Methods and devices for monitoring the condition of wind turbine blades

Abstract

This disclosure relates to a method and apparatus for monitoring the condition of wind turbine blades. The method includes: predicting index data of blades from multiple wind turbines using a data-driven model; grouping the blades of the multiple wind turbines based on the predicted index data; and determining that any blade or its onboard sensor for detecting index data is abnormal if the difference between the actual measured value of the index data of any blade and the actual measured value of the index data of the center blade within the group to which that blade belongs exceeds a predetermined threshold. By employing this disclosure, effective monitoring of blade condition can be achieved while reducing the accuracy requirements of the data-driven model, avoiding power generation losses caused by directly reducing load based on sensor results and ignoring the causes of load increases, and enabling early detection of blade anomalies to prevent the problem from developing into an uncontrollable situation.

Description

Technical Field

[0001] This disclosure relates to the field of wind power generation technology, and more specifically, to a method and device for monitoring the blade condition of a wind turbine generator set. Background Technology

[0002] The diameter of rotors in the current speculative market is rapidly increasing, and the larger blades lead to increased blade loads. This, coupled with intensified market competition, necessitates strict cost control and a greater focus on blade safety. Some manufacturers are currently installing load sensors on the blades to provide a basis for load reduction and avoid blade problems caused by excessive loads. However, this process cannot clearly determine whether the excessive load is caused by normal or abnormal factors, making it difficult to detect blade anomalies in a timely manner. With the rapid development of artificial intelligence, machine learning models are increasingly being used in the monitoring of wind turbine generators. However, due to the complexity of wind power project sites and the limited amount of available fault or anomaly data, when common data-driven predictive models are directly applied to wind turbine generator monitoring, it is difficult to achieve good predictive accuracy, and high costs are required to correct the model, making implementation difficult and with high uncertainty. Summary of the Invention

[0003] To address the aforementioned issues, this disclosure proposes a method and device for monitoring the blade condition of a wind turbine generator, a computing system, and a computer-readable storage medium.

[0004] According to one aspect of this disclosure, a method for monitoring the blade condition of a wind turbine generator set is provided. The method includes: predicting index data of blades from multiple wind turbine generator sets using a data-driven model; grouping the blades of the multiple wind turbine generator sets based on the predicted values ​​of the index data; and determining that any blade or a sensor installed on it for detecting index data is abnormal if the difference between the actual measured value of the index data of any blade and the actual measured value of the index data of the center blade within the group to which the blade belongs exceeds a predetermined threshold. Specifically, among all blades within the group to which the blade belongs, the predicted value of the index data of the center blade is closest to the average value of the predicted values ​​of the index data of the blades within the group to which the blade belongs, and the difference between the actual measured value of the index data of the center blade and the predicted value of the index data of the center blade does not exceed the confidence interval width of the data-driven model.

[0005] Optionally, the index data includes load, vibration displacement, or vibration acceleration.

[0006] Optionally, the predetermined threshold is the sum of the maximum value of the difference between the predicted value of the index data of each leaf in the group to which any leaf belongs and the predicted value of the central leaf, and the confidence interval width of the data-driven model.

[0007] Optionally, the leaf index data includes index data of the leaf root and index data of the leaf.

[0008] Optionally, the wind turbine blade condition monitoring method further includes: for any blade, in response to determining that the difference between the actual measured value of only one indicator data among the indicator data at the blade root and the indicator data in the blade and the actual measured value of the corresponding indicator data of the central blade in the group to which the blade belongs exceeds a predetermined threshold, determining that the sensor installed on the blade corresponding to the one indicator data is abnormal; for any blade, in response to determining that the difference between the actual measured values ​​of the indicator data at the blade root and the indicator data in the blade and the actual measured value of the corresponding indicator data of the central blade in the group to which the blade belongs exceeds a predetermined threshold, determining that the blade itself is abnormal.

[0009] Optionally, the inputs to the data-driven model include on-site environmental characteristics and wind turbine operating status data.

[0010] Optionally, the step of grouping the blades of the multiple wind turbine generator sets based on the predicted values ​​of the indicator data includes: using a clustering algorithm to group blades with similar predicted values ​​of the indicator data into the same group.

[0011] According to another aspect of this disclosure, a blade condition monitoring device for a wind turbine generator set is provided. The blade condition monitoring device includes: a prediction unit configured to predict index data of blades of a plurality of wind turbine generator sets using a data-driven model; a grouping unit configured to group the blades of the plurality of wind turbine generator sets based on the predicted values ​​of the index data; and an anomaly detection unit configured to determine that any blade or a sensor installed on it for detecting index data is abnormal in response to the difference between the actual measured value of the index data of any blade and the actual measured value of the index data of the center blade in the group to which the blade belongs exceeding a predetermined threshold. The anomaly is further specified in the following context: among all blades in the group to which the blade belongs, the predicted value of the index data of the center blade is closest to the average value of the predicted values ​​of the index data of the blades in the group to which the blade belongs, and the difference between the actual measured value of the index data of the center blade and the predicted value of the index data of the center blade does not exceed the confidence interval width of the data-driven model.

[0012] Optionally, the index data includes load, vibration displacement, or vibration acceleration.

[0013] Optionally, the predetermined threshold is the sum of the maximum value of the difference between the predicted value of the index data of each leaf in the group to which any leaf belongs and the predicted value of the central leaf, and the confidence interval width of the data-driven model.

[0014] Optionally, the leaf index data includes index data of the leaf root and index data of the leaf.

[0015] Optionally, the anomaly detection unit is further configured to: for any one blade, in response to determining that the difference between the actual measured value of only one indicator data among the indicator data of the leaf root and the indicator data of the leaf and the actual measured value of the corresponding indicator data of the central blade in the group to which the one blade belongs exceeds a predetermined threshold, determine that the sensor installed on the one blade corresponding to the one indicator data is abnormal; for any one blade, in response to determining that the difference between the actual measured values ​​of the indicator data of the leaf root and the indicator data of the leaf and the actual measured value of the corresponding indicator data of the central blade in the group to which the one blade belongs exceeds a predetermined threshold, determine that the one blade itself is abnormal.

[0016] Optionally, the inputs to the data-driven model include on-site environmental characteristics and wind turbine operating status data.

[0017] Optionally, the grouping unit is configured to: use a clustering algorithm to group leaves with similar predicted values ​​of indicator data into the same group.

[0018] According to another aspect of this disclosure, a computing system is provided that includes at least one computing device and at least one storage device for storing instructions, wherein the instructions, when executed by the at least one computing device, cause the at least one computing device to perform the blade condition monitoring method for a wind turbine generator as described above.

[0019] According to another aspect of this disclosure, a computer-readable storage medium is provided that stores instructions, wherein when the instructions are executed by at least one computing device, the at least one computing device causes the at least one computing device to perform the blade condition monitoring method for a wind turbine generator as described above.

[0020] By adopting this disclosure, it is possible to effectively monitor the blade condition while reducing the accuracy requirements of the data-driven model. It can avoid power generation loss caused by directly reducing load based on sensor results and ignoring the reasons for load increase. Furthermore, it can detect blade anomalies early and prevent blade anomalies from developing to an uncontrollable level. Attached Figure Description

[0021] The above and / or other objects and advantages of this disclosure will become clearer from the following description of embodiments in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart illustrating a method for monitoring the blade condition of a wind turbine generator according to an exemplary embodiment of the present disclosure; Figure 2 This is a flowchart illustrating an example of a blade condition monitoring method for a wind turbine generator according to the present disclosure; Figure 3 This is a block diagram illustrating a blade condition monitoring device for a wind turbine generator set according to an exemplary embodiment of the present disclosure; Figure 4 This is a block diagram illustrating a computing system including at least one computing device and at least one storage device of storage instructions according to an exemplary embodiment of the present disclosure. Detailed Implementation

[0022] The following description, in conjunction with the accompanying drawings, provides specific embodiments to aid the reader in gaining a comprehensive understanding of the methods, apparatus, and / or systems described herein. However, upon understanding this disclosure, various changes, modifications, and equivalents of the methods, apparatus, and / or systems described herein will become apparent. For example, the order of operations described herein is merely illustrative and is not limited to those orders set forth herein, but may be altered as will become clear upon understanding this disclosure, except for operations that must occur in a specific order. Furthermore, for clarity and conciseness, descriptions of features known in the art may be omitted.

[0023] This disclosure proposes a method and device for monitoring the blade condition of wind turbines. By grouping blades and identifying outliers, it indirectly uses the prediction results of a data-driven model to provide early warning of blade anomalies. In the proposed method, firstly, a data-driven model is used to predict blade performance data (including but not limited to load, vibration displacement, and vibration acceleration, and any field-collected information reflecting blade condition can be incorporated as blade performance data) for all wind turbine blades in the entire wind farm. Then, the blades in the entire wind farm are grouped based on the predicted performance data. Finally, the presence of outliers within the same group is determined to identify any blade anomalies, and an early warning of turbine anomalies is issued to maintenance personnel if an anomaly is detected. The proposed method for monitoring the blade condition of wind turbines is logically simple and efficient. Instead of directly using the results of the data-driven model as the indicator for blade anomalies, it introduces a blade grouping method, which enables effective real-time monitoring of blade condition while reducing the accuracy requirements of the data-driven model, allowing for faster practical application of the data-driven model in the field.

[0024] Figure 1This is a flowchart illustrating a method for monitoring the blade condition of a wind turbine generator according to an exemplary embodiment of the present disclosure.

[0025] like Figure 1 As shown, in step S101, the indicator data of multiple wind turbine blades are predicted using a data-driven model. In this example, the blade indicator data includes, but is not limited to, load, vibration displacement, or vibration acceleration. In this example, the data-driven model may include, but is not limited to, linear models, classification models, or deep learning models. The advantage of a data-driven model is its ability to handle large amounts of data, automatically learn the characteristics and patterns of the data, and is suitable for the analysis and prediction of complex problems. In this example, the input to the data-driven model includes on-site environmental characteristic values ​​and the operating status data of the wind turbine. For example, the input data can be derived from prototype data corresponding to the turbine model, because using historical on-site data might introduce too many uncertainties.

[0026] In step S102, the blades of multiple wind turbine generators are grouped based on the predicted values ​​of the indicator data. For example, a clustering algorithm is used to group blades with similar predicted values ​​of the indicator data into the same group. By using a clustering algorithm, single-dimensional or multi-dimensional features can be grouped similarly, ensuring that the characteristics within the same group are as similar as possible and the characteristics between groups are as different as possible, and avoiding the influence of human experience errors on the grouping results. Here, the blades include the three blades of the wind turbine generator. When grouping the blades, the number of groups can be limited or determined based on factors such as the number of wind turbine generators in the entire wind farm and the software computing power. For example, if there are only 10 wind turbine generators (including 30 blades) in the entire wind farm, the number of groups can be limited to no more than 3. If the number of groups is too large, there will be too few blades in each group, which is not conducive to the identification of outliers. On the other hand, if there are hundreds of wind turbine generators in the entire wind farm and the software computing power resources are limited, the number of groups can be limited. By introducing blade grouping, the blade condition can be effectively monitored without using the results of the data-driven model as an indicator of blade anomalies. This reduces the accuracy requirements of the data-driven model and enables faster practical application of the data-driven model in the field.

[0027] In step S103, in response to the difference between the actual measured value of the index data of any blade and the actual measured value of the index data of the central blade within the group to which any blade belongs exceeding a predetermined threshold, it is determined that any blade or the sensor installed on it for detecting index data is abnormal. Specifically, among all blades within the group to which any blade belongs, the predicted value of the index data of the central blade is closest to the average of the predicted values ​​of the index data of the blades within the group to which any blade belongs, and the difference between the actual measured value and the predicted value of the index data of the central blade does not exceed the confidence interval width of the data-driven model. In this example, since it is assumed that the actual value of the blade should be close to the predicted value, the distribution range of the actual value of any blade in each group around the actual value of the central blade should be consistent with the distribution range of the predicted value of any blade in each group around the predicted value of the central blade. This distribution range is a one-dimensional or multi-dimensional range formed by the absolute value of the maximum difference between each group of blades and the central blade, such as a linear region range, a circular region range, a spherical region range, etc. Meanwhile, considering that the deviation between the actual and predicted values ​​of the central leaf will shift the entire distribution range, but this deviation will not exceed the confidence interval width of the data-driven model's deviation between the training and test datasets, the predetermined threshold is the sum of the maximum difference between the predicted values ​​of the indicator data of each leaf within any leaf's group and the predicted value of the central leaf, and the confidence interval width of the data-driven model. In the data-driven model, the confidence interval is used to determine whether the estimation of model parameters is reliable and the reliability of the prediction results. By calculating the confidence interval, the stability and generalization ability of the model under different conditions can be evaluated. Here, different confidence levels can be selected according to actual needs. For example, the confidence level can be preset to 90%. Therefore, a 90% confidence interval for the output signal of the data-driven model can be determined, and the probability that the true value falls within this interval can be considered to be 90%. For example, based on the combined statistical calculation of the bias in the training dataset and the bias in the test dataset, the confidence interval corresponding to the 90% confidence level is [predicted value of leaf index data × (-0.08), predicted value of leaf index data × 0.08]. The width of the confidence interval is equal to the difference between the upper boundary and the lower boundary of the confidence interval. That is, the confidence interval width of the bias in the data-driven model is predicted value of leaf index data × 0.16.

[0028] In the example, the blade's performance data includes data from the blade root and data from the blade center. Here, the blade center can be understood as the region near the center point of the blade along its length. For example, for any blade, if the difference between the actual measured value of only one performance data point from the blade root and the blade center and the actual measured value of the corresponding performance data point of the central blade within the group to which that blade belongs exceeds a predetermined threshold, it is determined that the sensor corresponding to that performance data point installed on that blade is abnormal; if the difference between the actual measured values ​​of both the blade root and the blade center and the actual measured values ​​of the corresponding performance data point of the central blade within the group to which that blade belongs exceeds a predetermined threshold, it is determined that the blade itself is abnormal. By calculating whether there are outliers in the actual measured values ​​within the same group to determine whether there are blade anomalies, blade anomalies can be detected as early as possible, and anomaly warnings can be issued to maintenance personnel to prevent blade problems from worsening due to increased load.

[0029] By utilizing the blade condition monitoring method for wind turbine generators according to exemplary embodiments of the present disclosure, it is possible to effectively monitor the blade condition while reducing the accuracy requirements of the data-driven model, and avoid power generation loss caused by directly reducing load based on sensor results and ignoring the reasons for load increase.

[0030] Figure 2 This is a flowchart illustrating an example of a blade condition monitoring method for a wind turbine generator according to the present disclosure.

[0031] The blade condition monitoring method for wind turbine generators disclosed herein identifies abnormal blades based on differences between blades within the same group after comparison and similarity grouping, thereby achieving blade condition detection. Its logic is simple and efficient, and it is highly user-friendly for field implementation. In the blade condition monitoring method for wind turbine generators disclosed herein, a high similarity grouping approach is used to leverage the prediction results of a data-driven model, including determining the center blade and applying relevant loads to the center blade; the grouping characteristics of the prediction results are used as the basis for identifying anomalies in the actual results.

[0032] exist Figure 2 The example shown assumes that a root load sensor and a mid-blade load sensor are available on-site. However, it should be understood that blade performance data may also include vibration displacement or vibration acceleration, and therefore the following can also be applied to other types of data (e.g., vibration displacement or vibration acceleration) and sensors (e.g., vibration displacement sensor or vibration acceleration sensor).

[0033] like Figure 2As shown, in step S201, the blade root load value is predicted, and the blades are grouped according to the prediction results. For example, the blade root load data-driven model is used to periodically obtain the blade root load prediction results, and the blades are grouped according to the similarity principle. In the example, the input of the blade root load data-driven model may include on-site environmental feature values ​​and wind turbine generator operating status data, which may come from prototype data corresponding to the model. The similarity principle means that if the characteristics of some blades are relatively similar to those of other blades, then these blades are considered similar blades. In the example, the blade root loads of some blades are relatively similar, so these blades can be grouped together. The specific method for judging similarity can use clustering algorithms or manual partitioning methods. Clustering algorithms can group single / multi-dimensional features similarly, ensuring that the characteristics within the same group are as similar as possible and the differences between characteristics between groups are as large as possible. Manual partitioning methods are only suitable for single-dimensional features, but engineer experience can be introduced, such as using the load boundary as one of the partition boundaries. Generally, clustering algorithms are used to avoid the influence of human experience errors on the grouping results. In addition, as mentioned above, the blades include the three blades of the wind turbine generator set. When grouping the blades, the number of groups can be limited or determined based on factors such as the total number of wind turbine generator sets in the field and the software computing power.

[0034] In step S202, the center blade of each group is determined. For example, the average predicted load of each group of blades is calculated, and the predicted load closest to this average is identified as the center predicted load. The blade corresponding to the center predicted load is the candidate center blade. Additionally, it can be calculated whether the difference between the actual load and the predicted load of the candidate center blade exceeds the confidence interval width of the data-driven model (i.e., the difference between the upper and lower limits of the confidence interval). If the difference between the actual load and the predicted load of the candidate center blade exceeds the confidence interval width of the data-driven model, it can be determined that the candidate center blade is abnormal, and the blade is removed. The average predicted load of the remaining blades in the group after removing the blade is recalculated, and the predicted load closest to the new average is re-identified as the new center predicted load. The blade corresponding to the new center predicted load is the new candidate center blade. Then, it is determined whether the new candidate center blade satisfies the condition that the difference between the actual load and the predicted load of the blade does not exceed the confidence interval width of the data-driven model, until a blade satisfying the above conditions is selected as the center blade of the corresponding group.

[0035] In step S203, the difference between the actual root load of each blade in the group and the actual root load of the central blade is determined. In step S204, it is determined whether the difference between the actual root load of each blade in the group and the actual root load of the central blade exceeds a boundary value. For example, the difference between the actual root load of each blade in the group and the actual root load of the central blade is calculated. If the difference between the actual root load of each blade and the actual root load of the central blade exceeds the boundary value, the root load of the corresponding blade is considered abnormal. The boundary value is equal to the sum of the maximum value of the difference between the predicted load of each blade in the group and the predicted load of the central blade, and the confidence interval width of the data-driven model. If step S204 determines that the difference between the actual root load of each blade in the group and the actual root load of the central blade exceeds the boundary value, then in step S205, an early warning is issued for the abnormal root load of the corresponding blade.

[0036] In step S206, the blade load value is predicted, and the blades are grouped according to the prediction results. For example, the blade load data-driven model is used to periodically obtain the blade load prediction results, and the blades are grouped according to the similarity principle. In the example, the input of the blade load data-driven model may include on-site environmental feature values ​​and wind turbine generator operating status data, which may come from prototype data corresponding to the turbine model. The similarity principle means that if the characteristics of some blades are similar to those of other blades, then these blades are considered similar blades. In the example, the blade loads of some blades are similar, so these blades can be grouped together. The specific method for judging similarity can use clustering algorithms or manual partitioning methods. Clustering algorithms can group single / multi-dimensional features similarly, ensuring that the characteristics within the same group are as similar as possible and the differences between characteristics between groups are as large as possible. Manual partitioning methods are only suitable for single-dimensional features, but engineer experience can be introduced, such as using the load boundary as one of the partition boundaries. Generally, clustering algorithms are used to avoid the influence of human experience errors on the grouping results. In addition, as mentioned above, the blades include the three blades of the wind turbine generator set. When grouping the blades, the number of groups can be limited or determined based on factors such as the total number of wind turbine generator sets in the field and the software computing power.

[0037] In step S207, the center blade of each group is determined. For example, the average predicted load of each group of blades is calculated, and the predicted load closest to this average is identified as the center predicted load. The blade corresponding to the center predicted load is identified as a candidate center blade. Additionally, it can be calculated whether the difference between the actual load and the predicted load of the candidate center blade exceeds the confidence interval width of the data-driven model (i.e., the difference between the upper and lower limits of the confidence interval). If the difference between the actual load and the predicted load of the candidate center blade exceeds the confidence interval width of the data-driven model, it can be determined that the candidate center blade is abnormal, and the blade is removed. The average predicted load of the remaining blades in the group after removing the removed blade is recalculated, and the predicted load closest to the new average is identified as the new center predicted load. The blade corresponding to the new center predicted load is identified as a new candidate center blade. Then, it is determined whether the new candidate center blade satisfies the condition that the difference between the actual load and the predicted load of the blade does not exceed the confidence interval width of the data-driven model, until a blade satisfying the above conditions is selected as the center blade of the corresponding group.

[0038] In step S208, the difference between the actual load in the blades within each group and the actual load in the center blade is determined. In step S209, it is determined whether the difference between the actual load in the blades within each group and the actual load in the center blade exceeds a boundary value. For example, the difference between the actual load in the blades of each blade within the group and the actual load in the center blade is calculated. If the difference between the actual load in the blades of each blade and the actual load in the center blade exceeds the boundary value, it is considered that the load in the corresponding blade is abnormal. The boundary value is equal to the sum of the maximum value among the differences between the predicted load of each blade in each group and the predicted load of the center blade, and the confidence interval width of the data-driven model. If it is determined in step S209 that the difference between the actual load in the blades within each group and the actual load in the center blade exceeds the boundary value, an early warning is issued in step S205 for the abnormal load condition of the corresponding blade.

[0039] Furthermore, for any given blade, in response to the determination that the difference between the actual measured value of only one of the indicator data (e.g., load, vibration displacement, or vibration acceleration) at the blade root and the actual measured value of the corresponding indicator data of the central blade within the group to which any given blade belongs exceeds a predetermined threshold, it is determined that the sensor corresponding to the stated indicator data installed on any given blade is abnormal. Similarly, in response to the determination that the difference between the actual measured values ​​of both the indicator data at the blade root and the indicator data within the blade and the actual measured value of the corresponding indicator data of the central blade within the group to which any given blade belongs exceeds a predetermined threshold, it is determined that the given blade itself is abnormal. For example, in this example, if both the load at the blade root and the load at the blade's mid-section are abnormal, it can be determined that the blade itself is abnormal; if only one of the loads at the blade's mid-section and the load at the blade's root is abnormal, it may be that the load sensor installed at the corresponding location (blade root or blade mid-section) of the blade is faulty. Furthermore, if it has been determined that the load sensor on the blade is not faulty, steps S201 to S204 can be used alone to determine whether the root load of the corresponding blade is in an abnormal state and thereby confirm whether the blade itself is abnormal. Alternatively, steps S206 to S209 can be used alone to determine whether the middle load of the corresponding blade is in an abnormal state and thereby confirm whether the blade itself is abnormal. Alternatively, all steps can be performed to more accurately determine whether the blade is in an abnormal state and the specific location of the load abnormality.

[0040] By using the blade condition monitoring method of the wind turbine generator according to this disclosure, it is possible to effectively monitor the blade condition while reducing the accuracy requirements of the data-driven model, and avoid power generation loss caused by directly reducing load based on sensor results and ignoring the reasons for load increase.

[0041] Figure 3 This is a block diagram illustrating a blade condition monitoring device for a wind turbine generator set according to an exemplary embodiment of the present disclosure.

[0042] like Figure 3As shown, a wind turbine blade condition monitoring device 300 according to an exemplary embodiment of the present disclosure includes: a prediction unit 301 configured to predict index data of multiple wind turbine blades using a data-driven model; a grouping unit 302 configured to group the multiple wind turbine blades based on the predicted values ​​of the index data; and an anomaly detection unit 303 configured to determine that any blade or a sensor installed on it for detecting index data is abnormal in response to the difference between the actual measured value of the index data of any blade and the actual measured value of the index data of the center blade in the group to which the blade belongs exceeding a predetermined threshold, wherein, among all blades in the group to which the blade belongs, the predicted value of the index data of the center blade is closest to the average value of the predicted values ​​of the index data of the blades in the group to which the blade belongs, and the difference between the actual measured value of the index data of the center blade and the predicted value of the index data of the center blade does not exceed the confidence interval width of the data-driven model.

[0043] In the example, the metrics data include load, vibration displacement, or vibration acceleration.

[0044] In the example, the predetermined threshold is the sum of the maximum difference between the predicted values ​​of the index data of each leaf in the group to which any leaf belongs and the predicted value of the central leaf, and the confidence interval width of the data-driven model.

[0045] In the example, the leaf index data includes index data from the leaf root and index data from within the leaf.

[0046] In the example, the anomaly detection unit 303 is further configured to: for any blade, in response to determining that the difference between the actual measured value of only one indicator data among the indicator data of the leaf root and the indicator data of the leaf and the actual measured value of the corresponding indicator data of the central blade in the group to which any blade belongs exceeds a predetermined threshold, determine that the sensor installed on any blade corresponding to the one indicator data is abnormal; for any blade, in response to determining that the difference between the actual measured values ​​of the indicator data of the leaf root and the indicator data of the leaf and the actual measured value of the corresponding indicator data of the central blade in the group to which any blade belongs exceeds a predetermined threshold, determine that any blade itself is abnormal.

[0047] In the example, the inputs to the data-driven model include on-site environmental characteristics and the operating status data of the wind turbine generator.

[0048] In the example, grouping unit 302 is configured to use a clustering algorithm to group leaves with similar predicted values ​​of index data into the same group.

[0049] The above combination Figures 1 to 2 The specific operations shown are respectively by Figure 3The corresponding unit in the blade condition monitoring device 300 of the wind turbine generator set shown is responsible for performing this operation. The specific operational details will not be elaborated here.

[0050] By utilizing the blade condition monitoring device for a wind turbine generator set according to an exemplary embodiment of the present disclosure, it is possible to effectively monitor the blade condition while reducing the accuracy requirements of the data-driven model, and avoid power generation loss caused by directly reducing load based on sensor results and ignoring the reasons for load increase.

[0051] Figure 4 This is a block diagram illustrating a computing system including at least one computing device and at least one storage device of storage instructions according to an exemplary embodiment of the present disclosure.

[0052] like Figure 4 As shown, the computing system 400 provided according to an exemplary embodiment of the present invention includes a computing device 401 and a storage device 402. The storage device 402 stores computer-executable instructions. When the computer-executable instructions are executed by the computing device 401, the blade condition monitoring method of the wind turbine generator set described in any of the foregoing embodiments is executed.

[0053] The computing device 401 can be deployed in a server or client, or on a node device in a distributed network environment. Furthermore, the computing device 401 can be a PC, tablet, personal digital assistant, smartphone, web application, or other device capable of executing the aforementioned set of instructions. Here, the computing device is not necessarily a single computing device; it can be any collection of devices or circuits capable of executing the aforementioned instructions (or instruction sets) individually or in combination. The computing device can also be part of an integrated control system or system manager, or can be configured to interconnect with a portable electronic device locally or remotely (e.g., via wireless transmission). In the computing device, the processor includes a central processing unit (CPU), a graphics processing unit (GPU), a programmable logic device, a dedicated processor system, a microcontroller, or a microprocessor. By way of example and not limitation, the processor also includes analog processors, digital processors, microprocessors, multi-core processors, processor arrays, network processors, etc.

[0054] According to another aspect of this disclosure, a computer-readable storage medium is provided that stores instructions, which, when executed by at least one computing device, cause the at least one computing device to perform the blade condition monitoring method for a wind turbine generator described in any of the foregoing embodiments. The computer-readable storage medium includes magnetic media such as floppy disks and magnetic tapes, optical media (including optical disc (CD) ROMs and DVD ROMs), magneto-optical media such as floppy discs, hardware devices such as ROMs and RAMs designed for storing and executing program commands, and flash memory. The instructions may include language code executable by a computer using an interpreter and machine language code generated by a compiler.

[0055] By adopting this disclosure, it is possible to effectively monitor the blade condition while reducing the accuracy requirements of the data-driven model. It can avoid power generation loss caused by directly reducing load based on sensor results and ignoring the reasons for load increase. Furthermore, it can detect blade anomalies early and prevent blade anomalies from developing to an uncontrollable level.

[0056] The processes, methods, or algorithms disclosed herein can be transmitted to, or implemented by, a processing device, controller, or computer, which may include any existing programmable electronic control unit or a dedicated electronic control unit. Similarly, the processes, methods, or algorithms can be stored in various forms as data and instructions executable by a controller or computer, including but not limited to information permanently stored on non-writable storage media (such as ROM devices) and information variablely stored on writable storage media (such as floppy disks, magnetic tapes, CDs, RAM devices, and other magnetic and optical media). The processes, methods, or algorithms can also be implemented in a software executable object. Optionally, the processes, methods, or algorithms can be implemented wholly or partially using suitable hardware components (such as ASICs, FPGAs, state machines, controllers, or other hardware components or devices) or a combination of hardware components, software components, and firmware components.

[0057] Although this disclosure includes specific examples, it will be apparent to those skilled in the art that various changes in form and detail may be made to these examples without departing from the spirit and scope of the claims and their equivalents. The examples described herein are to be considered merely for descriptive purposes and not for limiting purposes. The description of features or aspects in each example is to be considered applicable to similar features or aspects in other examples. Suitable results may be obtained if the described techniques are performed in a different order, and / or if components in the described system, architecture, apparatus, or circuit are combined in a different manner and / or if components in the described system, architecture, apparatus, or circuit are replaced or supplemented with other components or their equivalents. Therefore, the scope of this disclosure is not limited by the specific embodiments but by the claims and their equivalents, and all variations within the scope of the claims and their equivalents shall be construed as included in this disclosure.

Claims

1. A method for monitoring the blade condition of a wind turbine generator set, characterized in that, The method for monitoring the blade condition of the wind turbine generator set includes: Predicting the index data of blades for multiple wind turbine generator sets using a data-driven model; Based on the predicted values ​​of the indicator data, the blades of the multiple wind turbine generator sets are grouped. If the difference between the actual measured value of the index data of any blade and the actual measured value of the index data of the central blade in the group to which the blade belongs exceeds a predetermined threshold, it is determined that there is an anomaly in the blade or the sensor installed on it for detecting the index data. In this case, among all blades in the group to which the blade belongs, the predicted value of the index data of the central blade is closest to the average value of the predicted values ​​of the index data of the blades in the group to which the blade belongs, and the difference between the actual measured value of the index data of the central blade and the predicted value of the index data of the central blade does not exceed the confidence interval width of the data-driven model.

2. The method for monitoring the blade condition of a wind turbine generator set according to claim 1, characterized in that, The index data includes load, vibration displacement, or vibration acceleration.

3. The method for monitoring the blade condition of a wind turbine generator set according to claim 1, characterized in that, The predetermined threshold is the sum of the maximum value of the difference between the predicted value of the index data of each leaf in the group to which any leaf belongs and the predicted value of the central leaf, and the confidence interval width of the data-driven model.

4. The method for monitoring the blade condition of a wind turbine generator set according to claim 1 or 2, characterized in that, The leaf index data includes index data of the leaf root and index data of the leaf.

5. The method for monitoring the blade condition of a wind turbine generator set according to claim 4, characterized in that, The method for monitoring the blade condition of the wind turbine generator set also includes: For any of the blades, in response to the fact that the difference between the actual measured value of only one of the index data in the index data of the leaf root and the index data in the leaf and the actual measured value of the corresponding index data of the central blade in the group to which the blade belongs exceeds a predetermined threshold, it is determined that the sensor installed on the blade corresponding to the index data is abnormal. For any given leaf, if the difference between the actual measured values ​​of the index data at the leaf root and the index data in the leaf and the actual measured values ​​of the corresponding index data of the central leaf in the group to which the given leaf belongs exceeds a predetermined threshold, it is determined that the given leaf itself is abnormal.

6. The method for monitoring the blade condition of a wind turbine generator set according to claim 1, characterized in that, The inputs to the data-driven model include on-site environmental characteristics and the operating status data of the wind turbine generator set.

7. The method for monitoring the blade condition of a wind turbine generator set according to claim 1, characterized in that, The step of grouping the blades of the multiple wind turbine generator sets based on the predicted values ​​of the indicator data includes: using a clustering algorithm to group blades with similar predicted values ​​of the indicator data into the same group.

8. A blade condition monitoring device for a wind turbine generator set, characterized in that, The blade condition monitoring device for the wind turbine generator set includes: The prediction unit is configured to predict the index data of the blades of multiple wind turbine generators using a data-driven model. A grouping unit is configured to group the blades of the plurality of wind turbine generator sets based on predicted values ​​of indicator data. An anomaly detection unit is configured to determine that any blade or its sensor for detecting index data is abnormal in response to a difference between the actual measured value of the index data of any blade and the actual measured value of the index data of the center blade in the group to which the blade belongs, exceeding a predetermined threshold. This occurs when the difference between the actual measured value of the index data of any blade and the actual measured value of the index data of the center blade in the group to which the blade belongs is greater than or equal to the average of the predicted values ​​of the index data of the blades in the group to which the blade belongs, and the difference between the actual measured value of the index data of the center blade and the predicted value of the index data of the center blade does not exceed the confidence interval width of the data-driven model.

9. A computing system comprising at least one computing device and at least one storage device for storing instructions, characterized in that, When the instruction is executed by the at least one computing device, it causes the at least one computing device to perform the blade condition monitoring method for wind turbine generators according to any one of claims 1 to 7.

10. A computer-readable storage medium for storing instructions, characterized in that, When the instruction is executed by at least one computing device, it causes the at least one computing device to perform the blade condition monitoring method for wind turbine generators according to any one of claims 1 to 7.

Images