A wind turbine generator fault diagnosis method, system, device and storage medium

Abstract

The application relates to a wind turbine fault diagnosis method, system, device and storage medium, which comprises the following steps: acquiring sample data to be measured, the sample data to be measured comprising multiple groups of data, each group of data comprising vibration data, vibration data spectrum, vibration data demodulation spectrum and temperature data; determining a feature vector to be measured according to the vibration data, the vibration data spectrum and the vibration data demodulation spectrum based on a feature determination model, each group of data corresponding to a feature vector to be measured, and the feature vector to be measured being an array comprising multiple feature values; inputting the feature vector to be measured into a preset vibration diagnosis model to obtain a vibration risk parameter; inputting the temperature data into a preset temperature diagnosis model to obtain a temperature risk parameter; and outputting a risk prompt information based on a preset risk judgment rule, the vibration risk parameter and the temperature risk parameter. The application has the effect of improving the accuracy of judging the main bearing fault of a wind turbine.

Description

Technical Field

[0001] This application relates to the technical field of wind turbine fault diagnosis, and in particular to a wind turbine fault diagnosis method, system, device and storage medium. Background Technology

[0002] With the depletion of traditional mineral resources, wind energy, a renewable resource, is receiving increasing attention, leading to a continuous increase in the installed capacity of wind turbines to meet electricity demand. However, due to complex and variable environmental factors, the performance of various components in wind turbines gradually declines during operation, eventually causing component failures. The main bearing is a core component of a wind turbine, and its health directly affects the normal operation of the turbine. Therefore, routine inspection and maintenance of the main bearing are necessary, resulting in high inspection and maintenance costs. Consequently, methods for assessing the health status of the main bearing by monitoring its operating data have emerged.

[0003] Currently, fault diagnosis methods for main bearings mainly fall into two categories. One category involves acquiring temperature or vibration data from the main bearing and then determining whether the data exceeds a threshold for fault diagnosis. This method is simple to implement, but setting the threshold requires extensive human experience and is prone to false alarms and missed alarms. The other category utilizes existing fault data to train models for fault diagnosis and classification. This method relies on a large amount of fault data for model training, but obtaining fault samples in actual operation is difficult. Both methods, depending on human experience or a large amount of fault sample data, result in low accuracy in diagnosing main bearing faults. Summary of the Invention

[0004] To improve the accuracy of diagnosing main bearing faults in wind turbines, this application provides a method, system, equipment, and storage medium for diagnosing wind turbine faults.

[0005] In a first aspect of this application, a method for diagnosing faults in wind turbine generators is provided. The method includes:

[0006] Acquire the sample data to be tested, which includes multiple sets of data, each set of data including vibration data, vibration data spectrum, vibration data demodulation spectrum and temperature data;

[0007] Based on the feature determination model, the feature vector to be measured is determined according to the vibration data, vibration data spectrum and vibration data demodulation spectrum. Each set of data corresponds to a feature vector to be measured, and the feature vector to be measured is an array containing multiple feature values.

[0008] The feature vector to be measured is input into a preset vibration diagnosis model to obtain vibration risk parameters, which are used to reflect the risk status of main bearing vibration.

[0009] The temperature data is input into a preset temperature diagnostic model to obtain temperature risk parameters, which are used to reflect the risk status of the main bearing temperature risk.

[0010] Based on the preset risk assessment rules, the vibration risk parameters, and the temperature risk parameters, risk warning information is output.

[0011] As can be seen from the above technical solution, by analyzing the test sample data from both vibration and temperature data, corresponding vibration risk parameters and temperature risk parameters are obtained. Then, based on the vibration risk parameters, temperature risk parameters, and preset risk judgment rules, risk warning information under corresponding conditions is output. For vibration-related data, extracting the target feature vector from the vibration data, vibration data spectrum, and vibration data demodulation spectrum enables more accurate analysis of the vibration-related data of the main bearing. To a certain extent, this improves the low accuracy of diagnosing wind turbine main bearing faults and effectively enhances the accuracy of diagnosing wind turbine main bearing faults.

[0012] In one possible implementation, inputting the feature vector to be measured into a preset vibration diagnosis model to obtain vibration risk parameters includes:

[0013] Kernel principal component analysis is performed on multiple feature vectors to be tested to obtain multiple diagnostic values. Each feature vector to be tested corresponds to one of the diagnostic values. It is then determined whether all of the multiple diagnostic values ​​are within a preset threshold range.

[0014] If so, the maximum value of the multiple diagnostic values ​​is obtained, and the maximum value of the diagnostic values ​​is input into the diagnostic score function to obtain the vibration risk parameter.

[0015] One possible implementation also includes:

[0016] If not, acquire the diagnostic values ​​that are not within the threshold range;

[0017] Calculate the reconstruction error of the data, where the reconstruction error corresponds to the feature value in the feature vector to be tested, and multiple feature values ​​correspond to multiple reconstruction errors. Obtain the maximum value among the multiple reconstruction errors.

[0018] The vibration risk parameter is calculated based on the eigenvalue corresponding to the maximum value in the reconstruction error.

[0019] In one possible implementation, inputting the temperature data into a preset temperature diagnostic model to obtain temperature risk parameters includes:

[0020] The temperature risk parameters include absolute temperature difference constants, abnormal temperature rise values, and temperature difference constants.

[0021] The temperature data includes the full temperature data of the main bearing and the temperature data of the engine compartment;

[0022] The full temperature data is divided into bins according to the preset binning field threshold.

[0023] Based on the preset absolute temperature anomaly model, the full temperature data after binning is linearly mapped to determine the absolute temperature anomaly value.

[0024] The full temperature data is a temperature data sequence obtained by sorting according to a preset sorting rule. The temperature data sequence contains n data points, where n is a natural number greater than 1.

[0025] Calculate the difference between any two adjacent data points in the temperature data sequence to obtain a difference sequence, which contains (n-1) differences;

[0026] Based on the difference sequence, a difference sum sequence is determined, the difference sum sequence containing (n-1) difference sum data, the q-th difference sum data being the sum of the first q differences in the difference sequence, where q is a natural number and q≤n-1;

[0027] Obtain the maximum value in the difference sum sequence, and perform a linear mapping on the maximum value according to the different values ​​in the difference sum sequence to determine the abnormal temperature rise value;

[0028] Calculate the difference between the full temperature data and the cabin temperature data, and determine the temperature difference abnormality value based on the difference between the full temperature data and the cabin temperature data and the preset temperature difference abnormality model.

[0029] In one possible implementation, a method for obtaining the threshold range is also included:

[0030] Acquire historical vibration data, historical vibration spectrum, and historical vibration demodulation spectrum;

[0031] The vibration intensity value is calculated based on the historical vibration data and the preset intensity calculation model.

[0032] The shaft value is calculated based on the historical vibration spectrum, the historical vibration demodulation spectrum, and the preset shaft value calculation model.

[0033] According to the demodulation spectrum processing rules, some frequencies in the historical vibration demodulation spectrum are removed, and the signal-to-noise ratio of the historical vibration demodulation spectrum after removal is calculated.

[0034] The vibration intensity value, the shaft rotation value, and the signal-to-noise ratio are standardized to determine a standard vector;

[0035] Kernel principal component analysis was performed on the standard vector to obtain the analysis vector;

[0036] The threshold range is formed by combining the 99th percentile and 5th percentile of the analysis vector.

[0037] In one possible implementation, the step of outputting risk warning information based on preset risk assessment rules, vibration risk parameters, and temperature risk parameters includes:

[0038] When the abnormal temperature rise value is greater than or equal to the target preset value, or when the vibration risk parameter is greater than or equal to the target preset value and the absolute temperature difference abnormal value is greater than or equal to the target preset value and the vibration risk parameter is of the first type, a vibration lubrication prompt message is output.

[0039] When the abnormal temperature rise value is greater than or equal to the target preset value, or when the vibration risk parameter is greater than or equal to the target preset value and the absolute temperature difference abnormal value is greater than or equal to the target preset value and the vibration risk parameter is of the second type, the spindle vibration abnormality prompt information is output.

[0040] When the vibration risk parameter is less than the target preset value and the temperature difference abnormal value is greater than or equal to the target preset value, a heat dissipation abnormality prompt message is output.

[0041] In one possible implementation, the maximum value of the temperature risk parameter is obtained, and the maximum value is the final temperature anomaly detection value;

[0042] When the final temperature anomaly detection value is less than the target preset value and the temperature difference anomaly value is greater than or equal to the target preset value, a temperature attention prompt message is output.

[0043] When the final abnormal temperature detection value is greater than or equal to the target preset value and the abnormal temperature difference value is less than or equal to the target preset value, a vibration attention prompt message is output.

[0044] In a second aspect of this application, a wind turbine fault diagnosis system is provided. The system includes:

[0045] The data acquisition module is used to acquire the data of the sample to be tested. The data of the sample to be tested includes multiple sets of data, each set of data including vibration data, vibration data spectrum, vibration data demodulation spectrum and temperature data.

[0046] The vector extraction module is used to determine the model based on features, and to determine the feature vector to be measured according to the vibration data, vibration data spectrum and vibration data demodulation spectrum. Each set of data corresponds to a feature vector to be measured, and the feature vector to be measured is an array containing multiple feature values.

[0047] The vibration parameter determination module is used to input the feature vector to be measured into a preset vibration diagnosis model to obtain vibration risk parameters, which are used to reflect the risk status of the main bearing vibration risk.

[0048] The temperature parameter determination module is used to input the temperature data into a preset temperature diagnostic model to obtain temperature risk parameters, which are used to reflect the risk status of the main bearing temperature risk.

[0049] The information output module is used to output risk warning information based on preset risk assessment rules, the vibration risk parameters, and the temperature risk parameters.

[0050] In a third aspect of this application, an electronic device is provided. The electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the method described above.

[0051] In a fourth aspect of this application, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the method according to the first aspect of this application.

[0052] In summary, this application includes at least one of the following beneficial technical effects:

[0053] By analyzing the test sample data from both vibration and temperature perspectives, and extracting features from the vibration data to obtain the test feature vector, the corresponding vibration risk parameters and temperature risk parameters are obtained. Then, based on the vibration risk parameters, temperature risk parameters, and preset risk judgment rules, risk warning information is output under the corresponding conditions, which can improve the accuracy of judging the main bearing failure of wind turbine units. Attached Figure Description

[0054] Figure 1 This is a flowchart illustrating the wind turbine fault diagnosis method provided in this application.

[0055] Figure 2 This is a schematic diagram of the wind turbine fault diagnosis system provided in this application.

[0056] Figure 3 This is a schematic diagram of the structure of the electronic device provided in this application.

[0057] In the diagram, 200 is the wind turbine fault diagnosis system; 201 is the data acquisition module; 202 is the vector extraction module; 203 is the vibration parameter determination module; 204 is the temperature parameter determination module; 205 is the information output module; 301 is the CPU; 302 is the ROM; 303 is the RAM; 304 is the I / O interface; 305 is the input section; 306 is the output section; 307 is the storage section; 308 is the communication section; 309 is the driver; and 310 is the removable medium. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, 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, 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.

[0059] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0060] Current fault diagnosis methods for main bearings often rely on human experience or large amounts of fault sample data. This can lead to reduced accuracy in diagnosing main bearing faults due to inaccurate fault samples, insufficient data volume, or human error in setting the parameters. This application addresses this issue by extracting features from the vibration data of the main bearing under specified operating conditions, and then analyzing and diagnosing the fault based on relevant temperature data and the specific fault condition of the main bearing.

[0061] The embodiments of this application will now be described in further detail with reference to the accompanying drawings.

[0062] This application provides a method for diagnosing faults in wind turbine generators. The main process of the method is described below.

[0063] like Figure 1 As shown:

[0064] Step S101: Obtain the data of the sample to be tested.

[0065] Specifically, the sample data to be tested includes multiple sets of data, each set of data including vibration data, vibration data spectrum, vibration data demodulation spectrum, and temperature data.

[0066] Vibration data of the wind turbine is acquired daily, with vibration data segments sampled at fixed time intervals at a sampling frequency of no less than 250,000 Hz. The vibration data includes axial and horizontal acceleration sampling data for the front and rear bearings. FFT spectral analysis is performed on the acquired vibration data to obtain the vibration data spectrum, and envelope spectral analysis is performed to obtain the vibration data demodulation spectrum. The temperature data is obtained through a supervisory control and data acquisition (SCADA) system. The data obtained from the SCADA system is further filtered to obtain the aforementioned temperature data. The filtering process includes: first, filtering data from periods when the wind turbine power is greater than 80% of its full power output; then, further verification of the data, issuing data anomaly alarms and providing different status codes to identify the type of data anomaly for cases such as empty data, a column value of NaN, a data column not being a preset field, a data type not being float, duplicate data, and the number of data points being less than a preset number. NaN (Not a Number) is a type of value in computer science's numerical data type, representing an undefined or unrepresentable value. The missing data is filled in. The methods for filling in the missing data described above can include other methods such as the K-Nearest Neighbor (KNN) algorithm and the Ordered Nearest Neighbor Missing Value Filling (SKNN) algorithm. The KNN algorithm, SKNN algorithm, and the data judgment process described above are techniques well-known to those skilled in the art and will not be elaborated upon here. The temperature data mentioned above includes the full temperature data of the main bearing of the wind turbine and the nacelle temperature data of the wind turbine.

[0067] Step S102: Based on the feature determination model, determine the feature vector to be measured according to the vibration data, the vibration data spectrum and the vibration data demodulation spectrum.

[0068] Specifically, each set of data corresponds to a feature vector to be measured, and each feature vector is an array containing multiple feature values. These feature values ​​specifically include the intensity feature to be measured, the shaft rotation feature to be measured, and the signal-to-noise ratio feature to be measured. The intensity feature to be measured is obtained by calculating the vibration intensity of the vibration data. The vibration intensity represents the degree of vibration, and is usually expressed by the maximum value, average value, or root mean square value of parameters characterizing the vibration level (such as displacement, velocity, and acceleration). In this embodiment, the vibration intensity is represented by the root mean square value of the vibration data; that is, the intensity feature to be measured is the root mean square value of the vibration data.

[0069] The maximum value within the frequency range four times the rotational frequency of the aforementioned vibration data spectrum is searched and summed, and denoted as the first fault factor. Similarly, the maximum value within the frequency range four times the rotational frequency of the aforementioned vibration data demodulation spectrum is searched and summed, and denoted as the second fault factor. The summation of the first and second fault factors yields the characteristics of the shaft under test. The aforementioned "frequency range four times the rotational frequency" refers to fault frequencies within four times the rotational frequency.

[0070] The frequency-dependent components in the demodulated spectrum of the vibration data are removed. The frequency refers to the rotational frequency of the motor itself. This removal involves eliminating components near this rotational frequency (i.e., 0.5 Hz to the left and right of the rotational frequency). The signal-to-noise ratio (SNR) of the demodulated spectrum of the vibration data after removal is calculated; this SNR is the SNR feature to be measured. The intensity feature, the shaft feature, and the SNR feature to be measured together form a complete feature vector to be measured.

[0071] Step S103: Input the feature vector to be tested into the preset vibration diagnosis model to obtain vibration risk parameters.

[0072] Specifically, the dimensionality reduction of the above-mentioned feature vector to be tested is performed by the kernel principal component analysis method. Since the feature vector to be tested includes multiple feature values, the diagnostic value obtained after the data dimensionality reduction corresponds to the above-mentioned feature vector to be tested, that is, the feature vector to be tested and the diagnostic value correspond one-to-one. It is then determined whether the multiple diagnostic values ​​are all within the preset threshold range.

[0073] If multiple diagnostic values ​​are all within a preset threshold range, the maximum value among the multiple diagnostic values ​​corresponding to each feature value is obtained. This maximum value is then input into the diagnostic score function to obtain the vibration risk parameter. The kernel principal component analysis method described above is well-known to those skilled in the art and will not be elaborated upon here.

[0074] In this embodiment, the maximum value among the above diagnostic values ​​is denoted as `max_value`, and the threshold range is represented as [min_threshold, threshold], meaning that when m ≥ min_threshold and m ≤ threshold, m is within the threshold range. `max_value` is linearly mapped to [0~0.2] based on the threshold range. The specific mapping relationship is as follows: x is the independent variable, y is the dependent variable; when x = min_threshold, y = 0; when x = threshold, y = 0.2. For example, if min_threshold = 1 and threshold = 2, then y = 0.2x - 0.2. Substituting `max_value = 1.5` into the above linear function, we get y = 0.2 * 1.5 - 0.2 = 0.1, so the vibration risk parameter is 0.1.

[0075] If one or more diagnostic values ​​are outside the preset threshold range, the data outside the threshold range of the diagnostic values ​​are obtained, and the reconstruction error of the data outside the threshold range is calculated. The reconstruction error corresponds to the feature value in the feature vector to be measured. Multiple feature values ​​correspond to multiple reconstruction errors. The maximum value among the multiple reconstruction errors is obtained. Based on the feature value corresponding to the maximum value among the reconstruction errors, the vibration risk parameter corresponding to different feature values ​​is calculated.

[0076] When the ratio of the amount of data outside the threshold range to the total amount of data in all the feature vectors to be tested is greater than the fault threshold, historical data from the previous preset number of days is traced back, and it is determined whether there is any abnormal data in the historical data. The ratio of the amount of data outside the threshold range to the total amount of data in all the feature vectors to be tested is recorded as the fault ratio. For example, if the fault ratio for the day is 50% and the fault threshold is 60%, that is, the fault ratio is less than the fault threshold, it indicates that there is no abnormality; if the fault ratio for the day is 70% and the fault threshold is 60%, that is, the fault ratio exceeds the fault threshold, then historical data from the previous 3 days (preset number 3) is traced back, and it is determined whether the fault ratio for each day exceeds the fault threshold. If there are cases where the fault ratio does not exceed the fault threshold, it indicates that there is no abnormality; if the fault ratio for all 3 days exceeds the fault threshold, then:

[0077] When the feature corresponding to the maximum value of the above reconstruction error is the intensity feature to be measured, the intensity ratio is calculated. The intensity ratio is the ratio of the maximum value of the intensity feature to be measured to the preset initial value of intensity, that is, the intensity ratio = the maximum value of the intensity feature to be measured / the initial value of intensity.

[0078] When the feature corresponding to the maximum value of the above-mentioned reconstruction error is the feature of the shaft under test, the shaft ratio is calculated. This shaft ratio is the ratio of the maximum value of the feature under test to the preset initial value of the shaft, i.e., the shaft ratio = maximum value of the feature under test / initial value of the shaft. The fault type corresponding to the shaft ratio is determined by the relative magnitude of the rotational frequency and harmonics corresponding to the data in the above-mentioned abnormal dataset. The correspondence between the rotational frequency, harmonics, and fault type is preset. The fault types include shaft imbalance, shaft misalignment, and shaft looseness. The fault type with the highest probability is taken as the final fault type.

[0079] When the feature corresponding to the maximum value of the above reconstruction error is the signal-to-noise ratio feature to be tested, the signal-to-noise ratio value is calculated. The above signal-to-noise ratio value is the ratio of the maximum value of the above signal-to-noise ratio feature to the preset initial signal-to-noise value, that is, the above signal-to-noise ratio value = the maximum value of the above signal-to-noise ratio feature to be tested / the above initial signal-to-noise value.

[0080] When bearing component parameters are provided, the fault type corresponding to the aforementioned signal-to-noise ratio (SNR) value is determined by identifying the fault frequencies in the corresponding spectrum of the abnormal dataset, and the fault category with the highest probability is taken as the final fault category. When bearing component parameters are not provided, the fault type corresponding to the aforementioned SNR value is determined by the relative position relationship between the fundamental frequency and the rotational frequency of the harmonic frequencies in the corresponding spectrum of the abnormal dataset, and the fault category with the highest probability is taken as the final fault category. The aforementioned final fault categories include inner ring faults, outer ring faults, rolling element faults, and cage faults. The aforementioned "providing bearing component parameters" refers to the relevant data of the main bearing, whether the manufacturer provides it at the time of shipment. If the manufacturer provides it, it corresponds to "providing bearing component parameters" as described above; if the manufacturer does not provide it, it corresponds to "not providing bearing component parameters".

[0081] The intensity ratio, shaft rotation ratio, and signal-to-noise ratio calculated above are scored. In this embodiment, when the ratio is in the range of [0,1], the corresponding score is [0,0.2]; when the ratio is in the range of [1,1.25], the corresponding score is [0.2,0.4]; when the ratio is in the range of [1.25,1.5], the corresponding score is [0.4,0.6]; and when the ratio is in the range of [1.5,1.75], the corresponding score is [0.6,1]. The specific range group to which the boundary values ​​of the above ranges and corresponding scores belong is set according to actual needs. There is a linear mapping relationship between the above ratio ranges and corresponding scores. The specific calculation method is illustrated in the calculation of the "vibration risk parameter" when "all diagnostic values ​​are within the preset threshold range", and will not be repeated here. The score obtained after scoring the above ratios is the vibration risk parameter.

[0082] Step S104: Input the temperature data into the preset temperature diagnostic model to obtain the temperature risk parameters.

[0083] Specifically, the aforementioned temperature risk parameters include the aforementioned absolute temperature difference constant, the aforementioned temperature rise anomaly, and the aforementioned temperature difference constant. The total temperature data is binned according to preset binning thresholds. For example, if the total temperature data ranges from 10℃ to 50℃, 10℃ to 20℃ is grouped into one bin. Similarly, data from 20℃ to 30℃, 30℃ to 40℃, and 40℃ to 50℃ are grouped. The determination of boundary values ​​for each group is set according to actual needs. The aforementioned 10℃ to 20℃ represents one binning threshold. These binning thresholds are set by staff based on actual needs.

[0084] Based on a preset absolute temperature anomaly model, the full temperature data after binning is processed to determine the absolute temperature anomaly value. For the full temperature data after binning, the number of data points in the last bin and the second-to-last bin are obtained. In this embodiment, the last bin refers to the data group with the largest binning threshold, for example, the bin with temperatures between 40°C and 50°C is the last bin. The second-to-last bin refers to the bin with the next largest binning threshold, i.e., the bin with temperatures between 30°C and 40°C. In other embodiments, the definitions of the last bin and the second-to-last bin can be modified according to actual needs. Based on the number of data points in the last bin and a preset quantity value, the number of data points in the last bin is scored, linearly mapping the number of data points to the range of [0.6, 1], and the number of data points in the second-to-last bin is scored, linearly mapping the number of data points to the range of [0.4, 0.6]. For the last bin, when the number of data points is 1, the corresponding score is 0.6; when the number of data points is the preset quantity value, the corresponding score is 1. Based on the correspondence between the number of data points and their corresponding scores, a linear function relationship can be determined. When the number of data points is between 1 and a preset value, the corresponding score is calculated according to the above linear function relationship. When the number of data points is 0, the corresponding score is 0; when the number of data points is greater than the preset value, the corresponding score is 1. When the number of data points in the second-to-last box is greater than the preset value, the corresponding score is 0.6. When the number of data points in the second-to-last box is between 10 and the preset value, the method for determining the score for the second-to-last box is the same as that for the last box, and will not be elaborated here. The maximum value between the two corresponding scores is obtained; this maximum value is the absolute temperature difference constant. The above preset value is set by the staff according to actual needs.

[0085] The aforementioned full temperature data is a temperature data sequence obtained by sorting according to a preset sorting rule. This temperature data sequence contains n data points, where n is a natural number greater than 1. The difference between any two adjacent data points in the temperature data sequence is calculated to obtain a difference sequence containing (n-1) differences. Based on this difference sequence, a difference sum sequence is determined, containing (n-1) difference sum data points. The q-th difference sum data point is the sum of the first q differences in the difference sequence, where q is a natural number and q ≤ n-1. The maximum value in the difference sum sequence is obtained, and based on the different maximum values ​​in the difference sum sequence, a linear mapping is performed on the maximum value to determine the abnormal temperature rise value. The preset sorting rule is based on the data acquisition time.

[0086] It is understood that the full temperature data is sorted according to the acquisition time of the temperature, and then the difference between any two adjacent data in the above temperature data sequence is calculated to obtain a difference sequence. In this embodiment, the above difference is backward difference; in other embodiments, it can be forward difference or center difference. The cumulative sum of the above differences is calculated to form a difference sum sequence. For example, with ten full temperature data, nine difference values ​​can be obtained. The difference sum sequence is the first difference, the sum of the first two differences, the sum of the first three differences, the sum of the first four differences, ..., the sum of the first nine differences. The maximum value in the above difference sum sequence is obtained, and the position of the maximum value in the difference sum sequence is determined. When the position is within a preset position range in the difference sum sequence, the maximum value in the difference sum sequence is linearly mapped to the range [0, 0.6] according to the difference setting threshold. The method of linear mapping has been specifically described in the above process and will not be repeated here. For example, when the maximum value in the difference sum sequence is located in the last two-thirds of the sequence, i.e., between the fourth and ninth positions, the maximum value is linearly mapped to the range [0, 0.6]. That is, the minimum value of the difference setting threshold corresponds to 0, and the maximum value corresponds to 0.6. A relational function is determined, and the maximum value in the difference sum sequence is input into the relational function to obtain the corresponding score, i.e., the abnormal temperature rise value. The aforementioned difference setting threshold and preset position range are set by staff according to actual needs.

[0087] Calculate the difference between the above full temperature data and the cabin temperature data, and determine the temperature difference abnormality value based on the above difference and the preset temperature difference abnormality model.

[0088] It is understandable that there is a correspondence between the full-volume temperature data and the nacelle temperature data; that is, at the same time, the full-volume temperature data of the main bearing and the nacelle temperature data of the wind turbine can be acquired simultaneously. Based on this correspondence, the difference between the full-volume temperature data and the corresponding nacelle temperature data is calculated. If one or more differences are less than a preset difference value, the temperature difference constant is assigned a value of 0.4; if none of the differences are less than the preset difference value, the temperature difference constant is assigned a value of 0. These preset difference values ​​are set by the staff according to actual needs.

[0089] Step S105: Based on the preset risk assessment rules, vibration risk parameters, and temperature risk parameters, output risk warning information.

[0090] Specifically, when the above abnormal temperature rise value is greater than or equal to the target preset value, or when the above vibration risk parameter is greater than or equal to the target preset value and the above absolute temperature difference abnormal value is greater than or equal to the target preset value and the above vibration risk parameter is an intensity ratio, the above intensity ratio is the first type, and a vibration lubrication prompt message is output to remind the staff that the main bearing is not properly lubricated and needs to be repaired in time.

[0091] When the aforementioned abnormal temperature rise value is greater than or equal to the target preset value, or when the aforementioned vibration risk parameter is greater than or equal to the target preset value and the aforementioned absolute temperature difference abnormal value is greater than or equal to the target preset value, and the aforementioned vibration risk parameter is a shaft ratio value, the aforementioned shaft ratio value is of the second type, and a spindle vibration abnormality warning message is output to alert the operator that the vibration of the main bearing is abnormal. When the aforementioned vibration risk parameter is less than the target preset value and the aforementioned temperature difference abnormality is greater than or equal to the target preset value, a heat dissipation abnormality warning message is output to alert the operator that the heat dissipation of the main bearing is abnormal.

[0092] Obtain the maximum value of the above temperature risk parameters; this maximum value is the final temperature anomaly detection value. When the final temperature anomaly detection value is less than the target preset value and the above temperature difference anomaly value is greater than or equal to the target preset value, a temperature monitoring alert is output to inform the operator that the main bearing temperature is abnormal and requires continuous monitoring. When the final temperature anomaly detection value is greater than or equal to the target preset value and the above temperature difference anomaly value is less than or equal to the target preset value, a vibration monitoring alert is output to inform the operator that the main bearing vibration is abnormal and requires continuous monitoring. The above target preset values ​​are set by the operator according to actual needs.

[0093] The wind turbine fault diagnosis method also includes the method for obtaining the above-mentioned threshold range:

[0094] Historical vibration data, historical vibration spectra, and historical vibration demodulation spectra are acquired. The aforementioned historical vibration data refers to offline vibration data of the wind turbine generator set, specifically vibration data from within the past month when the turbine's power output exceeds 80% of its full capacity. Vibration data segments are sampled daily at fixed time intervals with a sampling frequency of at least 250,000 Hz. Measurement points should include axial and horizontal acceleration sampling data for both the front and rear bearings. FFT spectral analysis is performed on the acquired historical vibration data to obtain the historical vibration spectrum, and envelope spectral analysis is performed on the historical vibration data to obtain the historical vibration demodulation spectrum.

[0095] Based on the aforementioned historical vibration data and the preset intensity calculation model, vibration intensity values ​​are calculated. These values ​​represent the intensity of vibration corresponding to the historical vibration data. Vibration intensity values ​​are typically expressed using the maximum, average, or root mean square (RMS) values ​​of parameters characterizing vibration levels (such as displacement, velocity, and acceleration). In this embodiment, the vibration intensity value is represented by the RMS value of the historical vibration data; that is, the aforementioned vibration intensity value is the RMS value of the aforementioned historical vibration data. Based on the aforementioned historical vibration spectrum, the aforementioned historical vibration demodulation spectrum, and the preset shaft value calculation model, shaft values ​​are calculated. The calculation method for shaft values ​​is the same as the calculation method for the aforementioned shaft characteristics to be measured, and will not be elaborated here. According to the demodulation spectrum processing rules, a portion of frequencies in the aforementioned historical vibration demodulation spectrum are removed, and the signal-to-noise ratio (SNR) of the demodulation spectrum after removal is calculated. The calculation method for the SNR values ​​is the same as the calculation method for the aforementioned signal-to-noise ratio characteristics to be measured, and will not be elaborated here.

[0096] The vibration intensity values, shaft rotation values, and signal-to-noise ratio (SNR) values ​​are standardized to determine standard vectors. Kernel principal component analysis (KPCA) is then used to reduce the dimensionality of these standard vectors, yielding analysis vectors. The 99th and 5th quantiles of these analysis vectors are used to form threshold ranges. It is understood that multiple analysis vectors are obtained by processing historical vibration data. Each analysis vector includes a vibration analysis value corresponding to the vibration intensity value, a shaft rotation analysis value corresponding to the shaft rotation value, and a signal-to-noise ratio (SNR) analysis value corresponding to the SNR value. The 99th and 5th quantiles of these multiple vibration analysis values ​​are used to form vibration threshold ranges, the 99th and 5th quantiles of these multiple shaft rotation analysis values ​​are used to form shaft rotation threshold ranges, and the 99th and 5th quantiles of these multiple SNR analysis values ​​are used to form signal-to-noise threshold ranges.

[0097] The preset initial intensity value is the 95th quantile of the vibration analysis value. The preset initial shaft value is the 95th quantile of the shaft analysis value. The preset initial signal-to-noise ratio value is the 95th quantile of the signal-to-noise analysis value.

[0098] A percentile is a value that corresponds to a specific percentile in a dataset. If a set of data is sorted from smallest to largest and the cumulative percentiles are calculated, the value at that particular percentile is called the p-th percentile. It can be represented as: a set of n observations arranged numerically, for example, the value at the p% position is called the p-th percentile. Therefore, the 99th percentile refers to the value at the 99th percentile, and the 5th percentile refers to the value at the 5th percentile.

[0099] Wind turbine fault diagnosis methods also include:

[0100] The risk factor values ​​are defined as follows: [0, 0.2]: Normal; [0.2, 0.4]: Attention; [0.4, 0.6]: Warning; [0.6, 1]: Alarm. When the risk factor value is in the range [0, 0.2], it indicates that the corresponding risk factor value is normal; when it is in the range [0.2, 0.4], it indicates that the corresponding risk factor value is at attention; when it is in the range [0.4, 0.6], it indicates that the corresponding risk factor value is under warning; and when it is in the range [0.6, 1], it indicates that the corresponding risk factor value is under alarm.

[0101] When any one or more of the above-mentioned abnormal temperature rise value, absolute temperature difference value, vibration risk parameter, or temperature difference value falls within the above-mentioned scope of concern, warning range, or alarm range, a corresponding prompt message will be output.

[0102] It is understood that the aforementioned acquisition of sample data is daily data, meaning each day's data corresponds to a vibration risk parameter, a temperature risk parameter, and risk warning information. The results display section includes daily vibration data spectrum, temperature data, and cabin temperature data presented as graphs. It also displays daily temperature rise anomalies, absolute temperature difference anomalies, vibration risk parameters, and temperature difference anomalies. Furthermore, a health score is calculated based on preset weights: Health Score = Temperature Rise Anomaly * Temperature Rise Weight + Absolute Temperature Difference Anomaly * Absolute Temperature Difference Weight + Vibration Risk Parameter * Vibration Weight + Temperature Difference Anomaly * Temperature Difference Weight. Under different fault conditions, the weights corresponding to the temperature rise anomaly, absolute temperature difference anomaly, vibration risk parameter, and temperature difference anomaly are different. A weight mapping table is pre-stored in the database, storing the correspondence between the aforementioned temperature rise weight, absolute temperature difference weight, vibration weight, temperature difference weight, and fault condition.

[0103] This application provides a wind turbine fault diagnosis system 200, referring to... Figure 2 The wind turbine fault diagnosis system 200 includes:

[0104] The data acquisition module 201 is used to acquire the sample data to be tested, which includes multiple sets of data, each set of data including vibration data, vibration data spectrum, vibration data demodulation spectrum and temperature data;

[0105] The vector extraction module 202 is used to determine the model based on features, and to determine the feature vector to be measured according to the vibration data, vibration data spectrum and vibration data demodulation spectrum. Each set of data corresponds to a feature vector to be measured, and the feature vector to be measured is an array containing multiple feature values.

[0106] The vibration parameter determination module 203 is used to input the feature vector to be measured into a preset vibration diagnosis model to obtain vibration risk parameters, which are used to reflect the risk status of the main bearing vibration risk.

[0107] The temperature parameter determination module 204 is used to input the temperature data into a preset temperature diagnostic model to obtain temperature risk parameters, which are used to reflect the risk status of the main bearing temperature risk.

[0108] The information output module 205 is used to output risk warning information based on the preset risk judgment rules, the vibration risk parameters and the temperature risk parameters.

[0109] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the described module can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0110] This application discloses an electronic device. (Refer to...) Figure 3 The electronic device includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes based on a program stored in a read-only memory (ROM) 302 or a program loaded from a storage section 307 into a random access memory (RAM) 303. The RAM 303 also stores various programs and data required for system operation. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus. An input / output (I / O) interface 304 is also connected to the bus.

[0111] The following components are connected to I / O interface 304: an input section 305 including a keyboard, mouse, etc.; an output section 306 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 307 including a hard disk, etc.; and a communication section 308 including a network interface card such as a LAN card, modem, etc. The communication section 308 performs communication processing via a network such as the Internet. A drive 309 is also connected to I / O interface 304 as needed. A removable medium 310, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 309 as needed so that computer programs read from it can be installed into storage section 307 as needed.

[0112] Specifically, according to embodiments of this application, the flowchart above refers to... Figure 1The described process can be implemented as a computer software program. For example, embodiments of this application include a computer program product comprising a computer program carried on a machine-readable medium, the computer program containing program code for performing the methods shown in the flowchart. In such embodiments, the computer program can be downloaded and installed from a network via communication section 308, and / or installed from removable medium 310. When the computer program is executed by central processing unit (CPU) 301, it performs the functions defined in the apparatus of this application.

[0113] It should be noted that the computer-readable medium shown in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0114] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the foregoing application concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions claimed in this application.

Claims

1. A method for diagnosing faults in wind turbine generators, characterized in that, include: Acquire the sample data to be tested, which includes multiple sets of data, each set of data including vibration data, vibration data spectrum, vibration data demodulation spectrum and temperature data; Based on the feature determination model, a feature vector to be measured is determined according to the vibration data, vibration data spectrum, and vibration data demodulation spectrum. Each set of data corresponds to a feature vector to be measured, and the feature vector to be measured is an array containing multiple feature values. The feature vector to be measured is input into a preset vibration diagnosis model to obtain vibration risk parameters, which are used to reflect the vibration risk of the main bearing. The temperature data is input into a preset temperature diagnosis model to obtain temperature risk parameters, which are used to reflect the temperature risk of the main bearing. Based on the preset risk assessment rules, the vibration risk parameters, and the temperature risk parameters, risk warning information is output. The process of inputting the temperature data into a preset temperature diagnostic model to obtain temperature risk parameters includes: the temperature risk parameters include absolute temperature difference constants, temperature rise anomalies, and temperature difference constants; the temperature data includes full temperature data of the main bearing and engine room temperature data; the full temperature data is divided into bins according to preset binning thresholds; based on a preset absolute temperature anomaly model, the binned full temperature data is linearly mapped to determine the absolute temperature difference constant; the full temperature data is a temperature data sequence obtained by sorting according to preset sorting rules, and the temperature data sequence contains n data points, where n is a natural number greater than 1; calculating any phase in the temperature data sequence... The difference between two adjacent data points is used to obtain a difference sequence containing (n-1) differences. Based on the difference sequence, a difference sum sequence is determined, containing (n-1) difference sum data points, where the q-th difference sum data point is the sum of the first q differences in the difference sequence, where q is a natural number and q ≤ n-1. The maximum value in the difference sum sequence is obtained, and a linear mapping is performed on the maximum value based on its different values ​​to determine the temperature rise anomaly value. The difference between the full temperature data and the cabin temperature data is calculated, and the temperature difference anomaly value is determined based on the difference between the full temperature data and the cabin temperature data and a preset temperature difference anomaly model. The risk warning information output based on the preset risk judgment rules, the vibration risk parameter, and the temperature risk parameter includes: outputting vibration lubrication warning information when the abnormal temperature rise value is greater than or equal to the target preset value, or when the vibration risk parameter is greater than or equal to the target preset value and the absolute temperature difference abnormal value is greater than or equal to the target preset value and the vibration risk parameter is of the first type; outputting spindle vibration abnormality warning information when the abnormal temperature rise value is greater than or equal to the target preset value, or when the vibration risk parameter is greater than or equal to the target preset value and the absolute temperature difference abnormal value is greater than or equal to the target preset value and the vibration risk parameter is of the second type; and outputting heat dissipation abnormality warning information when the vibration risk parameter is less than the target preset value and the temperature difference abnormal value is greater than or equal to the target preset value.

2. The wind turbine fault diagnosis method according to claim 1, characterized in that, The step of inputting the feature vector to be measured into a preset vibration diagnosis model to obtain vibration risk parameters includes: performing kernel principal component analysis on multiple feature vectors to be measured to obtain multiple diagnostic values, wherein each feature vector to be measured corresponds one-to-one with a diagnostic value, and determining whether all of the multiple diagnostic values ​​are within a preset threshold range; if so, obtaining the maximum value of the multiple diagnostic values, and inputting the maximum value of the diagnostic values ​​into a diagnostic score function to obtain vibration risk parameters.

3. The wind turbine fault diagnosis method according to claim 2, characterized in that, Also includes: If not, acquire the diagnostic values ​​that are not within the threshold range; Calculate the reconstruction error of the data in the diagnostic values ​​that are not within the threshold range. The reconstruction error corresponds to the feature value in the feature vector to be measured. Multiple feature values ​​correspond to multiple reconstruction errors. Obtain the maximum value among the multiple reconstruction errors. Calculate the vibration risk parameter based on the feature value corresponding to the maximum value among the reconstruction errors.

4. The wind turbine fault diagnosis method according to claim 2, characterized in that, The method also includes a method for obtaining the threshold range: acquiring historical vibration data, historical vibration spectrum, and historical vibration demodulation spectrum; calculating vibration intensity values ​​based on the historical vibration data and a preset intensity calculation model; calculating rotational axis values ​​based on the historical vibration spectrum, the historical vibration demodulation spectrum, and a preset rotational axis value calculation model; removing some frequencies from the historical vibration demodulation spectrum according to demodulation spectrum processing rules, and calculating the signal-to-noise ratio (SNR) of the removed historical vibration demodulation spectrum; standardizing the vibration intensity values, the rotational axis values, and the SNR values ​​to determine a standard vector; performing kernel principal component analysis on the standard vector to obtain an analysis vector; and using the 99th percentile and 5th percentile of the analysis vector to form the threshold range.

5. The wind turbine fault diagnosis method according to claim 1, characterized in that, Also includes: Obtain the maximum value of the temperature risk parameter, which is the final temperature anomaly detection value; When the final temperature anomaly detection value is less than the target preset value and the temperature difference anomaly value is greater than or equal to the target preset value, a temperature attention prompt message is output. When the final abnormal temperature detection value is greater than or equal to the target preset value and the abnormal temperature difference value is less than or equal to the target preset value, a vibration attention prompt message is output.

6. A wind turbine fault diagnosis system that executes the wind turbine fault diagnosis method as described in any one of claims 1-5, characterized in that, include: The data acquisition module is used to acquire the sample data to be tested, which includes multiple sets of data, each set of data including vibration data, vibration data spectrum, vibration data demodulation spectrum, and temperature data; the vector extraction module is used to determine the feature vector to be tested based on the feature determination model, according to the vibration data, vibration data spectrum, and vibration data demodulation spectrum, where each set of data corresponds to one feature vector to be tested, and the feature vector to be tested is an array containing multiple feature values; the vibration parameter determination module is used to input the feature vector to be tested into a preset vibration diagnosis model to obtain vibration risk parameters, which are used to reflect the risk status of the main bearing vibration. The temperature parameter determination module is used to input the temperature data into a preset temperature diagnostic model to obtain temperature risk parameters, which are used to reflect the risk status of the main bearing temperature risk. The information output module is used to output risk warning information based on preset risk assessment rules, the vibration risk parameters, and the temperature risk parameters.

7. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executed according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer program is stored that can be loaded by a processor and executed according to any one of claims 1 to 5.

Images