State detection method and device of wind generating set and storage medium

Abstract

The embodiment of the invention discloses a state detection method and device of a wind generating set and a storage medium. The method comprises the following steps: acquiring actual fan operation data acquired by a wind generating set in a target wind field; according to M reference power distribution intervals determined based on a target power curve learning model, obtaining actual power distribution and actual confidence of actual fan operation data in each reference power distribution interval, the target power curve learning model is a model in which a prediction process of an output power prediction result selected from the N candidate power curve models reaches a preset prediction index, and the M reference power distribution intervals respectively correspond to M wind speed values; and determining the current running state of the wind generating set according to the comparison result. The technical problem that the detection result is inaccurate due to the fact that an existing wind generating set detection method cannot be dynamically matched with the operation environment is solved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of wind power generation control, and in particular, to a state detection method and device for a wind power generator set and a storage medium. Background Art

[0002] In order to achieve low carbon and environmental protection, more and more industries have begun to apply renewable energy to production and life. Among them, wind energy is a clean and renewable energy source that has received widespread attention in recent years. However, due to the complex structure and harsh operating environment of wind turbines, their failure rate is high, especially key components such as gearboxes, generators and blades are prone to failure. These failures are not only costly to repair, but also lead to long downtime. Therefore, the status detection of wind turbines has become the key to maintaining their efficient operation.

[0003] At present, the common methods for detecting the status of wind turbines include: signal-based methods, physical model methods and data-driven methods. Among them, the implementation cost of the signal-based method is relatively high, the physical model method requires the participation of experts with rich professional knowledge, and the data-driven method relies on the supervisory control and data acquisition system (SCADA) of the wind turbine. However, the existing data-driven models not only have complex data preprocessing and poor model generalization capabilities, but also because the operating conditions of the wind turbines (such as aging, sensor recalibration and replacement of key components) have changed, it will also cause the data-driven models that have been developed and put into use to gradually become invalid, and thus require frequent model updates and retraining.

[0004] That is to say, in the state detection method of the wind turbine provided by the related art, the adopted method cannot dynamically adapt to the constantly changing operating environment of the wind turbine, resulting in the problem of inaccurate state detection results of the wind turbine.

[0005] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0006] The embodiments of the present application provide a method and device for detecting the state of a wind turbine generator set, and a storage medium, so as to at least solve the technical problem that the current wind turbine generator set detection method cannot dynamically adapt to the operating environment, resulting in inaccurate detection results.

[0007] According to one aspect of an embodiment of the present application, a state detection method for a wind turbine generator set is provided, including: obtaining actual wind turbine operation data collected from a wind turbine generator set in a target wind farm; obtaining the actual power distribution and actual confidence of the actual wind turbine operation data in each reference power distribution interval according to M reference power distribution intervals determined based on a target power curve learning model, wherein the target power curve learning model is a model whose prediction process of the output power prediction result selected from N candidate power curve models reaches a preset prediction index, and the M reference power distribution intervals correspond to M wind speed values, respectively, and M is a natural number greater than 1; traversing the M reference power distribution intervals, and comparing the reference power distribution and the actual power distribution in each reference power distribution interval respectively; and determining the current operating state of the wind turbine generator set according to the comparison result.

[0008] In an optional implementation, the above-mentioned traversal of M reference power distribution intervals and respectively comparing the reference power distribution and the actual power distribution in each reference power distribution interval include: traversing M reference power distribution intervals, and respectively performing the following operations: obtaining the reference power distribution and reference confidence in the i-th reference power distribution interval; comparing the reference power distribution and the actual power distribution in the i-th reference power distribution interval to obtain the i-th power distribution difference, and comparing the reference confidence and the actual confidence in the i-th reference power distribution interval to obtain the i-th confidence difference; obtaining M power distribution differences and M confidence differences.

[0009] In an optional embodiment, the above-mentioned determination of the current operating state of the wind turbine generator set based on the comparison results includes: when the i-th power distribution difference indicates that there is a power difference between the reference power distribution and the actual power distribution in the i-th reference power distribution interval, and the i-th confidence difference indicates that the power difference is greater than a preset threshold, determining the i-th reference power distribution interval as a candidate fault power distribution interval; counting the number of candidate fault power distribution intervals in the results of M comparisons; when the number of candidate fault power distribution intervals is greater than a target number threshold, determining that the current operating state of the wind turbine generator set is a fault state; when the number of candidate fault power distribution intervals is less than or equal to the target number threshold, determining that the current operating state of the wind turbine generator set is a non-fault state.

[0010] In an optional embodiment, before obtaining the actual wind turbine operation data collected from the wind turbine generator sets in the target wind farm, the above also includes: obtaining processing result data after filtering the wind turbine operation data collected from the wind turbine generator sets in the target wind farm; based on the processing result data, determining a target power curve learning model from N candidate power curve learning models, wherein N is a natural number greater than or equal to 2; obtaining M reference power distribution intervals predicted by the target power curve learning model and the reference confidence corresponding to each reference power distribution interval.

[0011] In an optional embodiment, the above-mentioned acquisition of processing result data after filtering the wind turbine operation data collected from the wind turbine generator sets in the target wind farm includes: performing quantile calculation on the wind turbine operation data to obtain a specific quantile value of the power output in each wind speed range; filtering the wind turbine operation data based on the specific quantile value to obtain processing result data, wherein the processing result data is the wind turbine operation data after the abnormal data has been filtered out.

[0012] In an optional implementation, the above-mentioned determination of the target power curve learning model from N candidate power curve learning models based on the processing result data includes: obtaining learning process evaluation parameters obtained by each candidate power curve learning model after processing the processing result data, wherein the learning process evaluation parameters include: calculation process evaluation parameters, parameter adjustment complexity; comparing the learning process evaluation parameters corresponding to each candidate power curve learning model; and determining the target power curve learning model according to the comparison results.

[0013] In an optional implementation, after determining the current operating state of the wind turbine generator set according to the comparison result, the method further includes: outputting a fault report when the current operating state indicates that the wind turbine generator set is in a fault state.

[0014] According to another aspect of an embodiment of the present application, a state detection device for a wind turbine generator set is also provided, including: a first acquisition unit, used to acquire actual wind turbine operation data collected from wind turbine generator sets in a target wind farm; a second acquisition unit, used to acquire the actual power distribution and actual confidence of the actual wind turbine operation data in each reference power distribution interval according to M reference power distribution intervals determined based on a target power curve learning model, wherein the target power curve learning model is a model in which the prediction process of the output power prediction result selected from N candidate power curve models reaches a preset prediction index, and the M reference power distribution intervals correspond to M wind speed values, respectively, and M is a natural number greater than 1; a comparison unit, used to traverse the M reference power distribution intervals, and compare the reference power distribution and the actual power distribution in each reference power distribution interval respectively; a detection unit, used to determine the current operating state of the wind turbine generator set according to the comparison result.

[0015] In an optional embodiment, the comparison unit includes: a processing module, used to traverse M reference power distribution intervals, and perform the following operations respectively: obtain the reference power distribution and reference confidence within the i-th reference power distribution interval; compare the reference power distribution and the actual power distribution within the i-th reference power distribution interval to obtain the i-th power distribution difference, and compare the reference confidence and the actual confidence within the i-th reference power distribution interval to obtain the i-th confidence difference; obtain M power distribution differences and M confidence differences.

[0016] In an optional embodiment, the detection unit includes: a first determination module, used to determine that the i-th reference power distribution interval is a candidate fault power distribution interval when the i-th power distribution difference indicates that there is a power difference between the reference power distribution and the actual power distribution within the i-th reference power distribution interval, and the i-th confidence difference indicates that the power difference is greater than a preset threshold; a statistical module, used to count the number of candidate fault power distribution intervals in the results of M comparisons; a second determination module, used to determine that the current operating state of the wind turbine is a fault state when the number of candidate fault power distribution intervals is greater than a target number threshold; and to determine that the current operating state of the wind turbine is a non-fault state when the number of candidate fault power distribution intervals is less than or equal to the target number threshold.

[0017] In an optional embodiment, the above-mentioned device also includes: a third acquisition unit, used to obtain processing result data after filtering the wind turbine operation data collected from the wind turbine generator sets in the target wind farm before obtaining the actual wind turbine operation data collected from the wind turbine generator sets in the target wind farm; a first determination unit, used to determine the target power curve learning model from N candidate power curve learning models based on the processing result data, wherein N is a natural number greater than or equal to 2; and a fourth acquisition unit, used to obtain M reference power distribution intervals predicted by the target power curve learning model and the reference confidence corresponding to each reference power distribution interval.

[0018] In an optional embodiment, the third acquisition unit includes: a quantile calculation module, which is used to perform quantile calculation on the wind turbine operation data to obtain a specific quantile value of the power output in each wind speed range; a filtering module, which filters the fan operation data based on the specific quantile value to obtain processing result data, wherein the processing result data is the fan operation data after the abnormal data has been filtered out.

[0019] In an optional embodiment, the first determination unit includes: a first acquisition module, used to obtain the learning process evaluation parameters obtained by each candidate power curve learning model after processing the processing result data, wherein the learning process evaluation parameters include: calculation process evaluation parameters, parameter adjustment complexity; a comparison module, used to compare the learning process evaluation parameters corresponding to each candidate power curve learning model; a determination module, which determines the target power curve learning model according to the comparison results.

[0020] In an optional embodiment, the above-mentioned device further includes: an output unit for outputting a fault report after determining the current operating state of the wind turbine generator set according to the comparison result, if the current operating state indicates that the wind turbine generator set is in a fault state.

[0021] According to another aspect of the embodiments of the present application, a computer-readable storage medium is provided, which includes a stored program, wherein the program is executed by a processor when running to implement the state detection method of the wind turbine generator set.

[0022] According to another aspect of the embodiments of the present application, a computer program product is provided, including a computer program / instruction, which is used to implement the above-mentioned state detection method of the wind turbine generator set when executed by a processor.

[0023] In an embodiment of the present application, after obtaining the actual wind turbine operation data collected from the wind turbine generator set in the target wind farm, the actual power distribution and actual confidence of the actual wind turbine operation data in each reference power distribution interval are obtained according to the M reference power distribution intervals determined based on the target power curve learning model. Among them, the target power curve learning model here is a model in which the prediction process of the output power prediction result selected from the N candidate power curve models reaches the preset prediction index, and the M reference power distribution intervals correspond to M wind speed values ​​respectively. Afterwards, traverse the M reference power distribution intervals, compare the reference power distribution and the actual power distribution in each reference power distribution interval respectively, and determine the current operating status of the wind turbine generator set according to the comparison results. A target power curve learning model is selected from N candidate power curve models, and a reference power distribution interval is obtained based on the target power curve learning model, and then the current operating state of the wind turbine is judged by comparing the reference power distribution and the actual power in the reference power distribution interval. That is, a method of selecting a suitable power curve model to complete the state detection process of the wind turbine in the wind farm is achieved to dynamically adapt to the constantly changing operating environment of the wind turbine, thereby achieving the effect of improving the accuracy of the dynamic detection results of the wind turbine, and further solving the technical problem that the current wind turbine detection method cannot achieve dynamic adaptation to the operating environment, resulting in inaccurate detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The drawings described herein are used to provide a further understanding of the embodiments of the present application and constitute a part of the embodiments of the present application. The illustrative embodiments of the embodiments of the present application and their descriptions are used to explain the embodiments of the present application and do not constitute an improper limitation on the embodiments of the present application. In the drawings:

[0025] Figure 1 is a flow chart of an optional method for detecting the state of a wind turbine generator set according to an embodiment of the present application;

[0026] Figure 2 is a flow chart of another optional method for detecting the state of a wind turbine generator set according to an embodiment of the present application;

[0027] Figure 3 is a flow chart of another optional method for detecting the state of a wind turbine generator set according to an embodiment of the present application;

[0028] Figure 4 is a flow chart of another optional method for detecting the state of a wind turbine generator set according to an embodiment of the present application;

[0029] Figure 5 is a schematic structural diagram of an optional state detection device for a wind turbine generator set according to an embodiment of the present application;

[0030] Figure 6 It is a structural schematic diagram of an optional computer system for implementing the above-mentioned state detection method of the wind turbine generator set according to an embodiment of the present application. DETAILED DESCRIPTION

[0031] In order to enable those skilled in the art to better understand the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only embodiments of a part of the embodiments of the present application, not all embodiments. Based on the embodiments in the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the embodiments of the present application.

[0032] It should be noted that the terms "first", "second", etc. in the specification and claims of the embodiments of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0033] It should be noted that the relevant information and data involved in this disclosure are all information and data authorized by the user or fully authorized by all parties. The acquisition, storage, and processing of data involved in the embodiments of this application are in compliance with relevant regulations, and the collected information is information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure, and application of relevant data are in compliance with relevant laws, regulations, and standards in relevant regions, necessary confidentiality measures are taken, public order and good customs are not violated, and corresponding operation entrances are provided for users to choose to authorize or refuse.

[0034] The present invention is described below in conjunction with preferred implementation steps. Figure 1 is a flow chart of a method for detecting the state of a wind turbine generator set according to an embodiment of the present application, such as Figure 1 As shown, the method comprises the following steps:

[0035] S102, acquiring actual wind turbine operation data collected from wind turbine generator sets in the target wind farm;

[0036] Optionally, in this embodiment, the wind turbine generator set is a device that uses wind energy to generate electricity, which can be composed of, but not limited to, components such as a wind turbine generator, a tower, blades, a control system, and a converter. When the wind turbine generator set is acted upon by wind, the blades begin to rotate, driving the generator to rotate and generate electrical energy. After being processed by the converter, the electrical energy is fed into the power grid for power supply.

[0037] Optionally, in this embodiment, the actual wind turbine operation data is data collected during the actual operation of the wind turbine generator set, and can be exported from a software system for monitoring and controlling industrial processes (Supervisory Control and Data Acquisition, referred to as SCADA).

[0038] Optionally, the operating data in this embodiment is used to reflect the parameters of the operating state of the wind turbine generator set, and the operating data includes internal data and external data. The internal data is used to reflect the structure and movement of the wind turbine generator set itself, and the external data is used to reflect the impact of the environment on the operation of the wind turbine generator set. Among them, the external data includes but is not limited to wind speed, air density, and ambient temperature. The internal data includes but is not limited to pitch angle. Among them, the wind speed can be obtained by correcting the wind speed data source obtained through a meteorological station or a meteorological station, and the data format before and after correction is consistent.

[0039] Optionally, for the correction of the wind speed data source, it is necessary to collect the original wind speed through a meteorological observatory or a weather station, as well as the original height at which the original wind speed was measured. Determine the corrected height to be converted according to actual needs.

[0040] S104, according to the M reference power distribution intervals determined based on the target power curve learning model, obtain the actual power distribution and actual confidence of the actual wind turbine operation data in each reference power distribution interval, wherein the target power curve learning model is a model whose prediction process of the output power prediction result selected from N candidate power curve models reaches a preset prediction index, and the M reference power distribution intervals correspond to M wind speed values ​​respectively, and M is a natural number greater than 1.

[0041] It should be noted that the actual power distribution in this implementation can be obtained from, but is not limited to, SCADA.

[0042] Optionally, the target power curve learning model in this embodiment is a model for learning and predicting the power curve of a wind turbine. In wind power generation, there is a certain relationship between wind speed and power, that is, different wind speeds correspond to different output powers. The target power curve learning model is to establish a relationship model between wind speed and output power by learning historical data, which is used to predict the expected output power at different wind speeds.

[0043] Optionally, the above confidence level is used to indicate the degree of trust in the current operating state of the wind turbine generator set, and can also be used to indicate the degree of trust in the power distribution interval. In wind turbine generator set state detection, the confidence level can help determine the current operating state of the wind turbine generator set so that corresponding measures can be taken in time.

[0044] Optionally, the actual wind turbine operation data collected is input into the target power curve learning model, and can be divided into several wind speed intervals according to the set wind speed range, the set wind speed interval starting value, the wind speed interval end value and the wind speed interval step length, and M reference power distribution intervals are divided by the wind speed intervals. For example, with a wind speed of 10m / s as the interval starting value, a wind speed of 5m / s as the interval step length, and 30m / s as the interval end value, the wind speed intervals are 10m / s to 15m / s, 15m / s to 20m / s, 20m / s to 25m / s, and 25m / s to 30m / s, respectively, and the corresponding value of M is 4.

[0045] In an optional embodiment, the target power curve learning model is a model in which the prediction process of the output power prediction result selected from N candidate power curve models reaches a preset prediction index, and the candidate power curve model is a different prediction model or learning model used to predict the output power of the wind turbine generator set.

[0046] S106, traversing M reference power distribution intervals, and comparing the reference power distribution and the actual power distribution in each reference power distribution interval respectively;

[0047] Optionally, in this embodiment, the reference power distribution interval is used to compare the actual power distribution and confidence of the wind turbine operation data. The reference power distribution interval corresponds to different wind speed values. By comparing the reference power distribution and the actual power distribution in each reference power distribution interval, the current operating state of the wind turbine generator set can be determined. These reference power distribution intervals are determined based on a preset wind speed range and a corresponding output power curve model. The reference power distribution interval is a standard power interval for measuring the normal operating state of the wind turbine.

[0048] S108, determining the current operating state of the wind turbine generator set according to the comparison result.

[0049] Optionally, in this embodiment, when the actual power distribution value is less than the reference power distribution, the actual output power of the wind turbine is less than the power that the wind turbine should theoretically output; conversely, when the actual output power of the wind turbine is greater than or equal to the power that the wind turbine should theoretically output, the operating status of the wind turbine is judged by comparing the difference between the actual power distribution and the reference power distribution with the set threshold.

[0050] Through the above-mentioned embodiments of the present application, dynamic adaptation is achieved according to the constantly changing operating environment of the wind turbine generator set to ensure the accuracy of the detection results. By obtaining the actual power distribution and actual confidence of the actual wind turbine operating data in each reference power distribution interval, and comparing the reference power distribution and the actual power distribution respectively, the state of the wind turbine generator set can be comprehensively analyzed from multiple dimensions. By counting the number of candidate fault power distribution intervals in the comparison results, it is possible to determine whether the current operating state of the wind turbine generator set is a fault state, and to detect faults in time and output fault reports. The power curve learning model in the method can be predicted based on the actual wind turbine operating data, which can more accurately reflect the actual operating state of the wind turbine generator set and improve the accuracy and efficiency of the detection. The flexible selection of N candidate power curve models and M reference power distribution intervals can meet the detection needs of wind turbine generator sets in different wind farms.

[0051] As an optional solution, the above step S106, traversing M reference power distribution intervals, and comparing the reference power distribution and the actual power distribution in each reference power distribution interval respectively includes:

[0052] Traverse the M reference power distribution intervals and perform the following operations respectively:

[0053] S1061, obtaining a reference power distribution and a reference confidence level in an i-th reference power distribution interval;

[0054] S1062, comparing the reference power distribution and the actual power distribution in the i-th reference power distribution interval to obtain an i-th power distribution difference, and comparing the reference confidence and the actual confidence in the i-th reference power distribution interval to obtain an i-th confidence difference;

[0055] Optionally, in this embodiment, the confidence level is obtained:

[0056] Determine the distance between two points in space as the reference confidence R 1 , by setting the coordinates between the two points, the straight-line distance between the two points is obtained as the reference confidence R 1 .

[0057] Optionally, in this embodiment, R is determined based on the following formula: 1 :

[0058]

[0059] Where X 1 , X 2 The horizontal axis represents the set power, Y 1 , Y 2 The vertical axis represents the set power.

[0060] The actual power distribution and the reference power distribution are taken as two points in a two-dimensional coordinate system, and the distance between the two points is calculated as the actual confidence R 2 , convert the actual power distribution and the reference power distribution into coordinate representation, and calculate the straight-line distance between the two coordinate points as the actual confidence level R 2 .

[0061] Optionally, in this embodiment, R can be determined based on the following formula: 2 :

[0062]

[0063] Wherein, X represents the horizontal coordinate, Y represents the vertical coordinate, P2 represents the actual power, and P1 represents the reference power.

[0064] According to R 1 and R 2 Perform comparison and obtain confidence difference;

[0065] S1063: Obtain M power distribution differences and M confidence differences.

[0066] Through the above-mentioned embodiments of the present application, a dynamic comparison is performed based on the actual operating data of the wind turbine generator set and the reference power distribution within the reference power distribution interval, thereby achieving dynamic adaptation to the operating environment and making the detection result more accurate. Moreover, by comparing the reference power distribution with the actual power distribution and obtaining the power distribution difference and the confidence difference, the current operating state of the wind turbine generator set can be determined more accurately, especially in terms of fault diagnosis. In addition, by obtaining the reference power distribution and confidence within multiple reference power distribution intervals and comparing the actual power distribution, the reliability of state detection can be improved and the possibility of misjudgment can be reduced. Finally, the reference power distribution interval is determined based on the actual wind turbine operating data and the power curve learning model, and the state of the wind turbine generator set is detected in a data-driven manner, which can better adapt to different operating environments and wind field conditions.

[0067] As an optional solution, the above step S108, determining the current operating state of the wind turbine generator set according to the comparison result includes:

[0068] S1081, when the i-th power distribution difference indicates that there is a power difference between the reference power distribution and the actual power distribution in the i-th reference power distribution interval, and the i-th confidence difference indicates that the power difference is greater than a preset threshold, determining the i-th reference power distribution interval as a candidate fault power distribution interval;

[0069] S1082, counting the number of candidate fault power distribution intervals in the M comparison results;

[0070] S1083, when the number of candidate fault power distribution intervals is greater than the target number threshold, determine that the current operating state of the wind turbine generator set is a fault state; when the number of candidate fault power distribution intervals is less than or equal to the target number threshold, determine that the current operating state of the wind turbine generator set is a non-fault state.

[0071] Optionally, in this embodiment, the number of candidate fault power distribution intervals may be calculated by, but is not limited to, a set effect size and a set confidence interval.

[0072] Optionally, in this embodiment, the width of the candidate fault power confidence interval is determined, where the confidence interval width is the allowable error range; the effect size is determined, where the effect size is the degree of difference between the number of candidate fault power distribution intervals and the target number threshold distribution, and the number of candidate fault power distribution intervals is calculated by the candidate fault power confidence interval width and the effect size, and the formula is as follows:

[0073]

[0074] Where Z is the quantile of the standard normal distribution of the number of candidate fault power distribution intervals, P is the effect size, and E is the width of the confidence interval.

[0075] Through the above-mentioned embodiments of the present application, according to the comparison results of the reference power distribution interval and the actual power distribution, the power distribution interval where the fault may exist can be determined in a targeted manner, so as to more accurately judge the operating status of the wind turbine generator set. And by counting the number of candidate fault power distribution intervals in the comparison results, the current operating status of the wind turbine generator set can be automatically determined, which improves efficiency and reduces manual intervention. In addition, by setting preset thresholds and target number thresholds, misjudgments and false alarms can be avoided to a certain extent, and the reliability of the judgment results is improved. In addition, the method can make status judgments based on the wind turbine operation data collected in real time, and can timely discover potential faults and take corresponding measures to ensure the safe operation of the wind turbine generator set. Finally, the method can be flexibly applied to the status detection of different wind turbine generator sets, and has strong versatility and applicability.

[0076] As an alternative, Figure 2 As shown, before obtaining the actual wind turbine operation data collected from the wind turbine generator set in the target wind farm, it also includes:

[0077] S202, obtaining processing result data after filtering the wind turbine operation data collected from the wind turbine generator sets in the target wind farm;

[0078] Optionally, in this embodiment, the comparison result is obtained by comparing the collected fan operation data with the fan operation data quantile. According to the comparison result, the fan operation data is detected for abnormality, abnormal fan operation data and normal fan operation data are screened out, and the normal fan operation data is used as the filtered processing result data. Among them, the fan operation data quantile can be obtained by enumeration or cross-validation method, etc. The method for obtaining the quantile in this embodiment is only an example, and its selection is not limited.

[0079] S204, determining a target power curve learning model from N candidate power curve learning models based on the processing result data, where N is a natural number greater than or equal to 2;

[0080] When applying this embodiment, you can select a multilayer perceptron artificial neural network model (MultilayerPerceptron, MLP for short), a radial basis function network model (Radial Basis Function Network, RBFNetwork for short) and a gradient boosting network model (Radial Basis Function Network, LightGBM for short), and the value of N is 3. MLP has a strong fitting ability and is applicable to various types of data, so it can be used to process complex power characteristic relationships and wind turbine operation data. RBF has the characteristics of good performance on nonlinear problems, so it can well feedback the power curve. LightGBM has an efficient training speed and excellent performance, so it is suitable for a large number of features and requires an efficient training process in the embodiments of this application. The above-mentioned actual wind turbine operation data is input into these three models, and the model whose prediction process of the selected output power prediction result reaches the preset prediction index is the target power curve learning model. The wind speed range and candidate power curve models in this embodiment are all examples, and their selection is not limited.

[0081] S206, obtaining M reference power distribution intervals predicted by the target power curve learning model and reference confidences corresponding to each reference power distribution interval.

[0082] Optionally, in this embodiment, the above-mentioned obtaining of the processing result data after filtering the wind turbine operation data collected from the wind turbine generator sets in the target wind farm includes:

[0083] S2021, performing quantile calculation on the wind turbine operation data to obtain a specific quantile value of power output in each wind speed range;

[0084] S2022, filtering the fan operation data based on a specific quantile value to obtain processing result data, wherein the processing result data is the fan operation data after the abnormal data has been filtered out.

[0085] Optionally, in this embodiment, the above quantile calculation may, but is not limited to, adopt a quantile filtering processing method. That is, the quantile of the power output in each wind speed interval is calculated by the quantile filtering method for the wind turbine operation data.

[0086] Optionally, in this embodiment, the above specific quantile value can be used to measure the power output in different wind speed intervals. By calculating the specific quantile value, the power output in different wind speeds can be better understood, and then the operation status of the wind turbine generator set can be evaluated and monitored.

[0087] Optionally, in this embodiment, a specific quantile q is obtained through the above quantile filtering method. The wind turbine operation data is set as x. It is set that the wind turbine operation data x falling outside the specific quantile q is normal data, and the wind turbine operation data x falling within the specific quantile q is abnormal data. The abnormal data in the wind turbine operation data is filtered, and the obtained normal data is the processed result data. Among them, the probability that x falls within the quantile q is determined by P[X < x] ≤ k / q, and the probability that x falls outside the quantile q is determined by P[X < x] ≥ 1 - k / q, where q is the specific quantile, x is the k-th quantile of the variable X, and k is the quantile order.

[0088] Through the above embodiments of the present application, the quantile calculation method has strong robustness to outliers, can effectively process outliers, and improve the accuracy and reliability of data. The quantile calculation method is applicable to various types of data distributions, without assuming that the data follows a specific distribution, and has strong adaptability. By calculating the specific quantile value, the power output in different wind speeds can be intuitively understood, which is beneficial for engineers and operation and maintenance personnel to evaluate and monitor the operation status of the wind turbine generator set. The quantile calculation method has a fast calculation speed and can process a large amount of data in real time, which is beneficial for timely monitoring and analysis of the wind turbine operation data.

[0089] As an optional solution, based on the processed result data, determining the target power curve learning model from N candidate power curve learning models includes:

[0090] S1, obtaining the learning process evaluation parameters obtained by each candidate power curve learning model after processing the processed result data, where the learning process evaluation parameters include: calculation process evaluation parameters, tuning parameter complexity;

[0091] S2, comparing the learning process evaluation parameters corresponding to each candidate power curve learning model;

[0092] S3, determining the target power curve learning model according to the comparison result.

[0093] Optionally, in this embodiment, a learning process evaluation parameter is obtained after each candidate power curve learning model processes the processing result data, wherein the learning process evaluation parameter includes: a calculation process evaluation parameter and a parameter adjustment complexity;

[0094] Optionally, in this embodiment, the calculation process evaluation parameters are used to evaluate and compare the performance of different models, including but not limited to evaluating the calculation speed and accuracy of the candidate power curve learning model. The parameter tuning complexity is used to evaluate the complexity of hyperparameter tuning in the target power curve learning model.

[0095] Optionally, in this embodiment, obtaining a learning process evaluation parameter obtained after each candidate power curve learning model processes the processing result data includes: calculating a predicted average parameter of each candidate power curve learning model result. The learning process evaluation parameter is obtained by comparing the predicted average value with the predicted critical value to judge the accuracy of the calculation result of the candidate power curve learning model.

[0096] For example, specific combinations such as Figure 3 The process shown is to illustrate:

[0097] S301, define the null hypothesis and alternative hypothesis.

[0098] Optionally, in this embodiment, the null hypothesis is defined as assuming that the prediction capabilities of all candidate power curve learning models are the same, H 0 ; Define the alternative hypothesis that at least one candidate power curve learning model has a different predictive ability than the other models, H 1 .

[0099] S302, calculating the predicted average value;

[0100] Optionally, in this embodiment, the calculation process may include but is not limited to two parts: first, the calculation of the prediction error sequence {fit}, calculating the prediction results of n models at time t and the true value y t Second, define the loss function used by the model as g. For example, the formula can be:

[0101] Optionally, in this embodiment, the prediction results of each candidate power curve learning model are calculated to obtain a loss error sequence; a loss difference sequence is calculated based on the predicted loss error sequence; and a predicted average value parameter is calculated based on the loss difference sequence.

[0102] Optionally, in this embodiment, the prediction loss error sequence f of the candidate power curve learning model can be used but is not limited to it Calculate the loss difference sequence d it , as follows:

[0103] {d it}={g(f it )-g(f i+1，t )},i=1,2,3,...,n (5)

[0104] Among them, d is the loss difference sequence; g is the loss function; f is the prediction loss error sequence; i is the order of the candidate power curve learning model; t is the running time of the candidate power curve learning model.

[0105] Obtaining the loss error sequence can help analyze the prediction deviation of each model at different wind speeds and help discover the prediction performance of the model under specific conditions. Evaluation based on the loss difference sequence and the prediction mean parameter can objectively compare the differences between models and avoid the influence of subjective factors.

[0106] In addition, in this embodiment, the predicted average value d can be obtained by averaging the loss difference sequence of the candidate power curve learning model at time t, but is not limited to: s , as follows:

[0107]

[0108] Among them, d s is the predicted average; d is the loss difference sequence; i is the order of the candidate power curve learning model; t is the running time of the candidate power curve learning model.

[0109] By comparing the predicted mean value with the predicted critical value, the accuracy of the calculation results of the candidate models can be more intuitively evaluated, thereby selecting the optimal model. Combining the predicted mean value and the predicted critical value, the model's predictive ability and ability to handle uncertainty can be comprehensively considered, making the model selection more comprehensive and objective. By evaluating the predictive ability of the candidate power curve learning model through the predicted mean value parameter, the optimal model suitable for the scenario can be better selected in practical applications, improving the practicality and applicability of the technical solution.

[0110] Optionally, in this embodiment, the predicted power result of the candidate power curve learning model is obtained as follows: Get the actual power data y from the actual wind turbine operation data, and predict the power result as and the actual power data y to calculate the prediction loss error sequence {f it}. The following formula is used to obtain the loss error sequence:

[0111]

[0112] Where f is the prediction loss error sequence; is the predicted power result of the wind turbine state prediction; i is the order of the candidate power curve learning model; t is the running time of the candidate power curve learning model.

[0113] Through the above embodiments of the present application, the prediction loss error sequence can objectively reflect the prediction accuracy of the model without being affected by other factors. This helps to more accurately evaluate the performance of the model. By calculating the prediction loss error sequence at different times, the performance of the model under different conditions can be compared. This helps to discover the advantages and disadvantages of the model under different conditions, and then select the most suitable model.

[0114] S303, determining a prediction critical value;

[0115] Optionally, in this embodiment, the prediction critical value is obtained based on the following method: setting the significance level ∝ value; and calculating the prediction critical value based on the significance level and 1-∝ cumulative probability point.

[0116] Through the above-mentioned embodiments of the present application, by setting different values ​​of the significance level α, the strictness of the hypothesis test can be flexibly controlled to meet the needs in different practical scenarios.

[0117] S305, judging the prediction ability of the models.

[0118] Optionally, in this embodiment, the accuracy of the calculation result of the candidate power curve learning model is judged by comparing the predicted average value with the predicted critical value, including: if the predicted average value is less than or equal to the predicted critical value, it is the null hypothesis, and the candidate power curve learning model with the shortest prediction time parameter is selected as the target power curve learning model; if the predicted average value is greater than the predicted critical value, it is the alternative hypothesis, and the target power curve learning model is selected based on the determination coefficient. Among them, the prediction time parameter is the operation time of the candidate power curve learning model from the input of the actual wind turbine operation data to the calculation process of obtaining the evaluation parameters of the learning process. Among them, the determination coefficient is used to judge the degree of fit of the candidate power curve learning model to the actual data, that is, to judge the accuracy of the prediction of the candidate power curve learning model, so as to judge whether there is a fault in the wind turbine generator set.

[0119] Through the above-mentioned embodiments of the present application, the predicted average value is compared with the predicted critical value, and the prediction time parameter and the determination coefficient are combined to comprehensively consider the prediction accuracy, running time and fitting degree of the model to the actual data, so as to more comprehensively select the target power curve learning model. By comprehensively considering multiple parameters, including the prediction time parameter and the determination coefficient, the process of selecting the target power curve learning model is more flexible and can adapt to different situations and requirements.

[0120] S306-1, when the prediction capabilities of the models are equal, directly selecting the target power curve learning model;

[0121] S306-2, when the prediction capabilities of the models are not equal, selecting a target power curve learning model based on a first weight (determination coefficient) and a second weight (prediction time);

[0122] Optionally, in this embodiment, if the predicted average value is greater than the predicted critical value, the target power curve learning model is selected based on the determination coefficient, including: sorting the candidate power curve learning models from large to small according to the determination coefficient; if the difference in the determination coefficient between the first-rank candidate power curve learning model and the second-rank candidate power curve learning model is greater than or equal to 2%, then the first-rank candidate power curve learning model is selected as the target power curve learning model. If the difference in the determination coefficient between the first-rank candidate power curve learning model and the second-rank candidate power curve learning model is less than 2%, a weighted calculation is performed based on the running time to obtain a calculation process evaluation parameter, so as to select the candidate power curve learning model with the maximum value of the calculation process evaluation parameter as the target power curve learning model.

[0123] Through the above embodiment of the present application, the method of comparing the determination coefficient and the running time to select the best model is very efficient. This embodiment first sorts according to the determination coefficient, and then determines the best model based on whether the difference is greater than or equal to 2%, so that the best model can be selected quickly, improving the efficiency of the entire process.

[0124] Optionally, in this embodiment, the coefficient of determination of the candidate power curve learning model is The prediction time parameters of the candidate power curve learning model are used to obtain the calculation process evaluation parameters through weighted calculation. The formula is as follows:

[0125]

[0126] Among them: Score wi Evaluate parameters for computational processes; is the determination coefficient of the candidate power curve learning model i; W is the weight coefficient of the running time; T is the running time of the candidate power curve learning model i.

[0127] As an optional solution, after determining the current operating state of the wind turbine generator set according to the comparison result, the method further includes:

[0128] When the current operating status indicates that the wind turbine generator set is in a fault state, a fault report is output.

[0129] Through the above embodiments of the present application, the operating status of the wind turbine generator set can be monitored in real time, and faults can be discovered in time. Once a fault is discovered, the system will automatically output a fault report without manual intervention, saving human resource costs.

[0130] The solution in this embodiment is described below in conjunction with a specific application scenario. Assuming that the target wind farm includes three wind power stations, such as A1, A2 and A3, each of which includes a group of wind turbine generator sets, the status detection of the wind turbine generator sets in the wind farm can be achieved through the following process, such as Figure 4 As shown:

[0131] S401, start: start checking the status of the wind turbine generator set;

[0132] S402, obtaining actual wind turbine operation data and calculating the minimum sample size required for data verification;

[0133] The actual wind turbine operation data of A1, A2 and A3 are obtained through the SCADA system, and the data of each wind power station is checked to determine the required sample size. The actual data is obtained based on the minimum sample size, where the actual data includes the actual power distribution and the actual confidence.

[0134] S403, screening abnormal data;

[0135] The quantile filtering method is used to calculate the quantiles of the actual wind turbine operation data, and the specific quantile value of the power output in each wind speed range is obtained. The quantile filtering method is used to detect abnormalities in the actual wind turbine operation data, filter out abnormal data, and retain normal data as the processing result data.

[0136] S404, selecting a power curve learning model;

[0137] MLP, RBF Network, and LightGBM are selected as candidate power curve learning models.

[0138] S405, evaluating the prediction effect of the model;

[0139] Input the obtained processing result data input into the three candidate power curve learning models respectively, and obtain the output results of the candidate power curve learning models. Based on the output results of the candidate power curve learning models, the candidate power curve learning models are sorted from large to small according to the determination coefficient; if the difference in determination coefficient between the first-rank candidate power curve learning model and the second-rank candidate power curve learning model is greater than or equal to 2%, the first-rank candidate power curve learning model is selected as the target power curve learning model. If the difference in determination coefficient between the first-rank candidate power curve learning model and the second-rank candidate power curve learning model is less than 2%, a weighted calculation is performed based on the running time, and the formula is used. A calculation process evaluation parameter is obtained, and a candidate power curve learning model with a maximum value of the calculation process evaluation parameter is selected as a target power curve learning model.

[0140] S406-1, obtaining a reference power distribution interval;

[0141] S406-2, obtaining a reference confidence level;

[0142] Set the wind speed to 10-15m / s, 15-20m / s, 20-25m / s, and 25-30m / s, for a total of 4 wind speed intervals, input the wind speed interval values ​​into the target power curve learning model, and obtain the reference power distribution interval and reference confidence.

[0143] S407, inspection data;

[0144] Compare the power distribution and confidence output faulty data: compare the actual power distribution with the reference power distribution within the reference power distribution interval, and when the power difference between the actual power distribution and the reference power distribution interval, and the difference between the actual confidence and the reference confidence is greater than a preset threshold, determine that the actual power is abnormal and mark it as a candidate fault power distribution interval; count the number of candidate fault power distribution intervals in the results of the set number of comparisons; when the number of candidate fault power distribution intervals is greater than the target number threshold, determine that the current operating state of the wind turbine is a fault state; when the number of candidate fault power distribution intervals is less than or equal to the target number threshold, determine that the current operating state of the wind turbine is a non-fault state.

[0145] S408, outputting faulty data;

[0146] When the current operating status indicates that the wind turbine generator set is in a fault state, a fault report is output.

[0147] S409, end.

[0148] According to another aspect of the embodiment of the present application, a state detection device for a wind turbine generator set is provided for implementing the state detection method for a wind turbine generator set. Figure 5 As shown, the device comprises:

[0149] The first acquisition unit 502 is used to acquire actual wind turbine operation data collected from wind turbine generator sets in the target wind farm;

[0150] The second acquisition unit 504 is used to obtain the actual power distribution and the actual confidence of the actual wind turbine operation data in each reference power distribution interval according to the M reference power distribution intervals determined based on the target power curve learning model, wherein the target power curve learning model is a model whose prediction process of the output power prediction result selected from the N candidate power curve models reaches a preset prediction index, and the M reference power distribution intervals correspond to M wind speed values ​​respectively, and M is a natural number greater than 1;

[0151] A comparison unit 506, configured to traverse the M reference power distribution intervals and compare the reference power distribution and the actual power distribution in each reference power distribution interval respectively;

[0152] The detection unit 508 is used to determine the current operating state of the wind turbine generator set according to the comparison result.

[0153] Through the above-mentioned embodiments of the present application, the actual wind turbine operation data and actual power distribution are obtained, and the current operating status of the wind turbine generator set can be timely understood, which is conducive to timely discovery and resolution of problems. By comparing the reference power distribution and the actual power distribution, the current operating status of the wind turbine generator set can be accurately determined, and the accuracy and reliability of data analysis are improved. By automatically acquiring data and comparing and analyzing, manual intervention can be reduced and work efficiency can be improved. In addition, according to actual conditions, the reference power distribution interval can be flexibly set to adapt to different wind speed values, thereby improving applicability and flexibility. Finally, by presetting the model of prediction indicators, actual needs can be better met, and the accuracy and credibility of the prediction can be improved.

[0154] Optionally, in this embodiment, the comparison unit 506 includes:

[0155] The processing module is used to traverse the M reference power distribution intervals and perform the following operations respectively:

[0156] Obtain the reference power distribution and reference confidence within the i-th reference power distribution interval; compare the reference power distribution and actual power distribution within the i-th reference power distribution interval to obtain the i-th power distribution difference, and compare the reference confidence and actual confidence within the i-th reference power distribution interval to obtain the i-th confidence difference; obtain M power distribution differences and M confidence differences.

[0157] Through the above-mentioned embodiments of the present application, the actual wind turbine operation data can be obtained in real time, and the operation status of the wind turbine can be understood in time, which helps to find and solve problems in time. Moreover, based on the comparison between the reference power distribution interval and the actual power distribution, the operation status of the wind turbine can be determined more accurately, which improves the accuracy and reliability of the detection. In addition, by determining the reference power distribution interval through the target power curve learning model, the performance of the wind turbine can be more effectively analyzed and evaluated, and the status of the wind turbine can be identified and evaluated.

[0158] Optionally, in this embodiment, the detection unit 508 includes:

[0159] A first determination module is used to determine that the i-th reference power distribution interval is a candidate fault power distribution interval when the i-th power distribution difference indicates that there is a power difference between the reference power distribution and the actual power distribution in the i-th reference power distribution interval, and the i-th confidence difference indicates that the power difference is greater than a preset threshold;

[0160] A statistical module, used to count the number of candidate fault power distribution intervals in the M comparison results;

[0161] The second determination module is used to determine that the current operating state of the wind turbine generator set is a fault state when the number of candidate fault power distribution intervals is greater than the target number threshold; and to determine that the current operating state of the wind turbine generator set is a non-fault state when the number of candidate fault power distribution intervals is less than or equal to the target number threshold.

[0162] Optionally, in this embodiment, the above device further includes:

[0163] A third acquisition unit is used to obtain the processing result data after filtering the wind turbine operation data collected from the wind turbine generator sets in the target wind farm before obtaining the actual wind turbine operation data collected from the wind turbine generator sets in the target wind farm;

[0164] A first determining unit, configured to determine a target power curve learning model from N candidate power curve learning models based on the processing result data, wherein N is a natural number greater than or equal to 2;

[0165] The fourth acquisition unit is used to acquire the M reference power distribution intervals predicted by the target power curve learning model and the reference confidence corresponding to each reference power distribution interval.

[0166] Through the above-mentioned embodiments of the present application, a modular method is used for fault diagnosis, which can automatically judge the operating status of the wind turbine generator set, reduce the need for manual intervention, and improve the efficiency and accuracy of diagnosis. By setting the preset threshold and the target quantity threshold, it can be flexibly adjusted according to the actual situation, and is applicable to wind turbine generator sets of different types and sizes, with certain versatility and applicability. Moreover, it can quickly make judgments in the fault state, and take corresponding repair and maintenance measures in time, which effectively reduces the impact of the fault on the generator set and improves the power generation efficiency and reliability.

[0167] Optionally, in this embodiment, the third obtaining unit includes:

[0168] The quantile calculation module is used to perform quantile calculation on the wind turbine operation data to obtain a specific quantile value of the power output in each wind speed range;

[0169] The filtering module filters the fan operation data based on a specific quantile value to obtain processing result data, wherein the processing result data is the fan operation data after the abnormal data has been filtered out.

[0170] Through the above embodiments of the present application, quantile calculation and filtering can effectively eliminate abnormal data and improve the accuracy of fan operation data. Moreover, the filtered processing result data is cleaner and more reliable, which can improve the efficiency and accuracy of data analysis.

[0171] Optionally, in this embodiment, the first determining unit includes:

[0172] The first acquisition module is used to obtain the learning process evaluation parameters obtained after each candidate power curve learning model processes the processing result data, wherein the learning process evaluation parameters include: calculation process evaluation parameters and parameter adjustment complexity;

[0173] A comparison module, used for comparing the learning process evaluation parameters corresponding to each candidate power curve learning model;

[0174] The determination module is used to determine the target power curve learning model according to the comparison result.

[0175] Through the above embodiments of the present application, by comparing the evaluation parameters of the learning process, the most suitable target power curve learning model can be objectively determined, avoiding the influence of subjective factors on model selection. Accuracy: By comparing the evaluation parameters of the learning process, the most suitable learning model for processing the result data can be selected, thereby improving the accuracy of the model's prediction and analysis.

[0176] Optionally, in this embodiment, the device further includes: an output unit for outputting a fault report after determining the current operating state of the wind turbine generator set according to the comparison result, if the current operating state indicates that the wind turbine generator set is in a fault state.

[0177] Through the above-mentioned embodiments of the present application, the output unit can automatically determine the current operating status of the wind turbine generator set according to the comparison results, and output a fault report, thereby reducing manual intervention and improving efficiency.

[0178] In this embodiment, the embodiment of the state detection device of the wind turbine generator set can refer to the above method embodiment, which will not be described in detail here.

[0179] According to another aspect of the embodiments of the present application, a computer-readable storage medium is provided for implementing the above-mentioned state detection method of the wind turbine generator set.

[0180] Optionally, in this embodiment, the computer-readable storage medium may be configured to store a computer program for performing the following steps:

[0181] S1, obtaining actual wind turbine operation data collected from wind turbine generator sets in a target wind farm;

[0182] S2, according to the M reference power distribution intervals determined based on the target power curve learning model, obtaining the actual power distribution and the actual confidence of the actual wind turbine operation data in each reference power distribution interval, wherein the target power curve learning model is a model whose prediction process of the output power prediction result selected from the N candidate power curve models reaches a preset prediction index, and the M reference power distribution intervals correspond to M wind speed values ​​respectively, and M is a natural number greater than 1;

[0183] S3, traversing M reference power distribution intervals, and comparing the reference power distribution and the actual power distribution in each reference power distribution interval respectively;

[0184] S4, determining the current operating status of the wind turbine generator set according to the comparison result.

[0185] Optionally, in the embodiments of the present application, the term "module" or "unit" refers to a computer program or a part of a computer program with a predetermined function, and works together with other related parts to achieve a predetermined goal, and can be implemented in whole or in part by using software, hardware (such as processing circuits or memories), or a combination thereof. Similarly, a processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be part of an overall module or unit that includes the function of the module or unit.

[0186] Optionally, in this embodiment, a person of ordinary skill in the art may understand that all or part of the steps in the various methods of the above embodiments may be completed by instructing hardware related to the terminal device through a program, and the program may be stored in a computer-readable storage medium, and the storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, etc.

[0187] If the integrated units in the above embodiments are implemented in the form of software functional units and sold or used as independent products, they can be stored in the above computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for one or more computer devices (which can be personal computers, servers or network devices, etc.) to perform all or part of the steps of the methods of each embodiment of the embodiment of the present application.

[0188] According to one aspect of the embodiments of the present application, a computer program product is provided, the computer program product comprising a computer program / instruction, the computer program / instruction comprising a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be Figure 6 The communication part 609 in the computer system shown is downloaded and installed from the network, and / or installed from the removable medium 611. When the computer program is executed by the central processor 601, various functions provided by the embodiment of the present application are executed.

[0189] Among them, Figure 6 As shown, the computer system includes a central processing unit 601 (CPU), which can perform various appropriate actions and processes according to the program stored in the read-only memory 602 (ROM) or the program loaded from the storage part 608 to the random access memory 603 (RAM). Various programs and data required for system operation are also stored in the random access memory 603. The central processing unit 601, the read-only memory 602 and the random access memory 603 are connected to each other through a bus 604. The input / output interface 605 (Input / Output interface, i.e., I / O interface) is also connected to the bus 604.

[0190] The following components are connected to the input / output interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a local area network card, a modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the input / output interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed so that a computer program read therefrom is installed into the storage section 608 as needed.

[0191] In the above-mentioned embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0192] In the several embodiments provided in the embodiments of the present application, it should be understood that the disclosed client can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0193] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0194] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of software functional units.

[0195] The above is only a preferred implementation of the embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the embodiment of the present application. These improvements and modifications should also be regarded as the scope of protection of the embodiment of the present application.

Claims

1. A method for detecting the state of a wind turbine generator set, characterized in that: include: Acquire actual wind turbine operation data collected from wind turbine generator sets in the target wind farm; According to the M reference power distribution intervals determined based on the target power curve learning model, the actual power distribution and the actual confidence of the actual wind turbine operation data in each reference power distribution interval are obtained, wherein the target power curve learning model is a model in which the prediction process of the output power prediction result selected from N candidate power curve models reaches a preset prediction index, and the M reference power distribution intervals correspond to M wind speed values ​​respectively, and M is a natural number greater than 1; Traversing the M reference power distribution intervals, and comparing the reference power distribution and the actual power distribution in each reference power distribution interval respectively; The current operating state of the wind turbine generator set is determined according to the comparison result.

2. The method according to claim 1, characterized in that: The traversing the M reference power distribution intervals and comparing the reference power distribution and the actual power distribution in each reference power distribution interval respectively comprises: Traverse the M reference power distribution intervals and perform the following operations respectively: Obtaining a reference power distribution and a reference confidence level within an i-th reference power distribution interval; Comparing the reference power distribution within the i-th reference power distribution interval with the actual power distribution to obtain an i-th power distribution difference, and comparing the reference confidence within the i-th reference power distribution interval with the actual confidence to obtain an i-th confidence difference; M power distribution differences and M confidence differences are obtained.

3. The method according to claim 2, characterized in that Determining the current operating state of the wind turbine generator set according to the comparison result includes: When the i-th power distribution difference indicates that there is a power difference between the reference power distribution in the i-th reference power distribution interval and the actual power distribution, and the i-th confidence difference indicates that the power difference is greater than a preset threshold, determining the i-th reference power distribution interval as a candidate fault power distribution interval; Counting the number of candidate fault power distribution intervals in the M comparison results; When the number of the candidate fault power distribution intervals is greater than the target number threshold, the current operating state of the wind turbine generator set is determined to be a fault state; when the number of the candidate fault power distribution intervals is less than or equal to the target number threshold, the current operating state of the wind turbine generator set is determined to be a non-fault state.

4. The method according to claim 1, characterized in that: Before acquiring the actual wind turbine operation data collected from the wind turbine generator sets in the target wind farm, the method further includes: Obtaining processing result data after filtering the wind turbine operation data collected from the wind turbine generator sets in the target wind farm; Based on the processing result data, determining the target power curve learning model from N candidate power curve learning models, where N is a natural number greater than or equal to 2; The M reference power distribution intervals predicted by the target power curve learning model and the reference confidence corresponding to each of the reference power distribution intervals are obtained.

5. The method according to claim 4, characterized in that The obtaining of the processing result data after filtering the wind turbine operation data collected from the wind turbine generator set in the target wind farm includes: Calculating the quantiles of the wind turbine operation data to obtain a specific quantile value of the power output in each wind speed interval; The fan operation data is filtered based on the specific quantile value to obtain the processing result data, wherein the processing result data is the fan operation data after the abnormal data has been filtered out.

6. The method according to claim 4, characterized in that The determining the target power curve learning model from N candidate power curve learning models based on the processing result data includes: Obtaining a learning process evaluation parameter obtained after each candidate power curve learning model processes the processing result data, wherein the learning process evaluation parameter includes: a calculation process evaluation parameter and a parameter adjustment complexity; Comparing the learning process evaluation parameters corresponding to each of the candidate power curve learning models; The target power curve learning model is determined according to the comparison result.

7. The method according to any one of claims 1 to 6, characterized in that After determining the current operating state of the wind turbine generator set according to the comparison result, the method further includes: When the current operating state indicates that the wind turbine generator set is in a fault state, a fault report is output.

8. A state detection device for a wind turbine generator set, characterized in that: include: A first acquisition unit is used to acquire actual wind turbine operation data collected from wind turbine generator sets in a target wind farm; A second acquisition unit is used to obtain the actual power distribution and actual confidence of the actual wind turbine operation data in each reference power distribution interval according to the M reference power distribution intervals determined based on the target power curve learning model, wherein the target power curve learning model is a model in which the prediction process of the output power prediction result selected from N candidate power curve models reaches a preset prediction index, and the M reference power distribution intervals correspond to M wind speed values ​​respectively, and M is a natural number greater than 1; A comparison unit, used for traversing the M reference power distribution intervals, and comparing the reference power distribution and the actual power distribution in each reference power distribution interval respectively; The detection unit is used to determine the current operating state of the wind turbine generator set according to the comparison result.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein the program is executed by a processor to perform the method described in any one of claims 1 to 7.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images