A wind power generation system and an abnormality identification and early warning method for a generator winding

Abstract

The application provides a wind power generation system and an abnormality identification and early warning method for a generator winding, and comprises the following steps: S1, obtaining current temperature detection data of a to-be-identified wind turbine; S2, performing data processing on the current temperature detection data to obtain processed data; S3, performing dimension conversion on the processed data to obtain single-wind-turbine time-series dimension data, single-wind-turbine working condition dimension data and same-model working condition dimension data; S4, using a temperature abnormality identification model to identify the single-wind-turbine time-series dimension data, the single-wind-turbine working condition dimension data and the same-model working condition dimension data to obtain early warning risk features of the to-be-identified wind turbine; S5, obtaining historical early warning information and historical risk values of the generator winding in the wind power generation system to generate supervised learning early warning information; and S6, comparing the early warning risk features of the to-be-identified wind turbine with the supervised learning early warning information, and if the early warning risk features of the to-be-identified wind turbine contain the supervised learning early warning information, early warning is prompted.

Description

Technical Field

[0001] This invention relates to the field of wind power generation technology, and in particular to a method for identifying and warning of anomalies in wind power generation systems and generator windings. Background Technology

[0002] Wind power generation is receiving increasing attention due to the clean, environmentally friendly, and renewable nature of wind energy. Its basic working principle involves converting wind energy into mechanical energy through a wind turbine, then converting that mechanical energy into electrical energy through a generator, and finally feeding it into the power grid. A wind power generation system mainly consists of multiple components, including a wind turbine, a transmission system, a generator, and a supporting tower. The wind turbine has blades that, when rotated by the wind, convert wind energy into mechanical energy.

[0003] Stator thermal failures can cause winding insulation damage and structural component deformation, leading to electrical short circuits and abnormal vibrations in the generator unit, seriously jeopardizing normal generator operation and grid safety. To ensure the normal operation of each wind turbine, traditional generator stator winding temperature monitoring systems, relying solely on temperature inspection instruments, suffer from long inspection cycles and simple, limited diagnostic methods. These systems struggle to detect stator winding temperature anomalies in a timely manner, and are unable to analyze temperature rise trends or provide overheat warnings.

[0004] Chinese patent CN109458308B discloses a method for early warning of over-temperature of the main shaft of a fan. It constructs an intelligent temperature prediction model based on a signal reconstruction algorithm. The model predicts the temperature under normal operating conditions by using power, rotor speed, and historical bearing temperature information. The deviation between the predicted value and the measured value is called the model prediction residual. The system is judged to be abnormal based on the distribution of the residual.

[0005] However, the temperature prediction methods in the aforementioned technical solutions are overly dependent on the quantity and quality of samples, thus reducing their sensitivity to practical applications. With the increasing informatization of wind power generation and the ever-growing scale of data, the lack of an early warning mechanism for abnormal generator winding temperatures means that maintenance personnel only address the issue when a complete failure renders the turbine inoperable, failing to provide early warning of the fault. This leads to increased operating costs. Summary of the Invention

[0006] In view of this, the present invention proposes an abnormal identification and early warning method for generator windings. By processing and transforming the current temperature detection data and using a temperature abnormality identification model, it can promptly detect abnormal temperatures in the generator windings for maintenance, thereby improving maintenance efficiency.

[0007] The technical solution of this invention is implemented as follows:

[0008] In a first aspect, the present invention provides a method for identifying and warning of anomalies in generator windings, comprising the following steps:

[0009] S1. Obtain the current temperature detection data of the fan to be identified;

[0010] S2. Process the current temperature detection data to obtain processed data;

[0011] S3. Perform dimensional transformation on the processed data to obtain single fan time-series dimensional data, single fan operating condition dimensional data, and same model operating condition dimensional data.

[0012] S4. Use the temperature anomaly identification model to identify the time-series data of the single fan, the operating condition data of the single fan, and the operating condition data of the same model to obtain the early warning risk characteristics of the fan to be identified.

[0013] S5. Obtain historical early warning information and historical risk values ​​of the generator windings in the wind power generation system, and generate early warning information for supervised learning;

[0014] S6. Compare the early warning risk features of the wind turbine to be identified with the early warning information of the supervised learning. If the early warning risk features of the wind turbine to be identified contain the early warning information of the supervised learning, then issue an early warning.

[0015] Based on the above technical solutions, preferably, the temperature anomaly identification model includes a Bollinger evaluation model, a density matching model, and a correlation model, and the generator winding is equipped with a target temperature measuring point N. i i∈1,3,6, step S4 specifically includes:

[0016] S41. Use the Bollinger evaluation model to identify the time-series data of a single fan to obtain the fluctuation characteristics of the temperature at the measuring point of the fan to be identified.

[0017] S42. Use the density fit model to identify the single-fan operating condition dimension data to obtain the trend change characteristics of the temperature at the measuring point of the fan to be identified;

[0018] S43. Use the correlation model to identify the operating condition dimension data of the same model to obtain the difference characteristics of the temperature of the measuring point of the fan to be identified;

[0019] S44. Based on the fluctuation characteristics, trend characteristics, and difference characteristics, obtain the early warning risk characteristics of the wind turbine to be identified.

[0020] Based on the above technical solutions, preferably, step S41 specifically includes:

[0021] The generator winding temperatures at the target temperature measurement points within the operating period T form a sample dataset DFT, which can be expressed as:

[0022] DFT =t k N,P,C,t k ∈T∩Δt,

[0023] Where N represents the measuring point, P represents the power, C represents the ambient temperature, and t represents the temperature. k This represents the k-th time point within the period T, where Δt represents the time interval.

[0024] The sample dataset is input into the Bollinger evaluation model to obtain the Bollinger evaluation results. Based on the Bollinger evaluation results, the temperature fluctuation characteristics of the measuring point of the wind turbine to be identified are obtained.

[0025] Based on the above technical solutions, preferably, the calculation method of the Bollinger evaluation model specifically includes:

[0026] BW = UB - LB;

[0027] Where BW represents the bandwidth of the Bollinger Bands evaluation model, UB represents the upper band of the Bollinger Bands evaluation model, and LB represents the lower band of the Bollinger Bands evaluation model.

[0028] LB = LB tk =mint k N

[0029]

[0030]

[0031] Among them, LB tk The value of the lower bound of the Bollinger evaluation model represents the minimum value of the temperature measurement point, Q represents the Bollinger evaluation result, and t represents the minimum value of the lower bound of the model. k (N) represents t k The temperature of a moment.

[0032] Based on the above technical solutions, preferably, step S42 specifically includes:

[0033] The density mapping function is obtained by performing density mapping on the single-fan operating condition dimension data.

[0034] The density mapping function is input into the density matching model for calculation to obtain the density matching degree of single wind turbine operating condition dimension data.

[0035] The overlap value is calculated using the density matching degree, and the trend change characteristics of the temperature at the measuring point of the fan to be identified are determined based on the overlap value.

[0036] Based on the above technical solutions, the preferred method for calculating the density matching model is as follows:

[0037]

[0038] in,

[0039] in, This represents the density fit of the temperature data of the fan expected to be identified at TA and TB, where n is the number of subsets divided according to the period T, and S is the density fit. k Let f denote a subset set, k∈[1,n], and f be the density mapping function. Indicates in T A The target is to identify the temperature dataset of the fan. T represents B We need to identify the temperature dataset of the wind turbine.

[0040] Based on the above technical solutions, preferably, step S43 specifically includes:

[0041] The data of the same model operating conditions is transformed to obtain the transformed data;

[0042] The transformed data is input into the correlation model for calculation to obtain the correlation of data in the same model operating condition dimension.

[0043] The similarity of the operating condition data of the same model is determined by the correlation of the operating condition data of the same model.

[0044] The differences in temperature at the measuring points of the fan to be identified are determined based on the similarity.

[0045] Based on the above technical solutions, the preferred method for calculating the correlation model is as follows:

[0046]

[0047]

[0048] D FiT =t k N,P,C,F i ,t k ∈T∩Δt,i∈[1,j]

[0049] Wherein, p(F i F j ) indicates the fan to be identified, F i Operating condition data of the same model and other wind turbines F j Correlation of data on the same model's operating conditions, D”F i Indicates the fan to be identified, F i The collection of data after data transformation based on the same model's operating conditions, D”F j Indicates other wind turbines F j"D" represents the set of data after data transformation of the same model's operating condition data. Indicates the fan to be identified, F i Data set of the same model operating conditions during period T.

[0050] Furthermore, the anomaly identification and early warning method further includes the following steps:

[0051] S7. The user confirms the data according to the warning prompt and provides a warning receipt. Based on the warning receipt, an evaluation model is established to evaluate the temperature anomaly identification model, and the evaluation result is obtained. The temperature anomaly identification model is updated according to the evaluation result.

[0052] Secondly, the present invention also discloses a wind power generation system, the wind power generation system comprising at least one wind turbine, the wind turbine comprising at least one generator winding, and the wind power generation system further comprising: a memory, a processor, a communication bus, and a control program stored in the memory.

[0053] The communication bus is used to enable communication between the processor and the memory;

[0054] The processor is used to execute the control program to implement the steps of the generator winding anomaly identification and early warning method as described in any of the above statements.

[0055] The generator winding anomaly identification and early warning method of the present invention has the following advantages over the prior art:

[0056] (1) By processing the current temperature detection data and performing dimensional transformation on the processed data, a temperature anomaly identification model is set up. The temperature anomaly identification model is used to identify the time-series data of a single fan, the operating condition data of a single fan, and the operating condition data of the same model after dimensional transformation. The warning risk characteristics of the fan to be identified are obtained, and the warning status of the generator winding temperature anomaly of the fan to be identified is determined in a more refined manner, thereby improving the accuracy and refinement of the warning.

[0057] (2) By using the Bollinger evaluation model, density matching model and correlation model in the temperature anomaly identification model, the time series data of a single fan, the operating condition data of a single fan and the operating condition data of the same model are identified respectively, the fluctuation characteristics of the temperature of the fan to be identified are judged, the temperature anomaly is detected in time, the occurrence of fan failure is reduced, and the operating efficiency and life of the fan are improved.

[0058] (3) By using patterns and regularities in historical data to predict the risk of the wind turbine to be identified, the wind turbine to be identified can be set as a wind turbine that needs to be warned. Combined with the warning information from supervised learning, more accurate warning prompts can be provided to help maintenance personnel take corresponding measures for maintenance and repair, thereby improving the reliability and safety of the wind turbine. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0060] Figure 1 This is a flowchart of the generator winding anomaly identification and early warning method of the present invention;

[0061] Figure 2 This is a block diagram of the generator winding anomaly identification and early warning method of the present invention;

[0062] Figure 3 This is a schematic diagram of the measurement point distribution for the generator winding anomaly identification and early warning method of the present invention;

[0063] Figure 4 This is a schematic diagram of the temperature density distribution of the generator winding anomaly identification and early warning method of the present invention;

[0064] Figure 5 This is a schematic diagram of the temperature trend distribution of the generator winding anomaly identification and early warning method of the present invention;

[0065] Figure 6 This is a schematic diagram of the temperature correlation distribution of the generator winding anomaly identification and early warning method of the present invention;

[0066] Figure 7 This is a schematic diagram illustrating the evaluation of the generator winding anomaly identification and early warning method of the present invention;

[0067] Figure 8 This is a block diagram of the wind power generation system of the present invention. Detailed Implementation

[0068] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0069] As those skilled in the art will understand, the impact of abnormal generator winding temperature on wind turbines can vary. Some abnormalities may have a more severe impact on the wind turbine, such as causing a short-term shutdown, while others may have a milder impact and may not cause a shutdown for a long time, but the potential hazard always exists. Therefore, in order to promptly detect the possibility of abnormal generator winding temperature in the wind turbines to be identified, it is necessary to determine the generator winding temperature characteristics of each wind turbine based on its respective detection data, determine whether there are any abnormalities in the current temperature detection data of the wind turbine to be identified, and if so, promptly notify maintenance personnel for repair and maintenance to avoid direct shutdown of the wind turbine and affecting its operation.

[0070] like Figure 1 and Figure 2 As shown, the present invention provides a method for identifying and warning of abnormalities in generator windings, comprising the following steps:

[0071] S1. Obtain the current temperature detection data of the fan to be identified.

[0072] This application can obtain the current temperature data of the generator winding of the wind turbine to be identified through sensors or other monitoring devices. This application does not specifically limit the data acquisition method.

[0073] S2. Process the current temperature detection data to obtain processed data.

[0074] It is understandable that the current temperature detection data needs to be preprocessed, including data clearing, outlier handling, and missing value filling. This preprocessing ensures the accuracy and completeness of the processed data.

[0075] S3. Perform dimensional transformation on the processed data to obtain single fan time-series dimensional data, single fan operating condition dimensional data, and same model operating condition dimensional data.

[0076] By transforming the processed data, the temperature data of the fan to be identified can be analyzed and compared according to different dimensions, making it easier to observe the distribution and relationship of the data in a more intuitive way.

[0077] S4. Use the temperature anomaly identification model to identify the time-series data of the single fan, the operating condition data of the single fan, and the operating condition data of the same model to obtain the early warning risk characteristics of the fan to be identified.

[0078] A temperature anomaly identification model can be used to determine whether a fan exhibits an abnormal temperature. This model can employ machine learning algorithms, statistical methods, or rule engines to identify anomalies, thereby improving the accuracy and reliability of the temperature anomaly identification model.

[0079] S5. Obtain historical early warning information and historical risk values ​​of the generator windings in the wind power generation system, and generate early warning information through supervised learning. Utilize patterns and regularities in historical data to predict the risk situation of the wind turbine to be identified.

[0080] Furthermore, the warning status of the generator winding temperature of the wind turbine to be identified can be determined based on the risk situation. For example, a risk value range [0, 1] can be set, the risk value can be determined based on the risk situation, and the warning status of abnormal generator winding temperature of the wind turbine to be identified can be determined based on the risk value range. This refines the determination of the warning status of abnormal generator winding temperature of the wind turbine to be identified, thereby improving the accuracy and refinement of the warning.

[0081] In one embodiment of this application, the early warning information of supervised learning also includes setting different maintenance measures according to different abnormal early warnings. That is, the maintenance measures corresponding to historical early warning information can be found as reference maintenance measures, so that maintenance personnel can quickly and accurately check and repair the abnormal situation of the generator winding of the wind turbine to be identified by the above-mentioned reference maintenance measures, improve the efficiency of maintenance, and prevent the wind power generation system from shutting down due to abnormal wind turbines and thus causing losses.

[0082] Understandably, the same wind turbine may experience multiple anomalies, with an early warning issued before each one. New warnings, along with previous warnings, are combined to form historical warning information. This allows for adaptive adjustments based on the risk level of the generator winding temperature of the wind turbine under warning, thereby reducing the false alarm rate.

[0083] Specifically, historical early warning information includes historical early warning status, risk value, maintenance measures, a detailed description of the early warning and the warning time, as well as an early warning curve. The early warning curve is generated by the number of historical early warnings and the warning time, which helps maintenance personnel to view the overall temperature of the generator windings of the wind turbines under early warning.

[0084] S6. Compare the early warning risk characteristics of the wind turbine to be identified with the early warning information from the supervised learning. If the early warning risk characteristics of the wind turbine to be identified contain the early warning information from the supervised learning, an early warning is issued. This allows users to promptly detect abnormalities in the wind turbine and take corresponding measures for maintenance and repair, thereby improving the reliability and safety of the wind turbine.

[0085] In this embodiment, by processing and transforming the current temperature detection data of the wind turbine to be identified, the temperature data of the wind turbine to be identified can be better understood and compared. By identifying the transformed data through a temperature anomaly identification model, it is possible to quickly and accurately determine whether there is an anomaly in the wind turbine to be identified. If there is abnormal detection data, it indicates that the generator winding temperature of the wind turbine to be identified is abnormal. The wind turbine to be identified is set as a wind turbine that needs to be warned. Combined with the warning information from supervised learning, more accurate warning prompts can be provided to help maintenance personnel take corresponding measures for maintenance and repair, thereby improving the reliability and safety of the wind turbine.

[0086] In one embodiment of this application, the temperature anomaly identification model includes a Bollinger evaluation model, a density matching model, and a correlation model, and the generator winding is provided with a target temperature measuring point N. i i∈1,3,6, step S4 specifically includes:

[0087] S41. Use the Bollinger evaluation model to identify the time-series data of a single fan to obtain the temperature fluctuation characteristics of the measuring point of the fan to be identified.

[0088] Understandably, when i=1, meaning there is one measuring point on the generator winding, but a single measuring point cannot meet the usage conditions of the Bollinger evaluation model, the number of target temperature measuring points must be at least two. Preferably, there are three or six measuring points on the generator winding.

[0089] In this embodiment of the application, by identifying the time-series data of a single fan using the Boolean evaluation model, the fluctuation characteristics of the temperature of the fan to be identified can be determined, abnormal temperature conditions can be detected in a timely manner, the occurrence of fan failures can be reduced, and the operating efficiency and lifespan of the fan can be improved.

[0090] like Figure 3 As shown, specifically, step S41 includes:

[0091] S411, The generator winding temperature at the target temperature measuring point within the operating cycle T forms a sample dataset D. FT Sample dataset D FT This can be expressed as:

[0092] D FT =t k N,P,C,t k ∈T∩Δt,

[0093] Where N represents the measuring point, P represents the power, C represents the ambient temperature, and t represents the temperature. k This represents the k-th time point within the period T;

[0094] S412. Input the concentrated data of the sample dataset into the Bollinger evaluation model to obtain the Bollinger evaluation results, and obtain the temperature fluctuation characteristics of the wind turbine measurement point to be identified based on the Bollinger evaluation results.

[0095] In this embodiment, the temperature of the fan to be identified can be monitored and analyzed in detail using the time-series data of a single fan and the target temperature measurement point. This helps to understand the operating status and temperature changes of the fan to be identified. The abnormal fluctuation of the fan temperature is judged by the Boolean evaluation model. The warning risk characteristics of the fan to be identified are compared with the warning information of the supervised learning. If the warning risk characteristics of the fan to be identified contain the warning information of the supervised learning, a warning is issued. This allows for timely detection of abnormalities in the fan and the implementation of corresponding maintenance and repair measures, thereby improving the reliability and safety of the fan.

[0096] Furthermore, the calculation method of the Bollinger evaluation model specifically includes:

[0097] BW = UB - LB;

[0098] Where BW represents the bandwidth of the Bollinger Bands evaluation model, UB represents the upper band of the Bollinger Bands evaluation model, and LB represents the lower band of the Bollinger Bands evaluation model.

[0099] LB = LB tk =mint k N

[0100]

[0101]

[0102] Among them, LB tk The value of the lower bound of the Bollinger evaluation model represents the minimum value of the temperature measurement point, Q represents the Bollinger evaluation result, and t represents the minimum value of the lower bound of the model. k (N) represents t k The temperature of a moment.

[0103] In this embodiment of the application, the upper and lower rails of the Bollinger Bands are calculated using the Bollinger evaluation model to determine whether the temperature at the measuring point is within the normal range. If the temperature at the measuring point exceeds the upper rail or is lower than the lower rail, it indicates that there is an abnormal fluctuation in the temperature at the measuring point, which requires attention.

[0104] S42. Use the density fit model to identify the single-fan operating condition dimension data to obtain the trend change characteristics of the temperature at the measuring point of the fan to be identified.

[0105] In this embodiment, a density matching model is used to identify the operating condition dimension data of a single fan, which can accurately identify the trend change characteristics of the temperature at the measuring point of the fan to be identified, improve the accurate identification of the temperature trend change of the measuring point and the refined feature extraction, and improve the accuracy and reliability of the identification.

[0106] like Figure 4 and Figure 5 As shown, specifically, step S42 includes:

[0107] S421. Perform density mapping on the single fan operating condition dimension data to obtain the density mapping function.

[0108] It is understandable that the data of a single wind turbine operating condition is mapped according to the time dimension to ensure the accuracy and consistency of the data.

[0109] Specifically, density mapping function for:

[0110]

[0111] in, For the wind turbine to be identified at T1, ... T j Temperature data sets corresponding to different periods, T i This represents the i-th period, specifically...

[0112]

[0113] Where M is the generator temperature of the wind turbine to be identified, P is the power of the wind turbine to be identified, C is the ambient temperature, and t is the temperature of the wind turbine. k This represents the k-th time point within the period T, which is 15 seconds, and Δt represents the time interval.

[0114] In this embodiment of the application, by performing density mapping on the single fan operating condition dimension data, the data can be visualized and analyzed in the operating condition dimension, which can provide a more intuitive understanding of the operating condition characteristics of a single fan, including the temperature distribution and trend.

[0115] S422. The density mapping function is input into the density matching model for calculation to obtain the density matching degree of single wind turbine operating condition dimension data.

[0116] In this embodiment of the application, by density mapping and density fit calculation, the degree of fit between the operating data of a single fan and the ideal operating conditions is evaluated, and it is determined whether the temperature distribution of the target data is similar in different periods. By comparing the temperature distribution characteristics under different cycles, it is possible to identify whether there is a risk of trend change in temperature during periodic changes, which helps maintenance personnel to detect abnormal temperature changes in a timely manner and take corresponding measures for maintenance and repair, thereby improving the reliability and safety of the fan.

[0117] Furthermore, the calculation method for the density matching model is as follows:

[0118]

[0119] in,

[0120] in, Indicates in T A and T B The timeframe is to identify the density fit of the fan's temperature data, where n is the number of subsets divided according to the period T, and S is the density fit. k Let g represent a subset, k∈[1,n], and g be the density mapping function. Indicates in T A The target is to identify the temperature dataset of the fan. T represents B We need to identify the temperature dataset of the wind turbine.

[0121] In this embodiment, density fit is used to evaluate the similarity between the operating data of a single fan and the preset ideal operating data. A higher density fit indicates that the operating data of a single fan is more consistent with the ideal operating data, while a lower density fit indicates a greater difference.

[0122] S423. Calculate the overlap value using the density matching degree, and determine the trend change characteristics of the temperature at the measuring point of the fan to be identified based on the overlap value.

[0123] In this embodiment, the overlap value is calculated using the density matching degree to quantify the trend of temperature change at a single fan measuring point. The overlap value reflects the similarity between the temperature at a single fan measuring point and the ideal operating condition. A higher overlap value indicates that the temperature trend at the single fan measuring point is consistent with the ideal operating condition, while a lower overlap value indicates a significant difference.

[0124] Specifically, the formula for calculating the overlap value df is as follows:

[0125]

[0126] Among them U s(M) This represents the set formed by slicing the temperature data set M times. Indicates in T A and T B The density and consistency of the temperature data of the fan are expected to be identified.

[0127] Understandable. This indicates the existence of slices such that the two slices do not overlap at all. This indicates that there is no slice and the two slices completely overlap.

[0128] In this embodiment of the application, by inputting the target measuring point under the T-type cycle, the temperature distribution characteristics of the target measuring point under different cycles are obtained. By using the density matching model, the trend change characteristics of the temperature distribution of the generator winding temperature in the wind turbine to be identified under different cycles are obtained, thereby identifying whether there is a risk that the current temperature of the wind turbine to be identified will change with the cycle.

[0129] By conducting a comprehensive analysis of the temperature distribution of the generator windings in the wind turbine under different periods, we can gain a deeper understanding of the temperature change patterns and distribution characteristics, provide early warnings of potential faults or problems, and take timely and appropriate measures to avoid excessive damage to the wind turbine and other equipment.

[0130] S43. Use the correlation model to identify the operating condition dimension data of the same model to obtain the difference characteristics of the temperature of the measuring point of the fan to be identified.

[0131] By calculating the correlation between the wind turbine to be identified and other wind turbines, the similarity between the wind turbine to be identified and other wind turbines under the same operating conditions can be determined. This allows for the rapid identification of any abnormal risks in the same region, model, and cycle of the wind turbine to be identified, and timely implementation of corresponding measures for maintenance and repair, thereby improving the reliability and safety of the wind turbine to be identified.

[0132] like Figure 6 As shown, specifically, step S43 includes:

[0133] S431. The same model operating condition dimension data is converted to obtain the converted data.

[0134] In this embodiment of the application, data transformation is used to eliminate scale differences between different data, reduce the impact of noise and outliers, and extract more representative features.

[0135] S432. Input the converted data into the correlation model for calculation to obtain the correlation of the data of the same model operating condition dimension.

[0136] It is understandable that different correlation calculation methods can be used to calculate the similarity between the wind turbine to be identified and other wind turbines, such as Pearson correlation coefficient, graph intersection method, area overlap method, scatter density method, curve fitting method, etc., and this application does not impose specific limitations on this. Among them, a higher correlation indicates a strong linear relationship between the data, while a lower correlation indicates a weaker relationship between the data.

[0137] Specifically, the formula for calculating the correlation model is as follows:

[0138]

[0139]

[0140]

[0141] Wherein, p(F i F j ) indicates the fan to be identified, F i Operating condition data of the same model and other wind turbines F j Correlation of data on the same model's operating conditions, D”F i Indicates the fan to be identified, F i The collection of data after data transformation based on the same model's operating conditions, D”F j Indicates other wind turbines F j "D" represents the set of data after data transformation of the same model's operating condition data. Indicates the fan to be identified, F i Data set of the same model operating conditions during period T.

[0142] S433. Determine the similarity of the operating condition dimension data of the same model through the correlation of the operating condition dimension data of the same model.

[0143] It is understandable that similarity can be calculated using a formula, which includes the result of the correlation calculation and other parameters. The similarity can be defined and adjusted according to specific needs, and this application does not impose any specific restrictions on it.

[0144] S434. Determine the difference characteristics of the temperature measurement points of the fan to be identified based on the similarity.

[0145] By combining decision theory and discrete methods, we can quickly identify whether there are abnormal risks in the same region, model and cycle of the wind turbine to be identified. We can set thresholds or use other discrimination methods to determine whether abnormal risks exist.

[0146] In this application embodiment, by analyzing and comparing the operating condition data of the same model, it can help determine whether the temperature change of the fan to be identified is consistent with the operating condition of the same model, thereby timely detecting abnormalities and taking corresponding measures for maintenance and repair.

[0147] S44. Based on the fluctuation characteristics, trend characteristics, and difference characteristics, obtain the early warning risk characteristics of the wind turbine to be identified.

[0148] In this embodiment, by comprehensively considering the fluctuation characteristics, trend characteristics, and differences of the wind turbine to be identified, the early warning risk of the wind turbine can be assessed more comprehensively. This can better reflect the actual situation of the wind turbine, avoid missed or false alarms, and provide more accurate support for the normal operation and maintenance of the wind turbine. This improves the overall performance and user satisfaction of the early warning system. At the same time, it can help maintenance personnel better understand the performance and status of the wind turbine, provide more accurate support for preventive maintenance and optimized operation of the wind turbine, reduce the degree of equipment damage and maintenance costs, and improve the reliability and stability of the equipment.

[0149] In one embodiment of this application, step S6 further includes:

[0150] Determine whether the warning risk features of the wind turbine to be identified contain the warning information learned by supervised learning; if so, it indicates that there is abnormal data in the current temperature detection data of the wind turbine to be identified.

[0151] The warning risk features are confirmed to obtain the validity of the abnormal data in the warning risk features; that is, to determine whether the warning risk features are truly abnormal situations.

[0152] The invalid abnormal data in the early warning risk features are compared with the first target data interval to determine the position of the invalid abnormal data in the first target data interval; wherein, the first target data interval is the temperature range under normal operating conditions;

[0153] Based on the correspondence between the first target data interval and the first preset risk value, the first preset risk value corresponding to the first target data interval is determined; wherein the first preset risk value is the risk range corresponding to the abnormal temperature data;

[0154] The first preset risk value is determined as the risk level of the wind turbine to be warned.

[0155] In this embodiment of the application, by analyzing and processing the current temperature detection data of the wind turbine to be identified, abnormal conditions of generator winding temperature can be detected in a timely manner. The abnormal temperature data in the current temperature detection data is compared with historical data to determine the target data area where the abnormal data is located, thereby determining the degree of risk. The degree of risk is converted into a risk value, and the risk value is used to identify the degree of risk of the wind turbine to be identified, thereby providing early warning of the current temperature detection data of the wind turbine to be identified and taking corresponding measures.

[0156] In one embodiment of this application, the anomaly identification and early warning method further includes the following steps:

[0157] S7. The user confirms the data according to the warning prompt and provides a warning receipt. Based on the warning receipt, an evaluation model is established to evaluate the temperature anomaly identification model, and the evaluation result is obtained. The temperature anomaly identification model is updated according to the evaluation result.

[0158] Specifically, it is determined whether the early warning risk characteristics of the wind turbine to be identified are normal. If they are normal, an early warning is issued and an early warning receipt is provided.

[0159] The temperature anomaly identification model is evaluated based on the aforementioned early warning receipt, and an evaluation model is established.

[0160] The target parameters in the temperature anomaly identification model are evaluated using the evaluation model, and the temperature anomaly identification model is updated based on the evaluation results.

[0161] The evaluation results include the accuracy, false alarm rate, and false negative rate of the temperature anomaly identification model.

[0162] like Figure 7 As shown, Figure 7 This is a schematic diagram illustrating the evaluation results of a temperature anomaly identification model in one embodiment of this application. The target parameters of the temperature anomaly identification model are shown in Table 1.

[0163] Table 1

[0164]

[0165] As shown in Table 1, the formula for calculating accuracy is as follows:

[0166]

[0167] Where AT represents the number of correctly identified samples when the sample dataset is abnormal, and AF represents the number of incorrectly identified samples when the sample dataset is abnormal.

[0168] The false alarm rate (Misreport) is calculated using the following formula:

[0169]

[0170] The formula for calculating the false negative rate (Omission) is as follows:

[0171]

[0172] Where NF represents the number of identification errors when the sample dataset is normal.

[0173] Furthermore, the formula for the target parameter is:

[0174]

[0175] Where α represents the set of parameters for the evaluation model, and T j Represents a period at any given time. For α in T j The adaptive adjustment is given by ω, which represents the contribution of α to the evaluation model.

[0176] The formula for calculating the contribution ω is as follows:

[0177]

[0178]

[0179]

[0180] in, X is [T] j--1 T j Samples for interval identification This represents the receipt of the identified sample, and A is the set of parameter matrices.

[0181] Optimal solution for objective parameters: for Δα i The effectiveness of the attenuation is determined by evaluation indicators. That is:

[0182]

[0183]

[0184]

[0185] Where, α i This represents the i-th data point in the set of parameters for the evaluation model.

[0186] Understandably, stopping at T occurs when the following conditions are met. j Adaptive adjustment of α under periodicity:

[0187]

[0188]

[0189]

[0190]

[0191] After obtaining the target evaluation index through the above parameter adaptive mechanism, the target parameters can be identified based on the target evaluation index.

[0192] Generally, the triggering mechanisms for alpha adaptive adjustment include fixed-period mode and evaluation index mode. Fixed-period mode typically involves setting a time interval, with the task triggering alpha adaptive adjustment; evaluation index mode typically involves real-time monitoring of certain indicators to trigger alpha adaptive adjustment.

[0193] In this embodiment of the application, by using a parameter adaptive mechanism, target evaluation indicators can be obtained efficiently, thereby accurately identifying target parameters, reducing manual intervention, avoiding errors caused by human factors, saving a lot of time and resources, and improving identification efficiency and accuracy.

[0194] In one embodiment of this application, step S71 further includes:

[0195] When the warning risk characteristics of the wind turbine to be identified are normal, a warning prompt will be issued. The user will then confirm the data based on the warning prompt. Once the user confirms, the warning prompt will be cleared, and the historical warning information will be updated based on the warning prompt and the corresponding confirmation operation.

[0196] In this embodiment, the system monitors the early warning risk characteristics of the wind turbine to be identified, such as generator winding temperature, and compares the monitored data with historical data to determine if any anomalies exist. If the early warning risk characteristics are normal, i.e., no anomalies are detected, the system issues an early warning notification. The notification can be sent to the user via pop-up windows, SMS, email, etc. The user can view the monitoring data, historical data, and related alarm information to determine whether an anomaly exists. When the user confirms an anomaly, they perform a confirmation operation, such as clicking a confirmation button or replying to a confirmation email. The purpose of the confirmation operation is to clear the early warning notification and record the confirmed anomaly. Based on the early warning notification and the corresponding confirmation operation, the historical early warning information is updated. The update may include marking confirmed anomalies as processed and recording the processing results. This application can effectively avoid false alarms or missed alarms, increase user attention and participation in monitoring data, and take timely measures to prevent further deterioration of anomalies or losses. Updating and improving historical early warning information provides a more accurate basis for subsequent early warning judgments and analyses.

[0197] The present invention also provides a wind power generation system, the wind power generation system including at least one wind turbine, the wind turbine including at least one generator winding, the wind power generation system further including: a memory, a processor, a communication bus, and a control program stored in the memory: the communication bus is used to realize the connection communication between the processor and the memory; the processor is used to execute the control program to implement the steps of the abnormal identification and early warning method of the generator winding as described in any one of the above.

[0198] like Figure 8 As shown, this application also provides an anomaly identification and early warning system based on supervised learning, including a processor 1001, such as a CPU, a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005.

[0199] The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be high-speed RAM or stable, non-volatile memory, such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0200] Understandably, the memory 1005 may include an operating system, a network communication module, a user interface module, and a control program. The operating system is a program that manages and controls the wind power generation system and software resources, supporting the operation of the network communication module, user interface module, control program, and other programs or software. The network communication module manages and controls the network interface 1004; the user interface module manages and controls the user interface 1003.

[0201] The network interface 1004 is mainly used to connect to the backend server and communicate with the backend server; the user interface 1003 is mainly used to connect to the client (user end) and communicate with the client; the processor 1001 can call the control program stored in the memory 1005 and execute the generator winding abnormality identification and early warning method.

[0202] This application also provides a computer-readable storage medium storing computer instructions that cause the computer to implement the generator winding anomaly identification and early warning method as described in any of the preceding claims.

[0203] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for identifying and warning of anomalies in generator windings, characterized in that, Includes the following steps: S1. Obtain the current temperature detection data of the fan to be identified; S2. Process the current temperature detection data to obtain processed data; S3. Perform dimensional transformation on the processed data to obtain single fan time-series dimensional data, single fan operating condition dimensional data, and same model operating condition dimensional data. S4. Use the temperature anomaly identification model to identify the time-series data of the single fan, the operating condition data of the single fan, and the operating condition data of the same model to obtain the early warning risk characteristics of the fan to be identified. S5. Obtain historical early warning information and historical risk values ​​of the generator windings in the wind power generation system, and generate early warning information for supervised learning; S6. Compare the early warning risk features of the wind turbine to be identified with the early warning information of the supervised learning. If the early warning risk features of the wind turbine to be identified contain the early warning information of the supervised learning, then issue an early warning. The temperature anomaly identification model includes a Bollinger evaluation model, a density matching model, and a correlation model. The generator winding is equipped with target temperature measuring points. Step S4 specifically includes: S41. Use the Bollinger evaluation model to identify the time-series data of a single fan to obtain the fluctuation characteristics of the temperature at the measuring point of the fan to be identified. S42. Use the density matching model to identify the single-fan operating condition dimension data to obtain the trend change characteristics of the temperature at the measuring point of the fan to be identified; S43. Use the correlation model to identify the operating condition dimension data of the same model to obtain the difference characteristics of the temperature of the measuring point of the fan to be identified; S44. Based on the fluctuation characteristics, trend characteristics, and difference characteristics, obtain the early warning risk characteristics of the wind turbine to be identified.

2. The method for abnormal identification and early warning of generator windings as described in claim 1, characterized in that, Step S41 specifically includes: The generator winding temperatures at the target temperature measurement points within the operating cycle T form a sample dataset D. FT Sample dataset D FT This can be expressed as: ； Where N represents the measuring point, P represents the power, C represents the ambient temperature, and t represents the temperature. k This indicates the k-th time point within the period T. Indicates a time interval; The sample dataset is input into the Bollinger evaluation model to obtain the Bollinger evaluation results. Based on the Bollinger evaluation results, the temperature fluctuation characteristics of the measuring point of the wind turbine to be identified are obtained.

3. The method for abnormal identification and early warning of generator windings as described in claim 2, characterized in that, The calculation method of the Bollinger evaluation model specifically includes: ； Where BW represents the bandwidth of the Bollinger Bands evaluation model, UB represents the upper band of the Bollinger Bands evaluation model, and LB represents the lower band of the Bollinger Bands evaluation model. ； ； ； Among them, LB tk The value of the lower bound of the Bollinger evaluation model represents the minimum value of the temperature measurement point, Q represents the Bollinger evaluation result, and t represents the minimum value of the lower bound of the model. k (N) represents t k The temperature of a moment.

4. The generator winding anomaly identification and early warning method as described in claim 1, characterized in that, Step S42 specifically includes: The density mapping function is obtained by performing density mapping on the single-fan operating condition dimension data. The density mapping function is input into the density matching model for calculation to obtain the density matching degree of single wind turbine operating condition dimension data. The overlap value is calculated using the density matching degree, and the trend change characteristics of the temperature at the measuring point of the fan to be identified are determined based on the overlap value.

5. The generator winding anomaly identification and early warning method as described in claim 4, characterized in that, The specific calculation method for the density matching model is as follows: ； in, ; in, This represents the density fit of the temperature data of the fan expected to be identified at TA and TB, where n is the number of subsets divided according to the period T, and S is the density fit. k Let g represent a subset set, k∈[1,n], and g be the density mapping function. Indicates in T A The target is to identify the temperature dataset of the fan. T represents B We need to identify the temperature dataset of the wind turbine.

6. The method for abnormal identification and early warning of generator windings as described in claim 1, characterized in that, Step S43 specifically includes: The data of the same model operating conditions is transformed to obtain the transformed data; The transformed data is input into the correlation model for calculation to obtain the correlation of data in the same model operating condition dimension. The similarity of the operating condition data of the same model is determined by the correlation of the operating condition data of the same model. The differences in temperature at the measuring points of the fan to be identified are determined based on the similarity.

7. The generator winding anomaly identification and early warning method as described in claim 6, characterized in that, The formula for calculating the correlation model is as follows: ； ； ； Wherein, p(F) i F j ) indicates the fan to be identified, F i Operating condition data of the same model and other wind turbines F j The correlation of data on the same model and operating conditions. Indicates the fan to be identified, F i The collection after data transformation of operating condition data of the same model. Indicates other wind turbines F j The collection after data transformation of operating condition data of the same model. This represents the dataset after data transformation of operating condition data for the same model. Indicates the fan to be identified, F i Data set of the same model operating conditions during period T.

8. The generator winding anomaly identification and early warning method as described in claim 1, characterized in that, The anomaly identification and early warning method also includes the following steps: S7. The user confirms the data according to the warning prompt and provides a warning receipt. Based on the warning receipt, an evaluation model is established to evaluate the temperature anomaly identification model, and the evaluation result is obtained. The temperature anomaly identification model is updated according to the evaluation result.

9. A wind power generation system, characterized in that, The wind power generation system includes at least one wind turbine, and the wind turbine includes at least one generator winding. The wind power generation system also includes: a memory, a processor, a communication bus, and a control program stored in the memory. The communication bus is used to enable communication between the processor and the memory; The processor is used to execute the control program to implement the steps of the generator winding anomaly identification and early warning method as described in any one of claims 1-8.

Images