Fan temperature measurement data processing method and system

Abstract

The application relates to the technical field of wind turbine temperature measurement data processing, in particular to a wind turbine temperature measurement data processing method and system. The method comprises the following steps: acquiring real-time active power data and real-time temperature data; performing time alignment processing, constructing a power temperature synchronous data sequence, identifying a dynamic response working condition, extracting a dynamic response power input sequence and a dynamic response temperature output sequence, calculating a dynamic response characteristic parameter, and recording a reference benchmark value; constructing a dynamic response characteristic parameter sequence; judging whether the dynamic response characteristic parameter sequence is a change mode; after identifying a benign response change, updating the reference benchmark value of the dynamic response characteristic parameter; judging whether the dynamic response characteristic parameter sequence is a change mode, and identifying a real fault. The method has the advantages that the existing system cannot distinguish between benign response changes and real equipment faults, thereby causing misjudgment, unnecessary operation and maintenance costs, and reduced trust in the diagnosis system.

Description

Technical Field

[0001] This application relates to the field of wind turbine generator temperature measurement data processing technology, and more specifically, to a wind turbine temperature measurement data processing method and system. Background Technology

[0002] Modern wind power technology plays an increasingly important role in ensuring the stable operation of the power grid. To adapt to the grid's rapid response and peak-shaving demands, wind turbine generators need to frequently and drastically adjust their power output. During this process, the temperature of critical components inside the turbine changes dramatically, placing higher demands on the accuracy and reliability of the temperature measurement system. However, in actual operation and maintenance, even minor changes to the hardware characteristics of the temperature signal transmission link, such as replacing the signal cable with a new model, can lead to slight distortions in the dynamic characteristics of the temperature signal.

[0003] Existing fan temperature measurement data processing systems often rely on calibrating their diagnostic logic based on the characteristics of older hardware, making it difficult to distinguish between signal changes resulting from benign hardware modifications and actual equipment malfunctions. For example, when the signal cable in the temperature signal transmission link is replaced with a new model, even if the new cable itself is of good performance, its inherent electrical characteristics may cause a systematic deviation in the dynamic response characteristics of the temperature signal (such as the rate of change or the acceleration of the rate of change). This deviation causes the signal received by the system to be inconsistent with the "healthy" model built based on the data from the old cable, thus leading to misjudgments as "abnormal sensor response" or "abnormal signal transmission link."

[0004] This misjudgment leads the maintenance team to conduct frequent and unnecessary on-site inspections and troubleshooting, resulting in wind turbine downtime losses and wasting spare parts and human resources. More seriously, these recurring "pseudo-fault" alarms that cannot be resolved through conventional means severely affect the maintenance team's trust in the diagnostic system, and may even lead to the degradation or blocking of system functions, reducing the efficiency of wind turbine maintenance and contradicting the original intention of introducing the intelligent prediction system.

[0005] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0006] This application discloses a method and system for processing fan temperature measurement data, which aims to solve the problem that existing fan temperature measurement data processing systems have difficulty distinguishing between benign response changes caused by changes in hardware characteristics and real equipment failures, thereby leading to misjudgments, unnecessary operation and maintenance costs, and reduced trust in the diagnostic system.

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

[0008] In a first aspect, this application discloses a method for processing fan temperature measurement data, including:

[0009] Acquire real-time active power data and real-time temperature data of the wind turbine generator set;

[0010] The collected real-time active power data and real-time temperature data are time-aligned to construct a power-temperature synchronized data sequence.

[0011] Based on the power-temperature synchronization data sequence, dynamic response conditions with significant power fluctuation characteristics are identified, and the dynamic response power input sequence and dynamic response temperature output sequence are extracted from the corresponding power-temperature synchronization data sequence.

[0012] Based on the dynamic response power input sequence and the dynamic response temperature output sequence, the dynamic response characteristic parameters describing its response characteristics are calculated, and its reference values ​​are recorded.

[0013] Continuously record dynamic response feature parameters and construct a dynamic response feature parameter sequence;

[0014] Determine whether the dynamic response characteristic parameter sequence exhibits a one-time deviation and then stabilizes, in order to identify benign response changes caused by variations in hardware characteristics;

[0015] After identifying a benign change in response, update the reference baseline value of the dynamic response characteristic parameters;

[0016] Determine whether there is a persistent shift or increased fluctuation pattern in the dynamic response characteristic parameter sequence in order to identify real faults in the temperature measurement signal transmission link.

[0017] Through the above technical solution, this application can effectively distinguish between benign response changes caused by hardware characteristic changes and real faults in the fan temperature measurement signal transmission link, avoiding misjudgments caused by the inability of traditional methods to distinguish between the two, thereby reducing unnecessary operation and maintenance costs and downtime losses, and improving the accuracy and reliability of the diagnostic system.

[0018] Furthermore, determining whether the dynamic response characteristic parameter sequence exhibits a one-time deviation followed by a stabilization pattern includes:

[0019] After identifying the dynamic response condition, the dynamic response power input sequence and dynamic response temperature output sequence within a fixed time window are extracted to construct a response morphology segment.

[0020] The response pattern fragments are normalized and their difference from historical baseline patterns is calculated to obtain the deviation index.

[0021] Add the deviation index to the dynamic response feature parameter sequence;

[0022] When the deviation index shows a sudden change in its average value within a preset time period and the fluctuation amplitude is lower than a preset fluctuation threshold, it is identified as a benign response change caused by a change in hardware characteristics.

[0023] By normalizing response pattern fragments and calculating differences, and combining the abrupt changes and fluctuations of deviation indicators, the system can accurately identify benign response changes, avoid misjudging small and stable changes, and improve the system's adaptability to changes in hardware characteristics.

[0024] Furthermore, after identifying a benign response change, the reference baseline values ​​of the dynamic response characteristic parameters are updated, including:

[0025] After identifying benign response changes, multiple subsequent response pattern fragments are obtained and normalization is performed.

[0026] The stability of multiple normalized response form fragments is evaluated, and sample fragments that meet the preset stability conditions are selected.

[0027] The sample fragments are aggregated to generate an updated response baseline.

[0028] The updated reference baseline values ​​are calculated based on the dynamic response characteristic parameters using the response baseline shape.

[0029] The above technical solution enables the dynamic updating of reference baseline values ​​based on the aggregation and processing of subsequent stable sample fragments. This ensures that the system can quickly adapt and re-establish accurate health baselines after benign changes in hardware characteristics, thereby maintaining diagnostic accuracy.

[0030] Furthermore, determining whether the dynamic response characteristic parameter sequence exhibits a persistent shift or intensified fluctuation pattern is crucial for identifying actual faults in the temperature measurement signal transmission link, including:

[0031] The dynamic response feature parameter sequence is subjected to sliding segmentation to construct multiple dynamic response feature parameter segments;

[0032] Calculate the mean and fluctuation amplitude of each dynamic response characteristic parameter segment;

[0033] If the mean continuously deviates from the reference benchmark value, or if the fluctuation amplitude continuously exceeds the preset fluctuation threshold, a real fault is identified.

[0034] The above technical solution can effectively identify persistent signal anomalies by segmenting the dynamic response characteristic parameter sequence and analyzing its mean and fluctuation amplitude, thereby accurately judging the real fault in the temperature measurement signal transmission link and avoiding misjudgment of instantaneous fluctuations.

[0035] Furthermore, the dynamic response power input sequence and dynamic response temperature output sequence are extracted from the corresponding power-temperature synchronization data sequence, including:

[0036] Obtain peak-shaving control commands from wind turbine generators;

[0037] Calculate the rate of change of real-time active power data within a set time window;

[0038] When the peak shaving control command is active and the rate of change exceeds the preset change threshold, the current period is determined to be a dynamic response condition.

[0039] Extract the power input sequence and temperature output sequence corresponding to the current time period from the power and temperature synchronization data sequence, and use them as the dynamic response power input sequence and dynamic response temperature output sequence.

[0040] By combining the above technical solutions with peak-shaving control commands and power change rates, the dynamic response conditions of wind turbines during peak-shaving can be accurately identified, ensuring that the extracted power input and temperature output sequences can truly reflect the system's response characteristics under severe conditions, thus improving the accuracy of data extraction.

[0041] Further, determining if the mean continuously deviates from the reference benchmark value, or if the fluctuation amplitude continuously exceeds a preset fluctuation threshold, includes:

[0042] Statistical process monitoring is performed on the dynamic response characteristic parameter segments to construct control limits for reference baseline values;

[0043] Determine whether the mean or fluctuation range of each dynamic response characteristic parameter segment continuously exceeds the control limit, or whether it exhibits a non-random offset trend;

[0044] When any of the judgment conditions are met, it is determined that there is a pattern of continuous shift or increased fluctuation.

[0045] By introducing statistical process monitoring methods through the above technical solutions, and by constructing control limits and judging whether the mean or fluctuation range continuously exceeds or shows a non-random deviation trend, persistent anomalies can be identified more scientifically and sensitively, thus improving the reliability of fault diagnosis.

[0046] Furthermore, statistical process monitoring is performed on the dynamic response characteristic parameter segments to construct control limits for reference baseline values, including:

[0047] Extract dynamic response characteristic parameter fragments from multiple historical time periods and calculate their long-term mean and standard deviation of fluctuation.

[0048] The initial control limit interval is constructed based on the long-term mean and the standard deviation of the fluctuation.

[0049] Perform trend analysis on the dynamic response characteristic parameter segments of the current time period. When it is determined to be a stable drift pattern, update the long-term mean and standard deviation of fluctuation.

[0050] The upper and lower thresholds of the initial control limit interval are dynamically adjusted based on the updated long-term mean and standard deviation of fluctuation, and then used as the control limits.

[0051] The above technical solution can dynamically correct the control limits to adapt to the slow and stable drift that may occur during the operation of the wind turbine, avoiding false alarms or missed alarms caused by fixed limits and improving the adaptability of monitoring.

[0052] Furthermore, trend analysis is performed on the dynamic response characteristic parameter segments of the current time period, including:

[0053] Calculate the mean and fluctuation range of dynamic response characteristic parameter segments within the current time period;

[0054] A sliding window approach is used to compare the mean and fluctuation range of multiple consecutive time periods to identify trend evolution patterns;

[0055] When the mean continues to change slowly according to a preset time and the fluctuation range is within a preset stable range, the trend evolution pattern is identified as a drift pattern.

[0056] The above technical solution enables the analysis of continuous changes in mean and fluctuation amplitude through a sliding window, accurately identifying stable drift patterns and providing an accurate basis for the dynamic correction of subsequent control limits.

[0057] Furthermore, when a stable drift pattern is identified, the long-term mean and standard deviation of volatility are updated, including:

[0058] Based on multiple dynamic response characteristic parameter segments within the current time period, the long-term mean is recalculated using a moving average or exponential weighting method to generate an updated reference benchmark value.

[0059] Adjust the upper and lower thresholds of the initial control limit range based on the updated reference baseline and standard deviation of fluctuation.

[0060] The above technical solutions enable the smooth updating of long-term averages using methods such as moving averages or exponential weighting, ensuring that the updating process of the reference benchmark is stable and representative, and further enhancing the system's adaptability to long-term drift.

[0061] Secondly, this application also discloses a fan temperature measurement data processing system, the system comprising:

[0062] The data acquisition module is used to acquire real-time active power data and real-time temperature data of the generator windings of the wind turbine generator set.

[0063] The synchronization processing module is used to perform time alignment processing on the collected real-time active power data and real-time temperature data to construct a power and temperature synchronization data sequence.

[0064] The operating condition identification and recording module is used to identify dynamic response operating conditions with significant power fluctuation characteristics based on the power-temperature synchronization data sequence, and to extract the dynamic response power input sequence and dynamic response temperature output sequence from the corresponding power-temperature synchronization data sequence.

[0065] The parameter calculation module calculates the dynamic response characteristic parameters describing its response characteristics based on the dynamic response power input sequence and the dynamic response temperature output sequence, and records its reference benchmark value.

[0066] The parameter tracking module is used to continuously record dynamic response feature parameters and construct a dynamic response feature parameter sequence.

[0067] The benign change identification module is used to determine whether there is a one-time deviation and stabilization pattern in the dynamic response characteristic parameter sequence, so as to identify benign response changes caused by changes in hardware characteristics;

[0068] The baseline update module is used to update the reference baseline values ​​of the dynamic response characteristic parameters after a benign response change is detected.

[0069] The real fault identification module is used to determine whether there is a continuous shift or increased fluctuation pattern in the dynamic response characteristic parameter sequence, so as to identify real faults in the temperature measurement signal transmission link.

[0070] Through the above technical solutions, this application provides a system that can effectively distinguish between benign hardware characteristic changes and real faults. Through modular design, it realizes the complete process of data acquisition, synchronous processing, operating condition identification, parameter calculation, parameter tracking, benign change identification, benchmark update and real fault identification, providing a hardware foundation for intelligent diagnosis of fan temperature measurement systems.

[0071] Beneficial effects

[0072] This application effectively solves the problem of existing fan temperature measurement data processing systems struggling to distinguish between benign signal changes and actual equipment faults when faced with minor changes in hardware characteristics. Specifically, by introducing dynamic response characteristic parameters and performing refined analysis of their change patterns, this application can accurately identify one-off, stabilizing response changes caused by hardware characteristic changes such as replacing signal cables, and promptly update the reference baseline value, thus avoiding misjudging such benign changes as faults. Simultaneously, for change patterns of persistent deviation or increased fluctuations, this application can accurately identify them as actual temperature measurement signal transmission link faults. Therefore, the method of this application significantly improves the accuracy and reliability of fan temperature measurement system diagnosis, reduces unnecessary on-site inspections and downtime losses, saves maintenance costs, and enhances the maintenance team's trust in the diagnostic system, overcoming the shortcomings of existing technologies such as frequent "false fault" alarms and reduced maintenance efficiency. Attached Figure Description

[0073] Figure 1 This is a flowchart illustrating a method for processing fan temperature measurement data provided in this application.

[0074] Figure 2 A flowchart of a fan temperature measurement data processing system provided in this application.

[0075] In the diagram: 1. Data acquisition module; 2. Synchronization processing module; 3. Operating condition identification and recording module; 4. Parameter calculation module; 5. Parameter tracking module; 6. Beneficial change identification module; 7. Baseline update module; 8. Real fault identification module. Detailed Implementation

[0076] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0077] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0078] Traditional wind turbine temperature data processing systems struggle to accurately distinguish between dynamic temperature signal distortion caused by minor hardware changes in the temperature signal transmission link (such as replacing the signal cable with a newer model) and actual equipment malfunctions when wind turbines frequently adjust their power output. This diagnostic logic often relies on calibrating based on outdated hardware characteristics, leading to signals received by the system that don't match the "health" model built from older data. This results in misdiagnosis as "abnormal sensor response" or "abnormal signal transmission link." If these issues aren't addressed, such misdiagnosis will force maintenance teams to conduct frequent, unnecessary on-site inspections and troubleshooting, causing wind turbine downtime losses, wasting spare parts and manpower, and severely impacting the maintenance team's trust in the diagnostic system, ultimately reducing wind turbine maintenance efficiency.

[0079] Reference Figure 1 In response, this application proposes a method for processing fan temperature measurement data, including:

[0080] S1000: Acquires real-time active power data and real-time temperature data of the generator windings of the wind turbine generator set;

[0081] S2000: Performs time alignment processing on the collected real-time active power data and real-time temperature data to construct a power-temperature synchronized data sequence;

[0082] S3000: Based on the power-temperature synchronization data sequence, identify dynamic response conditions with significant power fluctuation characteristics, and extract the dynamic response power input sequence and dynamic response temperature output sequence from the corresponding power-temperature synchronization data sequence.

[0083] S4000: Based on the dynamic response power input sequence and the dynamic response temperature output sequence, calculate the dynamic response characteristic parameters that describe its response characteristics, and record its reference value.

[0084] S5000: Continuously records dynamic response characteristic parameters and constructs a dynamic response characteristic parameter sequence;

[0085] S6000: Determines whether the dynamic response characteristic parameter sequence has a one-time deviation and tends to stabilize change pattern, in order to identify benign response changes caused by changes in hardware characteristics;

[0086] S7000: After detecting a change in benign response, update the reference baseline value of the dynamic response characteristic parameters;

[0087] S8000: Determines whether there is a persistent shift or increased fluctuation pattern in the dynamic response characteristic parameter sequence in order to identify real faults in the temperature measurement signal transmission link.

[0088] Therefore, this method effectively solves the problem of misjudging benign hardware changes and real faults in existing technologies, and improves the accuracy of diagnosis and the efficiency of operation and maintenance.

[0089] In this embodiment, the real-time active power data of the wind turbine generator set typically refers to the electrical power data output by the wind turbine generator set in real time during operation, which can be collected by power sensors on the grid side or inside the generator set.

[0090] Real-time temperature data of generator windings usually refers to the real-time temperature of the windings inside the generator of a wind turbine generator set, which can be collected by an embedded temperature sensor (such as a PT100 resistance temperature detector).

[0091] Time alignment refers to synchronizing data from different sources and with different sampling frequencies (such as power and temperature data) along a timeline using methods such as timestamps or interpolation, to facilitate joint analysis. A power-temperature synchronized data sequence refers to a dataset where power and temperature data are arranged chronologically after time alignment.

[0092] Dynamic response conditions refer to the operating state of a wind turbine generator when its power output fluctuates significantly (e.g., peak shaving, start-up / shutdown, load abrupt changes). Dynamic response power input sequence refers to the sequence of power output changes of the wind turbine generator over time under dynamic response conditions. Dynamic response temperature output sequence refers to the sequence of generator winding temperature changes over time under dynamic response conditions; this change is a response to changes in power input. Dynamic response characteristic parameters are parameters that describe the dynamic response characteristics between power input and temperature output, such as response time, overshoot, damping ratio, rise time, settling time, or specific transfer function parameters. These parameters quantify the system's temperature response behavior to power changes.

[0093] A reference baseline value refers to the expected or standard value of a dynamic response characteristic parameter under normal system conditions or specific hardware configurations, used to subsequently determine whether the parameter has become abnormal. A one-time deviation and subsequent stabilization pattern refers to a one-time, non-continuous jump in the dynamic response characteristic parameter at a certain point in time, followed by stabilization at a new level, usually caused by hardware replacement or system parameter adjustments.

[0094] A persistent deviation or increased fluctuation pattern refers to a sustained deviation of the mean value of the dynamic response characteristic parameters from the reference value, or a sustained increase in the amplitude of its fluctuations, which usually indicates a system fault or performance degradation. Benign response changes refer to changes in system response characteristics caused by non-fault factors (such as hardware upgrades or component replacements) that do not affect the normal function of the system. A real fault in the temperature measurement signal transmission link refers to a hardware or software failure occurring on the signal transmission path from the sensor to the data processing unit, leading to signal distortion or loss and affecting the accuracy of temperature measurement. This method can typically be implemented in the monitoring center of a wind farm or in the Supervisory Control and Data Acquisition (SCADA) system of a wind turbine generator, utilizing industrial control computers or dedicated data processing servers for data processing and analysis.

[0095] The implementation principle of this application is as follows: First, it is necessary to acquire real-time active power data and real-time temperature data of the generator windings of the wind turbine. One implementation method is to manually read the field instrument data periodically and input it into the data processing system. Another implementation method is to use a simple analog signal acquisition card to convert the analog signals output by the sensors into digital signals; however, the acquisition card may not have high-precision or multi-channel synchronous acquisition capabilities. Yet another implementation method is to set a preset fixed time interval (e.g., every minute) to retrieve data from the SCADA system of the wind turbine, regardless of the real-time nature of data updates or event triggering.

[0096] Secondly, the collected real-time active power data and real-time temperature data are time-aligned to construct a power-temperature synchronization data sequence. One implementation method is to use a simple timestamp matching method; if the timestamps of different data sources are not completely consistent, the mismatched data points are discarded. Another implementation method is to use linear interpolation to fill in the data with inconsistent timestamps, but this method may not be suitable for data with non-linear changes. Yet another implementation method is to set a wider time window, considering data falling within this window as synchronized, but this method may introduce a large time deviation. Based on the power-temperature synchronization data sequence, dynamic response conditions with significant power fluctuation characteristics are identified, and the dynamic response power input sequence and dynamic response temperature output sequence are extracted from the corresponding power-temperature synchronization data sequence. One implementation method is to manually set a fixed power change threshold; when the absolute value change of power exceeds this threshold, the current state is considered to be in dynamic response condition. Another implementation method is to simply monitor the absolute value change of power data, without considering its rate of change or its correlation with specific control commands. As another implementation, a sliding window with a fixed time period can be used to identify power fluctuations, but this method may not be able to accurately capture the start and end points of the dynamic response. Once the dynamic response condition is identified, the corresponding power input sequence and temperature output sequence within that time period can be extracted.

[0097] Based on this, dynamic response characteristic parameters describing its response characteristics are calculated using the dynamic response power input sequence and the dynamic response temperature output sequence, and their reference values ​​are recorded. As one implementation method, the simple lag time or maximum temperature rise of the temperature response can be calculated as the dynamic response characteristic parameter. As another implementation method, historical data can be statistically averaged once to obtain a fixed reference value, without considering its variation with time or operating conditions. As yet another implementation method, data from several typical operating conditions can be manually selected, and their characteristic parameters calculated as a reference. Subsequently, the dynamic response characteristic parameters are continuously recorded to construct a dynamic response characteristic parameter sequence. As one implementation method, the dynamic response characteristic parameters obtained each time can be simply appended to the database to form a time series. As another implementation method, a fixed frequency (e.g., every hour) can be set to record the parameters, regardless of the frequency of occurrence of the dynamic response operating conditions.

[0098] Next, it is determined whether the dynamic response characteristic parameter sequence exhibits a one-time deviation followed by a stabilizing pattern, in order to identify benign response changes caused by hardware characteristic variations. A fixed threshold can be set; when the difference between the current value of the dynamic response characteristic parameter and the reference value exceeds this threshold, it is considered a deviation. In another implementation, the dynamic response characteristic parameter sequence graph can be visually inspected to determine if a one-time deviation exists. As yet another implementation, the average values ​​of two consecutive time periods can be compared; if the difference is significant, a deviation is considered to exist, but further determination of whether it has stabilized is not required.

[0099] After identifying a benign change in response, the reference baseline value of the dynamic response characteristic parameter is updated. As one implementation, the first stable parameter value identified after the benign change can be directly used as the new reference baseline value. As another implementation, after confirming the benign change, a new reference baseline value can be manually entered through manual intervention.

[0100] Finally, the dynamic response characteristic parameter sequence is analyzed to determine if there is a persistent shift or increased fluctuation pattern, thus identifying a real fault in the temperature signal transmission link. One implementation method is to set a fixed upper and lower limit; if the dynamic response characteristic parameter exceeds this limit N times consecutively, it is considered a persistent shift. Another implementation method is to simply calculate the standard deviation of the parameter sequence; if the standard deviation exceeds a preset threshold, it is considered an increase in fluctuation. Yet another implementation method is to manually review the trend chart of the parameter sequence periodically to determine if there is a persistent shift or increased fluctuation.

[0101] The method provided in this application works by refining and analyzing real-time active power data and real-time temperature data of the wind turbine windings to accurately identify anomalies in the temperature measurement signal transmission link. Specifically, firstly, the real-time active power data and real-time temperature data of the generator windings are acquired and time-aligned to construct a synchronized power-temperature data sequence, laying the foundation for subsequent joint analysis. Secondly, based on this synchronized data sequence, the system can identify dynamic response conditions with significant power fluctuations and extract the corresponding dynamic response power input sequence and dynamic response temperature output sequence. This step ensures that the analysis focuses on the critical operating conditions where the wind turbine's response characteristics best reflect its health status.

[0102] Subsequently, based on the extracted dynamic response power input sequence and dynamic response temperature output sequence, dynamic response characteristic parameters describing its response characteristics are calculated, and their reference values ​​are recorded. These parameters quantify the temperature response behavior of the fan under dynamic operating conditions, while the reference values ​​provide a basis for judging whether it is abnormal. Continuously recording these dynamic response characteristic parameters and constructing a dynamic response characteristic parameter sequence enables the system to track the long-term operating status of the fan temperature measurement link.

[0103] This application introduces two different judgment modes to distinguish anomaly types. On the one hand, by judging whether the dynamic response characteristic parameter sequence has a one-time deviation and tends to stabilize change pattern, benign response changes caused by hardware characteristic changes can be identified. For example, when a new type of signal cable is replaced, the dynamic response characteristics of the temperature measurement signal may experience a one-time, systematic shift, but will subsequently stabilize in a new "healthy" state. This method can identify such benign changes and update the reference value of the dynamic response characteristic parameters in a timely manner after identification, thereby avoiding misjudging such normal system adjustments as faults. On the other hand, by judging whether the dynamic response characteristic parameter sequence has a continuous shift or increased fluctuation change pattern, real faults in the temperature measurement signal transmission link can be accurately identified. For example, when there is poor contact or component aging in the signal transmission link, the dynamic response characteristic parameters may continuously deviate from their reference value, or their fluctuation amplitude may increase significantly.

[0104] Therefore, this method, by distinguishing between benign response changes and genuine faults, enables the diagnostic system to adapt to changes in the wind turbine's operating environment and hardware configuration, avoiding false alarms caused by hardware updates in traditional methods. The various technical features work together, from data acquisition, preprocessing, operating condition identification, parameter calculation and tracking, to final intelligent judgment and benchmark updates, forming a complete, closed-loop diagnostic process. This significantly improves the accuracy and reliability of fault diagnosis in wind turbine temperature measurement systems, thereby enhancing the operation and maintenance efficiency and economic benefits of wind turbine generator sets.

[0105] Furthermore, in another embodiment of this application, S6000 includes:

[0106] S6100: After identifying the dynamic response condition, it extracts the dynamic response power input sequence and dynamic response temperature output sequence within a fixed time window to construct a response morphology segment.

[0107] S6200: Normalizes the response morphology fragment and calculates the difference between it and the historical baseline morphology to obtain the deviation index.

[0108] S6300: Add the deviation index to the dynamic response characteristic parameter sequence;

[0109] S6400: When the deviation index shows a sudden change in the average value within a preset time period and the fluctuation amplitude is lower than the preset fluctuation threshold, it is identified as a benign response change caused by changes in hardware characteristics.

[0110] Specifically, after identifying the dynamic response condition of the wind turbine generator, it is necessary to extract the dynamic response power input sequence and dynamic response temperature output sequence within a specific time window from the power-temperature synchronization data sequence to construct a response pattern segment. This response pattern segment characterizes the transient response relationship between power input and temperature output under specific dynamic conditions. Subsequently, to eliminate the influence of dimensional differences and amplitude under different operating conditions or measurement conditions, the response pattern segment is normalized. The normalized response pattern segment is then compared with the pre-stored historical baseline pattern to calculate the degree of difference, thereby quantifying the deviation between the current response and the historical normal response, and obtaining a deviation index. This deviation index is then added to the dynamic response characteristic parameter sequence for continuous monitoring and analysis. Finally, by analyzing the deviation index, when it experiences a sudden change in average value within a preset time period, i.e., a sudden jump from one stable level to another, and its fluctuation amplitude is lower than a preset fluctuation threshold, it can be identified as a benign response change caused by a change in hardware characteristics. This change is usually caused by non-faulty factors such as sensor aging and minor changes in line impedance, rather than a true temperature signal transmission link failure.

[0111] This application's solution achieves accurate identification of benign response changes by introducing response pattern segments, normalization processing, deviation index, and abrupt change and fluctuation analysis based on the deviation index. Specifically, it extracts the dynamic response power input sequence and dynamic response temperature output sequence within a fixed time window to construct response pattern segments, aiming to capture the transient thermal response characteristics of the wind turbine under dynamic operating conditions. Normalizing this response pattern segment eliminates the influence of changes in external environment or operating conditions on the data amplitude, making the response patterns at different time points comparable. By calculating the difference with historical benchmark patterns, the degree of deviation between the current response and the normal response can be quantified, generating a deviation index. The introduction of this deviation index simplifies the complex dynamic response characteristics into a single quantifiable value, facilitating subsequent trend analysis. When the deviation index shows a sudden change in average value with low fluctuation amplitude within a preset time period, it indicates that the system response characteristics have undergone a one-time, stable change. This is usually due to a small but stable change in the physical characteristics of the hardware (such as sensor drift or line impedance changes), rather than intermittent or continuous faults in the signal transmission link. This mechanism enables the system to distinguish between normal hardware aging or minor adjustments and actual faults, thus avoiding false alarms.

[0112] Furthermore, in another embodiment of this application, S7000 includes:

[0113] S7100: After identifying a benign response change, it acquires multiple subsequent response pattern fragments and performs normalization processing;

[0114] S7200: Performs stability evaluation on multiple normalized response pattern fragments and selects sample fragments that meet the preset stability conditions.

[0115] S7300: Aggregates sample fragments to generate an updated response baseline.

[0116] S7400: Updated reference baseline values ​​for dynamic response characteristic parameters calculated based on the response baseline shape.

[0117] Specifically, after identifying benign response changes caused by variations in hardware characteristics, the system proactively acquires multiple response pattern segments after the benign change occurred and the system operation tends towards a stable state. These segments are instantaneous snapshots of the relationship between power input and temperature output of the wind turbine generator under dynamic response conditions. To eliminate dimensional differences under different operating conditions or measurement conditions, these response pattern segments are normalized to ensure they are compared and analyzed on a uniform scale. Normalization can employ various methods, such as max-min normalization and Z-score normalization, to ensure the accuracy of subsequent evaluations.

[0118] The stability assessment of multiple normalized response pattern fragments aims to ensure that the samples used to update the benchmark are stable and representative. Stability assessment can be based on indicators such as intra-fragment volatility, inter-fragment similarity, or deviation from short-term average patterns. For example, the variance or standard deviation of each normalized fragment can be calculated, and a preset stability threshold can be set; a fragment is considered stable only when its volatility is below this threshold. Furthermore, the similarity between different fragments can be assessed by calculating the correlation coefficient or Euclidean distance, thereby selecting sample fragments that are highly consistent with each other.

[0119] In practical applications, the selected sample fragments that meet preset stability conditions are aggregated to extract a new, representative response benchmark from multiple stable samples. Aggregation can employ statistical methods such as averaging, median, and weighted average to fuse information from multiple stable fragments into a single, more robust representation. For example, all selected sample fragments can be aligned on a time axis, and the normalized values ​​at each time point can be averaged to obtain a smooth benchmark curve representing the new response characteristics.

[0120] Therefore, based on the generated updated response baseline shape, updated reference baseline values ​​for dynamic response characteristic parameters can be calculated. This calculation process is consistent with the calculation method of the initial reference baseline value, ensuring that the new baseline value accurately reflects the dynamic response characteristics of the system after the hardware characteristics have changed. For example, if the dynamic response characteristic parameter is a response time constant or gain, the model will be refitted according to the new response baseline shape to obtain the updated time constant or gain as the new reference baseline value.

[0121] This application's solution, upon identifying a benign response change, does not simply reuse the old reference baseline value, but actively collects and processes multiple subsequent response pattern fragments. Specifically, by acquiring multiple subsequent response pattern fragments and performing normalization processing, data consistency and comparability are ensured. Furthermore, by performing stability assessments and screening on these normalized fragments, fragments affected by transient disturbances or abnormal data are excluded, ensuring the reliability of the samples used to update the baseline. Finally, the selected stable sample fragments are aggregated, effectively smoothing data noise and extracting the true and stable dynamic response characteristics of the system after hardware characteristic changes, thereby generating a response baseline pattern that accurately reflects the current system state. Based on this new response baseline pattern, the updated reference baseline value is calculated, enabling subsequent fault diagnosis to be based on the latest and most realistic system characteristics, thus avoiding misjudgments caused by changes in hardware characteristics.

[0122] Through the above technical solution, this application can effectively address the changes in the dynamic response characteristics of the temperature measurement signal transmission link when the hardware characteristics of a wind turbine generator undergo benign changes (such as sensor drift, component aging, or replacement). By dynamically updating the reference value of the dynamic response characteristic parameters, the system can continuously adapt to the long-term evolution of the wind turbine's operating state, ensuring the accuracy and robustness of fault diagnosis in the temperature measurement signal transmission link throughout the wind turbine's lifespan. This significantly improves the accuracy of fault identification, reduces unnecessary on-site inspections and maintenance, thereby lowering operating costs and enhancing the operational reliability of the wind turbine generator.

[0123] In some preferred embodiments, the following specific example illustrates the situation:

[0124] Suppose that after a period of operation, the temperature sensor of the generator winding of a wind turbine experiences a one-time, stable, and minute change in its response characteristics due to slight aging, which is recognized by the system as a benign change in response.

[0125] Upon recognizing this positive change, the system immediately initiates a baseline update process. Specifically, over the next few hours or days, whenever the wind turbine enters dynamic response conditions such as peak shaving, the system continuously captures and acquires response pattern fragments from multiple (e.g., 10) subsequent power-temperature synchronization data sequences. These fragments are then normalized to eliminate the influence of different power levels and temperature ranges.

[0126] Next, the stability of these 10 normalized response fragments is evaluated. For example, the root mean square error or standard deviation of each fragment is calculated, and a threshold is set. If the root mean square error of a fragment is too large, it is considered unstable and is removed. Suppose that after screening, 8 fragments are considered stable sample fragments.

[0127] Subsequently, these eight stable sample segments are aggregated. For example, a point-by-point averaging method can be used to align these eight segments on the time axis and average the normalized temperature values ​​at each time point, thereby generating a new, smooth response baseline curve.

[0128] Finally, based on this new response baseline curve, the dynamic response characteristic parameters are recalculated (e.g., if the characteristic parameter is a response time constant, a first-order or second-order system model is refitted) to obtain updated reference baseline values. Subsequently, the system will use these updated reference baseline values ​​to determine whether there is a persistent shift or increased fluctuation in the subsequent dynamic response characteristic parameter sequence, thereby more accurately identifying real faults in the temperature measurement signal transmission link and avoiding misjudging normal response changes due to sensor aging as faults.

[0129] Furthermore, in another embodiment of this application, step S8000 includes:

[0130] S8100: Performs sliding segmentation on the dynamic response feature parameter sequence to construct multiple dynamic response feature parameter segments;

[0131] S8200: Calculate the mean and fluctuation amplitude of each dynamic response characteristic parameter segment respectively;

[0132] S8300: If the mean value deviates continuously from the reference benchmark value, or if the fluctuation amplitude continuously exceeds the preset fluctuation threshold, a real fault is identified.

[0133] The sliding segmentation process for the dynamic response feature parameter sequence refers to dividing the continuously recorded dynamic response feature parameter sequence according to a preset time window or number of data points, thereby constructing a series of overlapping or non-overlapping dynamic response feature parameter segments. For example, a fixed-length sliding window can be set, moving one step at a time (e.g., one data point or a fixed time interval), thereby generating a series of segments containing recent dynamic response feature parameters.

[0134] Calculating the mean and fluctuation amplitude of each dynamic response characteristic parameter segment means that for each constructed dynamic response characteristic parameter segment, the arithmetic mean of all parameter values ​​within it is calculated as the mean, and the dispersion of its parameter values ​​is calculated as the fluctuation amplitude. For example, the fluctuation amplitude can be represented by standard deviation, variance, or range (the difference between the maximum and minimum values) to quantify the stability or drastic change of the parameters within the segment. Determining whether a true fault exists by continuously deviating from the reference value or continuously exceeding a preset fluctuation threshold involves comparing the mean of each dynamic response characteristic parameter segment with a preset reference value, and comparing its fluctuation amplitude with a preset fluctuation threshold to determine if an anomaly exists. "Continuously deviating" or "continuously exceeding" means that in a certain number of consecutive segments, the mean continuously exceeds the normal range of the reference value (e.g., exceeds statistical control limits), or the fluctuation amplitude continuously exceeds the allowable threshold. This determination of continuity helps to eliminate sporadic interference or instantaneous fluctuations, thereby more accurately identifying persistent problems caused by hardware failures, line aging, poor contact, etc., in the temperature measurement signal transmission link.

[0135] The proposed solution employs refined sliding segmentation of the dynamic response characteristic parameter sequence and quantifies the mean and fluctuation amplitude of each segment. This effectively captures persistent and systematic changes in the temperature measurement signal transmission link caused by real faults, avoiding misinterpretation of benign response changes as faults. This method significantly improves the sensitivity and specificity of fault identification, ensuring that fault alarms are only issued when a persistent problem exists in the temperature measurement signal transmission link, thereby reducing false alarm rates and improving the reliability and maintenance efficiency of wind turbine generator operation.

[0136] Furthermore, in another embodiment of this application, the sub-step of S3000 is proposed: extracting the dynamic response power input sequence and the dynamic response temperature output sequence from the corresponding power-temperature synchronization data sequence, including:

[0137] S3210: Obtain peak shaving control commands from wind turbine generator sets;

[0138] S3220: Calculates the rate of change of real-time active power data within a set time window;

[0139] S3230: When the peak shaving control command is active and the rate of change exceeds the preset change threshold, the current time period is determined to be a dynamic response condition.

[0140] S3240: Extract the power input sequence and temperature output sequence corresponding to the power-temperature synchronization data sequence in the current time period, and use them as the dynamic response power input sequence and dynamic response temperature output sequence.

[0141] Specifically, peak-shaving control commands refer to power regulation commands executed by wind turbine generators in response to changes in grid frequency or power demand under grid dispatch or their own control strategies. These commands can be obtained from the wind turbine generator's control system, SCADA system, or grid dispatch center. The rate of change of real-time active power data within a set time window can be obtained through differential calculation or regression analysis of continuously collected real-time active power data. For example, the average rate of change of active power can be calculated within a preset sliding time window, or determined by comparing the power values ​​at the beginning and end of the window. The length of the set time window can be adjusted according to the wind turbine generator's response characteristics and data sampling frequency to ensure that effective power fluctuations are captured. When peak-shaving control commands are active, it means that the wind turbine generator is actively participating in the grid's peak-shaving or frequency regulation services; at this time, its rapid changes in active power are usually a controlled and purposeful response.

[0142] The preset change threshold is an empirical value or a critical value determined through historical data analysis, used to distinguish between normal power fluctuations and significant power changes caused by peak shaving commands. When the peak shaving control command is active and the rate of change of real-time active power data exceeds the preset change threshold, it can be accurately determined that the wind turbine generator is in dynamic response mode during the current period. Once the current period of dynamic response mode is determined, the corresponding power data within that period can be precisely extracted from the pre-constructed power-temperature synchronization data sequence as the dynamic response power input sequence, and the corresponding temperature data as the dynamic response temperature output sequence.

[0143] This application's solution introduces peak-shaving control commands as a prerequisite for judging dynamic response conditions and combines this with double confirmation using the rate of change of real-time active power data. This enables a more accurate identification of dynamic response conditions arising from wind turbine generators actively responding to grid demands. This method avoids misjudging power fluctuations caused by non-peak-shaving factors such as wind speed changes and fault tripping as dynamic response conditions, thus ensuring that the extracted dynamic response power input sequence and dynamic response temperature output sequence accurately reflect the thermal response characteristics of the generator windings under controlled power changes. This provides a high-quality, highly correlated data foundation for the subsequent calculation of dynamic response characteristic parameters.

[0144] Furthermore, in another embodiment of this application, S8300 includes:

[0145] S8310: Perform statistical process monitoring on dynamic response characteristic parameter segments and construct control limits for reference baseline values;

[0146] S8320: Determine whether the mean or fluctuation range of each dynamic response characteristic parameter segment continuously exceeds the control limit, or whether it exhibits a non-random offset trend;

[0147] S8330: When any judgment condition is met, it is determined that there is a pattern of continuous offset or increased fluctuation.

[0148] Specifically, statistical process monitoring of dynamic response characteristic parameter segments refers to using statistical methods to monitor data sequences in real time to identify whether the process is under statistical control. Its purpose is to determine whether data changes exceed the normal fluctuation range through quantitative statistical indicators, thereby more accurately identifying anomalies. Constructing control limits for reference benchmark values ​​can be understood as determining an upper and lower limit interval based on historical data or a pre-set statistical distribution to define the normal fluctuation range of dynamic response characteristic parameters. For example, the three-sigma rule can be used to set control limits, meaning that fluctuations within the range of the mean ± 3 standard deviations are considered normal. Determining whether the mean or fluctuation amplitude of each dynamic response characteristic parameter segment continuously exceeds the control limits means that when the mean or fluctuation amplitude falls outside the control limits for multiple consecutive sampling points or time periods, an anomaly is considered to exist. Alternatively, determining whether a non-random deviation trend is present means that even if data points do not continuously exceed the control limits, but their change patterns exhibit obvious non-randomness, such as continuous increases or decreases, periodic fluctuations, etc., these trends may also indicate potential faults. For example, non-random offset trends can be identified using run chart rules or Westgard rules, such as multiple consecutive points falling on one side of the mean line, or multiple consecutive points showing a monotonically rising / falling trend.

[0149] This application's solution introduces a statistical process monitoring method to perform a more refined analysis of dynamic response characteristic parameter segments. Traditional simple threshold judgments may not effectively distinguish between random noise and persistent changes caused by real faults. By constructing control limits for reference benchmark values ​​and combining this with judgments on whether the mean or fluctuation amplitude continuously exceeds the control limits and whether it exhibits a non-random offset trend, it is possible to more accurately identify persistent offsets or aggravated fluctuation patterns caused by real faults in the temperature measurement signal transmission link based on statistical principles. This method can effectively filter out random fluctuations and focus on statistically significant anomalies, thereby improving the accuracy and robustness of fault identification.

[0150] Through the above technical solution, this application can significantly improve the accuracy and reliability of fault identification in the wind turbine temperature measurement signal transmission link. Compared with simple threshold judgment, statistical process monitoring can more effectively identify persistent deviations or increased fluctuations caused by real faults, while reducing the false alarm rate caused by random fluctuations or changes in normal operating conditions. Therefore, it can ensure a more accurate assessment of the operating status of wind turbine generators, timely detection and handling of potential temperature measurement system faults, thereby ensuring the safe and stable operation of wind turbine generators and reducing unplanned downtime.

[0151] In some preferred embodiments, the following specific example illustrates the situation:

[0152] Assume that during the daily operation of a wind turbine generator, the dynamic response characteristic parameter sequence is continuously monitored. To determine whether there is a persistent shift or increased fluctuation, the system first calculates the long-term mean and standard deviation of the characteristic parameter based on historical data, and sets upper and lower control limits accordingly. For example, the mean ± 3 times the standard deviation is used as the control limit. During real-time monitoring, whenever a new segment of the dynamic response characteristic parameter is acquired, its mean and fluctuation amplitude are calculated. If the mean of five consecutive segments falls outside the upper control limit, or the fluctuation amplitude of five consecutive segments exceeds the preset fluctuation threshold and exceeds the upper control limit, the system will determine that there is a persistent shift or increased fluctuation pattern and trigger a fault alarm. Furthermore, even if the mean and fluctuation amplitude do not continuously exceed the control limits, if the mean of seven consecutive segments shows a monotonically increasing trend, or the mean of seven consecutive segments falls on the same side of the centerline (long-term mean), the system will also identify it as a non-random shift trend and similarly determine that there is a potential fault. This statistical process monitoring method can effectively avoid misjudgments caused by instantaneous fluctuations, ensuring that fault identification is only performed when there is a persistent abnormality in the temperature signal transmission link.

[0153] Furthermore, in another embodiment of this application, S8310 includes:

[0154] S8311: Extract dynamic response characteristic parameter fragments from multiple historical time periods and calculate their long-term mean and standard deviation of fluctuation;

[0155] S8312: Construct the initial control limit interval based on the long-term mean and the standard deviation of fluctuation;

[0156] S8313: Perform trend analysis on the dynamic response characteristic parameter segments of the current time period. When it is determined to be a stable drift pattern, update the long-term mean and standard deviation of fluctuation.

[0157] S8314: Based on the updated long-term mean and standard deviation of fluctuation, the upper and lower thresholds of the initial control limit interval are dynamically adjusted and used as the control limits.

[0158] Specifically, extracting dynamic response characteristic parameter segments from multiple historical time periods refers to selecting sufficiently long and representative time periods, such as several months or years of operating data, from the long-term historical data of wind turbine generators, and extracting dynamic response characteristic parameter segments calculated under dynamic response conditions. The purpose is to obtain enough sample data to accurately reflect the long-term statistical characteristics of the parameter. Based on this, the long-term mean and standard deviation of fluctuation of these historical segments are calculated. The long-term mean can be understood as the average level of the parameter under normal operating conditions, while the standard deviation of fluctuation reflects its fluctuation range around the mean. These statistics form the basis for constructing control limits. Constructing an initial control limit interval based on the long-term mean and standard deviation of fluctuation involves using statistical process control (SPC) principles, such as the 3σ principle, with the long-term mean as the center line and the long-term mean plus or minus a certain multiple of the standard deviation of fluctuation as the upper and lower control limits, forming an initial, static control limit interval. This initial control limit interval is used to preliminarily determine whether the dynamic response characteristic parameter is within the normal range.

[0159] In practical applications, trend analysis of dynamic response characteristic parameter segments for the current time period refers to monitoring and analyzing the mean and fluctuation amplitude of dynamic response characteristic parameter segments over a recent period (e.g., several days or weeks). When the mean shows a slow, continuous upward or downward trend, and the fluctuation amplitude remains within a preset stable range, it can be identified as a stable drift pattern. This drift is usually caused by benign factors such as sensor aging, minor changes in line impedance, or slow changes in the internal physical characteristics of the generator set. When a stable drift pattern is identified, the long-term mean and standard deviation of fluctuation need to be updated so that the control limits can adapt to this benign change. Based on the updated long-term mean and standard deviation of fluctuation, the upper and lower thresholds of the initial control limit interval are dynamically adjusted and used as the control limits. This means that once a stable drift is identified and the baseline statistics are updated, the control limits will be adjusted accordingly, thus forming a dynamically adaptive control limit. This dynamic adjustment ensures that the control limits can always accurately reflect the current normal operating range of the system, avoiding misjudging normal system drift as a fault.

[0160] This application's solution introduces trend analysis of dynamic response characteristic parameter segments and dynamically updates the long-term mean and standard deviation of fluctuation based on the analysis results, thereby achieving adaptive adjustment of the control limits. Specifically, when the system exhibits a stable drift pattern, traditional fixed control limits may cause parameter values ​​to continuously exceed the limits, frequently triggering false alarms. This application identifies such benign drifts and updates the reference baseline and control limits accordingly, enabling the control limits to "follow" the system's normal drift, thus avoiding misjudging normal system characteristic changes as faults. Simultaneously, the new dynamic control limits can still effectively identify non-drift, abnormal offsets, or increased fluctuations, ensuring the sensitivity of fault diagnosis.

[0161] In some preferred embodiments, the following specific example illustrates the situation:

[0162] Suppose that in the initial stage of wind turbine operation, the long-term mean of the dynamic response characteristic parameter calculated from historical data is X, and the standard deviation of fluctuation is Y. The initial control limits constructed from this are [X-3Y, X+3Y]. As the unit operates for longer periods, due to slight sensor aging, the mean of this characteristic parameter may slowly increase, for example, by 0.01 units per month, but its fluctuation range remains stable. If a fixed control limit is used, after several months, even without a real fault, the mean of this parameter may continuously exceed the upper limit of X+3Y, causing the system to frequently issue fault alarms. When the system recognizes this trend of a slowly increasing mean with stable fluctuation, it determines it as a stable drift mode. At this point, the system will recalculate the long-term mean and standard deviation of fluctuation based on new data; for example, the updated long-term mean becomes X', and the standard deviation of fluctuation remains Y. Subsequently, the control limits will be dynamically adjusted to [X'-3Y, X'+3Y]. In this way, even if the mean of the parameter undergoes benign drift, as long as it remains within the new, dynamically adjusted control limits, it will not be misjudged as a fault. Only when the mean or fluctuation range of the parameters exceeds this dynamically adjusted control limit, or exhibits non-random, abrupt abnormal changes, will it be identified as a real temperature measurement signal transmission link fault, thereby avoiding false alarms and improving the accuracy of fault diagnosis.

[0163] Furthermore, in another embodiment of this application, a sub-step of S8313 is proposed: performing trend analysis on the dynamic response characteristic parameter segment of the current time period, including:

[0164] A1: Calculate the mean and fluctuation range of the dynamic response characteristic parameter segment within the current time period;

[0165] A2: Use a sliding window approach to compare the mean and fluctuation range of multiple consecutive time periods to identify trend evolution patterns;

[0166] A3: When the mean continues to change slowly according to a preset time and the fluctuation range is within a preset stable range, the trend evolution pattern is identified as a drift pattern.

[0167] Specifically, when performing trend analysis on a segment of dynamic response characteristic parameters within the current time period, it is first necessary to calculate the mean and fluctuation range of the dynamic response characteristic parameter segment within the current time period. The mean can be understood as the arithmetic mean of all dynamic response characteristic parameters within the segment, used to characterize the central trend of the parameters within that time period. The fluctuation range can be measured using statistics such as standard deviation, variance, or range, used to characterize the dispersion or stability of the parameters within the segment. To identify the trend evolution pattern of the parameters, a sliding window approach is used to compare the mean and fluctuation range of multiple consecutive time periods. Specifically, a fixed-size sliding window can be set, which slides forward on the dynamic response characteristic parameter sequence, calculating the mean and fluctuation range of the parameter segment within the window at each slide. By comparing the mean and fluctuation range of adjacent or consecutive windows, the changing trend of the parameters over time can be observed. For example, if the mean shows a continuous small increase or decrease, while the fluctuation range remains relatively stable, it may indicate a certain trend.

[0168] Therefore, when the mean value consistently exhibits a preset slow change pattern, and the fluctuation range remains within a preset stable range, the trend evolution pattern is identified as a drift pattern. Here, "preset slow change pattern" means that the rate of change of the mean value is below a certain preset threshold, indicating that the baseline value of the parameter is shifting slowly and gradually, rather than changing abruptly; "preset stable range" means that the fluctuation range remains within normal operating or expected noise levels, indicating that the intrinsic variability of the parameter has not increased significantly. When both conditions are met simultaneously, it can be determined that the system response characteristics have undergone a benign, gradual drift.

[0169] This application's solution, through refined analysis of the mean and fluctuation amplitude of dynamic response characteristic parameter segments and the introduction of a sliding window comparison mechanism, effectively captures subtle trends in parameter changes. In particular, by explicitly defining the identification conditions for "drift modes"—namely, slow changes in the mean and stable fluctuation amplitude—the system can distinguish between benign, gradual response changes caused by hardware aging or environmental variations, and abnormal fluctuations or offsets caused by failures in the temperature signal transmission link. This discriminatory ability is crucial for subsequent updates to the reference baseline, ensuring the adaptability and accuracy of the monitoring system.

[0170] Furthermore, in another embodiment of this application, a sub-step of S8313 is proposed: when a stable drift pattern is determined, updating the long-term mean and the standard deviation of fluctuation includes:

[0171] B1: Based on multiple dynamic response characteristic parameter segments within the current time period, the long-term mean is recalculated using a moving average or exponential weighting method to generate an updated reference benchmark value.

[0172] B2: Adjust the upper and lower thresholds of the initial control limit range based on the updated reference baseline and standard deviation of fluctuation.

[0173] "Moving average" is a commonly used data smoothing technique that eliminates short-term fluctuations by calculating the average of data within a fixed-length window, thereby revealing the long-term trend of the data. For example, a sliding window containing N segments of the latest dynamic response characteristic parameters can be set; each time a new segment is added, the oldest segment is removed, and the mean of all segments within the window is recalculated. "Exponential weighting" is an averaging method that assigns higher weight to recent data. Its characteristic is that it can respond more quickly to the latest changes in data while retaining the influence of historical data. For example, a decay factor can be used to weight historical data so that newer data has a greater impact on the mean. Both methods aim to make the calculated long-term mean more accurately reflect the true baseline state of the current system under stable drift mode.

[0174] Therefore, by recalculating the long-term mean using a moving average or exponential weighting method, the reference value can dynamically adapt to the slow and stable drift of the temperature response characteristics of the wind turbine generator caused by factors such as component aging and environmental changes during long-term operation. This adaptive reference update mechanism ensures that the control limits always remain near the current normal operating state of the system, thereby avoiding misjudging normal system characteristic drift as a fault.

[0175] This application also discloses a fan temperature measurement data processing system, which includes:

[0176] Data acquisition module 1 is used to acquire real-time active power data and real-time temperature data of the wind turbine generator set.

[0177] Synchronization processing module 2 is used to perform time alignment processing on the collected real-time active power data and real-time temperature data to construct a power-temperature synchronization data sequence;

[0178] The operating condition identification and recording module 3 is used to identify dynamic response operating conditions with significant power fluctuation characteristics based on the power-temperature synchronization data sequence, and to extract the dynamic response power input sequence and dynamic response temperature output sequence from the corresponding power-temperature synchronization data sequence.

[0179] Parameter calculation module 4 calculates dynamic response characteristic parameters describing its response characteristics based on the dynamic response power input sequence and dynamic response temperature output sequence, and records its reference benchmark value.

[0180] Parameter tracking module 5 is used to continuously record dynamic response feature parameters and construct a dynamic response feature parameter sequence;

[0181] The benign change identification module 6 is used to determine whether there is a one-time deviation and a tendency to stabilize change pattern in the dynamic response characteristic parameter sequence, so as to identify benign response changes caused by changes in hardware characteristics;

[0182] The baseline update module 7 is used to update the reference baseline value of the dynamic response characteristic parameters after a benign response change is detected.

[0183] The real fault identification module 8 is used to determine whether there is a continuous shift or increased fluctuation pattern in the dynamic response characteristic parameter sequence, so as to identify real faults in the temperature measurement signal transmission link.

[0184] This system, through its modular design, achieves efficient processing and intelligent diagnosis of temperature measurement data from wind turbine generators, aiming to solve the problem in existing technologies of distinguishing between benign hardware changes and actual equipment failures. Specifically, the system can acquire and simultaneously process real-time active power data and real-time temperature data of the generator windings, thereby identifying dynamic response conditions and extracting the corresponding power input and temperature output sequences. Based on this, the system calculates and tracks dynamic response characteristic parameters, and intelligently distinguishes between benign response changes caused by hardware characteristic variations and actual faults in the temperature measurement signal transmission link by judging their change patterns. When a benign response change is identified, the system can promptly update the reference value, thereby avoiding false alarms and ensuring the accuracy of diagnosis and the adaptability of the system.

[0185] The above are merely embodiments of this application and are not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for processing temperature measurement data of a fan, characterized in that, The method comprises: acquiring real-time active power data of a wind turbine generator set and real-time temperature data of a generator winding; time-aligning the collected real-time active power data and real-time temperature data to construct a power-temperature synchronous data sequence; based on the power-temperature synchronous data sequence, identifying a dynamic response working condition with a significant power fluctuation characteristic and extracting a dynamic response power input sequence and a dynamic response temperature output sequence in the power-temperature synchronous data sequence corresponding to the dynamic response working condition; based on the dynamic response power input sequence and the dynamic response temperature output sequence, calculating a dynamic response characteristic parameter describing the response characteristic thereof and recording a reference benchmark value thereof; continuously recording the dynamic response characteristic parameter to construct a dynamic response characteristic parameter sequence; judging whether the dynamic response characteristic parameter sequence has a one-time deviation and a change mode tending to be stable to identify a benign response change caused by a hardware characteristic change; after identifying the benign response change, updating the reference benchmark value of the dynamic response characteristic parameter; judging whether the dynamic response characteristic parameter sequence has a persistent deviation or a change mode of fluctuation aggravation to identify a real fault in a temperature measurement signal transmission link; the judging whether the dynamic response characteristic parameter sequence has a one-time deviation and a change mode tending to be stable comprises: after identifying the dynamic response working condition, intercepting the dynamic response power input sequence and the dynamic response temperature output sequence within a fixed time window to construct a response mode segment; performing normalization processing on the response mode segment and performing difference calculation with a historical benchmark mode to obtain a deviation index; adding the deviation index to the dynamic response characteristic parameter sequence; when the deviation index has an average value mutation within a preset time period and a fluctuation amplitude lower than a preset fluctuation threshold, judging that the benign response change is caused by a hardware characteristic change; the updating the reference benchmark value of the dynamic response characteristic parameter after identifying the benign response change comprises: after identifying the benign response change, acquiring a plurality of subsequent response mode segments and performing normalization processing thereon; performing stability evaluation on a plurality of normalized response mode segments to screen out sample segments satisfying a preset stability condition; performing aggregation processing on the sample segments to generate an updated response benchmark mode; based on the response benchmark mode, calculating an updated reference benchmark value of the dynamic response characteristic parameter.

2. The temperature measurement data processing method of claim 1, wherein, the judging whether the dynamic response characteristic parameter sequence has a persistent deviation or a change mode of fluctuation aggravation to identify a real fault in a temperature measurement signal transmission link comprises: performing sliding segmentation processing on the dynamic response characteristic parameter sequence to construct a plurality of dynamic response characteristic parameter segments; respectively calculating a mean value and a fluctuation amplitude of each dynamic response characteristic parameter segment; when the mean value continuously deviates from the reference benchmark value or when the fluctuation amplitude continuously exceeds a preset fluctuation threshold, judging that the real fault exists.

3. The fan temperature measurement data processing method of claim 1, wherein, the extracting a dynamic response power input sequence and a dynamic response temperature output sequence in the power-temperature synchronous data sequence corresponding to the dynamic response working condition comprises: acquiring a peak shaving control instruction of a wind turbine generator system; calculating a change rate of the real-time active power data within a set time window; determining a current time period as the dynamic response working condition when the peak shaving control instruction is in an activated state and the change rate exceeds a preset change threshold; extracting a power input sequence and a temperature output sequence corresponding to the current time period from the power temperature synchronous data sequence as the dynamic response power input sequence and the dynamic response temperature output sequence.

4. The fan temperature measurement data processing method of claim 2, wherein, The determining whether the mean value continuously deviates from the reference benchmark value or whether the fluctuation amplitude continuously exceeds the preset fluctuation threshold value comprises: performing statistical process monitoring on the dynamic response characteristic parameter segment to construct a control limit of the reference benchmark value; determining whether the mean value or the fluctuation amplitude of each dynamic response characteristic parameter segment continuously exceeds the control limit or whether a non-random deviation trend is presented; determining that there is a persistent deviation or fluctuation intensification change mode when any of the determination conditions is met.

5. The temperature measurement data processing method of claim 4, wherein, The performing statistical process monitoring on the dynamic response characteristic parameter segment to construct a control limit of the reference benchmark value comprises: extracting the dynamic response characteristic parameter segment in a plurality of historical time periods to calculate a long-term mean value and a fluctuation standard deviation; constructing an initial control limit interval according to the long-term mean value and the fluctuation standard deviation; performing trend analysis on the dynamic response characteristic parameter segment of a current time period, and updating the long-term mean value and the fluctuation standard deviation when a stable drift mode is determined; dynamically correcting upper and lower threshold values of the initial control limit interval as the control limit according to the updated long-term mean value and fluctuation standard deviation.

6. The temperature measurement data processing method of claim 5, wherein, The performing trend analysis on the dynamic response characteristic parameter segment of a current time period comprises: calculating a mean value and a fluctuation amplitude of the dynamic response characteristic parameter segment in the current time period; comparing the mean values and the fluctuation amplitudes of a plurality of continuous time periods in a sliding window manner to identify a trend evolution mode; identifying the trend evolution mode as a drift mode when the mean value continuously is in a preset slow change state and the fluctuation amplitude is in a preset stable range.

7. The fan temperature measurement data processing method of claim 5, wherein, The updating the long-term mean value and the fluctuation standard deviation when a stable drift mode is determined comprises: recomputing a long-term mean value of the dynamic response characteristic parameter segment in a plurality of current time periods in a sliding average or exponential weighting manner to generate an updated reference benchmark value based on the dynamic response characteristic parameter segment; adjusting upper and lower threshold values of the initial control limit interval according to the updated reference benchmark value and fluctuation standard deviation.

8. A fan temperature measurement data processing system, characterized in that, The system for executing the method of claim 1 comprises: a data acquisition module configured to acquire real-time active power data of a wind turbine generator system and real-time temperature data of a generator winding; a synchronous processing module configured to perform time alignment processing on the acquired real-time active power data and real-time temperature data to construct a power temperature synchronous data sequence; The working condition recognition and recording module is configured to recognize a dynamic response working condition with a significant power fluctuation feature based on the power-temperature synchronous data sequence, and extract a dynamic response power input sequence and a dynamic response temperature output sequence in the power-temperature synchronous data sequence corresponding to the dynamic response working condition. The parameter calculation module is configured to calculate a dynamic response characteristic parameter describing a response characteristic of the dynamic response working condition based on the dynamic response power input sequence and the dynamic response temperature output sequence, and record a reference baseline value of the dynamic response characteristic parameter. The parameter tracking module is configured to continuously record the dynamic response characteristic parameter, and construct a dynamic response characteristic parameter sequence. The benign change recognition module is configured to judge whether the dynamic response characteristic parameter sequence has a one-time deviation and a change mode tending to be stable, so as to recognize a benign response change caused by a hardware characteristic change. The baseline updating module is configured to update the reference baseline value of the dynamic response characteristic parameter after the benign response change is recognized. The real fault recognition module is configured to judge whether the dynamic response characteristic parameter sequence has a persistent deviation or a change mode of fluctuation aggravation, so as to recognize a real fault in a temperature measurement signal transmission link.

Images