A wind turbine generator cabin sliding prediction method, device, equipment and storage medium

Abstract

The application discloses a wind turbine generator set cabin sliding prediction method, device, equipment and storage medium, the method comprises the following steps: obtaining the operation parameter of the unit in the non-yaw state; the cabin sliding frequency ratio and the proximity switch change frequency ratio are constructed by using the operation parameter; according to two ratios and prediction current fault model, the prediction result of whether the cabin exists sliding fault is obtained; the model construction process is: according to the cabin sliding fault information, the historical operation parameter fault is labeled; the historical operation parameter within the preset time length before the cabin sliding fault corresponding historical operation parameter is labeled as fault; two historical ratios are constructed, and the prediction current fault model is obtained by training. The technical scheme disclosed by the application, by predicting the current fault model construction, the operation parameters within the preset time length before the fault are all labeled as faults, and the two ratios constructed and the prediction current fault model are used to realize early detection of faults in the early stage of faults, so that proactive maintenance protection is realized.

Description

Technical Field

[0001] This application relates, more specifically, to a method, apparatus, equipment, and storage medium for predicting the sliding of a wind turbine nacelle. Background Technology

[0002] As an important component of wind turbine generator sets, the yaw brake functions to provide sufficient braking force to prevent the nacelle from shifting position under operating conditions such as power generation, turbulence, wind shear, and yaw failure, thereby ensuring the reliability and safety of the generator set operation.

[0003] Current wind turbine nacelle slip detection methods constantly calculate the difference between the nacelle's current position and the position recorded when the yaw drive stops. If the difference exceeds a set threshold in the non-yaw state, the main controller issues a shutdown command to protect the turbine to the greatest extent possible. However, this is a passive protection measure, only triggered when the yaw brake fails. Furthermore, the existing solution uses a single-point judgment; when the difference exceeds the threshold in the non-yaw state, it indicates that the nacelle slip is already very severe, and in most cases, the equipment is damaged beyond repair. This detection method results in high maintenance costs, long repair times, and significant power generation losses.

[0004] In summary, how to predict the nacelle slippage hazard of wind turbine units in advance is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the purpose of this application is to provide a method, apparatus, equipment and storage medium for predicting nacelle slippage in wind turbines, for making advance predictions of nacelle slippage hazards in wind turbines.

[0006] To achieve the above objectives, this application provides the following technical solution:

[0007] A method for predicting nacelle slip in a wind turbine includes:

[0008] Obtain the operating parameters of the wind turbine under test in non-yaw condition from historical time to the current time;

[0009] The percentage of first cabin sliding times and the percentage of first proximity switch changes are constructed using the aforementioned operating parameters;

[0010] Based on the proportion of sliding times of the first nacelle, the proportion of changes of the first proximity switch, and the pre-built prediction model of the current fault, the prediction result of whether the nacelle of the wind turbine under test has a sliding fault is obtained.

[0011] The process of constructing the current fault prediction model is as follows:

[0012] Historical operating parameters and nacelle sliding fault information of the wind turbine under non-yaw conditions are obtained. Faults are marked on the historical operating parameters based on the nacelle sliding fault information. Specifically, historical operating parameters within a preset time period before the historical operating parameter corresponding to the nacelle sliding fault are all marked as faults. The historical operating parameters are used to construct the percentage of historical nacelle sliding times and the percentage of historical proximity switch changes. The model is trained using the percentage of historical nacelle sliding times, the percentage of historical proximity switch changes, and the corresponding fault markings to obtain the current fault prediction model.

[0013] Preferably, before obtaining the prediction result of whether the nacelle of the wind turbine under test has a sliding fault based on the proportion of nacelle sliding times, the proportion of proximity switch changes, and the pre-built prediction model of the current fault, the method further includes:

[0014] Obtain the historical operating parameters of the wind turbine under test in a non-yaw state, and use the historical operating parameters to construct the proportion of the second nacelle sliding times and the proportion of the second proximity switch changes.

[0015] The percentage of the second cabin sliding times and the percentage of the second proximity switch changes are input into a pre-built prediction model for future cabin sliding, so as to obtain a future prediction result of whether the cabin will have a sliding failure in the future time after the time corresponding to the historical operating parameters.

[0016] Obtain the prediction result corresponding to the current time from the future prediction results;

[0017] The process of constructing the predicted future cabin sliding model is as follows:

[0018] The wind turbine is labeled with a tag indicating whether it will fail in the future based on the nacelle sliding fault information; the model is trained using the percentage of historical nacelle sliding times, the percentage of historical proximity switch changes, and the tag indicating whether it will fail in the future, to obtain the model predicting future nacelle sliding.

[0019] All historical operating parameters within a preset time period preceding the historical operating parameters corresponding to the cabin sliding fault are marked as faults, including:

[0020] The historical operating parameters corresponding to the aforementioned cabin sliding fault are marked as faults;

[0021] Based on the percentage of sliding events of the first nacelle, the percentage of changes in the first proximity switch, and a pre-built prediction model for the current fault, a prediction result is obtained regarding whether the nacelle of the wind turbine under test has a sliding fault, including:

[0022] Based on the percentage of sliding times of the first nacelle, the percentage of changes of the first proximity switch, the prediction result corresponding to the current time, and the prediction model for the current fault, the prediction result for whether the nacelle of the wind turbine under test has a sliding fault is obtained.

[0023] Preferably, labeling the wind turbine with a tag indicating whether it will fail in the future based on the nacelle sliding fault information includes:

[0024] For any given moment of the wind turbine, query at equal intervals according to the optimal sliding step size until the most recent fault occurs;

[0025] Calculate the future failure percentage based on the number of queries and the number of cabin sliding failures found.

[0026] If the proportion of future failures is greater than the preset proportion, then a failure label is marked at the specified time.

[0027] If the percentage of future failures is not greater than the preset percentage, then the time point is labeled as normal.

[0028] Preferably, the model for predicting future cabin sliding is trained using the historical cabin sliding frequency ratio, the historical proximity switch change frequency ratio, and the label indicating whether a malfunction is imminent, including:

[0029] The model predicting future cabin sliding is obtained by training any one of Bayesian network, random forest, and neural network using the historical cabin sliding frequency ratio, the historical proximity switch change frequency ratio, and the label of whether a future malfunction will occur.

[0030] Preferably, the historical cabin sliding frequency ratio and the historical proximity switch change frequency ratio are constructed using the historical operating parameters, including:

[0031] The number of different cabin positions, the number of different proximity switch states, the average wind speed within a set time period before and after the cabin position changes, and the difference between two adjacent cabin positions are counted within the interval between two adjacent yaw times.

[0032] If the difference between two adjacent nacelle positions in the current statistics is greater than the width of one tooth pitch of the wind turbine, and / or if the average wind speed within a set time period before and after the change of nacelle position in the current statistics is greater than the design load wind speed of the turbine, then the current statistics are discarded.

[0033] Based on the optimal sliding step size, the sliding statistics are performed on the number of times the wind turbine's yaw interval is between two adjacent yaw times, the number of times the number of different nacelle positions within the interval between two adjacent yaw times is greater than 1, and the number of times the number of different proximity switch states within the interval between two adjacent yaw times is greater than 1.

[0034] The percentage of historical proximity switch changes is obtained based on the number of times the number of different proximity switch states is greater than 1 within the interval between two adjacent yaw times and the number of intervals between two adjacent yaw times of the wind turbine.

[0035] The percentage of historical cabin sliding times is obtained based on the number of times the number of different cabin positions is greater than 1 within two adjacent yaw intervals and the number of times the number of different proximity switch states is greater than 1 within two adjacent yaw intervals.

[0036] Preferably, the historical operating parameters of the wind turbine in non-yaw condition are obtained, including:

[0037] Collect historical real-time operating parameters of the wind turbine; the historical real-time operating parameters include the yaw state machine.

[0038] The yaw time is selected by the yaw state machine, and the non-yaw state of the wind turbine is obtained based on the yaw time; wherein, the time between two adjacent yaw times is the non-yaw time, and the non-yaw time corresponds to the non-yaw state.

[0039] Obtain the historical operating parameters of the wind turbine in non-yaw state.

[0040] Preferably, after collecting the historical real-time operating parameters of the wind turbine, the method further includes:

[0041] Based on linear interpolation data along the time axis;

[0042] After obtaining the historical operating parameters of the wind turbine in a non-yaw state, the method further includes:

[0043] If the non-yaw time is less than a set time length threshold, then the non-yaw time and the corresponding historical real-time operating parameters are deleted.

[0044] A wind turbine nacelle sliding prediction device, comprising:

[0045] The first acquisition module is used to acquire the operating parameters of the wind turbine under test in a non-yaw state from historical time to the current time.

[0046] The module is used to construct the percentage of first cabin sliding times and the percentage of first proximity switch changes using the operating parameters;

[0047] The prediction result module is used to obtain a prediction result of whether the nacelle of the wind turbine under test has a sliding fault based on the proportion of the first nacelle sliding times, the proportion of the first proximity switch change times, and the pre-built prediction current fault model.

[0048] It also includes a first building module for constructing the predictive current fault model, the first building module comprising:

[0049] The first construction unit is used to acquire historical operating parameters and nacelle slip fault information of the wind turbine in non-yaw state, and to label the historical operating parameters with faults based on the nacelle slip fault information; wherein, all historical operating parameters within a preset time period before the historical operating parameter corresponding to the nacelle slip fault are labeled as faults; the historical operating parameters are used to construct the historical nacelle slip frequency ratio and the historical proximity switch change frequency ratio, and the historical nacelle slip frequency ratio, the historical proximity switch change frequency ratio and the corresponding fault labels are used for training to obtain the current fault prediction model.

[0050] A wind turbine nacelle sliding prediction device includes:

[0051] Memory, used to store computer programs;

[0052] A processor for executing the computer program to implement the steps of the wind turbine nacelle sliding prediction method as described in any of the preceding claims.

[0053] A readable storage medium storing a computer program that, when executed by a processor, implements the steps of the wind turbine nacelle sliding prediction method as described in any of the preceding claims.

[0054] This application provides a method, apparatus, device, and storage medium for predicting nacelle slippage in wind turbines. The method includes: acquiring the operating parameters of the wind turbine under test from historical time to the current time under non-yaw conditions; constructing a first nacelle slippage frequency ratio and a first proximity switch change frequency ratio using the operating parameters; and obtaining a prediction result of whether the nacelle of the wind turbine under test has a slippage fault based on the first nacelle slippage frequency ratio, the first proximity switch change frequency ratio, and a pre-constructed current fault prediction model. The construction process of the current fault prediction model is as follows: acquiring historical operating parameters and nacelle slippage fault information of the wind turbine under non-yaw conditions; labeling historical operating parameters with faults based on the nacelle slippage fault information; wherein historical operating parameters within a preset time period preceding the historical operating parameters corresponding to the nacelle slippage fault are all labeled as faults; constructing historical nacelle slippage frequency ratio and historical proximity switch change frequency ratio using the historical operating parameters; and training the model using the historical nacelle slippage frequency ratio, historical proximity switch change frequency ratio, and corresponding fault labels to obtain the current fault prediction model.

[0055] The technical solution disclosed in this application pre-acquires historical operating parameters and nacelle slip fault information of the wind turbine in a non-yaw state. Based on the nacelle slip fault information, faults are labeled on the historical operating parameters. During fault labeling, all historical operating parameters within a preset time period preceding the corresponding nacelle slip fault are labeled as faults, enabling fault detection before the nacelle slip fault occurs, thus achieving early fault prediction. Then, the historical operating parameters are used to construct the percentage of historical nacelle slip occurrences and the percentage of historical proximity switch changes. These percentages, along with the corresponding fault labels, are used to train a model predicting the current fault. After obtaining the model, the operating parameters of the wind turbine under test in a non-yaw state from historical time to the current time are acquired. The operating parameters are then used to construct the percentage of first nacelle slip occurrences and the percentage of first proximity switch changes, allowing for process judgment using operating parameters over a period of time. This enables early fault detection and improves the accuracy of fault judgment. Subsequently, based on the proportion of first nacelle sliding occurrences, the proportion of first proximity switch changes, and the pre-built current fault prediction model, a prediction result is obtained regarding whether the nacelle of the wind turbine under test has a sliding fault. Through the above process, this application achieves early fault detection by marking all operating parameters within a preset time period before the fault as faults in the current fault prediction model construction, constructing the proportion of first nacelle sliding occurrences and the proportion of first proximity switch changes, and utilizing the proportion of first nacelle sliding occurrences and the proportion of first proximity switch changes, along with the current fault prediction model. This enables proactive maintenance and protection, reducing damage to the wind turbine, maintenance costs, and improving the power generation and operational reliability of the wind turbine. Attached Figure Description

[0056] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0057] Figure 1 A flowchart illustrating a wind turbine nacelle sliding prediction method provided in this application embodiment;

[0058] Figure 2 A schematic diagram illustrating the construction and verification of the future cabin sliding model and the current fault prediction model provided in the embodiments of this application;

[0059] Figure 3 This is a schematic diagram illustrating the implementation of labeling wind turbines with tags indicating whether they will fail in the future based on nacelle sliding fault information, as provided in an embodiment of this application.

[0060] Figure 4 A flowchart illustrating the construction of the percentage of cabin sliding times and the percentage of proximity switch changes provided in embodiments of this application;

[0061] Figure 5 A schematic diagram of a wind turbine nacelle sliding prediction device provided in an embodiment of this application;

[0062] Figure 6 This is a schematic diagram of the structure of a wind turbine nacelle sliding prediction device provided in an embodiment of this application. Detailed Implementation

[0063] Currently, when performing nacelle sliding protection, the unit protects itself through a control program: the difference between the current position of the nacelle and the value recorded when the yaw drive stops is constantly calculated. If the difference is greater than a set threshold in the non-yaw state, the main controller issues a shutdown command.

[0064] However, this is a passive protection measure, triggering the protection logic only when the yaw brake fails. Furthermore, the existing solution uses a single-point judgment; when the difference in the non-yaw state exceeds a threshold, it indicates that the nacelle slippage is already very severe, and in most cases, the equipment is damaged beyond repair. This detection method results in high maintenance costs, long repair times, and significant power generation losses.

[0065] Therefore, this application provides a method, apparatus, equipment and storage medium for predicting nacelle slippage in wind turbines, which can be used to predict the nacelle slippage hazard of wind turbines in advance.

[0066] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0067] See Figure 1 The document illustrates a flowchart of a wind turbine nacelle sliding prediction method. The wind turbine nacelle sliding prediction method provided in this application embodiment may include:

[0068] S11: Obtain historical operating parameters and nacelle slip fault information of the wind turbine in non-yaw state, and mark the historical operating parameters as faults based on the nacelle slip fault information; among them, all historical operating parameters within a preset time period before the historical operating parameters corresponding to the nacelle slip fault are marked as faults.

[0069] In this application, historical operating parameters and nacelle slip fault information of multiple wind turbines under non-yaw conditions can be obtained first (there is a temporal correspondence between historical operating parameters and nacelle slip fault information, meaning that historical operating parameters and nacelle slip fault information are obtained simultaneously at the same time). Specifically, the historical operating parameters may include nacelle position, the status of yaw proximity switches 1-2 (i.e., yaw proximity switch 1 and yaw proximity switch 2), and wind speed information. Yaw proximity switches 1 and 2 are two sensors, and changes in nacelle position can be obtained through these two sensors.

[0070] Then, based on the cabin sliding fault information, the corresponding historical operating parameters can be labeled as faults. When labeling faults, in addition to marking the historical operating parameter corresponding to the cabin sliding fault as a fault (e.g., labeled as 1), all historical operating parameters within a preset time period preceding the historical operating parameter corresponding to the cabin sliding fault are also marked as faults (e.g., labeled as 1). The preset time period can be set according to actual needs. By marking all historical operating parameters within the preset time period preceding the historical operating parameter corresponding to the cabin sliding fault as faults, it is easier to predict the cabin sliding fault before it occurs, thus achieving early prediction of cabin sliding faults.

[0071] It should be noted that, apart from the above historical operating parameters being marked as faults, all other historical operating parameters are marked as not faulty (or marked as normal, for example, all marked as 0).

[0072] In addition, after labeling historical operating parameters for faults, all historical operating parameters can be divided into training and validation sets at a certain ratio (e.g., 7:3). Specifically, the division of the training and validation sets can be done chronologically.

[0073] S12: Construct the historical cabin sliding frequency ratio and the historical proximity switch change frequency ratio using historical operating parameters. Use the historical cabin sliding frequency ratio, the historical proximity switch change frequency ratio and the corresponding fault labels to train and obtain the current fault prediction model.

[0074] After obtaining historical operating parameters of multiple wind turbines under non-yaw conditions, two feature variables can be constructed using the historical operating parameters of each wind turbine in the training set under non-yaw conditions: the percentage of historical nacelle slides and the percentage of historical proximity switch changes. The proximity switches refer to the aforementioned yaw proximity switches 1 and 2. The constructed percentage of historical nacelle slides and the percentage of historical proximity switch changes are two process variables. That is, these percentages are constructed from historical operating parameters over a period of time, allowing for process detection based on these two feature variables, thereby improving the reliability and accuracy of nacelle slide detection.

[0075] After constructing the historical cabin sliding frequency ratio and the historical proximity switch change frequency ratio, the current fault model can be obtained by using the historical cabin sliding frequency ratio, the historical proximity switch change frequency ratio and the fault labels corresponding to the historical operating parameters in the training set, which are constructed using historical operating data in the training set.

[0076] After training the current fault model, the effectiveness of the current fault model can be validated using a partitioned validation set, and the model can be evaluated using precision, recall, F1 score, and ROC_AUC. F1 is a statistical metric used to measure the accuracy of a binary classification model; the F1 score can be seen as a weighted average of the model's precision and recall, with a maximum value of 1 and a minimum value of 0. ROC is the receiver operating characteristic curve, a comprehensive indicator reflecting the sensitivity and specificity of continuous variables. AUC is the area under the ROC curve.

[0077] It should be noted that steps S11-S12 are the process of pre-constructing a model to predict the current fault. These two steps can be performed only once in actual execution to obtain the model, which can then be directly used to predict cabin sliding faults. Alternatively, the model can be updated according to steps S11-S12 based on new or historical operating parameters to improve its accuracy and precision.

[0078] S13: Obtain the operating parameters of the wind turbine under test in non-yaw condition from historical time to the current time.

[0079] When performing slip prediction on a wind turbine nacelle, the operating parameters of the wind turbine under test in a non-yaw state from historical time to the present time can be obtained. The wind turbine under test can be one of the aforementioned multiple wind turbines, or it can be any other than one of them; this application does not impose any limitation on this. The historical time can be determined based on the current time and the conditions during the training of the current fault prediction model. Furthermore, the operating parameters mentioned here can specifically include the nacelle position and the state variables of yaw proximity switches 1 and 2.

[0080] S14: Construct the percentage of the first cabin sliding times and the percentage of the first proximity switch changes using operating parameters.

[0081] After obtaining the operating parameters of the wind turbine under test in the non-yaw state from the historical time to the current time, the operating parameters of the wind turbine under test in the non-yaw state from the historical time to the current time can be used to construct the percentage of the first nacelle sliding times and the percentage of the first proximity switch changes.

[0082] S15: Based on the proportion of sliding times of the first nacelle, the proportion of changes of the first proximity switch, and the pre-built prediction model of the current fault, the prediction result of whether the nacelle of the wind turbine under test has a sliding fault is obtained.

[0083] Based on step S14, the constructed percentage of the first nacelle sliding times and the percentage of the first proximity switch changes can be input into the pre-constructed current fault prediction model to obtain the prediction result of whether the nacelle of the wind turbine under test currently has a sliding fault. This enables the prediction of nacelle sliding faults in advance, thereby facilitating relevant personnel to proactively carry out maintenance and protection of the nacelle in advance, reducing the damage and maintenance costs of nacelle sliding to the wind turbine, and improving the power generation and operational reliability of the wind turbine.

[0084] The technical solution disclosed in this application pre-acquires historical operating parameters and nacelle slip fault information of the wind turbine in a non-yaw state. Based on the nacelle slip fault information, faults are labeled on the historical operating parameters. During fault labeling, all historical operating parameters within a preset time period preceding the corresponding nacelle slip fault are labeled as faults, enabling fault detection before the nacelle slip fault occurs, thus achieving early fault prediction. Then, the historical operating parameters are used to construct the percentage of historical nacelle slip occurrences and the percentage of historical proximity switch changes. These percentages, along with the corresponding fault labels, are used to train a model predicting the current fault. After obtaining the model, the operating parameters of the wind turbine under test in a non-yaw state from historical time to the current time are acquired. The operating parameters are then used to construct the percentage of first nacelle slip occurrences and the percentage of first proximity switch changes, allowing for process judgment using operating parameters over a period of time. This enables early fault detection and improves the accuracy of fault judgment. Subsequently, based on the proportion of first nacelle sliding occurrences, the proportion of first proximity switch changes, and the pre-built current fault prediction model, a prediction result is obtained regarding whether the nacelle of the wind turbine under test has a sliding fault. Through the above process, this application achieves early fault detection by marking all operating parameters within a preset time period before the fault as faults in the current fault prediction model construction, constructing the proportion of first nacelle sliding occurrences and the proportion of first proximity switch changes, and utilizing the proportion of first nacelle sliding occurrences and the proportion of first proximity switch changes, along with the current fault prediction model. This enables proactive maintenance and protection, reducing damage to the wind turbine, maintenance costs, and improving the power generation and operational reliability of the wind turbine.

[0085] The wind turbine nacelle slip prediction method provided in this application embodiment may further include, before obtaining the prediction result of whether the nacelle of the wind turbine under test has a slip fault based on the proportion of nacelle slip occurrences, the proportion of proximity switch change occurrences, and a pre-constructed prediction model of the current fault, the following:

[0086] Obtain historical operating parameters of the wind turbine under test in non-yaw state, and use the historical operating parameters to construct the proportion of the second nacelle sliding times and the proportion of the second proximity switch changes;

[0087] The proportion of the second cabin sliding times and the proportion of the second proximity switch changes are input into the pre-built prediction model of future cabin sliding to obtain the future prediction result of whether the cabin will have a sliding failure in the future time after the time corresponding to the historical operating parameters.

[0088] Obtain the prediction result corresponding to the current time from the future prediction result;

[0089] The process of constructing a model predicting the future flight deck sliding mechanism is as follows:

[0090] The wind turbine is labeled with a tag indicating whether it will fail in the future based on the nacelle sliding fault information; the model is trained using the historical nacelle sliding frequency ratio, the historical proximity switch change frequency ratio, and the future failure label to obtain a model predicting future nacelle sliding.

[0091] All historical operating parameters within a preset time period preceding the historical operating parameters corresponding to the cabin sliding fault are marked as faults, which may include:

[0092] Mark the historical operating parameters corresponding to the nacelle sliding fault as the fault;

[0093] Based on the proportion of sliding events in the first nacelle, the proportion of changes in the first proximity switch, and the pre-built prediction model for the current fault, the prediction result of whether the nacelle of the wind turbine under test has a sliding fault is obtained, which may include:

[0094] Based on the proportion of sliding times of the first nacelle, the proportion of changes of the first proximity switch, the prediction results corresponding to the current time, and the current fault prediction model, the prediction results of whether the nacelle of the wind turbine under test has a sliding fault are obtained.

[0095] In this application, a model predicting future nacelle slippage can also be pre-constructed. Specifically, the wind turbine can be labeled with a future failure indicator based on nacelle slippage fault information (specifically, historical operating parameters in the training set can be labeled with future failure indicators). Then, the model is trained using the historical operating parameters in the training set, corresponding to the percentage of historical nacelle slippage occurrences, the percentage of historical proximity switch changes, and the future failure indicator labels, to obtain the model predicting future nacelle slippage. This allows the model to participate in nacelle slippage prediction, improving the accuracy of the current failure prediction model. See [link to details]. Figure 2 It shows a schematic diagram of the construction and verification of the future cabin sliding model and the current fault prediction model provided in the embodiments of this application. After the future cabin sliding model is constructed, the future cabin sliding model and the current fault prediction model can be verified using the verification set.

[0096] Based on the construction of a model predicting future cabin sliding failures, during the construction of a model predicting current failures, when all historical operating parameters within a preset time period preceding the historical operating parameters corresponding to the cabin sliding failure are marked as failures, only the historical operating parameters corresponding to the cabin sliding failure can be marked as failures. This allows for the early prediction of cabin sliding failures through the participation of the model predicting future cabin sliding failures. In other words, when using only the model predicting current failures for cabin sliding failure prediction, the model is constructed by marking all historical operating parameters within a preset time period preceding the historical operating parameters corresponding to the cabin sliding failure as failures, where the preset time period is greater than 0 and can be in days. When using both the model predicting current failures and the model predicting future cabin sliding failures for cabin sliding failure prediction, the model is constructed by marking only the historical operating parameters corresponding to the cabin sliding failure as failures, meaning the preset time period is equal to 0.

[0097] After constructing the future nacelle slip model, before obtaining the prediction result of whether the nacelle of the wind turbine under test has a slip fault based on the proportion of nacelle slip occurrences, the proportion of proximity switch changes, and the pre-constructed prediction model of the current fault, the historical operating parameters of the wind turbine under test in non-yaw state (historical operating parameters corresponding to the time before the current time) can be obtained. The obtained historical operating parameters are used to construct a second proportion of nacelle slip occurrences and a second proportion of proximity switch changes. Then, the second proportion of nacelle slip occurrences and the second proportion of proximity switch changes are input into the pre-constructed prediction model of the future nacelle slip to obtain the future prediction result of whether the nacelle has a slip fault in the future time after the time corresponding to the historical operating parameters. Afterwards, the prediction result corresponding to the current time can be obtained from the future prediction result of whether the nacelle has a slip fault in the future time after the time corresponding to the historical operating parameters. Furthermore, if the future prediction results obtained by the wind turbine under test based on multiple historical operating parameters and the prediction model of the future nacelle slip all contain the prediction result corresponding to the current time, all of them are obtained, and these multiple prediction results corresponding to the current time are integrated (e.g., by averaging, weighted averaging, etc.) to obtain the integrated prediction result corresponding to the current time. For example, if the current time is December 25th, the future prediction results obtained from the historical operating parameters from December 1st to December 7th and the future cabin sliding model include the prediction results for December 25th; the future prediction results obtained from the historical operating parameters from December 8th to December 14th and the future cabin sliding model include the prediction results for December 25th; and the future prediction results obtained from the historical operating parameters from December 15th to December 21st and the future cabin sliding model include the prediction results for December 25th. Therefore, the three prediction results corresponding to December 25th can be integrated to obtain the integrated prediction result for December 25th.

[0098] Based on the above, when obtaining the prediction result of whether the nacelle of the wind turbine under test has a sliding fault according to the proportion of the first nacelle sliding frequency, the proportion of the first proximity switch change frequency, and the pre-constructed prediction current fault model, the proportion of the first nacelle sliding frequency, the proportion of the first proximity switch change frequency, and the prediction result corresponding to the current time (or the integrated prediction result if integrated) can be input into the prediction current fault model to obtain the prediction result of whether the nacelle of the wind turbine under test has a sliding fault. That is, it realizes the prediction result corresponding to the current time using the operating parameters before the current time and the prediction future nacelle sliding model, and predicts whether the nacelle of the wind turbine under test has a sliding fault based on the operating parameters corresponding to the current time, the prediction result corresponding to the current time, and the prediction current fault model, thereby improving the accuracy of the prediction.

[0099] As can be seen from the above, this application uses wind turbine operating parameters for modeling and prediction, which is low-cost, efficient, and highly interpretable. It predicts future unit failures based on current data and predicts current failures based on current data and historical data, thereby enhancing the model's sensitivity to the changing trends of unit nacelle sliding failures.

[0100] It should be noted that after constructing the percentage of the first cabin sliding times and the percentage of the first proximity switch changes at the current time, the percentage of the first cabin sliding times and the percentage of the first proximity switch changes at the current time can be input into the future cabin sliding prediction model to predict whether the cabin will have a sliding failure in the future time after the current time, so that it can participate in the prediction of whether the cabin will slide in the future time.

[0101] This application provides a method for predicting nacelle slippage in wind turbines, which labels wind turbines with potential future failures based on nacelle slippage fault information. This method may include:

[0102] For any given moment of the wind turbine, query at equal intervals according to the optimal sliding step size until the most recent fault occurs;

[0103] Calculate the future failure percentage based on the number of queries and the number of cabin sliding failures found.

[0104] If the proportion of future failures exceeds the preset proportion, then a failure label will be marked at that time.

[0105] If the future failure rate is not greater than the preset rate, then the time period will be labeled as normal.

[0106] In this application, the goal is to train the model to predict the current fault with an F1 score greater than or equal to 0.95. The Expectation-Maximization (EM) algorithm is used to obtain the specific number of days for the optimal sliding step, which is 7 days in this case. Of course, the value of F1 greater than or equal to can be changed according to actual needs, in which case the optimal sliding step will also be changed.

[0107] See Figure 3 The illustration shows a schematic diagram of how the wind turbine is labeled with a future failure indicator based on nacelle sliding fault information, according to an embodiment of this application. Specifically, the process of labeling the wind turbine with a future failure indicator based on nacelle sliding fault information can be as follows:

[0108] 1) At any given moment, query the wind turbine generator at equal intervals using the optimal sliding step size until the most recent fault occurs;

[0109] 2) Obtain the number of queries N and the number of cabin sliding failures n corresponding to step 1), and then calculate the future failure rate P according to the formula P = n / N;

[0110] 3) Compare the future failure rate P with the preset rate. The preset rate can be set according to the actual situation, for example, it can be set to 0.3.

[0111] 4) If the future failure rate P is greater than the preset rate, then the wind turbine will be labeled with a failure tag (e.g., labeled as 1) at the time mentioned above. If the future failure rate P is not greater than the preset rate, then the wind turbine will be labeled with a normal tag (e.g., labeled as 0) at the time mentioned above.

[0112] By implementing the above method, wind turbines can be labeled with potential future failures based on nacelle slippage fault information, thus enabling early detection of faults in the early stages of nacelle slippage.

[0113] This application provides a wind turbine nacelle slip prediction method, which uses the historical nacelle slip frequency ratio, the historical proximity switch change frequency ratio, and future fault labels for training to obtain a model predicting future nacelle slip, which may include:

[0114] By using the percentage of historical cabin sliding events, the percentage of historical proximity switch changes, and labels indicating future malfunctions, a Bayesian network, random forest, or neural network can be trained to obtain a model predicting future cabin sliding events.

[0115] In this application, when training a model to predict future cabin sliding by using the historical cabin sliding frequency ratio, the historical proximity switch change frequency ratio, and the label of whether the cabin will malfunction in the future, any one of Bayesian network, random forest, and neural network can be used to train the model to predict future cabin sliding.

[0116] Furthermore, the model for predicting the current fault can be trained by using the historical cabin sliding frequency ratio, the historical proximity switch change frequency ratio, and the corresponding fault labels. Alternatively, the model can be trained using any of the following methods: Bayesian network, random forest, and neural network.

[0117] The above methods can improve the accuracy of the model, thereby improving the accuracy of cabin sliding prediction.

[0118] See Figure 4This document illustrates a flowchart illustrating the construction process of the percentage of nacelle sliding occurrences and the percentage of proximity switch changes provided in an embodiment of this application. The wind turbine nacelle sliding prediction method provided in this application, which utilizes historical operating parameters to construct the percentage of historical nacelle sliding occurrences and the percentage of historical proximity switch changes, may include:

[0119] The number of different cabin positions, the number of different proximity switch states, the average wind speed within a set time period before and after the cabin position changes, and the difference between two adjacent cabin positions are counted within the interval between two adjacent yaw times.

[0120] If the difference between two adjacent nacelle positions in the current statistics is greater than the width of one tooth pitch of the wind turbine, and / or if the average wind speed within a set time period before and after the change of nacelle position in the current statistics is greater than the wind speed of the turbine's design load, then the current statistics will be discarded.

[0121] Based on the optimal sliding step size, the sliding statistics are performed on the number of times between two adjacent yaw times of the wind turbine, the number of times the number of different nacelle positions within two adjacent yaw times is greater than 1, and the number of times the number of different proximity switch states within two adjacent yaw times is greater than 1.

[0122] The percentage of historical proximity switch changes is obtained based on the number of times the number of different proximity switch states is greater than 1 within the interval between two adjacent yaw times and the number of intervals between two adjacent yaw times of the wind turbine.

[0123] The percentage of historical cabin sliding times is obtained by counting the number of times the number of different cabin positions is greater than 1 within two adjacent yaw intervals and the number of times the number of different proximity switch states is greater than 1 within two adjacent yaw intervals.

[0124] In this application, the specific implementation process of constructing the historical cabin sliding frequency ratio and the historical proximity switch change frequency ratio using historical operating parameters is as follows:

[0125] a) Count the number of different nacelle positions, the number of different proximity switch states (i.e., the number of times the state data corresponding to two proximity switches changes) within the interval between two adjacent yaw times, the average wind speed within a set time period before and after the nacelle position change, and the difference between two adjacent nacelle positions. The interval between two adjacent yaw times is the non-yaw time. Finding the non-yaw time by using two adjacent yaw times can reduce interference and improve the accuracy of non-yaw determination. The set time period can be set according to actual conditions, for example, it can be 10 seconds.

[0126] b) If the difference between two adjacent nacelle positions in the current statistical analysis is greater than the width of one tooth pitch of the wind turbine, and / or if the average wind speed within a set time period before and after the nacelle position change in the current statistical analysis is greater than the design load wind speed of the turbine, then the data is considered dirty. To avoid affecting the model accuracy, the current statistical analysis will be discarded in at least one of the above situations to improve model accuracy and thus improve the accuracy of nacelle slip prediction.

[0127] c) Based on the optimal sliding step size, the number of times (num) the wind turbine has two adjacent yaw intervals, the number of times (yaw_num) the number of times (jjkg ...

[0128] d) Based on the number of times the number of different proximity switch states is greater than 1 within the interval between two adjacent yaw times (jjkg_num) and the number of times the wind turbine unit has an interval between two adjacent yaw times (num), the percentage of historical proximity switch changes (jjkg_rate) is obtained using the formula jjkg_rate=jjkg_num / num.

[0129] e) Based on the number of times the number of different cabin positions is greater than 1 within two adjacent yaw time intervals (yaw_num) and the number of times the number of different proximity switch states is greater than 1 within two adjacent yaw time intervals (jjkg_num), the historical cabin sliding rate (yaw_rate) is obtained using the formula yaw_rate=yaw_num / jjkg_num.

[0130] It should be noted that the statistics for the percentage of cabin sliding times and the percentage of proximity switch changes are similar to those above, and will not be repeated here.

[0131] The above methods can improve the accuracy and reliability of calculating the percentage of nacelle sliding events and the percentage of proximity switch changes, thereby improving the accuracy of wind turbine nacelle sliding prediction.

[0132] This application provides a method for predicting nacelle slip of a wind turbine, which obtains historical operating parameters of the wind turbine in a non-yaw state, and may include:

[0133] Collect historical real-time operating parameters of the wind turbine; historical real-time operating parameters may include the yaw state machine.

[0134] The yaw time is selected by the yaw state machine, and the non-yaw state of the wind turbine is obtained based on the yaw time; the time between two adjacent yaw times is the non-yaw time, and the non-yaw time corresponds to the non-yaw state.

[0135] Obtain historical operating parameters of the wind turbine in non-yaw condition.

[0136] In this application, the specific process for obtaining historical operating parameters of the wind turbine in a non-yaw state can be as follows:

[0137] 101) Collect historical real-time operating parameters of the wind turbine; historical real-time operating parameters include yaw status machine, nacelle position, status variables of yaw proximity switches 1-2 and wind speed.

[0138] 102) The yaw time is selected based on the yaw state machine, and the non-yaw state of the wind turbine is obtained based on the yaw time; the time between two adjacent yaw times is the non-yaw time, which corresponds to the non-yaw state. Directly finding the non-yaw time may introduce interference, while determining the non-yaw time through the yaw time can improve the accuracy of determining the non-yaw time and non-yaw state.

[0139] 103) Obtain historical operating parameters of the wind turbine in non-yaw state.

[0140] The above methods can improve the accuracy of historical operating parameters obtained under non-yaw conditions, thereby improving the accuracy of wind turbine nacelle sliding prediction.

[0141] The wind turbine nacelle sliding prediction method provided in this application embodiment may further include, after collecting historical real-time operating parameters of the wind turbine, the following:

[0142] Based on linear interpolation data along the time axis;

[0143] After obtaining the historical operating parameters of the wind turbine in non-yaw condition, it may also include:

[0144] If the non-yaw time is less than the set time length threshold, then delete the non-yaw time and the corresponding historical real-time operating parameters.

[0145] In this application, after collecting historical real-time operating parameters of the wind turbine, these parameters can be processed. Specifically, the data can be separated according to the time axis linearity difference to ensure the continuity of the data over time. After obtaining the historical operating parameters of the wind turbine in a non-yaw state, if the non-yaw time is determined to be less than a set time length threshold (e.g., 30 seconds or other values ​​set based on practical experience), it indicates that the non-yaw time is too short. In this case, the non-yaw time shorter than the set time length threshold and the corresponding historical real-time operating parameters can be deleted to avoid adverse effects on model predictions, thereby improving the accuracy of wind turbine nacelle sliding prediction.

[0146] This application also provides a wind turbine nacelle sliding prediction device. See [link to relevant documentation] Figure 5It shows a structural schematic diagram of a wind turbine nacelle sliding prediction device provided in an embodiment of this application, which may include:

[0147] The first acquisition module 51 is used to acquire the operating parameters of the wind turbine under test in a non-yaw state from historical time to the current time.

[0148] Module 52 is used to construct the percentage of sliding times of the first cabin and the percentage of changes in the first proximity switch using operating parameters;

[0149] The prediction result module 53 is used to obtain the prediction result of whether the nacelle of the wind turbine under test has a sliding fault based on the proportion of the first nacelle sliding times, the proportion of the first proximity switch change times, and the pre-built prediction current fault model.

[0150] It may also include a first building block 54 for constructing a model to predict the current failure, the first building block may include:

[0151] The first construction unit is used to acquire historical operating parameters and nacelle slip fault information of the wind turbine in non-yaw state, and to label the historical operating parameters with faults based on the nacelle slip fault information. Specifically, all historical operating parameters within a preset time period before the historical operating parameter corresponding to the nacelle slip fault are labeled as faults. The historical operating parameters are used to construct the historical nacelle slip frequency ratio and the historical proximity switch change frequency ratio. The historical nacelle slip frequency ratio, the historical proximity switch change frequency ratio and the corresponding fault labels are used for training to obtain the current fault prediction model.

[0152] The wind turbine nacelle sliding prediction device provided in this application embodiment may further include:

[0153] The second acquisition module is used to acquire the historical operating parameters of the wind turbine under test in a non-yaw state before obtaining the prediction result of whether the nacelle of the wind turbine under test has a sliding fault based on the proportion of nacelle sliding times, the proportion of proximity switch changes, and the pre-built prediction current fault model. The historical operating parameters are used to construct the second proportion of nacelle sliding times and the second proportion of proximity switch changes.

[0154] The input module is used to input the percentage of the second cabin sliding times and the percentage of the second proximity switch changes into the pre-built prediction model of future cabin sliding, so as to obtain the future prediction result of whether the cabin will have a sliding failure in the future time after the time corresponding to the historical operating parameters;

[0155] The third acquisition module is used to obtain the prediction result corresponding to the current time from the future prediction results;

[0156] It may also include a second building block for constructing a predictive model of future cabin sliding, the second building block may include:

[0157] The second building unit is used to label the wind turbine with a label indicating whether it will fail in the future based on the nacelle sliding fault information; the model is trained using the historical nacelle sliding frequency ratio, the historical proximity switch change frequency ratio, and the label indicating whether it will fail in the future to obtain a model predicting future nacelle sliding.

[0158] The first building unit may include:

[0159] The annotation sub-unit is used to annotate the historical operating parameters corresponding to the engine room sliding fault as the fault.

[0160] The module 53 for obtaining prediction results may include:

[0161] The prediction result unit is used to obtain the prediction result of whether the nacelle of the wind turbine under test has a sliding fault based on the proportion of the first nacelle sliding times, the proportion of the first proximity switch change times, the prediction result corresponding to the current time, and the prediction current fault model.

[0162] This application provides a wind turbine nacelle sliding prediction device, wherein the second building unit may include:

[0163] The query sub-unit is used to query the wind turbine at any time according to the optimal sliding step size until the most recent fault.

[0164] The calculation subunit is used to calculate the future failure percentage based on the number of queries and the number of cabin sliding failures found.

[0165] The first labeling subunit is used to label a time with a fault label if the proportion of future faults is greater than the preset proportion.

[0166] The second labeling subunit is used to label a time as normal if the future failure rate is not greater than the preset rate.

[0167] This application provides a wind turbine nacelle sliding prediction device, wherein the second building unit may include:

[0168] A sub-unit is constructed to train any one of Bayesian network, random forest, and neural network using the historical cabin sliding frequency ratio, the historical proximity switch change frequency ratio, and the label of whether a malfunction will occur in the future, so as to obtain a model for predicting future cabin sliding.

[0169] This application provides a wind turbine nacelle sliding prediction device, the first building unit of which may include:

[0170] The statistics subunit is used to count the number of different cabin positions within the time interval between two adjacent yaws, the number of different proximity switch states, the average wind speed within a set time length before and after the cabin position changes, and the difference between two adjacent cabin positions.

[0171] Discard sub-units are used to discard the current statistics if the difference between two adjacent nacelle positions in the current statistics is greater than the width of one tooth pitch of the wind turbine, and / or if the average wind speed within a set time period before and after the change of nacelle position in the current statistics is greater than the design load wind speed of the turbine.

[0172] The sliding statistics subunit is used to perform sliding statistics on the number of times between two adjacent yaw times of the wind turbine, the number of times the number of different nacelle positions is greater than 1 within the interval between two adjacent yaw times, and the number of times the number of different proximity switch states is greater than 1 within the interval between two adjacent yaw times, based on the optimal sliding step size.

[0173] The first sub-unit is used to obtain the percentage of historical proximity switch changes based on the number of times the number of different proximity switch states is greater than 1 within the interval between two adjacent yaw times and the number of intervals between two adjacent yaw times of the wind turbine.

[0174] The second sub-unit is used to obtain the percentage of historical cabin sliding times based on the number of times the number of different cabin positions is greater than 1 within two adjacent yaw time intervals and the number of times the number of different proximity switch states is greater than 1 within two adjacent yaw time intervals.

[0175] This application provides a wind turbine nacelle sliding prediction device, the first building unit of which may include:

[0176] The collection subunit is used to collect historical real-time operating parameters of the wind turbine; the historical real-time operating parameters may include the yaw state machine.

[0177] The filtering subunit is used to filter out the yaw time based on the yaw state machine, and obtain the non-yaw state of the wind turbine based on the yaw time; wherein, the time between two adjacent yaw times is the non-yaw time, and the non-yaw time corresponds to the non-yaw state.

[0178] The acquisition sub-unit is used to acquire historical operating parameters of the wind turbine in non-yaw mode.

[0179] The wind turbine nacelle sliding prediction device provided in this application embodiment may further include the following in its first building unit:

[0180] The interpolation subunit is used to linearly interpolate data based on the time axis after collecting historical real-time operating parameters of the wind turbine.

[0181] The delete sub-unit is used to delete the non-yaw time and the corresponding historical real-time operating parameters after obtaining the historical operating parameters of the wind turbine in non-yaw state. If the non-yaw time is less than a set time length threshold, the delete sub-unit is used.

[0182] This application also provides a wind turbine nacelle sliding prediction device. See [link to relevant documentation] Figure 6 It shows a structural schematic diagram of a wind turbine nacelle sliding prediction device, which may include:

[0183] Memory 61 is used to store computer programs;

[0184] When processor 62 executes a computer program stored in memory 61, it can perform the following steps:

[0185] The system acquires the operating parameters of the wind turbine under test in a non-yaw state from historical time to the current time. It then constructs the percentage of first nacelle sliding occurrences and the percentage of first proximity switch changes using these operating parameters. Based on the percentage of first nacelle sliding occurrences, the percentage of first proximity switch changes, and a pre-built current fault prediction model, it obtains a prediction result indicating whether the nacelle of the wind turbine under test has a sliding fault. The construction process of the current fault prediction model is as follows: It acquires the historical operating parameters and nacelle sliding fault information of the wind turbine in a non-yaw state, and labels the historical operating parameters with faults based on the nacelle sliding fault information. Specifically, historical operating parameters within a preset time period preceding the historical operating parameters corresponding to the nacelle sliding fault are all labeled as faults. The system then constructs the historical nacelle sliding occurrence percentage and the historical proximity switch change percentage using the historical operating parameters, and trains the model using the historical nacelle sliding occurrence percentage, the historical proximity switch change percentage, and the corresponding fault labels to obtain the current fault prediction model.

[0186] This application embodiment also provides a readable storage medium storing a computer program, which, when executed by a processor, can perform the following steps:

[0187] The system acquires the operating parameters of the wind turbine under test in a non-yaw state from historical time to the current time. It then constructs the percentage of first nacelle sliding occurrences and the percentage of first proximity switch changes using these operating parameters. Based on the percentage of first nacelle sliding occurrences, the percentage of first proximity switch changes, and a pre-built current fault prediction model, it obtains a prediction result indicating whether the nacelle of the wind turbine under test has a sliding fault. The construction process of the current fault prediction model is as follows: It acquires the historical operating parameters and nacelle sliding fault information of the wind turbine in a non-yaw state, and labels the historical operating parameters with faults based on the nacelle sliding fault information. Specifically, historical operating parameters within a preset time period preceding the historical operating parameters corresponding to the nacelle sliding fault are all labeled as faults. The system then constructs the historical nacelle sliding occurrence percentage and the historical proximity switch change percentage using the historical operating parameters, and trains the model using the historical nacelle sliding occurrence percentage, the historical proximity switch change percentage, and the corresponding fault labels to obtain the current fault prediction model.

[0188] The readable storage medium may include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0189] For a description of the relevant parts of the wind turbine nacelle sliding prediction device, equipment and readable storage medium provided in the embodiments of this application, please refer to the detailed description of the relevant parts of the wind turbine nacelle sliding prediction method provided in the embodiments of this application, and will not be repeated here.

[0190] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that the elements inherent in a process, method, article, or apparatus that includes a list of elements are included. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. Additionally, portions of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of corresponding technical solutions in the prior art have not been described in detail to avoid excessive elaboration.

[0191] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting nacelle slippage in a wind turbine generator, characterized in that, include: Obtain the operating parameters of the wind turbine under test in non-yaw condition from historical time to the current time; The percentage of first cabin sliding times and the percentage of first proximity switch changes are constructed using the aforementioned operating parameters; Based on the proportion of sliding times of the first nacelle, the proportion of changes of the first proximity switch, and the pre-built prediction model of the current fault, the prediction result of whether the nacelle of the wind turbine under test has a sliding fault is obtained. The process of constructing the current fault prediction model is as follows: Historical operating parameters and nacelle sliding fault information of the wind turbine under non-yaw conditions are obtained. Faults are labeled on the historical operating parameters based on the nacelle sliding fault information. Specifically, historical operating parameters within a preset time period before the historical operating parameter corresponding to the nacelle sliding fault are all labeled as faults. The historical operating parameters are used to construct the historical nacelle sliding frequency ratio and the historical proximity switch change frequency ratio. The historical nacelle sliding frequency ratio, the historical proximity switch change frequency ratio, and the corresponding fault labels are used for training to obtain the current fault prediction model. Before obtaining the prediction result of whether the nacelle of the wind turbine under test has a sliding fault based on the proportion of nacelle sliding times, the proportion of proximity switch changes, and the pre-built prediction model of the current fault, the process also includes: Obtain the historical operating parameters of the wind turbine under test in a non-yaw state, and use the historical operating parameters to construct the proportion of the second nacelle sliding times and the proportion of the second proximity switch changes. The percentage of the second cabin sliding times and the percentage of the second proximity switch changes are input into a pre-built prediction model for future cabin sliding, so as to obtain a future prediction result of whether the cabin will have a sliding failure in the future time after the time corresponding to the historical operating parameters. Obtain the prediction result corresponding to the current time from the future prediction results; The process of constructing the predicted future cabin sliding model is as follows: The wind turbine is labeled with a tag indicating whether it will fail in the future based on the nacelle sliding fault information; the model is trained using the percentage of historical nacelle sliding times, the percentage of historical proximity switch changes, and the tag indicating whether it will fail in the future, to obtain the model predicting future nacelle sliding. All historical operating parameters within a preset time period preceding the historical operating parameters corresponding to the cabin sliding fault are marked as faults, including: The historical operating parameters corresponding to the aforementioned cabin sliding fault are marked as faults; Based on the percentage of sliding events of the first nacelle, the percentage of changes in the first proximity switch, and a pre-built prediction model for the current fault, a prediction result is obtained regarding whether the nacelle of the wind turbine under test has a sliding fault, including: Based on the proportion of the first nacelle sliding times, the proportion of the first proximity switch changes, the prediction result corresponding to the current time, and the prediction current fault model, the prediction result of whether the nacelle of the wind turbine under test has a sliding fault is obtained. Using the historical operating parameters, we can construct the percentage of historical cabin sliding events and the percentage of historical proximity switch changes, including: The number of different cabin positions, the number of different proximity switch states, the average wind speed within a set time period before and after the cabin position changes, and the difference between two adjacent cabin positions are counted within the interval between two adjacent yaw times. If the difference between two adjacent nacelle positions in the current statistics is greater than the width of one tooth pitch of the wind turbine, and / or if the average wind speed within a set time period before and after the change of nacelle position in the current statistics is greater than the design load wind speed of the turbine, then the current statistics are discarded. Based on the optimal sliding step size, the sliding statistics are performed on the number of times the wind turbine's yaw interval is between two adjacent yaw times, the number of times the number of different nacelle positions within the interval between two adjacent yaw times is greater than 1, and the number of times the number of different proximity switch states within the interval between two adjacent yaw times is greater than 1. The percentage of historical proximity switch changes is obtained based on the number of times the number of different proximity switch states is greater than 1 within the interval between two adjacent yaw times and the number of intervals between two adjacent yaw times of the wind turbine. The percentage of historical cabin sliding times is obtained based on the number of times the number of different cabin positions is greater than 1 within two adjacent yaw intervals and the number of times the number of different proximity switch states is greater than 1 within two adjacent yaw intervals.

2. The wind turbine nacelle sliding prediction method according to claim 1, characterized in that, Based on the nacelle sliding fault information, the wind turbine is labeled with a tag indicating whether it will fail in the future, including: For any given moment of the wind turbine, query at equal intervals according to the optimal sliding step size until the most recent fault occurs; Calculate the future failure percentage based on the number of queries and the number of cabin sliding failures found. If the proportion of future failures is greater than the preset proportion, then a failure label is marked at the specified time. If the percentage of future failures is not greater than the preset percentage, then the time point is labeled as normal.

3. The wind turbine nacelle sliding prediction method according to claim 1, characterized in that, The model for predicting future cabin sliding is trained using the historical cabin sliding frequency percentage, the historical proximity switch change frequency percentage, and the label indicating whether a malfunction is imminent. This model includes: The model predicting future cabin sliding is obtained by training any one of Bayesian network, random forest, and neural network using the historical cabin sliding frequency ratio, the historical proximity switch change frequency ratio, and the label of whether a future malfunction will occur.

4. The wind turbine nacelle sliding prediction method according to claim 1, characterized in that, Obtain historical operating parameters of the wind turbine in non-yaw condition, including: Collect historical real-time operating parameters of the wind turbine; the historical real-time operating parameters include the yaw state machine. The yaw time is selected by the yaw state machine, and the non-yaw state of the wind turbine is obtained based on the yaw time; wherein, the time between two adjacent yaw times is the non-yaw time, and the non-yaw time corresponds to the non-yaw state. Obtain the historical operating parameters of the wind turbine in non-yaw state.

5. The wind turbine nacelle sliding prediction method according to claim 4, characterized in that, After collecting the historical real-time operating parameters of the wind turbine, the following is also included: Based on linear interpolation data along the time axis; After obtaining the historical operating parameters of the wind turbine in a non-yaw state, the method further includes: If the non-yaw time is less than a set time length threshold, then the non-yaw time and the corresponding historical real-time operating parameters are deleted.

6. A wind turbine nacelle sliding prediction device, characterized in that, include: The first acquisition module is used to acquire the operating parameters of the wind turbine under test in a non-yaw state from historical time to the current time. The module is used to construct the percentage of first cabin sliding times and the percentage of first proximity switch changes using the operating parameters; The prediction result module is used to obtain a prediction result of whether the nacelle of the wind turbine under test has a sliding fault based on the proportion of the first nacelle sliding times, the proportion of the first proximity switch change times, and the pre-built prediction current fault model. It also includes a first building module for constructing the predictive current fault model, the first building module comprising: The first construction unit is used to acquire historical operating parameters and nacelle sliding fault information of the wind turbine in a non-yaw state, and to label the historical operating parameters with faults based on the nacelle sliding fault information; wherein, all historical operating parameters within a preset time period before the historical operating parameter corresponding to the nacelle sliding fault are labeled as faults; the historical operating parameters are used to construct the historical nacelle sliding frequency ratio and the historical proximity switch change frequency ratio, and the historical nacelle sliding frequency ratio, the historical proximity switch change frequency ratio and the corresponding fault labels are used for training to obtain the current fault prediction model; The wind turbine nacelle sliding prediction device also includes: The second acquisition module is used to acquire the historical operating parameters of the wind turbine under test in a non-yaw state before obtaining the prediction result of whether the nacelle of the wind turbine under test has a sliding fault based on the proportion of nacelle sliding times, the proportion of proximity switch changes, and the pre-built prediction current fault model. The historical operating parameters are used to construct the second proportion of nacelle sliding times and the second proportion of proximity switch changes. The input module is used to input the percentage of the second cabin sliding times and the percentage of the second proximity switch changes into the pre-built prediction model of future cabin sliding, so as to obtain the future prediction result of whether the cabin will have a sliding failure in the future time after the time corresponding to the historical operating parameters; The third acquisition module is used to obtain the prediction result corresponding to the current time from the future prediction results; It may also include a second building block for constructing a predictive model of future cabin sliding, the second building block may include: The second building unit is used to label the wind turbine with a label indicating whether it will fail in the future based on the nacelle sliding fault information; the model is trained using the historical nacelle sliding frequency ratio, the historical proximity switch change frequency ratio, and the label indicating whether it will fail in the future to obtain a model predicting future nacelle sliding. The first building unit includes: The annotation sub-unit is used to annotate the historical operating parameters corresponding to the engine room sliding fault as the fault. The module for obtaining prediction results includes: The prediction result unit is used to obtain the prediction result of whether the nacelle of the wind turbine under test has a sliding fault based on the proportion of the first nacelle sliding times, the proportion of the first proximity switch change times, the prediction result corresponding to the current time, and the prediction current fault model. The first building block includes: The statistics subunit is used to count the number of different cabin positions within the time interval between two adjacent yaws, the number of different proximity switch states, the average wind speed within a set time length before and after the cabin position changes, and the difference between two adjacent cabin positions. The discard sub-unit is used to discard the current statistics if the difference between the positions of two adjacent nacelles in the current statistics is greater than the width of one tooth pitch of the wind turbine, and / or if the average wind speed within a set time period before and after the change of nacelle position in the current statistics is greater than the design load wind speed of the turbine. The sliding statistics subunit is used to perform sliding statistics on the number of times the wind turbine's yaw interval is between two adjacent yaw times, the number of times the number of different nacelle positions is greater than 1 within the interval between two adjacent yaw times, and the number of times the number of different proximity switch states is greater than 1 within the interval between two adjacent yaw times, based on the optimal sliding step size. The first sub-unit is used to obtain the percentage of historical proximity switch changes based on the number of times the number of different proximity switch states is greater than 1 within two adjacent yaw intervals and the number of times the wind turbine unit has two adjacent yaw intervals. The second sub-unit is used to obtain the percentage of historical cabin sliding times based on the number of times the number of different cabin positions is greater than 1 within two adjacent yaw time intervals and the number of times the number of different proximity switch states is greater than 1 within two adjacent yaw time intervals.

7. A wind turbine nacelle sliding prediction device, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the wind turbine nacelle sliding prediction method as described in any one of claims 1 to 5.

8. A readable storage medium, characterized in that, The readable storage medium stores a computer program that, when executed by a processor, implements the steps of the wind turbine nacelle sliding prediction method as described in any one of claims 1 to 5.

Images