained aging prediction sub-model.

[0124] The second extraction unit 305 is used to extract a second training sample from the training sample dataset. The second training sample includes: lubricating oil physicochemical parameters, wind turbine operating parameters, wear particles, aging information and fault information.

[0125] The second training unit 306 is used to iteratively train the initial second network sub-model with the lubricating oil physicochemical parameters, wind turbine operating parameters, wear particles and aging information in the second training sample as inputs and the fault information as outputs, so as to obtain the trained fault prediction sub-model.

[0126] Output unit 307 is used to connect the aging prediction sub-model and the fault prediction sub-model, and to merge and output the output results of the aging prediction sub-model and the fault prediction sub-model through the fully connected layer to obtain the risk trend prediction model.

[0127] Furthermore, the building unit 302 is also used for:

[0128] Align the lubricating oil physicochemical parameters and wind turbine operating parameters in the lubricating oil monitoring sample data with the lubricating oil physicochemical parameters and test parameters in the test data to establish the correlation between the test data and the wind turbine lubricating oil monitoring sample data. Then, merge the test data and the wind turbine lubricating oil monitoring sample data to construct a training sample dataset.

[0129] The step of aligning the lubricating oil physicochemical parameters and wind turbine operating parameters in the lubricating oil monitoring sample data with the lubricating oil physicochemical parameters and test parameters in the test data includes:

[0130] The physical and chemical parameters of the lubricating oil, the test parameters, and the operating parameters of the wind turbine were sequentially cleaned and formatted.

[0131] After data cleaning and formatting, the characteristics of each parameter in the lubricating oil physicochemical parameters, test parameters, and wind turbine operating parameters are obtained, and parameter range analysis is performed.

[0132] Based on the parameter range, the lubricating oil physicochemical parameters, test parameters, and wind turbine operating parameters are decentered to obtain the data difference distribution.

[0133] Alignment rules are determined based on the data difference distribution, and the lubricating oil physicochemical parameters, test parameters, and wind turbine operating parameters are aligned according to the alignment rules.

[0134] The step of determining the alignment rule based on the data difference distribution includes:

[0135] When the range of the data difference distribution is within a preset range, it is aligned according to the actual value.

[0136] When the range of the data difference distribution is outside the preset range, fuzzy alignment is performed according to the preset alignment interval.

[0137] Furthermore, the acquisition unit 301 is also used for:

[0138] Collect wind turbine fault information, decompose the wind turbine fault information into wind turbine operating parameters, and determine the wind turbine operating environment parameters, including speed, temperature, and friction.

[0139] Using the aforementioned rotational speed, temperature, and friction as independent variables, and fault information as the dependent variable, a relationship fitting was performed to determine the influence relationship of the wind turbine's operating environment parameters.

[0140] The test parameters were configured based on the influence relationship of the wind turbine operating environment parameters.

[0141] Test samples were obtained based on the lubricating oil composition information, and wind turbine operation simulation tests were conducted on the test samples based on the test parameters. The physicochemical parameters and aging information of the lubricating oil were detected and recorded to construct test data.

[0142] In this embodiment of the disclosure, the prediction module 200 is further configured to:

[0143] The pre-processed operating data of the wind turbine and the physicochemical parameters of the lubricating oil are input into the pre-trained aging prediction sub-model to obtain the lubricating oil aging prediction results.

[0144] The aging prediction results, the pre-processed physicochemical parameters of the lubricating oil, and the wear particles are input into the pre-trained fault prediction sub-model to obtain the fault probability prediction results of the wind turbine.

[0145] The risk prediction result is obtained by outputting the lubricating oil aging prediction result and the wind turbine failure probability prediction result as the output result of the risk trend prediction model through the fully connected layer.

[0146] In summary, the risk prediction system based on wind turbine lubricating oil proposed in this embodiment predicts risks based on wind turbine lubricating oil, ensuring equipment safety and improving the technical effects of wind turbine operation efficiency, real-time tracking, and predictive maintenance.

[0147] Example 3

[0148] To implement the above embodiments, this disclosure also proposes an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements the method described in Embodiment 1.

[0149] Example 4

[0150] To implement the above embodiments, this disclosure also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in Embodiment 1.

[0151] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0152] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0153] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A risk prediction method based on wind turbine lubricating oil, characterized in that, The method includes: The system acquires the operating parameters of the wind turbine and the physical and chemical parameters of the lubricating oil, as well as the wear particles, in real time, and preprocesses the operating parameters, the physical and chemical parameters of the lubricating oil, and the wear particles. The preprocessed operating data of the wind turbine, the physicochemical parameters of the lubricating oil, and the wear particles are input into a pre-trained risk trend prediction model to obtain the risk prediction results output by the risk trend prediction model. The risk trend prediction model includes: an aging prediction sub-model, a failure prediction sub-model, and a fully connected layer. The risk prediction results include: lubricating oil aging prediction results and wind turbine failure probability prediction results.

2. The method as described in claim 1, characterized in that, The training process of the risk trend prediction model includes: Acquire lubricating oil monitoring sample data and test data of wind turbine generators. The lubricating oil monitoring sample data includes: lubricating oil physicochemical parameters, wear particles and wind turbine generator operating parameters, and fault information. The test data includes: lubricating oil composition information, test parameters and corresponding lubricating oil physicochemical parameters, and aging information. A training sample dataset is constructed based on the lubricating oil monitoring sample data and test data of the wind turbine. The training sample dataset includes: lubricating oil physicochemical parameters, aging information, wind turbine operating parameters, wear particles, and fault information. Extract a first training sample from the training sample dataset. The first training sample includes: lubricating oil physicochemical parameters, wind turbine operating parameters and aging information. Using the lubricating oil physicochemical parameters and wind turbine operating parameters in the first training sample as inputs and aging information as outputs, the initial first network sub-model is iteratively trained to obtain a trained aging prediction sub-model. Extract a second training sample from the training sample dataset. The second training sample includes: lubricating oil physicochemical parameters, wind turbine operating parameters, wear particles, aging information and fault information. Using the lubricating oil physicochemical parameters, wind turbine operating parameters, wear particles, and aging information from the second training sample as inputs and fault information as outputs, the initial second network sub-model is iteratively trained to obtain a trained fault prediction sub-model. The aging prediction sub-model and the fault prediction sub-model are connected, and the outputs of the aging prediction sub-model and the fault prediction sub-model are merged and output through the fully connected layer to obtain the risk trend prediction model.

3. The method as described in claim 2, characterized in that, The training sample dataset constructed based on the lubricating oil monitoring sample data and test data of the wind turbine includes: Align the lubricating oil physicochemical parameters and wind turbine operating parameters in the lubricating oil monitoring sample data with the lubricating oil physicochemical parameters and test parameters in the test data to establish the correlation between the test data and the wind turbine lubricating oil monitoring sample data. Then, merge the test data and the wind turbine lubricating oil monitoring sample data to construct a training sample dataset.

4. The method as described in claim 2, characterized in that, The process of acquiring the experimental data includes: Collect wind turbine fault information, decompose the wind turbine fault information into wind turbine operating parameters, and determine the wind turbine operating environment parameters, including speed, temperature, and friction. Using the aforementioned rotational speed, temperature, and friction as independent variables, and fault information as the dependent variable, a relationship fitting was performed to determine the influence relationship of the wind turbine's operating environment parameters. The test parameters were configured based on the influence relationship of the wind turbine operating environment parameters. Test samples were obtained based on the lubricating oil composition information, and wind turbine operation simulation tests were conducted on the test samples based on the test parameters. The physicochemical parameters and aging information of the lubricating oil were detected and recorded to construct test data.

5. The method as described in claim 3, characterized in that, The process of aligning the lubricating oil physicochemical parameters and wind turbine operating parameters in the lubricating oil monitoring sample data with the lubricating oil physicochemical parameters and test parameters in the test data includes: The physical and chemical parameters of the lubricating oil, the test parameters, and the operating parameters of the wind turbine were sequentially cleaned and formatted. After data cleaning and formatting, the characteristics of each parameter in the lubricating oil physicochemical parameters, test parameters, and wind turbine operating parameters are obtained, and parameter range analysis is performed. Based on the parameter range, the lubricating oil physicochemical parameters, test parameters, and wind turbine operating parameters are decentered to obtain the data difference distribution. Alignment rules are determined based on the data difference distribution, and the lubricating oil physicochemical parameters, test parameters, and wind turbine operating parameters are aligned according to the alignment rules.

6. The method as described in claim 5, characterized in that, The step of determining the alignment rule based on the data difference distribution includes: When the range of the data difference distribution is within a preset range, it is aligned according to the actual value. When the range of the data difference distribution is outside the preset range, fuzzy alignment is performed according to the preset alignment interval.

7. The method as described in claim 2, characterized in that, The process involves inputting the pre-processed operating data of the wind turbine, the physicochemical parameters of the lubricating oil, and wear particles into a pre-trained risk trend prediction model to obtain the risk prediction results output by the risk trend prediction model, including: The pre-processed operating data of the wind turbine and the physicochemical parameters of the lubricating oil are input into the pre-trained aging prediction sub-model to obtain the lubricating oil aging prediction results. The aging prediction results, the pre-processed physicochemical parameters of the lubricating oil, and the wear particles are input into the pre-trained fault prediction sub-model to obtain the fault probability prediction results of the wind turbine. The risk prediction result is obtained by outputting the lubricating oil aging prediction result and the wind turbine failure probability prediction result as the output result of the risk trend prediction model through the fully connected layer.

8. A risk prediction system based on wind turbine lubricating oil, characterized in that, The system includes: The acquisition module is used to acquire the operating parameters of the wind turbine and the physical and chemical parameters of the lubricating oil and wear particles in real time, and to preprocess the operating parameters, the physical and chemical parameters of the lubricating oil and wear particles. The prediction module is used to input the preprocessed operating data of the wind turbine and the physicochemical parameters of the lubricating oil and wear particles into the pre-trained risk trend prediction model to obtain the risk prediction result output by the risk trend prediction model. The risk trend prediction model includes: an aging prediction sub-model, a failure prediction sub-model, and a fully connected layer. The risk prediction results include: lubricating oil aging prediction results and wind turbine failure probability prediction results.

9. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-7.

Images