A wind turbine generator electric energy meter state detection method, device and equipment

Abstract

The application discloses a wind turbine generator set electric energy meter state detection method, device and equipment. When multiple electric energy meters in the same wind turbine generator set detect the same measuring point of the wind turbine generator set, the theoretically obtained detection results are basically consistent. Based on this, the electric energy meters of the wind turbine generator set are used to detect a target measuring point, and a state early warning model is constructed by using the correlation between the detection data of the electric energy meters on the target measuring point. Based on this, detection data of n electric energy meters of a to-be-detected wind turbine generator set on the target measuring point are obtained. Wherein, n is a positive integer. The detection data of the n electric energy meters of the to-be-detected wind turbine generator set on the target measuring point are input into the constructed state early warning model, so that the state index of the electric energy meter can be obtained. Based on the state index, it can be determined whether the electric energy meter is abnormal. In this way, based on the state early warning model and the detection data of the electric energy meter on the target measuring point, the accurate detection of the state of the electric energy meter of the wind turbine generator set can be realized.

Description

Technical Field

[0001] This application relates to the field of wind power generation technology, specifically to a method, device, and equipment for detecting the status of a wind turbine power meter. Background Technology

[0002] Wind turbines are devices that convert wind energy into electrical energy. The electricity meter is a crucial monitoring component within the wind turbine's converter cabinet. Feedback from the grid side regarding current, voltage, active power, reactive power, and frequency primarily comes from the electricity meter. Deviations in the electricity meter readings will cause control deviations in the wind turbine; therefore, monitoring the status of the wind turbine's electricity meter is of paramount importance. Summary of the Invention

[0003] In view of this, embodiments of this application provide a method, apparatus, and equipment for detecting the status of generator set energy meters, so as to achieve accurate detection of the status of generator set energy meters.

[0004] To address the above problems, the technical solutions provided in this application are as follows:

[0005] This application provides a method for detecting the status of a wind turbine energy meter, the method comprising:

[0006] Obtain the test data of n energy meters of the wind turbine generator set under test for the target measuring point, where n is a positive integer;

[0007] The detection data of n energy meters of the wind turbine to be tested at the target measuring point are input into the status early warning model to obtain the status index of the energy meter. The status index is used to indicate whether the energy meter is abnormal.

[0008] The state warning model is constructed based on the correlation between the detection data of the wind turbine's energy meter at the target measuring point.

[0009] Optionally, the correlation between the detection data of the energy meters of the wind turbine for the target measuring point is obtained by linearly fitting the detection data of the energy meters of any one of the multiple wind turbines for the target measuring point, and then clustering the corresponding fitting coefficients of the multiple wind turbines according to the clustering categories.

[0010] Optionally, the target measurement points include the first phase current, the second phase current, the third phase current, the first phase voltage, the second phase voltage, and the third phase voltage.

[0011] Optionally, the step of inputting the detection data of the n energy meters of the wind turbine to be tested at the target measuring point into the state early warning model to obtain the state indicators of the energy meters includes:

[0012] The detection data of n energy meters of the wind turbine under test for each target measuring point are respectively input into the corresponding state early warning model to obtain the state index of the energy meter for each target measuring point.

[0013] The maximum value among the status indicators of the energy meter for each target measuring point is taken as the detection status level of the energy meter of the wind turbine to be tested.

[0014] Optionally, the status indicator includes multiple status levels, with different status levels representing the degree of abnormality of the electricity meter; the method further includes:

[0015] When the status level of the energy meter of the wind turbine under test is greater than the level threshold, the system outputs that the energy meter of the wind turbine under test has an abnormality and / or an abnormal measurement point.

[0016] This application embodiment also provides a training method for a state warning model, the method comprising:

[0017] For any one of multiple wind turbine units, obtain the detection data of p energy meters of the wind turbine unit for the target measuring point, where p is a positive integer;

[0018] Linear fitting is performed on the detection data of any q energy meters out of p energy meters to obtain the fitting coefficients corresponding to the wind turbine, where q is a positive integer and p is greater than or equal to q.

[0019] The fitting coefficients corresponding to each of the multiple wind turbine units are clustered to obtain m cluster categories, where m is a positive integer;

[0020] Based on the mapping between the m cluster categories and the state of the energy meters of the multiple wind turbine units, the state indicators corresponding to each of the multiple wind turbine units are obtained.

[0021] A state early warning model is trained based on the detection data of the target measurement points from the multiple wind turbine units and the corresponding state indicators of the multiple wind turbine units.

[0022] Optionally, the step of linearly fitting the detection data of any q energy meters out of p energy meters to obtain the fitting coefficients corresponding to the wind turbine includes:

[0023] The target energy meter of the wind turbine is linearly fitted with the detection data of each other energy meter for the target measurement point to obtain the fitting coefficient corresponding to the wind turbine; the target energy meter is the preset energy meter of the wind turbine, and the other energy meters are the energy meters of the wind turbine other than the target energy meter.

[0024] Optionally, the status early warning model is specifically used to obtain the status indicators of the target energy meter of the wind turbine to be tested.

[0025] Optionally, the step of linearly fitting the detection data of any q energy meters out of p energy meters to obtain the fitting coefficients corresponding to the wind turbine includes:

[0026] The other energy meters of the wind turbine are used to perform pairwise linear fitting of the detection data of the target measuring point to obtain the fitting coefficients corresponding to the wind turbine; the other energy meters are the energy meters of the wind turbine other than the target energy meter, and the target energy meter is the preset energy meter of the wind turbine.

[0027] Optionally, the status warning model is specifically used to obtain the status indicators of other energy meters of the wind turbine to be tested.

[0028] This application embodiment also provides a wind turbine energy meter status detection device, the device comprising:

[0029] The first acquisition unit is used to acquire the detection data of n energy meters of the wind turbine generator set under test for the target measuring point, where n is a positive integer;

[0030] The second acquisition unit is used to input the detection data of n energy meters of the wind turbine to be tested at the target measuring point into the status early warning model to obtain the status index of the energy meter, and the status index is used to indicate whether the energy meter is abnormal.

[0031] The state warning model is constructed based on the correlation between the detection data of the wind turbine's energy meter at the target measuring point.

[0032] This application embodiment also provides a training device for a state warning model, the device comprising:

[0033] The first acquisition unit is used to acquire the detection data of p energy meters of any one of the multiple wind turbine units for the target measuring point, where p is a positive integer;

[0034] The second acquisition unit is used to perform linear fitting on the detection data of any q energy meters among p energy meters to obtain the fitting coefficients corresponding to the wind turbine, where q is a positive integer and p is greater than or equal to q.

[0035] The third acquisition unit is used to cluster the fitting coefficients corresponding to each of the multiple wind turbine units to obtain m cluster categories, where m is a positive integer;

[0036] The fourth acquisition unit is used to obtain the status indicators corresponding to each of the multiple wind turbines based on the state mapping between the m cluster categories and the energy meters of the multiple wind turbines.

[0037] The training unit is used to train a state early warning model based on the detection data of the multiple wind turbines at the target measurement points and the corresponding state indicators of the multiple wind turbines.

[0038] This application also provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the wind turbine energy meter status detection method or the status early warning model training method described in any of the above claims.

[0039] This application also provides a computing device, including:

[0040] At least one processor;

[0041] At least one memory stores a computer program, which, when executed by the at least one processor, implements the wind turbine energy meter status detection method or the status early warning model training method described above.

[0042] Therefore, the embodiments of this application have the following beneficial effects:

[0043] This application provides a method, apparatus, and device for detecting the status of electricity meters in wind turbine generator sets. Theoretically, when multiple electricity meters in the same wind turbine generator set detect the same measuring point, the detection results should be essentially consistent. Based on this, the target measuring point is detected using the electricity meters of the wind turbine generator set beforehand, and a status warning model is constructed using the correlation between the detection data of the electricity meters at the target measuring point. The status warning model can input the detection data of the electricity meters at the target measuring point and output the status indicators of the electricity meters. Detection data of n electricity meters in the wind turbine generator set to be tested at the target measuring point are obtained, where n is a positive integer. The detection data of the n electricity meters in the wind turbine generator set to be tested at the target measuring point are input into the constructed status warning model to obtain the status indicators of the electricity meters. Based on the status indicators, it can be determined whether there is an anomaly in the electricity meter. Thus, based on the status warning model and the detection data of the electricity meters at the target measuring point, accurate detection of the status of the generator set's electricity meters can be achieved. Attached Figure Description

[0044] Figure 1 A schematic diagram illustrating an exemplary application scenario provided in this application embodiment;

[0045] Figure 2 A flowchart illustrating a training method for a state warning model provided in an embodiment of this application;

[0046] Figure 3 A flowchart of a method for detecting the status of a wind turbine energy meter provided in this application embodiment;

[0047] Figure 4 This is a schematic diagram of an early warning result provided in an embodiment of this application;

[0048] Figure 5 A schematic diagram of the structure of a training device for a state early warning model provided in an embodiment of this application;

[0049] Figure 6 This is a schematic diagram of the structure of a wind turbine energy meter status detection device provided in an embodiment of this application. Detailed Implementation

[0050] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the embodiments of this application will be further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0051] To facilitate understanding and explanation of the technical solutions provided in the embodiments of this application, the background technology involved in the embodiments of this application will be introduced first.

[0052] Wind turbines are devices that convert wind energy into electrical energy. The electricity meter is a crucial monitoring component within the wind turbine's converter cabinet. Feedback from the grid side regarding current, voltage, active power, reactive power, and frequency primarily comes from the electricity meter. Deviations in the electricity meter readings will cause control deviations in the wind turbine; therefore, monitoring the status of the wind turbine's electricity meter is of paramount importance.

[0053] Based on this, this application provides a method, apparatus, and device for detecting the status of electricity meters in wind turbine generator sets. Theoretically, when multiple electricity meters in the same wind turbine generator set detect the same measuring point, the detection results obtained are basically consistent, meaning that the correlation between the detection data of multiple electricity meters in the same wind turbine generator set for the same measuring point is very high. Based on this, the target measuring point can be detected in advance using the electricity meters of the wind turbine generator set, and a status warning model can be constructed using the correlation between the detection data of the electricity meters for the target measuring point. The status warning model can input the detection data of the electricity meters for the target measuring point and output the status indicators of the electricity meters. When it is necessary to detect the status of the electricity meters of the wind turbine generator set to be tested, the detection data of n electricity meters of the wind turbine generator set to be tested for the target measuring point are obtained, where n is a positive integer. The detection data of the n electricity meters of the wind turbine generator set to be tested for the target measuring point are input into the constructed status warning model to obtain the status indicators of the electricity meters. Based on the status indicators, it can be determined whether there is an anomaly in the electricity meter. Thus, based on the status warning model and the detection data of the electricity meters for the target measuring point, accurate detection of the status of the electricity meters of the generator set can be achieved.

[0054] The wind turbine energy meter status detection method provided in this application is implemented based on a constructed status warning model. For ease of understanding and explanation, the training method of the status warning model provided in this application will be described first. It is understood that the wind turbine energy meter status detection method provided in this application can be applied to the main controller of a wind turbine or other computing devices. The training method of the status warning model provided in this application can be applied to any computing device. This application does not limit the executing entity of the wind turbine energy meter status detection method or the status warning model training method. The wind turbine energy meter status detection method or the status warning model training method provided in this application can be executed by the same executing entity or by different executing entities.

[0055] To facilitate understanding of the training method for the state warning model provided in the embodiments of this application, the following is combined with... Figure 1 The example scenario is shown below. See also... Figure 1 As shown in the figure, this figure is a schematic diagram of an exemplary application scenario provided in the embodiments of this application.

[0056] In practical applications, one wind turbine is randomly selected from multiple turbine units, and p energy meters within that turbine are used to measure the target measurement point. The measurement data from the p energy meters for the target measurement point are obtained, where p is a positive integer. Linear fitting is then performed on the measurement data from any q energy meters among the p energy meters to obtain the fitting coefficients for the wind turbine unit, where q is a positive integer and p is greater than or equal to q.

[0057] The above operations are performed on each of the multiple wind turbine units to obtain the fitting coefficients for each unit. Then, the fitting coefficients for each wind turbine unit are clustered to obtain m clusters, where m is a positive integer. Based on the mapping between the m clusters and the state of the energy meters of the multiple wind turbine units, state indicators for each wind turbine unit are obtained. These state indicators can be viewed as labels for the detection data of the p energy meters in the wind turbine unit.

[0058] Based on this, by training the state warning model using the detection data of multiple wind turbine units targeting the measurement points and the corresponding state indicators of multiple wind turbine units as labels, a trained state warning model can be obtained.

[0059] Those skilled in the art will understand that Figure 1 The schematic diagram shown is merely one example in which embodiments of this application can be implemented. The scope of application of the embodiments of this application is not limited by any aspect of this framework.

[0060] To facilitate understanding of this application, the following description, in conjunction with the accompanying drawings, illustrates a training method for a state warning model provided in an embodiment of this application.

[0061] See Figure 2 As shown, this figure is a flowchart of a training method for a state warning model provided in an embodiment of this application. Figure 2 As shown, the method may include S201-S205:

[0062] S201: For any one of the multiple wind turbine units, obtain the detection data of p energy meters of the wind turbine unit for the target measuring point, where p is a positive integer.

[0063] As an optional example, the wind turbine in this embodiment is an offshore wind turbine. Offshore wind turbines have a large rated output power, and the converter adopts a cabinet design. The energy meter is an important detection component in the converter cabinet. Furthermore, this embodiment does not limit the specific scenario of the wind turbine, nor does it limit the specific type of wind turbine, the number of wind turbines, or the number of energy meters within the wind turbine. For example, the wind turbine can also be a direct-drive megawatt-class unit, a doubly-fed megawatt-class unit, etc.

[0064] In this embodiment, the state warning model is trained based on the correlation of detection data from the electricity meters of multiple wind turbines at the target measuring point. Based on this, multiple wind turbines are first identified, and for any one of these turbines, detection data from p electricity meters at the target measuring point are obtained. Here, p electricity meters can be all the electricity meters in the wind turbine or a subset of electricity meters within that turbine. In practical applications, the detection data from the p electricity meters at the target measuring point can be the sampling data collected over one day.

[0065] The target measuring point is the object detected by the electricity meter. There can be one or more target measuring points. As an optional example, the target measuring points include the first phase current, the second phase current, the third phase current, the first phase voltage, the second phase voltage, and the third phase voltage. The first phase can be phase A, the second phase can be phase B, and the third phase can be phase C.

[0066] In this scenario, the system can obtain the detection data of p energy meters for the first phase current, the second phase current, the third phase current, the first phase voltage, the second phase voltage, and the third phase voltage of the wind turbine. When the energy meters detect current, the data is current data. When the energy meters detect voltage, the data is voltage data.

[0067] Taking the detection data of the first phase current of a wind turbine unit by p electricity meters as an example, where the first phase current is phase A, if p = 4, and the four electricity meters are denoted as 1#, 2#, 3#, and 4#, then the detection data of the first phase current by the four electricity meters are the current data obtained by electricity meter 1# from detecting the phase A current, the current data obtained by electricity meter 2# from detecting the phase A current, the current data obtained by electricity meter 3# from detecting the phase A current, and the current data obtained by electricity meter 4# from detecting the phase A current.

[0068] S202: Perform linear fitting on the detection data of any q energy meters among the p energy meters to obtain the fitting coefficients corresponding to the wind turbine, where q is a positive integer and p is greater than or equal to q.

[0069] When multiple energy meters in the same wind turbine are used to detect the same measuring point, the theoretically obtained detection data will be basically consistent. Based on this, after acquiring the detection data of p energy meters from any one of the multiple wind turbines at the target measuring point, the detection data of any q energy meters among the p energy meters are linearly fitted to obtain the fitting coefficients corresponding to the wind turbine. Specifically, pairwise linear fitting is performed on the detection data of the q energy meters at the same measuring point in the target area to obtain the fitting coefficients of that wind turbine at the target measuring point.

[0070] Linear fitting can be represented by the formula y = kx + b. y represents the data from one electricity meter at one measuring point, and x represents the data from another electricity meter at the same measuring point. k is the proportionality coefficient, and b is the intercept; k and b are collectively referred to as the fitting coefficients.

[0071] By performing pairwise linear fitting on the detection data of q energy meters at the same measuring point within the target area, correlation analysis can be performed on the detection data obtained by q energy meters from the same measuring point under the same wind turbine. The fitting coefficient represents the correlation between the detection data obtained by two energy meters from the same measuring point. Theoretically, when the correlation between two sets of data is extremely high, the more similar the two sets of data are, the better the fitting coefficient should be k=1 and b=0. Therefore, the fitting coefficient can be used to indicate whether there is an abnormal energy meter among the two energy meters.

[0072] As one possible example, p equals q. As another possible example, q is less than p, and q is greater than or equal to 2.

[0073] For example, when p = q = 4, the four energy meters of the wind turbine unit are denoted as 1#, 2#, 3#, and 4#, with one of the target measuring points being the A-phase current. When performing pairwise fitting on the A-phase current data measured by the four energy meters, linear fitting is performed on the A-phase current data measured by 1# and 2#, 1# and 3#, 1# and 4#, 2# and 3#, 2# and 4#, and 3# and 4# respectively. This yields six fitting coefficients. These six fitting coefficients are the fitting coefficients for the A-phase current measuring point corresponding to this wind turbine unit.

[0074] In one or more embodiments, the average of multiple fitting coefficients at the same measuring point corresponding to the wind turbine is calculated, and the average value is used as the fitting coefficient of the wind turbine at that measuring point. This fitting coefficient can indicate whether there is an anomaly in the energy meter of the wind turbine based on the analysis of the obtained detection data at that measuring point.

[0075] Similarly, the fitting coefficients of the fan unit at the B-phase current measurement point, the C-phase current measurement point, the A-phase voltage measurement point, the B-phase voltage measurement point, and the C-phase voltage measurement point can also be obtained.

[0076] Understandably, by executing S202 for multiple wind turbines, the fitting coefficients for each wind turbine at the target measurement point can be obtained.

[0077] In some possible implementations, embodiments of this application provide two specific implementation methods for linearly fitting the detection data of any q energy meters among p energy meters to obtain the fitting coefficients corresponding to the wind turbine, as detailed below.

[0078] S203: Cluster the fitting coefficients of multiple wind turbine units to obtain m cluster categories, where m is a positive integer.

[0079] Each wind turbine has a corresponding fitting coefficient, which is the fitting coefficient for the target measurement point. The fitting coefficient can indicate whether there is an anomaly in the wind turbine's electricity meter under the analysis of the detection data at a specific measurement point. Alternatively, it can be understood as indicating whether the deviation of the detection data obtained from the wind turbine's electricity meter at a specific measurement point is large.

[0080] However, in practical applications, fluctuations in the detection data of the electricity meter at the target measuring point may cause the fitting coefficient (k,b) to deviate from the theoretical value (1,0). Therefore, the fitting coefficients for the target measuring points of multiple wind turbines are used as clustering data. Instead of the fitting coefficients, the cluster categories obtained from the clustering are used to represent whether the electricity meter under the wind turbine is abnormal. As an optional example, the points represented by (k,b) are clustered with the k value of the fitting coefficient on the x-axis and the b value of the fitting coefficient on the y-axis.

[0081] In practice, the fitting coefficients of multiple wind turbine units at the same measuring point are clustered to obtain m cluster categories at the same measuring point.

[0082] For example, there are 100 wind turbine units. Each of the 100 wind turbine units has a fitting coefficient for phase A current measurement point, phase B current measurement point, phase C current measurement point, phase A voltage measurement point, phase B voltage measurement point, and phase C voltage measurement point.

[0083] Therefore, all the fitting coefficients of wind turbine units 1-100 at the A-phase current measurement point are clustered together, resulting in m clusters. Similarly, all the fitting coefficients of wind turbine units 1-100 at the B-phase current measurement point are clustered together, also resulting in m clusters. The fitting coefficients of wind turbine units 1-100 at the C-phase current measurement point are clustered together, also resulting in m clusters. The phase voltage measurement points are similar to those for the phase current measurement points, and will not be described in detail here.

[0084] As an alternative example, the K-means unsupervised clustering algorithm can be chosen for clustering. The number of clusters m is determined based on the actual situation; for example, m can be 5 clusters.

[0085] S204: Based on the mapping between m cluster categories and the state of the energy meters of multiple wind turbine units, obtain the state indicators corresponding to each of the multiple wind turbine units.

[0086] After clustering, the m cluster categories are mapped to the energy meters of multiple wind turbine units to obtain the corresponding state indicators for each wind turbine unit. State mapping can mitigate the problem of excessive data loss during linear fitting, thereby improving the integrity of the data. State indicators can include multiple state levels, with different state levels representing the degree of energy meter anomalies. The state indicators can be a specific representation of the cluster categories, which can be understood as cluster labels. For example, if there are 5 cluster categories and state indicators 0, 1, 2, 3, and 4, the state indicators 0, 1, 2, 3, and 4 represent the 5 cluster categories. State indicators can be directly correlated with state levels; for example, state indicators 0, 1, 2, 3, and 4 can correspond to state levels 0, 1, 2, 3, and 4 respectively. Alternatively, multiple state indicators can correspond to a single state level; for example, state indicators 0, 1, and 2 can correspond to state level 1, and state indicators 3 and 4 can correspond to state level 2.

[0087] In a specific implementation, as an optional example, this application provides a specific implementation method for obtaining the corresponding state indicators of multiple wind turbines based on the state mapping between m cluster categories and the energy meters of multiple wind turbines, including:

[0088] By mapping the m cluster categories at the same measuring point to the state of the energy meters of multiple wind turbine units, the state indicators of each wind turbine unit at the same measuring point can be obtained.

[0089] For example, if the same measuring point is the A-phase current measuring point, there are 5 cluster categories, represented by state indices 0, 1, 2, 3, and 4. Mapping the cluster categories to which the fitting coefficients of wind turbine units 1-100 belong under the A-phase current measuring point to their respective energy meters, the resulting state indices for wind turbine units 1-100 under the A-phase current measuring point are any values ​​from 0 to 4. Thus, the state indices for each of the multiple wind turbine units under the A-phase current measuring point are obtained.

[0090] Then, the status indicators are tagged with services to obtain corresponding service tags. The service tags obtained after tagging with services are sufficient to indicate whether the energy meter under the wind turbine is abnormal. See Table 1, which is a schematic table of status indicators, service tags, and status descriptions.

[0091] Table 1 Status Indicators, Business Tags, and Status Descriptions

[0092]

[0093] As shown in Table 1, the service tags include Normal, Caution, Warning, Alarm, and Danger. These service tags represent status levels, including a 5-level warning system. Furthermore, the status level indicators mapped from the service tags represent the degree of abnormality in the wind turbine's energy meter. In a practical implementation, as an optional example, a level threshold is set. When the status level corresponding to the status indicator mapped from the service tag is greater than the level threshold, it indicates that the wind turbine's energy meter is abnormal; otherwise, it indicates normal. For example, if the level threshold is 2 (Warning), then the status levels corresponding to the status indicators 0 (Normal), 1 (Caution), and 2 (Warning) indicate that the wind turbine's energy meter is normal, while 3 (Alarm) and 4 (Danger) indicate that the wind turbine's energy meter is abnormal.

[0094] Based on this, the state index is used as the label for the wind turbine. This label can be used for training the subsequent state warning model. For example, after clustering, the state index of the fitting coefficient for the A-phase current measurement point of wind turbine 1 is 0. Then, the label corresponding to the detection data of the q energy meters of wind turbine 1 for the A-phase current measurement point is 0. The other phase current measurement points and voltage measurement points are similar and will not be described in detail here.

[0095] Furthermore, in practical applications, when using the trained state warning model for prediction, if the obtained state index of the wind turbine's energy meter is 1, the turbine's operating status can be monitored, and daily inspections and maintenance can be strengthened. If the obtained state index of the wind turbine's energy meter is 2, the energy meter can be included in the monitoring system, and inspections can be carried out as appropriate, followed by maintenance or repairs based on the inspection results. If the obtained state index of the wind turbine's energy meter is 3, inspections can be arranged as soon as possible, followed by maintenance or repairs based on the inspection results. If the obtained state index of the wind turbine's energy meter is 4, the turbine can be immediately shut down for inspection, followed by repairs based on the inspection results.

[0096] S205: A state early warning model is trained based on the detection data of multiple wind turbine units targeting the measurement points and the corresponding state indicators of multiple wind turbine units.

[0097] After obtaining the status indicators for multiple wind turbines, the detection data of these turbines at the target measurement points becomes labeled data, with the labels representing the corresponding status indicators. Then, the status early warning model is trained based on the detection data and status indicators from the multiple wind turbines at the target measurement points. Specifically, the detection data from the multiple wind turbines at the target measurement points serves as the input to the status early warning model, and the output is the predicted status indicator.

[0098] It is understandable that the detection data for each wind turbine at the target measurement point consists of the detection data from q electricity meters within each wind turbine at the target measurement point. For example, the detection data from q electricity meters in wind turbine 1 for phase A current measurement point, phase B current measurement point, phase C current measurement point, phase A voltage measurement point, phase B voltage measurement point, and phase C voltage and current measurement point are respectively. The other wind turbines are similar, and will not be elaborated here.

[0099] As an optional example, the state warning model is the model corresponding to the target measurement point. As another optional example, when the target measurement points are phase A current, phase B current, phase C current, phase A voltage, phase B voltage, and phase C voltage, the state warning model specifically consists of state warning models corresponding to the phase A current measurement point, phase B current measurement point, phase C current measurement point, phase A voltage measurement point, phase B voltage measurement point, and phase C voltage and current measurement point, respectively. That is, a corresponding state warning model can be established for different target measurement points. For example, when training the state warning model corresponding to the phase A current measurement point, the training data consists of the detection data of multiple wind turbine units for the phase A current measurement point and the corresponding state indicators of multiple wind turbine units under the phase A current measurement point. The state warning models corresponding to other measurement points are similar and will not be elaborated here.

[0100] In practical applications, based on the status indicators corresponding to multiple wind turbines as data labels, a data selection ratio is determined according to the status indicators. A certain amount of data is selected from the detection data of multiple wind turbines targeting the target measurement points for training and validation of the status early warning model. For example, the ratio of data with status indicators 1-4 to data with status indicator 0 is 2:1. If at least 10 wind turbines with status indicators 1-4 are selected, the detection data of these 10 wind turbines targeting the target measurement points is used. If at least 5 wind turbines with status indicator 0 are selected, the detection data of these 5 wind turbines targeting the target measurement points is used.

[0101] Based on the above, the detection data of multiple wind turbine units targeting the target measurement points and the corresponding state indicators of each wind turbine unit are further divided according to a certain ratio to obtain training data and validation data for the state warning model. As an optional example, the training data and validation data are divided using the Pareto principle (80 / 20 rule).

[0102] After model training, the trained state warning model is evaluated by inputting validation data. As an example, the amount of validation data corresponding to each state indicator is shown in Table 2. Table 2 shows the prediction results of the validation data, where the total amount of validation data is 66,626.

[0103] Table 2 Validation Data Prediction Results

[0104] Classification Actual data volume Predicted data volume Forecast / Actual 0 24695 26064 1.055436 1 12377 11508 0.929789 2 12425 13027 1.048451 3 12212 11443 0.937029 4 4917 4584 0.932276

[0105] As shown in Table 2, after inputting validation data for state warning model prediction, the deviation between the predicted data and the actual predicted data is within ±10%, indicating good prediction accuracy. Through multiple training sessions, the prediction accuracy of the state warning model can be maintained at around 80%.

[0106] The trained status warning model can be used to input the detection data of the energy meter of the wind turbine under test for the target measuring point, and output the status indicators of the energy meter. The output status indicators can indicate whether there is an anomaly in the energy meter. If an anomaly is found, an anomaly warning is output to the production management system, and a warning work order is dispatched for on-site investigation.

[0107] Furthermore, the trained status warning model can be deployed on a device health management platform, enabling scheduled, automatic, and distributed computation of the model in the cloud. Compared to traditional single-platform deployment, this method improves computation speed, allows for faster model iteration, handles greater computational load, and offers significant practical application value.

[0108] As an optional example, the state warning model is a model employing a supervised gradient boosting tree algorithm. Based on this, in one possible implementation, this application provides a specific method for determining the model parameters of the state warning model, as detailed below.

[0109] Based on the content of S201-S205, it is known that the detection data obtained by multiple energy meters detecting the same phase current or voltage are theoretically basically consistent. Therefore, for any one of the multiple wind turbines, the detection data of p energy meters for the target measurement point are obtained. Then, the detection data of any q energy meters among the p energy meters are linearly fitted to obtain the fitting coefficients corresponding to the wind turbine. If an abnormality is detected by one energy meter, the fitting coefficients of the detection data of that energy meter will deviate significantly from those of the other energy meters. Next, the fitting coefficients corresponding to each of the multiple wind turbines are clustered to obtain m cluster categories. Based on the mapping of the m cluster categories to the state of the energy meters of the multiple wind turbines, the state indicators corresponding to each of the multiple wind turbines are obtained. Using the state indicators corresponding to each of the multiple wind turbines as labels, a state warning model is trained based on the detection data of the multiple wind turbines for the target measurement point and the corresponding state indicators. In this way, the trained state warning model can perform regular health assessments of the wind turbine's electricity meter and reduce the incidence of production accidents such as high / low voltage mis-triggered circuits and prolonged turbine shutdowns caused by large deviations in electricity meter readings.

[0110] As an optional example, the state warning model is a model employing a supervised gradient boosting tree algorithm. Based on this, in one possible implementation, this application provides a specific method for determining the model parameters of the state warning model, as shown below.

[0111] First, set the model parameters for the experiment. Model parameters include step size, number of iterations, maximum depth of the weak classifier, subsampling, minimum number of samples required for internal node repartitioning, and minimum number of samples required for leaf nodes. Determine the range or specific values ​​for the model parameters. For example, the step size parameter range is 0.3, 0.4, 0.5, and 0.6. The iteration number parameter range is 100, 110, 120, 130, 140, 150, 150, 170, 180, 190, and 200. The maximum depth of the weak classifier parameter range is 5 and 8.

[0112] Furthermore, a grid search was performed on three model parameters: step size, number of iterations, and maximum depth of the weak classifier. Other parameters were assigned relatively optimal values ​​based on experience and multiple tests; for example, subsampling was set to 0.7. The minimum number of samples required for further partitioning of internal nodes was set to 200. The minimum number of samples required for leaf nodes was set to 500.

[0113] The model is evaluated based on the selected model parameters. The model performance is primarily evaluated using three metrics: accuracy, recall, and F1-score. Examples of specific metric evaluations are shown in Table 3, which also presents examples of the model's evaluation output results.

[0114] Table 3 Model Evaluation Output Results

[0115] Evaluation indicators 0 1 2 3 4 accuracy Macro average weighted average accuracy 0.974785 0.826239 0.826039 0.782708 0.741487 0.859748 0.830251 0.858997 Recall rate 0.985029 0.807539 0.850241 0.768163 0.732363 0.859748 0.828667 0.859748 F1 score 0.979881 0.816782 0.837965 0.775367 0.736896 0.859748 0.829378 0.859293 Data volume 20173 12283 12253 12539 4876 0.859748 62124 62124

[0116] The model's generalization ability is evaluated by analyzing metrics such as accuracy, recall, and F1-score, and the final model parameters are determined. For example, Table 4 shows the model parameters of the state warning model corresponding to the A-phase current measurement point.

[0117] Table 4 shows the model parameters of the state warning model corresponding to the phase A current measuring point.

[0118] parameter value Step length 0.6 Number of iterations 190 Maximum depth of weak classifier 8 Subsampling 0.7 Minimum number of samples required for internal node repartition 200 Minimum number of samples for leaf nodes 500

[0119] Similarly, the model parameters of the state warning models corresponding to other phase current and phase voltage measurement points are determined in the same way.

[0120] Because there may be static deviations between a target energy meter and other energy meters in the wind turbine unit, for example, the detection result of energy meter #1 may differ from that of energy meters #2, #3, and #4. Therefore, when constructing the status warning model, two scenarios can be implemented: establishing a status warning model for the target energy meter and separate status warning models for other energy meters. For example, establishing a status warning model for energy meter #1 and status warning models for energy meters #2, #3, and #4.

[0121] Based on the above, in one possible implementation, this application embodiment provides a specific implementation method for linearly fitting the detection data of any q energy meters out of p energy meters in step S202 to obtain the fitting coefficients corresponding to the wind turbine, including:

[0122] Linear fitting is performed on the target energy meter of the wind turbine and the detection data of each other energy meter for the target measurement point to obtain the fitting coefficient corresponding to the wind turbine. The target energy meter is the preset energy meter of the wind turbine, and the other energy meters are the energy meters of the wind turbine other than the target energy meter.

[0123] The preset energy meter is one that exhibits a static deviation from other energy meters. For example, the preset energy meter is #1, and the other energy meters are #2, #3, and #4. The detection data from energy meter #1 at the target measuring point is then fitted with the detection data from energy meters #2, #3, and #4 at the target measuring point, respectively, yielding three fitting coefficients. The average of these three fitting coefficients is taken as the fitting coefficient for the wind turbine unit.

[0124] Therefore, after obtaining the fitting coefficients for each wind turbine, the fitting coefficients for each wind turbine are clustered to obtain m clusters. Then, based on the mapping between the m clusters and the state of the energy meters of the wind turbines, the state indices for each wind turbine are obtained. Finally, a state warning model is trained using the detection data of the target measurement points from the wind turbines and the corresponding state indices. The trained state warning model is specifically used to obtain the state indices of the target energy meters of the wind turbines under test; that is, the state indices output by the state warning model are the state indices of the target energy meters of the wind turbines under test.

[0125] For example, when the target energy meter is #1, this status warning model is specifically used to obtain the status indicators of energy meter #1 of the wind turbine under test, and then determine the warning level corresponding to energy meter #1 of the wind turbine under test and whether energy meter #1 is abnormal. Furthermore, if the status warning model is specifically a status warning model under the A-phase current measurement point, then when the detection data of energy meter #1 of the wind turbine under test for the A-phase current measurement point is input, the output status indicator is the status indicator of energy meter #1 obtained based on the detection of the A-phase current measurement point. The other phase current measurement points and phase voltage measurement points are similar and will not be elaborated here.

[0126] It is understood that the number of target energy meters is not limited in the embodiments of this application, and can be flexibly set according to the actual situation. For example, the target energy meters can also be energy meters 1# and 2# that have static deviations.

[0127] In another possible implementation, this application embodiment also provides a specific implementation method for linearly fitting the detection data of any q energy meters among p energy meters in step S202 to obtain the fitting coefficients corresponding to the wind turbine, including:

[0128] The other energy meters of the wind turbine are used to perform pairwise linear fitting of the detection data of the target measuring point to obtain the fitting coefficients of the wind turbine. The other energy meters are the energy meters of the wind turbine other than the target energy meter, and the target energy meter is the preset energy meter of the wind turbine.

[0129] The preset energy meter is one that exhibits a static deviation from other energy meters. For example, the preset energy meter is #1, and the other energy meters are #2, #3, and #4. The detection data from energy meter #2, energy meter #3, and energy meter #4 at the target measurement point are then linearly fitted pairwise to obtain three fitting coefficients. The average of these three fitting coefficients is taken as the fitting coefficient for the wind turbine unit.

[0130] Therefore, after obtaining the fitting coefficients for each wind turbine, the fitting coefficients for each wind turbine are clustered to obtain m clusters. Then, based on the mapping between the m clusters and the state of the energy meters of the wind turbines, the state indices for each wind turbine are obtained. Finally, a state warning model is trained using the detection data of the target measurement points from the wind turbines and the corresponding state indices. The trained state warning model is specifically used to obtain the state indices of other energy meters of the wind turbine under test; that is, the state indices output by the state warning model are the state indices of other energy meters of the wind turbine under test.

[0131] For example, when the other electric energy meters are 2#, 3#, and 4#, the state early warning model is specifically configured to obtain the state indicators of the electric energy meter 2#, electric energy meter 3#, and electric energy meter 4# of the wind turbine to be detected, and further determine the corresponding early warning levels of the electric energy meter 2#, electric energy meter 3#, and electric energy meter 4# of the wind turbine to be detected, as well as whether there is any abnormal electric energy meter among the electric energy meter 2#, electric energy meter 3#, and electric energy meter 4#.

[0132] In addition, further, if the state early warning model is specifically the state early warning model under the phase A current measurement point, when the detection data of the electric energy meters 2#, 3#, and 4# of the wind turbine to be detected for the phase A current measurement point is input, the output state indicators are the state indicators of the electric energy meters 2#, 3#, and 4# obtained on the basis of detecting the phase A current measurement point.

[0133] It can be seen from the above two specific implementations of S202 that when the detection data for the target measurement point used to obtain the fitting coefficients are different, the obtained fitting coefficients are different, and finally the electric energy meters detected by the trained state early warning model are different, which can realize abnormal detection of different electric energy meters.

[0134] Based on the training method for the state early warning model provided in the above embodiments, after obtaining the trained state early warning model, the embodiments of the present application further provide a wind turbine electric energy meter state detection method.

[0135] In order to facilitate the understanding of the wind turbine electric energy meter state detection method provided in the embodiments of the present application, the following describes the wind turbine electric energy meter state detection method provided in the embodiments of the present application with reference to the accompanying drawings.

[0136] See Figure 3 as shown in the figure, which is a flowchart of the wind turbine electric energy meter state detection method provided in the embodiment of the present application, as Figure 3 shown, the method may include S301 to S302:

[0137] S301: Acquire detection data of n electric energy meters of the wind turbine to be detected for the target measurement point, where n is a positive integer.

[0138] The wind turbine to be detected is any wind turbine. The n electric energy meters of the wind turbine to be detected are all or several electric energy meters of the wind turbine to be detected. That is, as an optional example, n=p, n<p, or n>p.

[0139] In one or more embodiments, the target measurement points include a first phase current, a second phase current, a third phase current, a first phase voltage, a second phase voltage, and a third phase voltage. As an optional example, the first phase current, the second phase current, the third phase current, the first phase voltage, the second phase voltage, and the third phase voltage may be phase A current, phase B current, phase C current, phase A voltage, phase B voltage, and phase C voltage.

[0140] Based on this, the detection data of the first phase current, the second phase current, the third phase current, the first phase voltage, the second phase voltage, and the third phase voltage of the n energy meters of the wind turbine to be tested are obtained.

[0141] S302: Input the detection data of n energy meters of the wind turbine to be tested at the target measuring point into the status early warning model to obtain the status index of the energy meter. The status index is used to indicate whether there is an abnormality in the energy meter.

[0142] The condition warning model is constructed based on the correlation between the detection data of the wind turbine's energy meters at the target measuring point. The correlation is obtained by linearly fitting the detection data of any one of multiple wind turbines' energy meters to the target measuring point, obtaining the fitting coefficients, and then clustering the fitting coefficients for each wind turbine. The specific training process of the condition warning model can be found in the above embodiment and will not be repeated here.

[0143] In practice, after acquiring the detection data of n energy meters of the wind turbine to be tested for the target measuring points, the state warning model corresponding to the target measuring points is first determined based on the target measuring points. For example, if the target measuring points are phase A current, phase B current, phase C current, phase A voltage, phase B voltage, and phase C voltage, then the state warning model is specifically the state warning model corresponding to the phase A current measuring point, phase B current measuring point, phase C current measuring point, phase A voltage measuring point, phase B voltage measuring point, and phase C voltage and current measuring point, respectively.

[0144] Based on this, in one possible implementation, this application provides a specific implementation method for inputting the detection data of n energy meters of the wind turbine to be tested at the target measuring point into a state early warning model to obtain the state indicators of the energy meters, including:

[0145] The detection data of n energy meters of the wind turbine to be tested for each target measuring point are input into the corresponding state early warning model to obtain the state indicators of the energy meters for each target measuring point.

[0146] That is, after determining the target measuring points, the corresponding state warning models for each target measuring point are determined. Then, the detection data from n energy meters of the wind turbine for each target measuring point are input into the corresponding state warning models to obtain the state indicators of the energy meters for each target measuring point. For example, the retrieval data from n energy meters of the wind turbine for the A-phase current measuring point are input into the state warning model corresponding to the A-phase current measuring point to obtain the state indicators of the energy meters for the A-phase current measuring point. The state indicators of the A-phase current measuring point can be understood as the state indicators of the wind turbine under test under the A-phase current measuring point, specifically the state indicators of the n energy meters of the wind turbine under test obtained based on the detection of the A-phase current measuring point.

[0147] In addition, in one or more embodiments, when the state warning model is specifically a model for obtaining the state index of the target energy meter of the wind turbine to be tested, the detection data of the target energy meter of the wind turbine to be tested for each target measuring point are respectively input into the corresponding state warning model, and the obtained state index of the target energy meter for each target measuring point is the state index of the target energy meter in the wind turbine to be tested obtained based on each target measuring point.

[0148] For example, if the target measuring point is the phase A current measuring point and the target energy meter is energy meter 1#, the detection data of energy meter 1# of the wind turbine to be tested for the phase A current measuring point is input into the state early warning model corresponding to the phase A current measuring point to obtain the state index of energy meter 1# of the wind turbine to be tested based on the phase A current measuring point.

[0149] In one or more embodiments, when the state warning model is specifically a model for obtaining the state indicators of other energy meters of the wind turbine to be tested, the detection data of n energy meters of the wind turbine to be tested for each target measuring point are respectively input into the corresponding state warning model, and the obtained state indicators of other energy meters for each target measuring point are the state indicators of other energy meters in the wind turbine to be tested obtained based on each target measuring point.

[0150] For example, if the target measuring point is the A-phase current measuring point and the other energy meters are energy meters 2#, 3#, and 4#, the detection data of the other energy meters of the wind turbine to be tested for the A-phase current measuring point are input into the status warning model corresponding to the A-phase current measuring point. This yields the status indicators of energy meters 2#, 3#, and 4# of the wind turbine to be tested, based on the A-phase current measuring point. Understandably, these status indicators can indicate whether any of energy meters 2#, 3#, and 4# exhibit abnormalities.

[0151] As an optional example, the detection status level of the energy meter of the wind turbine under test is determined based on the status indicators of each target measuring point. If the detection status level of any measuring point of the wind turbine is abnormal, it is determined that the energy meter of the wind turbine under test is abnormal, and an early warning is issued. The early warning describes the abnormality of the energy meter at a certain measuring point and provides the detection status level.

[0152] The maximum value among the status indicators of each target measuring point is taken as the detection status level of the energy meter of the wind turbine unit under test.

[0153] As an optional example, the status indicator includes multiple status levels, with different status levels representing the degree of abnormality of the energy meter. When the status level corresponding to the status indicator of the energy meter of the wind turbine under test is greater than the level threshold, the system outputs that the energy meter of the wind turbine under test has an abnormality and / or an abnormal measurement point. The abnormal measurement point can be the target measurement point corresponding to the maximum value.

[0154] It is understood that the embodiments of this application do not limit the level threshold.

[0155] Based on the content of S301-S302, this application provides a method, apparatus, and device for detecting the status of electricity meters in wind turbine generator sets. Theoretically, when multiple electricity meters in the same wind turbine generator set detect the same measuring point, the detection results obtained are basically consistent. Therefore, the target measuring point is detected using the electricity meters of the wind turbine generator set beforehand, and a status warning model is constructed using the correlation between the detection data of the electricity meters at the target measuring point. The status warning model can input the detection data of the electricity meters at the target measuring point and output the status indicators of the electricity meters. Based on this, the detection data of n electricity meters of the wind turbine generator set to be tested for the target measuring point are obtained. Here, n is a positive integer. By inputting the detection data of the n electricity meters of the wind turbine generator set to be tested for the target measuring point into the constructed status warning model, the status indicators of the electricity meters can be obtained. Based on the status indicators, it can be determined whether there is an anomaly in the electricity meter. Thus, based on the status warning model and the detection data of the electricity meter at the target measuring point, it is possible to accurately detect the status of the electricity meter of the generator set and reduce the incidence of production accidents such as high and low voltage false triggering and long-term unit failure shutdown caused by large deviations in the electricity meter detection of wind turbine sets.

[0156] Based on the above, this application provides a specific implementation method for detecting the status of a wind turbine energy meter.

[0157] In practical applications, 45 wind turbine units were selected for testing, with the data sample spanning 50 days. Each wind turbine unit had four electricity meters.

[0158] Considering the real-time requirements of power grid-related data, the operating cycle of the status early warning model is set to 1 day, and the data collection cycle is also 1 day. Specifically, for each wind turbine, the data collected from the four electricity meters of the wind turbine at each measuring point within 1 day is sampled. Theoretically, the amount of data collected at each measuring point within 1 day is 12,085 records. Considering the reliability of the status early warning model, the initial data limit is set to 8,000 records.

[0159] After acquiring the detection data, data preprocessing is performed. Data preprocessing mainly includes removing empty columns, removing constant value columns, removing missing values, and determining constant values. Specifically, values ​​with the highest frequency exceeding a certain proportion of the total data volume and few unique value categories are identified as constant values. Simultaneously, empty columns (samples containing missing values) are removed. After data preprocessing, each measurement point has at least 6000 data entries.

[0160] Further, data reduction processing is performed, mainly setting reasonable ranges for current, voltage, and power. Specifically, the power range is set to [-50, 7500], the current range to [100, 2000], and the voltage range to [50, 450]. The data reduction processing for each measurement point involves no less than 4000 data entries.

[0161] Finally, the processed detection data is input into the state early warning model to predict the state indicators of each measuring point. If the amount of detection data corresponding to the current state indicator of the current measuring point exceeds 80% of the total data volume corresponding to that measuring point, then the state indicator of the current measuring point of the wind turbine is determined as the current state indicator. For example, if the current measuring point is the A-phase current measuring point, and the current state indicator is 0, and the amount of detection data corresponding to the state indicator of 0 exceeds the total amount of detection data under the A-phase current measuring point, then the state indicator of the wind turbine under the A-phase current measuring point is 0.

[0162] Additionally, see Figure 4 , Figure 4 This illustration, provided as an embodiment of the present application, shows the prediction results of the state indicators of different energy meters when the state warning models of the target energy meter and other energy meters are different. For example... Figure 4 As shown, the target measuring points include phase A current, phase B current, phase C current, phase A voltage, phase B voltage, and phase C voltage. The target energy meter is energy meter 1#, and the other energy meters are 2#, 3#, and 4#.

[0163] The state early warning model used to obtain the state indicators of electricity meter #1 is used to predict the state indicators of electricity meter #1. Similarly, the state early warning models used to obtain the state indicators of electricity meters #2, #3, and #4 are used to predict the state indicators of #2, #3, and #4.

[0164] Taking the state warning model used to obtain the state indicators of electricity meter 1# as an example, the detection data of the A-phase current measuring point detected by electricity meter 1# is input into the state warning model corresponding to the A-phase current measuring point to obtain the state indicators of electricity meter 1# under the A-phase current measuring point. Similarly, the state indicators of electricity meter 1# under the B-phase current measuring point, the C-phase current measuring point, the A-phase voltage measuring point, the B-phase voltage measuring point, and the C-phase voltage measuring point can also be obtained. The maximum value (MAX) among the six state indicators is taken as the detection state level of electricity meter 1#.

[0165] Similarly, the detection status levels of electricity meters 2#, 3#, and 4# can also be obtained. Finally, the detection status level of electricity meter 1#, as well as the detection status levels of electricity meters 2#, 3#, and 4#, are used as the early warning output for abnormal electricity meter detection of this wind turbine unit.

[0166] As an example, see Table 5, which is a schematic table of test results for the state warning model. After actual testing, it was found that the A-phase current detected by the power meter #1 of the three wind turbines (55#, 57#, and 24#) had a large deviation. The test results were issued as an early warning to facilitate subsequent handling by the staff.

[0167] Table 5. Test Results of the State Early Warning Model

[0168] Fan number Warning time Abnormal characteristics 55# 20211013000000 The current detection deviation of phase A of the #1 electricity meter is large. 57# 20211013000000 The current detection deviation of phase A of the #1 electricity meter is large. 24# 20211013000000 The current detection deviation of phase A of the #1 electricity meter is large.

[0169] Based on the training method for a state warning model provided in the above-described method embodiments, this application also provides a training device for a state warning model. The training device for the state warning model will be described below with reference to the accompanying drawings.

[0170] See Figure 5 As shown in the figure, this is a schematic diagram of the structure of a training device for a state early warning model provided in an embodiment of this application. Figure 5 As shown, the training device for this state warning model includes:

[0171] The first acquisition unit 501 is used to acquire the detection data of p energy meters of any one of the multiple wind turbine units for the target measuring point, where p is a positive integer;

[0172] The second acquisition unit 502 is used to perform linear fitting on the detection data of any q energy meters among p energy meters to obtain the fitting coefficients corresponding to the wind turbine, where q is a positive integer and p is greater than or equal to q.

[0173] The third acquisition unit 503 is used to cluster the fitting coefficients corresponding to the multiple wind turbine units to obtain m cluster categories, where m is a positive integer.

[0174] The fourth acquisition unit 504 is used to obtain the status indicators corresponding to each of the multiple wind turbines based on the state mapping between the m cluster categories and the energy meters of the multiple wind turbines.

[0175] Training unit 505 is used to train a state early warning model based on the detection data of the multiple wind turbine units targeting the target measurement points and the corresponding state indicators of the multiple wind turbine units.

[0176] In one possible implementation, the second acquisition unit 502 is specifically used for:

[0177] The target energy meter of the wind turbine is linearly fitted with the detection data of each other energy meter for the target measurement point to obtain the fitting coefficient corresponding to the wind turbine; the target energy meter is the preset energy meter of the wind turbine, and the other energy meters are the energy meters of the wind turbine other than the target energy meter.

[0178] In one possible implementation, the state warning model is specifically used to obtain the state indicators of the target energy meter of the wind turbine to be tested.

[0179] In one possible implementation, the second acquisition unit 502 is specifically used for:

[0180] The other energy meters of the wind turbine are used to perform pairwise linear fitting of the detection data of the target measuring point to obtain the fitting coefficients corresponding to the wind turbine; the other energy meters are the energy meters of the wind turbine other than the target energy meter, and the target energy meter is the preset energy meter of the wind turbine.

[0181] In one possible implementation, the status warning model is specifically used to obtain the status indicators of other energy meters of the wind turbine to be tested.

[0182] Based on the wind turbine energy meter status detection method provided in the above method embodiments, this application also provides a wind turbine energy meter status detection device, which will be described below with reference to the accompanying drawings.

[0183] See Figure 6 As shown in the figure, this is a structural schematic diagram of a wind turbine energy meter status detection device provided in an embodiment of this application. Figure 6 As shown, the wind turbine energy meter status detection device includes:

[0184] The first acquisition unit 601 is used to acquire the detection data of n energy meters of the wind turbine generator set under test for the target measuring point, where n is a positive integer;

[0185] The second acquisition unit 602 is used to input the detection data of n energy meters of the wind turbine to be tested for the target measuring point into the status early warning model to obtain the status index of the energy meter. The status index is used to indicate whether the energy meter is abnormal. The status early warning model is constructed based on the correlation between the detection data of the energy meters of the wind turbine for the target measuring point.

[0186] In one possible implementation, the correlation between the detection data of the energy meters of the wind turbine for the target measuring point is obtained by linearly fitting the detection data of the energy meters of any one of the multiple wind turbines for the target measuring point to obtain the fitting coefficient, and then clustering the fitting coefficients corresponding to each of the multiple wind turbines according to the cluster category.

[0187] In one possible implementation, the target measurement points include the first phase current, the second phase current, the third phase current, the first phase voltage, the second phase voltage, and the third phase voltage.

[0188] In one possible implementation, the second acquisition unit 602 is specifically used for:

[0189] The detection data of n energy meters of the wind turbine under test for each target measuring point are respectively input into the corresponding state early warning model to obtain the state index of the energy meter for each target measuring point.

[0190] The selection unit is used to take the maximum value from the status indicators of the energy meter for each target measuring point, and use it as the detection status level of the energy meter of the wind turbine to be tested.

[0191] In one possible implementation, the status indicator includes multiple status levels, with different status levels representing the degree of abnormality of the electricity meter; the device further includes:

[0192] The output unit is used to output that the energy meter of the wind turbine under test has an abnormality and / or an abnormal measurement point when the status level corresponding to the status index of the energy meter of the wind turbine under test is greater than the level threshold.

[0193] In addition, this application also provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the wind turbine energy meter status detection method described in any of the above embodiments, or the status early warning model training method described in any of the above embodiments.

[0194] In addition, embodiments of this application also provide a computing device, including:

[0195] At least one processor;

[0196] At least one memory stores a computer program, which, when executed by the at least one processor, implements the wind turbine energy meter status detection method or the status early warning model training method described in any of the above embodiments.

[0197] This application provides a wind turbine energy meter status detection device. Theoretically, when multiple energy meters in the same wind turbine detect the same measuring point, the detection results should be essentially consistent. Based on this, the target measuring point is pre-detected using the wind turbine's energy meters, and a status warning model is constructed using the correlation between the detection data from the energy meters at the target measuring point. The status warning model can input the energy meter detection data for the target measuring point and output the energy meter's status indicators. Based on this, detection data from n energy meters of the wind turbine to be tested for the target measuring point are obtained, where n is a positive integer. By inputting the detection data from the n energy meters of the wind turbine to be tested for the target measuring point into the constructed status warning model, the energy meter's status indicators can be obtained. Based on the status indicators, it can be determined whether the energy meter is abnormal. Thus, based on the status warning model and the energy meter detection data for the target measuring point, accurate detection of the generator set's energy meter status can be achieved.

[0198] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems or apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and relevant parts can be referred to the method section.

[0199] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0200] It should also be noted that, in this document, relational terms such as "first" and "second" are used only 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 a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. 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.

[0201] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0202] 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 detecting the status of a wind turbine energy meter, characterized in that, The method includes: Obtain the test data of n energy meters of the wind turbine generator set under test for the target measuring point, where n is a positive integer; The detection data of n energy meters of the wind turbine to be tested at the target measuring point are input into the status early warning model to obtain the status index of the energy meter. The status index is used to indicate whether the energy meter is abnormal. The state warning model is constructed based on the correlation between the detection data of the energy meters of the wind turbines at the target measuring points. The correlation between the detection data of the energy meters of the wind turbines at the target measuring points is obtained by linearly fitting the detection data of the energy meters of each wind turbine in multiple wind turbines to the target measuring points, and then clustering the corresponding fitting coefficients of the multiple wind turbines according to the clustering categories.

2. The method according to claim 1, characterized in that, The target measurement points include the first phase current, the second phase current, the third phase current, the first phase voltage, the second phase voltage, and the third phase voltage.

3. The method according to claim 2, characterized in that, The step involves inputting the detection data from n energy meters of the wind turbine under test at the target measuring point into the status early warning model to obtain the status indicators of the energy meters, including: The detection data of n energy meters of the wind turbine under test for each target measuring point are respectively input into the corresponding state early warning model to obtain the state index of the energy meter for each target measuring point. The maximum value among the status indicators of the energy meter for each target measuring point is taken as the detection status level of the energy meter of the wind turbine to be tested.

4. The method according to claim 1, characterized in that, The status indicators include multiple status levels, with different status levels representing the degree of abnormality of the energy meter; the method also includes: When the status level of the energy meter of the wind turbine under test is greater than the level threshold, the system outputs that the energy meter of the wind turbine under test has an abnormality and / or an abnormal measurement point.

5. A training method for a state early warning model, characterized in that, The method includes: For any one of multiple wind turbine units, obtain the detection data of p energy meters of the wind turbine unit for the target measuring point, where p is a positive integer; Linear fitting is performed on the detection data of any q energy meters out of p energy meters to obtain the fitting coefficients corresponding to the wind turbine, where q is a positive integer greater than or equal to 2, and p is greater than or equal to q. The fitting coefficients corresponding to each of the multiple wind turbine units are clustered to obtain m cluster categories, where m is a positive integer; Based on the mapping between the m cluster categories and the state of the energy meters of the multiple wind turbine units, the state indicators corresponding to each of the multiple wind turbine units are obtained. A state early warning model is trained based on the detection data of the target measurement points from the multiple wind turbine units and the corresponding state indicators of the multiple wind turbine units.

6. The method according to claim 5, characterized in that, The step of linearly fitting the detection data from any q energy meters out of p energy meters to obtain the fitting coefficients corresponding to the wind turbine includes: The target energy meter of the wind turbine is linearly fitted with the detection data of each other energy meter for the target measurement point to obtain the fitting coefficient corresponding to the wind turbine; the target energy meter is the preset energy meter of the wind turbine, and the other energy meters are the energy meters of the wind turbine other than the target energy meter.

7. The method according to claim 6, characterized in that, The aforementioned state early warning model is specifically used to obtain the state indicators of the target energy meter of the wind turbine unit to be tested.

8. The method according to claim 5, characterized in that, The step of linearly fitting the detection data from any q energy meters out of p energy meters to obtain the fitting coefficients corresponding to the wind turbine includes: The detection data of the other energy meters of the wind turbine unit, excluding the target energy meter, are linearly fitted pairwise to the target measuring point to obtain the fitting coefficients corresponding to the wind turbine unit. The target energy meter is the preset energy meter of the wind turbine unit.

9. The method according to claim 8, characterized in that, The aforementioned status early warning model is specifically used to obtain the status indicators of other energy meters of the wind turbine unit under test, excluding the target energy meter.

10. A wind turbine energy meter status detection device, characterized in that, The device includes: The first acquisition unit is used to acquire the detection data of n energy meters of the wind turbine generator set under test for the target measuring point, where n is a positive integer; The second acquisition unit is used to input the detection data of n energy meters of the wind turbine to be tested at the target measuring point into the status early warning model to obtain the status index of the energy meter, and the status index is used to indicate whether the energy meter is abnormal. The state warning model is constructed based on the correlation between the detection data of the energy meters of the wind turbines at the target measuring points. The correlation between the detection data of the energy meters of the wind turbines at the target measuring points is obtained by linearly fitting the detection data of the energy meters of each wind turbine in multiple wind turbines to the target measuring points, and then clustering the corresponding fitting coefficients of the multiple wind turbines according to the clustering categories.

11. A training device for a state early warning model, characterized in that, The device includes: The first acquisition unit is used to acquire the detection data of p energy meters of any one of the multiple wind turbine units for the target measuring point, where p is a positive integer; The second acquisition unit is used to perform linear fitting on the detection data of any q energy meters among p energy meters to obtain the fitting coefficients corresponding to the wind turbine, where q is a positive integer greater than or equal to 2, and p is greater than or equal to q. The third acquisition unit is used to cluster the fitting coefficients corresponding to each of the multiple wind turbine units to obtain m cluster categories, where m is a positive integer; The fourth acquisition unit is used to obtain the status indicators corresponding to each of the multiple wind turbines based on the state mapping between the m cluster categories and the energy meters of the multiple wind turbines. The training unit is used to train a state early warning model based on the detection data of the multiple wind turbines at the target measurement points and the corresponding state indicators of the multiple wind turbines.

12. A computer-readable storage medium storing a computer program, wherein, When the computer program is executed by the processor, it implements the wind turbine energy meter status detection method according to any one of claims 1-4, or the status early warning model training method according to any one of claims 5-9.

13. A computing device, comprising: At least one processor; At least one memory stores a computer program that, when executed by the at least one processor, implements the wind turbine energy meter status detection method according to any one of claims 1-4, or the status early warning model training method according to any one of claims 5-9.

Images