Method for evaluating residual life of wind driven generator based on time sequence prediction and related equipment

Abstract

The invention discloses a wind driven generator residual life evaluation method based on time sequence prediction and related equipment, and the method comprises the steps: obtaining multiple types of operation data, capable of reflecting the degradation information of a wind driven generator, of each part of the wind driven generator in a set time period, and each type of operation data comprises effective timestamp information; converting the multiple types of operation data into multivariate time series data with time synchronization, and preprocessing the multivariate time series data; intercepting multivariate time point sequence data of a first set time step from the preprocessed multivariate time sequence data, inputting the multivariate time point sequence data of the first set time step into the trained time sequence prediction model, and outputting predicted multivariate time sequence data of a second set time step; performing optimization processing on the multivariate time sequence data of the second set time step length, and performing fusion based on the optimized data so as to construct a health index curve capable of comprehensively reflecting the health state of the wind driven generator; and evaluating the remaining service life of the wind driven generator according to the constructed health index curve. The objective of the invention is to provide a more accurate and reliable means for health management and life prediction of the wind driven generator.

Description

Technical Field

[0001] The present invention belongs to the technical field of wind turbine fault prediction and health management, and in particular relates to a wind turbine remaining life assessment method based on time series prediction and related equipment. Background Art

[0002] As a complex mechatronic system, the operating performance of wind turbines is affected by a variety of factors, including but not limited to environmental conditions, changes in operating conditions, load fluctuations, and sudden large disturbances. Under long-term operation, these external factors interact with internal mechanical wear, material aging and other processes, resulting in the gradual degradation of key components and overall performance within the unit, which may in turn cause faults or even failures, seriously affecting the power generation efficiency and economic benefits of wind farms, and may bring safety risks and property losses. In view of this, it is very necessary to accurately predict the remaining life (i.e., remaining working time) of wind turbines based on real-time collected equipment status monitoring data, thereby providing a scientific basis for preventive maintenance, avoiding unplanned downtime, reducing maintenance costs, and improving the safety, reliability and economy of equipment operation.

[0003] At present, the industry mainly uses two types of methods to evaluate the remaining life of mechanical equipment: physics-based methods and data-driven methods. Although physics-based methods can deeply reveal the failure mechanism of materials, the modeling process is complex, the amount of calculation is large, and the model generalization ability is limited, making it difficult to adapt to changing working environments. Data-driven methods, especially intelligent prediction models based on deep learning theory, can use a large amount of historical data for training and quickly establish a mapping relationship between monitoring data and remaining life, but this method often ignores the time dependence of the equipment degradation process, resulting in unstable fluctuations in the prediction results, making it difficult to provide intuitive and clear support for operation and maintenance decisions.

[0004] Therefore, in view of the shortcomings of existing technologies, there is an urgent need for a new remaining life prediction method that comprehensively considers the time dependence of the equipment degradation process and combines the advantages of physics and data-driven, so as to improve the predictive maintenance level of major equipment such as wind turbines and ensure their long-term stable operation. Summary of the invention

[0005] In view of the problems existing in the prior art, the present invention provides a method for assessing the remaining life of a wind turbine based on time series prediction and related equipment, the purpose of which is to provide a more accurate and reliable means for health management and life prediction of wind turbines.

[0006] In order to solve the above technical problems, the present invention is implemented by the following technical solutions:

[0007] According to a first aspect of the present invention, a method for assessing the remaining life of a wind turbine based on time series prediction is provided, comprising:

[0008] Acquire multiple types of operating data of various components of the wind turbine within a set time period that can reflect the degradation information of the wind turbine, and each type of operating data contains valid timestamp information;

[0009] Converting the multiple types of operation data into time-synchronized multivariate time series data, and preprocessing the multivariate time series data;

[0010] Extracting multivariate time point sequence data of a first set time step from the preprocessed multivariate time series data, inputting the multivariate time point sequence data of the first set time step into a trained time series prediction model, and outputting predicted multivariate time series data of a second set time step;

[0011] Performing optimization processing on the multivariate time series data of the second set time step, and fusing the optimized data to construct a health index curve that can comprehensively reflect the health status of the wind turbine;

[0012] The remaining service life of the wind turbine is evaluated based on the constructed health index curve.

[0013] In a possible implementation manner of the first aspect, the training method of the time series prediction model is:

[0014] Obtain various types of operating data of wind turbine components during their entire life cycle that can reflect the degradation information of wind turbines;

[0015] Convert the multi-class operation data of the whole life cycle into time-synchronized multivariate time series data, and pre-process the multivariate time series data of the whole life cycle;

[0016] The preprocessed multivariate time series data of the whole life stage are intercepted according to the third set time step to obtain a plurality of training samples, wherein the multivariate time point series data of the whole life stage before the first set time step in each training sample is used as input, and the multivariate time point series data of the whole life stage after the second set time step is used as output, and the third set time step is the sum of the first set time step and the second set time step;

[0017] Multiple training samples are used to train a long time series prediction network based on the Transformer framework to obtain a time series prediction model.

[0018] In a possible implementation manner of the first aspect, converting the multiple types of operating data into time-synchronized multivariate time series data specifically includes:

[0019] The multi-class operating data are converted into time-synchronized multivariate time series data using an interpolation method.

[0020] In a possible implementation of the first aspect, the wind turbine generator components include a main bearing, a gearbox, and a generator;

[0021] The multiple types of operating data capable of reflecting the degradation information of the wind turbine generator include vibration data, current data and oil temperature data.

[0022] In a possible implementation manner of the first aspect, the preprocessing of the multivariate time series data is specifically:

[0023] The multivariate time series data are preprocessed using the maximum and minimum normalization processing method.

[0024] In a possible implementation of the first aspect, the multivariate time series data of the second set time step is optimized, and fused based on the optimized data to construct a health index curve that can comprehensively reflect the health status of the wind turbine, specifically:

[0025] Sorting the multidimensional multivariate time series data of the second set time step according to monotonicity, and selecting a feature sequence whose monotonicity meets the set requirements;

[0026] The principal component analysis method is used to reduce the dimension of the feature sequence whose monotonicity meets the set requirements to one dimension, and a health index curve that can comprehensively reflect the health status of the wind turbine is obtained.

[0027] In a possible implementation of the first aspect, the remaining service life of the wind turbine generator is evaluated according to the constructed health indicator curve, specifically:

[0028] Count the average or median of health indicators of wind turbines during their entire life cycle when a fault occurs, and set the value as the warning threshold for the occurrence of a fault;

[0029] If the health indicator curve does not exceed the warning threshold value within the prediction period, it is determined that the remaining service life of the wind turbine is greater than or equal to the second set time step;

[0030] If the health indicator curve exceeds the warning threshold during the forecast period, the remaining service life of the wind turbine is estimated based on the difference between the time when the health indicator curve first reaches the warning threshold and the current time.

[0031] According to a second aspect of the present invention, there is provided a device for evaluating the remaining life of a wind turbine generator based on time series prediction, comprising:

[0032] An acquisition module is used to acquire multiple types of operating data of various components of the wind turbine within a set time period that can reflect the degradation information of the wind turbine, and each type of operating data contains valid timestamp information;

[0033] A conversion module, used for converting the multiple types of operation data into time-synchronized multivariate time series data, and preprocessing the multivariate time series data;

[0034] A prediction module, used for intercepting multivariate time point sequence data of a first set time step from the preprocessed multivariate time series data, inputting the multivariate time point sequence data of the first set time step into a trained time series prediction model, and outputting predicted multivariate time series data of a second set time step;

[0035] A fusion module, used for optimizing the multivariate time series data of the second set time step, and fusing the optimized data to construct a health index curve that can comprehensively reflect the health status of the wind turbine;

[0036] The evaluation module is used to evaluate the remaining service life of the wind turbine according to the constructed health index curve.

[0037] According to a third aspect of the present invention, there is provided a device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the method for assessing the remaining life of a wind turbine based on time series prediction are implemented.

[0038] According to a fourth aspect of the present invention, there is provided a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of a method for evaluating the remaining life of a wind turbine based on time series prediction.

[0039] Compared with the prior art, the present invention has at least the following beneficial effects:

[0040] The present invention provides a method for assessing the remaining life of a wind turbine based on time series prediction, which focuses on predicting the degradation trends of multiple types of operating data of various components of the wind turbine that can reflect the degradation information of the wind turbine, rather than simple direct mapping. It not only has good interpretability, but also has relatively smooth prediction results and strong generalization ability, and is less sensitive to changes in operating conditions, showing high robustness. In addition, the present invention can combine and analyze multiple types of operating data (i.e., different data sources) that can reflect the degradation information of the wind turbine, breaking through the data barrier. In addition, the present invention evaluates the remaining life of the wind turbine, can timely understand the operating status of each component, and provide targeted maintenance suggestions, thereby saving wind farm operation and maintenance costs. It can be seen that the present invention can provide a more accurate and reliable means for the health management and life prediction of wind turbines.

[0041] The time series prediction model selected by the present invention has a strong self-representation ability and can effectively train the network even when the amount of data is limited.

[0042] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the specific implementation modes of the present invention, the drawings required for use in the description of the specific implementation modes will be briefly introduced below. Obviously, the drawings described below are some implementation modes of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0044] Figure 1 The present invention is a flow chart of a method for evaluating the remaining life of a wind turbine generator based on time series prediction.

[0045] Figure 2 Schematic diagram of vibration measurement points of WTG146-3300 wind turbine in an actual wind farm provided in Example 1 of the present invention.

[0046] Figure 3 A schematic diagram of the prediction results of the main bearing RMS time series provided in Example 1 of the present invention.

[0047] Figure 4 This is a health indicator degradation curve and a schematic diagram for calculating the remaining life provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] like Figure 1 As shown, an embodiment of the present invention provides a method for evaluating the remaining life of a wind turbine generator based on time series prediction, which specifically includes the following steps:

[0050] S1. Acquire multiple types of operation data of various components of the wind turbine within a set time period that can reflect the degradation information of the wind turbine, and each type of operation data contains valid timestamp information.

[0051] Specifically, various sensors installed on wind turbines collect multiple types of operating data of key components of wind turbines (such as blades, gearboxes, generators, etc.) in real time within a set time period. These data can fully reflect the operating status of wind turbines and their degradation trends, including but not limited to vibration signals, temperature, speed, current, voltage, etc. At the same time, ensure that each type of operating data contains valid timestamp information for subsequent time series analysis.

[0052] S2. Convert the multiple types of operating data into time-synchronized multivariate time series data, and preprocess the multivariate time series data.

[0053] Specifically, the acquired multi-class operating data are synchronously processed according to the timestamp information and converted into time-synchronized multivariate time series data. Subsequently, these multivariate time series data are preprocessed to ensure the accuracy and consistency of the data.

[0054] S3. Extract the multivariate time point sequence data of the first set time step from the preprocessed multivariate time series data, input the multivariate time point sequence data of the first set time step into the trained time series prediction model, and output the predicted multivariate time series data of the second set time step.

[0055] Specifically, the multivariate time point sequence data of the first set time step is intercepted from the preprocessed multivariate time series data as input and input into the trained time series prediction model. The model will output the predicted multivariate time series data of the second set time step, which represents the predicted operating status values ​​of various components of the wind turbine in the future.

[0056] S4. Optimizing the multivariate time series data of the second set time step, and fusing the optimized data to construct a health index curve that can comprehensively reflect the health status of the wind turbine.

[0057] Specifically, the multivariate time series data of the second set time step obtained by prediction is optimized, that is, the most critical data points for the remaining life assessment are selected according to the influence of each component on the overall performance of the wind turbine and the reliability of the predicted data. Then, based on the optimized data, fusion processing is performed, and a health index curve that can comprehensively reflect the health status of the wind turbine is constructed through weighted average, principal component analysis or other fusion algorithms.

[0058] S5. Evaluate the remaining service life of the wind turbine based on the constructed health index curve.

[0059] Specifically, the remaining service life of wind turbines is evaluated based on the constructed health indicator curve and the preset threshold. The evaluation results are used to guide the formulation of preventive maintenance plans for wind turbines, ensuring that the equipment operates in the best condition, extending its service life and reducing operation and maintenance costs.

[0060] In one possible implementation, the training method of the time series prediction model is:

[0061] a. Obtain various types of operating data of various components of wind turbines during their entire life cycle that can reflect the degradation information of wind turbines.

[0062] b. Convert the multi-category operation data of the whole life cycle into time-synchronized multivariate time series data, and pre-process the multivariate time series data of the whole life cycle.

[0063] c. The preprocessed multivariate time series data of the whole life stage are intercepted according to the third set time step to obtain multiple training samples, in which the multivariate time point sequence data of the whole life stage before the first set time step is used as input, and the multivariate time point sequence data of the whole life stage after the second set time step is used as output, and the third set time step is the sum of the first set time step and the second set time step.

[0064] d. Use multiple training samples to train a long time series prediction network based on the Transformer framework to obtain a time series prediction model.

[0065] That is to say, in order to train an accurate and reliable time series prediction model, it is necessary to obtain multiple types of operating data of various components of wind turbines throughout their life cycle. These data should cover the entire time period from the equipment being put into operation to before being scrapped or overhauled, ensuring that the entire process information of the equipment from healthy to gradually degraded is included. The multiple types of operating data in the full life stage are synchronously processed according to the timestamps and converted into time-synchronized multivariate time series data. Subsequently, these data are comprehensively preprocessed to ensure the integrity and consistency of the data and provide a high-quality data basis for the subsequent training process. The preprocessed multivariate time series data in the full life stage are intercepted according to the preset third set time step (the time step is the sum of the first set time step and the second set time step). In this way, each intercepted data segment can be used as a training sample, where the data of the first set time step is used as the input feature of the model, and the data of the second set time step is used as the output target of the model. In this way, a large number of training samples can be generated for training time series prediction models. Using the generated multiple training samples, a long time series prediction network based on the Transformer framework is trained. The Transformer network is a deep learning model based on the self-attention mechanism, which is particularly suitable for processing sequence data. During the training process, by adjusting the network structure, optimization algorithm and hyperparameters, the model can learn the complex mapping relationship between the input sequence and the output sequence. As the training progresses, the prediction accuracy of the model will gradually improve until it reaches a satisfactory level. During the training process, the model needs to be evaluated to check whether its prediction performance meets the requirements. The evaluation can be carried out by calculating the error, accuracy and other indicators between the prediction results and the actual data. If the model performance is found to be poor, it can be optimized by adjusting the training parameters, increasing the training data and so on. After multiple iterations of training and optimization, a stable and reliable time series prediction model is finally obtained. The time series prediction model obtained through the above training will be used in the subsequent wind turbine residual life assessment process to achieve accurate prediction of the future operating status of the wind turbine and effective assessment of the remaining life.

[0066] In one possible implementation, the multiple types of operating data are converted into time-synchronized multivariate time series data, specifically:

[0067] The multi-class operating data are converted into time-synchronized multivariate time series data using an interpolation method.

[0068] Specifically, in order to ensure that the multiple types of operating data obtained from various components of the wind turbine can be analyzed and predicted in the same time frame, these data need to be converted into time-synchronized multivariate time series data. This embodiment uses the interpolation method to achieve data synchronization. By using the interpolation method to convert multiple types of operating data into time-synchronized multivariate time series data, the integrity and consistency of the data are ensured, providing a reliable data basis for subsequent time series analysis and prediction.

[0069] Exemplarily, the time domain index and frequency domain index of each segment of vibration data are calculated as the features of the corresponding moment, thereby obtaining multiple feature sequences that vary with time, and then the timestamps of different feature sequences are unified and aligned by interpolation.

[0070] In one embodiment, the key components of the wind turbine mainly include but are not limited to:

[0071] ① Main bearing: As an important supporting component of the wind turbine, the health status of the main bearing directly affects the operating stability and life of the whole machine. The degradation of the main bearing is usually manifested as increased vibration, increased temperature, etc.

[0072] ② Gearbox: The gearbox is a key component in a wind turbine that converts the rotor speed into the speed required by the generator. Failures such as wear and tear of its internal gears will directly affect the power generation efficiency and service life. The degradation information of the gearbox can be reflected through various parameters such as vibration, oil temperature, and oil quality.

[0073] ③ Generator: The generator is the core component of the wind power generation system, responsible for converting mechanical energy into electrical energy. The operating status of the generator directly affects the output quality of electrical energy and the overall efficiency of the system. The degradation of the generator may manifest itself in current fluctuations, abnormal temperature rise, etc.

[0074] In order to fully reflect the degradation information of the wind turbine, this embodiment selects the following key operating data:

[0075] ①Vibration data: Vibration is a common phenomenon in the operation of various components of wind turbines and is also an important indicator of the health status of components. Vibration sensors installed on key components can collect vibration signals in real time and analyze their characteristic parameters such as frequency and amplitude to evaluate the degradation of components such as wear, looseness, and imbalance.

[0076] ②Current data: The current data of the generator directly reflects the quality and stability of its output power. By monitoring the changes in current, it is possible to promptly detect faults such as short circuits, open circuits, overloads, etc. inside the generator, as well as abnormal power output caused by degradation of mechanical components.

[0077] ③ Oil temperature data: Oil temperature is an important indicator of the lubrication status and heat dissipation performance of components such as gearboxes. Too high oil temperature may mean that the performance of the lubricant has declined or the heat dissipation system has failed, which in turn accelerates the wear and degradation of components. Therefore, by monitoring the oil temperature data, faults caused by poor lubrication or insufficient heat dissipation can be discovered and prevented in a timely manner.

[0078] The data were divided into three main component categories: main bearings, gearboxes and generators according to the location of the measuring points. Subsequent operations were performed on the data of each component category to ensure that the method can accurately capture the health status of each component and predict its life.

[0079] In summary, through real-time monitoring and analysis of vibration data, current data and oil temperature data of key components such as main bearings, gearboxes and generators, the health status and degradation trend of wind turbines can be comprehensively evaluated, thereby achieving an accurate prediction of their remaining service life.

[0080] In one achievable manner, the multivariate time series data is preprocessed, specifically: the multivariate time series data is preprocessed using a maximum and minimum normalization processing method. Through the maximum and minimum normalization processing, the differences caused by different dimensions and numerical ranges between different data types and components can be eliminated, so that the model can more accurately capture the patterns and trends in the data, thereby improving the accuracy and reliability of the remaining life assessment.

[0081] In one achievable manner, the multivariate time series data of the second set time step is optimized, and the optimized data is fused to construct a health index curve that can comprehensively reflect the health status of the wind turbine, specifically:

[0082] The multivariate time series data of the second set time step in the multidimensional space are sorted according to monotonicity, and a characteristic sequence whose monotonicity meets the set requirements is selected; the characteristic sequence whose monotonicity meets the set requirements is reduced to one dimension using the principal component analysis method, and a health index curve that can comprehensively reflect the health status of the wind turbine is obtained.

[0083] Specifically, first, a monotonicity analysis is performed on the multivariate time series data of the second set time step of the multidimensional data. Monotonicity refers to the trend of data changing over time, such as increasing, decreasing or stable. According to the characteristics of the degradation process of wind turbines, feature sequences that change monotonically over time (such as increasing or decreasing) are usually selected because these sequences are more likely to reflect the degradation trend of components. In specific implementation, a monotonicity test can be performed on each feature sequence, and then the feature sequences are sorted according to the test results, and the feature sequences whose monotonicity meets the set requirements are selected as candidate sets for subsequent analysis. Since the selected feature sequences may still contain multiple dimensions, directly using them to evaluate the health status of wind turbines will increase the complexity and computational complexity of the model. Therefore, the principal component analysis method (PCA) is used to reduce the dimensionality of these feature sequences. PCA can map multidimensional data to a low-dimensional space through linear transformation, while retaining the key information in the original data as much as possible.

[0084] In this embodiment, a feature sequence whose monotonicity meets the set requirements is used as the input of PCA, and the dimension is reduced to one dimension through PCA to obtain a health index curve that can comprehensively reflect the health status of the wind turbine. This curve will serve as an important basis for subsequent remaining life assessment.

[0085] It should be understood that the health index curve is a direct reflection of the health status of the wind turbine. As the wind turbine operates and degrades, the value of the health index curve will gradually change. By monitoring the changing trend of this curve, potential failures and degradation signs of the wind turbine can be discovered in time, and corresponding maintenance measures can be taken. In addition, the health index curve can also be compared with a preset threshold to determine whether the health status of the wind turbine is within the normal range. If the value of the health index curve exceeds the threshold, it indicates that the wind turbine may have serious degradation or failure problems and needs to be shut down for maintenance immediately.

[0086] In one achievable manner, the remaining service life of the wind turbine generator is evaluated according to the constructed health index curve, specifically:

[0087] a. Count the average or median of the health indicators of wind turbines during the entire life cycle when a fault occurs, and set this value as the warning threshold for the occurrence of a fault.

[0088] That is to say, it is necessary to count the average or median of the health indicators when a failure occurs during the entire life cycle (i.e., the entire life cycle of the wind turbine from being put into operation to being scrapped). This statistical value represents the general level of the health indicators of the wind turbine when it is close to a failure state. To be conservative, the median can be selected as the warning threshold because it is not sensitive to extreme values ​​and can better reflect most situations. The warning threshold obtained by statistics is set as the standard for judging whether the wind turbine is about to fail.

[0089] b. If the health indicator curve does not exceed the warning threshold within the prediction period, it is determined that the remaining service life of the wind turbine is greater than or equal to the second set time step;

[0090] c. If the health indicator curve exceeds the warning threshold during the forecast period, the remaining service life of the fan is estimated based on the difference between the time when the health indicator curve first reaches the warning threshold and the current time.

[0091] Specifically, the health index curve is compared with the set warning threshold. During the prediction period, if the value of the health index curve never exceeds the warning threshold, then it can be determined that the remaining service life of the wind turbine is greater than or equal to the second set time step (i.e., the length of the prediction period), because the health index curve does not trigger the warning, indicating that the wind turbine is still in a healthy or relatively healthy state during the prediction period. If the health index curve exceeds the warning threshold during the prediction period, it is necessary to estimate the remaining service life of the wind turbine based on the difference between the time when the health index curve first reaches the warning threshold and the current time.

[0092] Through the above steps, the remaining service life of the wind turbine can be evaluated according to the health index curve, providing an important reference for the operation and maintenance management of the wind farm.

[0093] The present invention is described in more detail below in conjunction with a specific embodiment.

[0094] S1. Data Collection

[0095] In this embodiment, the operation data of each key component (main bearing, gearbox, generator) of the WTG146-3300 wind turbine are collected, including vibration data, current data and oil temperature data. These data are all with valid timestamps to ensure the timeliness and traceability of the data. Figure 2 A schematic diagram showing the vibration measurement points of the fan.

[0096] S2. Data preprocessing

[0097] Data preprocessing is a key step to ensure data quality and analysis accuracy. In this step, the following operations are performed:

[0098] a. Eliminate external interference: For oil temperature data, the influence of external temperature is eliminated to more accurately reflect the internal working status of the wind turbine.

[0099] b. Feature extraction: For vibration data, since the data is short-term data collected every few hours, this embodiment calculates the time domain index and frequency domain index of each segment of data as the features of the corresponding moment, thereby obtaining multiple feature sequences that change with time, forming a multivariate time series.

[0100] c. Timestamp unification: The timestamps of different feature sequences are unified and aligned through interpolation to ensure the consistency of the data in the time dimension.

[0101] d. Normalization: In order to eliminate the differences between different data types and dimensions, each time series is subjected to maximum and minimum normalization.

[0102] S3. Time Series Forecasting

[0103] This embodiment intercepts the latest 16 days of observation data and performs preprocessing. This embodiment takes the main bearing component of the fan as an example. The component is provided with 1 vibration measurement point and 2 temperature measurement points, and 15 common indicators are calculated for each vibration measurement point. Therefore, the feature dimension of the gearbox is 17 (15 indicators of 1 vibration measurement point + 2 temperature measurement points). This embodiment inputs a matrix with a shape of 17×96 (17 features, 96 time steps for each feature) into the trained time series prediction model, and outputs a matrix with a shape of 17×288, which is the multivariate time series prediction result for the next 48 days. Figure 3 The prediction results of the effective value RMS of the main bearing vibration are shown. The solid line represents the actual sequence, and the dotted line represents the predicted sequence. Although there are certain errors in the details, the overall trend prediction is relatively accurate.

[0104] It should be noted that the time series prediction model needs to be trained in advance, and the training samples have the following characteristics: the existing wind turbine life cycle data is processed with a sliding window with a window width of 384 to obtain multiple training samples, and there will be overlap between adjacent samples.

[0105] S4. Construction of health index curve

[0106] In order to comprehensively reflect the health status of wind turbines, this embodiment selects and fuses the predicted multivariate time series. First, the n-dimensional feature sequences are sorted by monotony, and the top 10 feature sequences with the best monotony are selected. Then, the PCA method is used to reduce the dimension of the 10-dimensional sequence to one dimension, and a health index curve is obtained. This curve can comprehensively reflect the health status of the wind turbine. If the curve is less than the set threshold, no fault will occur within the predicted time period; if the curve exceeds the threshold, it indicates the occurrence of potential faults.

[0107] S5. Remaining useful life assessment

[0108] This embodiment calculates the remaining service life of the wind turbine based on the health index curve and the set threshold. If the health index does not exceed the threshold during the prediction period, it means that the remaining service life of the wind turbine is greater than or equal to the prediction period (48 days in this example), and it is estimated as 48 days. If the health index curve exceeds the threshold during the prediction period, the time when the curve first reaches the threshold is recorded as the time when the potential fault occurs, and the difference from the current time to this time is calculated as the estimated value of the remaining service life of the wind turbine. Figure 4 A schematic diagram of this process is shown, in which the diamond mark represents the current moment, the curve after the diamond is the predicted health indicator curve, and its intersection with the threshold line is the predicted moment when the wind turbine main bearing fails. The horizontal distance from the diamond mark to this intersection is the estimated remaining life of the main bearing.

[0109] In another embodiment of the present invention, a device for evaluating the remaining life of a wind turbine based on time series prediction is provided, comprising:

[0110] The acquisition module is used to acquire multiple types of operating data of various components of the wind turbine within a set time period that can reflect the degradation information of the wind turbine, and each type of operating data contains valid timestamp information.

[0111] The conversion module is used to convert the multiple types of operation data into time-synchronized multivariate time series data and pre-process the multivariate time series data.

[0112] The prediction module is used to intercept the multivariate time point sequence data of the first set time step from the preprocessed multivariate time series data, input the multivariate time point sequence data of the first set time step into the trained time series prediction model, and output the predicted multivariate time series data of the second set time step.

[0113] The fusion module is used to optimize the multivariate time series data of the second set time step, and fuse the optimized data to construct a health index curve that can comprehensively reflect the health status of the wind turbine.

[0114] All relevant contents of each step involved in the aforementioned embodiment of a method for assessing the remaining life of a wind turbine based on time series prediction can be referred to the functional description of the functional module corresponding to a device for assessing the remaining life of a wind turbine based on time series prediction in the embodiment of the present invention, and will not be repeated here. The division of modules in the embodiment of the present invention is schematic, and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional module in each embodiment of the present invention may be integrated into a processor, or may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated modules can be implemented in the form of hardware or in the form of software functional modules.

[0115] In another embodiment of the present invention, a computer device is provided, the computer device comprising a processor and a memory, the memory being used to store a computer program, the computer program comprising program instructions, and the processor being used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, which are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions in a computer storage medium to implement a corresponding method flow or corresponding function; the processor described in the embodiment of the present invention can be used for the operation of a method for evaluating the remaining life of a wind turbine generator based on time series prediction.

[0116] In another embodiment of the present invention, the present invention further provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It can be understood that the computer-readable storage medium here can include both built-in storage media in the computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the above-mentioned embodiment regarding a method for evaluating the remaining life of a wind turbine based on time series prediction.

[0117] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0118] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0119] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0120] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0121] In the present invention, the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" etc. mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the described specific features, structures, materials or characteristics may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0122] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for evaluating the remaining life of a wind turbine generator based on time series prediction, characterized in that: include: Acquire multiple types of operating data of various components of the wind turbine within a set time period that can reflect the degradation information of the wind turbine, and each type of operating data contains valid timestamp information; Converting the multiple types of operation data into time-synchronized multivariate time series data, and preprocessing the multivariate time series data; Extracting multivariate time point sequence data of a first set time step from the preprocessed multivariate time series data, inputting the multivariate time point sequence data of the first set time step into a trained time series prediction model, and outputting predicted multivariate time series data of a second set time step; Performing optimization processing on the multivariate time series data of the second set time step, and fusing the optimized data to construct a health index curve that can comprehensively reflect the health status of the wind turbine; The remaining service life of the wind turbine is evaluated based on the constructed health index curve.

2. A method for evaluating the remaining life of a wind turbine generator based on time series prediction according to claim 1, characterized in that: The training method of the time series prediction model is: Obtain various types of operating data of wind turbine components during their entire life cycle that can reflect the degradation information of wind turbines; Convert the multi-class operation data of the whole life cycle into time-synchronized multivariate time series data, and pre-process the multivariate time series data of the whole life cycle; The preprocessed multivariate time series data of the whole life stage are intercepted according to the third set time step to obtain a plurality of training samples, wherein the multivariate time point series data of the whole life stage before the first set time step in each training sample is used as input, and the multivariate time point series data of the whole life stage after the second set time step is used as output, and the third set time step is the sum of the first set time step and the second set time step; Multiple training samples are used to train a long time series prediction network based on the Transformer framework to obtain a time series prediction model.

3. The method for evaluating the remaining life of a wind turbine generator based on time series prediction according to claim 1, characterized in that: The converting of the multiple types of operation data into time-synchronized multivariate time series data is specifically: The multi-class operating data are converted into time-synchronized multivariate time series data using an interpolation method.

4. The method for evaluating the remaining life of a wind turbine generator based on time series prediction according to claim 1, characterized in that: The wind turbine components include a main bearing, a gearbox and a generator; The multiple types of operating data capable of reflecting the degradation information of the wind turbine generator include vibration data, current data and oil temperature data.

5. The method for evaluating the remaining life of a wind turbine generator based on time series prediction according to claim 1, characterized in that: The preprocessing of the multivariate time series data is specifically as follows: The multivariate time series data are preprocessed using the maximum and minimum normalization processing method.

6. The method for evaluating the remaining life of a wind turbine generator based on time series prediction according to claim 1, characterized in that: The method of performing optimal processing on the multivariate time series data of the second set time step and fusing the optimal data to construct a health index curve that can comprehensively reflect the health status of the wind turbine is specifically as follows: Sorting the multidimensional multivariate time series data of the second set time step according to monotonicity, and selecting a feature sequence whose monotonicity meets the set requirements; The principal component analysis method is used to reduce the dimension of the feature sequence whose monotonicity meets the set requirements to one dimension, and a health index curve that can comprehensively reflect the health status of the wind turbine is obtained.

7. The method for evaluating the remaining life of a wind turbine generator based on time series prediction according to claim 1, characterized in that: The remaining service life of the wind turbine generator is evaluated according to the constructed health index curve, specifically: Count the average or median of health indicators of wind turbines during their entire life cycle when a fault occurs, and set the value as the warning threshold for the occurrence of a fault; If the health indicator curve does not exceed the warning threshold value within the prediction period, it is determined that the remaining service life of the wind turbine is greater than or equal to the second set time step; If the health indicator curve exceeds the warning threshold during the forecast period, the remaining service life of the wind turbine is estimated based on the difference between the time when the health indicator curve first reaches the warning threshold and the current time.

8. A wind turbine remaining life assessment device based on time series prediction, characterized in that: include: An acquisition module is used to acquire multiple types of operating data of various components of the wind turbine within a set time period that can reflect the degradation information of the wind turbine, and each type of operating data contains valid timestamp information; A conversion module, used for converting the multiple types of operation data into time-synchronized multivariate time series data, and preprocessing the multivariate time series data; A prediction module, used for intercepting multivariate time point sequence data of a first set time step from the preprocessed multivariate time series data, inputting the multivariate time point sequence data of the first set time step into a trained time series prediction model, and outputting predicted multivariate time series data of a second set time step; A fusion module, used for optimizing the multivariate time series data of the second set time step, and fusing the optimized data to construct a health index curve that can comprehensively reflect the health status of the wind turbine; The evaluation module is used to evaluate the remaining service life of the wind turbine according to the constructed health index curve.

9. A device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method for assessing the remaining life of a wind turbine based on time series prediction as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of a method for evaluating the remaining life of a wind turbine based on time series prediction as described in any one of claims 1 to 7 are implemented.

Images