A weak fault diagnosis method based on chaotic signals of wind turbine stator current

Abstract

This invention discloses a weak fault diagnosis method based on chaotic signals of wind turbine stator current. The diagnostic method includes: acquiring the current signal of the wind turbine stator and preprocessing the current signal, the preprocessing including row-sequence component transformation, adaptive fundamental suppression, and detrending processing; calculating a chaos index based on the preprocessed signal; calculating a health index based on the chaos index and classifying the health status according to the health index; and outputting a diagnostic result including the health index, fault type, and graded early warning signal based on the aforementioned results.

Description

Technical Field

[0001] This invention relates to the field of wind turbine stator fault diagnosis technology, and in particular to a weak fault diagnosis method based on chaotic signals of wind turbine stator current. Background Technology

[0002] With the global trend of energy structure transitioning towards renewable energy, the importance of wind power generation is increasing. Wind turbine generators are the core equipment of the wind power industry, and the generator, as a key component for converting mechanical energy into electrical energy, directly determines the power generation efficiency and operation and maintenance costs of wind farms based on its operational reliability. Wind turbine generators operate for extended periods in variable speed, variable load, and high-noise environments, making their stators prone to early, subtle faults such as insulation aging, rotor bar breakage, and air gap eccentricity. These faults initially manifest as weak, aperiodic disturbances in the stator current, with low characteristic amplitudes and poor stability, easily masked by fundamental frequencies, grid noise, and operational fluctuations, making them difficult to identify effectively using traditional methods. If not detected in time, these faults will gradually develop into serious failures such as inter-turn short circuits and overheating damage, not only causing unplanned unit shutdowns but also potentially leading to safety accidents such as generator fires, seriously threatening operational safety and economic benefits.

[0003] In the field of wind turbine fault detection, existing monitoring methods mainly include frequency domain analysis, wavelet analysis, envelope demodulation, and model-driven methods. However, these methods generally suffer from limited noise immunity and strong parameter dependence, making it difficult to extract early fault characteristics of wind turbines under complex operating conditions. Furthermore, some detection methods rely on complex calculations and require additional equipment, limiting their widespread application in the wind power sector. Therefore, there is an urgent need in this field for a weak fault identification method that can be based on existing wind turbine acquisition devices, possessing high sensitivity and low operational barriers, to achieve real-time diagnosis and health status perception of early faults in wind turbine generator stators. Summary of the Invention

[0004] In view of the above-mentioned prior art, the present invention provides a weak fault diagnosis method based on the chaotic signal of wind turbine stator current, which mainly solves the technical problems existing in the background art.

[0005] To achieve the above objectives, the technical solution of this invention is implemented as follows: This invention discloses a weak fault diagnosis method based on chaotic signals of wind turbine stator current, the aforementioned diagnosis method comprising: The current signal of the wind turbine stator is acquired and the current signal is preprocessed, including row sequence component transformation, adaptive fundamental suppression and detrending processing. Calculate the chaos index based on the preprocessed signal; A health index is calculated based on the chaos index, and a health status classification is determined based on the health index. Based on the aforementioned results, the diagnostic results are output, including health index, fault type, and graded warning signals.

[0006] Preferably, the preprocessing process specifically includes: converting the current signal into positive-sequence, negative-sequence, and zero-sequence components through Clarke-Park transform; extracting the negative-sequence component and then using an adaptive filter to filter out the residual fundamental wave and its main harmonic interference from the negative-sequence current signal; and performing detrending processing on the filtered signal to obtain the final signal sequence to be analyzed.

[0007] Preferably, the chaos index is calculated using the following formula:

[0008] In the formula, The mean square shift is obtained by transforming the signal sequence to be analyzed, where n is the step size. This indicates taking the limit with respect to the step size.

[0009] Preferably, the health index is calculated using the following formula. :

[0010] In the formula, This is the fault threshold.

[0011] Preferably, the range of fault threshold values ​​is determined by manual judgment based on real-time collected unit speed and output power.

[0012] Preferably, the mean square shift is obtained after transforming the signal sequence to be analyzed. Specifically, this includes performing a 0-1 test on the preprocessed signal sequence to obtain the transformed variable. and ,based on and Calculate the mean square displacement.

[0013] Preferably, based on the health index The health status is classified and determined as follows: when H ≥ 85%, it is considered to be in a healthy state; when 70% ≤ H < 85%, it is considered to be slightly abnormal; when 50% ≤ H < 70%, it is considered to be moderately abnormal; and when H < 50%, it is considered to be seriously faulty.

[0014] Preferably, the method further includes determining the fault type based on the rate of increase of the chaotic index. If the chaotic index rises rapidly, the fault type is diagnosed as equipment faults such as rotor bar breakage and uneven air gap; if the chaotic index rises slowly, the fault type is diagnosed as insulation aging.

[0015] Preferably, different graded early warning signals are output according to different health status grading results, specifically including: outputting an operation and maintenance prompt signal when it is determined to be a minor abnormality, and outputting an emergency shutdown suggestion signal when it is determined to be a serious fault.

[0016] The beneficial effects of this invention are as follows: The weak fault diagnosis method based on chaotic signals of wind turbine stator current disclosed in this application can significantly improve the detectability of weak faults. By utilizing non-periodic micro-perturbations in the stator current to construct chaotic indices, it avoids the problem that traditional frequency domain and time-frequency domain methods are easily masked by fundamental waves, harmonics, and operating condition disturbances, thus achieving sensitive identification of early weak faults such as insulation aging, rotor bar breakage, and air gap unevenness. Furthermore, this application uses 0-1 testing to extract chaotic features, eliminating the need for phase space reconstruction and complex signal models, resulting in low computational load and suitability for online real-time monitoring under actual wind turbine operating conditions. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only preferred embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the weak fault diagnosis method based on chaotic signals of wind turbine stator current in an embodiment of this application. Detailed Implementation

[0019] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. In the following description, the expression "some embodiments" refers to a subset of all possible embodiments; however, it should be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments and can be combined with each other without conflict.

[0020] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention can be practiced without one or more of these details. In other instances, certain technical features well-known in the art have not been described in order to avoid obscuring the invention.

[0021] It should be understood that the present invention can be embodied in various forms and should not be construed as being limited to the embodiments set forth herein. Rather, providing these embodiments will make the disclosure thorough and complete, and will fully convey the scope of the invention to those skilled in the art. Furthermore, the terminology used herein is intended only to describe particular embodiments and is not intended to limit the invention. When used herein, the singular forms “a,” “an,” and “the” are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the terms “compose” and / or “comprising,” when used in this specification, identify the presence of the stated features, integers, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups. When used herein, the term “and / or” includes any and all combinations of the associated listed items.

[0022] It should also be noted that when an element is referred to as being "fixed to" another element, it can be directly attached to the other element or there may be an intervening element. When an element is referred to as being "connected to" another element, it can be directly connected to the other element or there may be an intervening element. The terms "vertical," "horizontal," "inner," "outer," "left," "right," and similar expressions used herein are for illustrative purposes only and do not represent the only possible implementation.

[0023] To fully understand this invention, a detailed structure will be presented in the following description to illustrate the technical solution proposed by this invention. Optional embodiments of the invention are described in detail below; however, in addition to these detailed descriptions, the invention may have other embodiments.

[0024] Please refer to the attached document. Figure 1 This application provides a weak fault diagnosis method based on chaotic signals of wind turbine stator current. The aforementioned diagnosis method includes: S1. Obtain the current signal of the wind turbine stator and preprocess the current signal. The preprocessing includes row sequence component transformation, adaptive fundamental suppression and detrending processing.

[0025] Specifically, the acquisition frequency of the wind turbine stator current signal collected by the original three-phase current transformer of the wind turbine generator needs to be adapted to the wind turbine operating characteristics, generally set to 2kHz~5kHz, to ensure complete capture of fault-related non-periodic weak disturbance signals. The sampling timestamp of the current signal is recorded synchronously during the acquisition process.

[0026] In the preprocessing process, the acquired three-phase stator current signal (denoted as...) is first processed. , , The Clarke-Park transformation is performed, specifically, the current signal in the three-phase stationary coordinate system is transformed. , , Convert to Two-phase stationary coordinate system Components and The component transformation process is based on the instantaneous value relationship of the three-phase currents. It is calculated using a preset transformation matrix, such as a matrix calculated based on the equal amplitude transformation principle. The Clarke transform, mentioned earlier, eliminates the coupling relationship of the three-phase current signals, simplifying the three-dimensional current information into two-dimensional planar information. Then, based on the Clarke transform... Components and The components undergo Park transformation, introducing a rotating phase angle synchronized with the grid frequency. Rotation phase angle Extracted from the grid synchronization signal, typically obtained through phase-locked loop (PLL) technology, ensuring real-time matching of the phase angle with the grid fundamental phase, and by rotating the phase angle... Will Two-phase stationary coordinate system Components and Component conversion Cartesian components in rotating coordinate system With cross axis components Through the and The signal undergoes spectral analysis and sequence component separation, further decomposing it into positive-sequence, negative-sequence, and zero-sequence components. The positive-sequence component mainly corresponds to the fundamental current component during normal operation of the wind turbine. The zero-sequence component is mostly generated by extreme conditions such as three-phase current asymmetry grounding (with extremely small amplitude under normal operating conditions). However, early weak faults in the wind turbine stator (such as partial discharge caused by insulation aging, current asymmetry caused by rotor bar breakage) will form characteristic non-periodic disturbances in the negative-sequence component. Therefore, it is necessary to focus on extracting the negative-sequence component as the core carrier for subsequent fault feature analysis. During the extraction process, the phase difference and amplitude ratio of the positive-sequence and negative-sequence components are compared to ensure the accuracy of the negative-sequence component extraction and avoid the mixing of positive-sequence and zero-sequence components.

[0027] After extracting the negative-sequence current component, an adaptive filter is used to filter it to remove residual fundamental waves and their main harmonic interference. This adaptive filter uses the Least Mean Square (LMS) algorithm as the coefficient update criterion, with the grid fundamental frequency (typically 50Hz or 60Hz) and the corresponding sinusoidal signal (determined based on the grid specifications and its main harmonic frequencies) as the reference input. The filter order is set to 8th to 12th. During the filtering process, the adaptive filter continuously compares the mean square error between the output signal and the desired signal, adjusting the filter coefficients in real time to reduce the amplitude of residual fundamental waves and main harmonics in the output negative-sequence current signal to below 5% of the original amplitude. This maximizes the preservation of fault-related aperiodic weak disturbance components and avoids distortion of the useful signal.

[0028] After adaptive filtering, the filtered negative-sequence current signal needs to be detrended to eliminate any linear or slowly changing trend terms that may exist in the signal. These trend terms mainly originate from the temperature drift of the current transformer. For example, temperature changes in the transformer during wind turbine operation can cause a slow shift in the output current baseline, and slight displacements in the sensor installation position caused by unit vibration can lead to slow fluctuations in the current acquisition value. If these trends are not eliminated, they will mask the subtle disturbance characteristics of the fault. The detrended signal sequence is obtained by calculating using linear regression.

[0029] After detrending, the processed signal needs to be validated. If the mean and standard deviation of the signal do not meet the requirements, the fitting interval of the linear regression model should be readjusted (e.g., a piecewise fitting method should be used to handle nonlinear trends) until the final signal sequence to be analyzed meets the requirements for subsequent calculation of chaos indicators. .

[0030] S2. Calculate the chaos index based on the preprocessed signal. The formula for calculating the chaos index is as follows:

[0031] In the formula, The mean square shift is obtained by transforming the signal sequence to be analyzed, where n is the step size. This indicates taking the limit with respect to the step size. The slope of the fitted curve is the final chaos index.

[0032] Furthermore, when calculating the mean square displacement, the signal sequence to be analyzed obtained after preprocessing is first defined as follows: Where N is the total number of sampling points in the signal sequence. In practical applications, N needs to be large enough to meet the approximate accuracy requirements of subsequent limit calculations. For example, in the diagnosis of doubly-fed asynchronous wind turbine generators, N=2000 can be set, and then the preprocessed signal sequence can be... Calculate the transformation variables and The calculation formula is as follows:

[0033]

[0034] In the formula, , where n is the index at which the accumulation ends. For sampling point index, The 0-1 test algorithm uses a built-in fixed constant that must be selected as a multiple of an irrational number within the range (0, π). For example, select... for The core function of this parameter is to construct the phase reference for trigonometric functions, through... Index the sampling points The signal is converted into a phase value, thereby mapping the weak current disturbance signal in the time domain to the phase domain, avoiding interference from the fundamental period of the power grid on the extraction of fault features.

[0035] After calculating the transformed variables, the corrected mean square displacement, which truly reflects the chaotic characteristics, is obtained by combining the mean square displacement and the correction term. The mean square displacement is calculated using the following formula. :

[0036] In the formula, This indicates taking the limit of the number of sampling points to reduce the calculation error caused by finite sampling. For the sampling point offset index, calculated and The displacement change of the transformation variable under a step size n can be quantified. Then, through the sum of squares and the mean, the average squared displacement of the signal in the phase domain is obtained. This value directly reflects the non-periodic variation characteristic of the signal, which is one of the core properties of chaotic signals. Under normal operating conditions, the non-periodicity of the signal is weak. Growth is moderate, and non-cyclical enhancement occurs during fault development. The growth rate has accelerated.

[0037] To eliminate signal mean pairs To mitigate interference, it is also necessary to calculate the expected value of the signal sequence. The formula is as follows:

[0038] Essentially, it is the limiting mean of the signal sequence. In actual calculations, it can be approximated by the arithmetic mean of a finite number of sampling points N. This parameter will be used to construct the correction term for the oscillation component. to compensate The periodic error introduced by the signal mean and phase constant c is corrected using the following formula:

[0039] By Subtracting from the correction term can effectively eliminate [the problematic items]. The irrelevant oscillations and disturbances are excluded, and the non-periodic disturbance characteristics related to the fault are retained. Finally, these are substituted into the formula for calculating the chaos index to obtain the chaos index.

[0040] The value of the chaos index K directly quantifies the chaotic characteristics of the stator current. For example, when the unit is in a healthy state, the calculated K value is usually small (e.g., K=0.035). When the stator experiences faults such as insulation aging or rotor bar breakage, the K value will gradually increase as the fault develops. Moreover, the rate of increase of the K value varies for different fault types, which provides a basis for subsequent fault type judgment.

[0041] S3. Calculate the health index based on the chaos index, and determine the health status level according to the health index. The health index is calculated using the following formula. :

[0042] in K is the fault threshold, and K is the chaotic index calculated in real time. Its value reflects the strength of the chaotic characteristics of the stator current signal. The value of K increases during the fault development stage.

[0043] Furthermore, based on the aforementioned health index The health status is graded as follows: when H ≥ 85%, it is considered to be in a healthy state, indicating that the stator has no obvious faults and the chaotic characteristics of the current signal are within the normal range; when 70% ≤ H < 85%, it is considered to be in a slightly abnormal state, at which point K is close to... The lower limit indicates a potential early, minor fault. When 50% ≤ H < 70%, it is considered a moderate anomaly, and K has already reached the threshold. In the middle of the value range, the fault characteristics gradually become more obvious; when H < 50%, it is judged as a serious fault, and K approaches or exceeds [the threshold value]. If the upper limit is reached, timely maintenance or shutdown measures should be taken to prevent the fault from escalating.

[0044] In some implementations, the fault threshold range is determined manually based on real-time collected unit speed and output power. Specifically, since the wind turbine is a variable speed and load device, the fundamental amplitude, harmonic components, and chaotic characteristics of the stator current differ under different speed and output power conditions. If a fixed... This can lead to misjudgments of the health status. For example, under low-speed and low-power operating conditions, the chaotic characteristics of the stator current during normal operation are relatively weak, and the K value itself is relatively small. If the value is too large, minor anomalies may be misjudged as healthy conditions; under high-speed and high-power conditions, the K value for normal operation is relatively high. If the value is too small, the health status may be misjudged as abnormal. Therefore, it is necessary to first divide the wind turbine into several typical operating condition ranges based on its rated operating parameters. For example, based on the unit speed, it can be divided into three ranges: below 30% of the rated speed, 30% to 70% of the rated speed, and above 70% of the rated speed; based on the output power, it can be divided into three ranges: below 20% of the rated power, 20% to 60% of the rated power, and above 60% of the rated power. These two ranges intersect to form multiple typical operating condition combinations, each corresponding to a specific operating condition. Reference range of values.

[0045] When human judgment is correct When determining the values, it is necessary to first obtain the real-time unit speed n and output power P to determine the combination range to which the current operating condition belongs. Then, retrieve the historical operating data under this operating condition, including the maximum value of the chaotic index of the same model unit under this operating condition during healthy operation and the minimum value of the chaotic index under slightly abnormal conditions. Based on this, the range is defined. The initial value range is usually set. The value range is [the maximum value of the chaos index during healthy operation under this operating condition, and the minimum value of the chaos index under slightly abnormal conditions]. During manual judgment, the value needs to be fine-tuned by considering additional factors such as the unit's operating years, historical fault records, and current grid voltage stability. For example, for units with an operating age exceeding 5 years, considering the drift in the chaos characteristic benchmark caused by equipment aging, the value can be... The value should be adjusted 5% to 10% towards the lower limit of the range; for units with no historical fault records and stable grid voltage, the midpoint of the range can be selected as the reference value. It balances sensitivity and anti-interference capabilities.

[0046] In some optional embodiments, the method further includes determining the fault type based on the rate of increase of the chaos index; if the chaos index rises rapidly, the fault type is diagnosed as equipment faults such as rotor bar breakage and uneven air gap; if the chaos index rises slowly, the fault type is diagnosed as insulation aging.

[0047] Specifically, when determining whether the rise is rapid or slow, The rise rate threshold needs to be pre-calibrated based on historical fault data of the same model of wind turbine. First, statistically analyze the rate of increase of chaotic indicators when a large number of equipment failures such as rotor bar breakage and uneven air gap occur, and take the minimum value as... The lower limit; simultaneously, the rate of increase of the chaotic index at the occurrence of insulation aging fault is statistically analyzed, and its maximum value is taken as the lower limit; The upper limit is used to ultimately determine the rate determination threshold. For example, by calibrating through historical data =0.02, when the calculated rate of ascent v ≥ When v < 0, it is determined that the chaos index is rising rapidly; when v < 0. At that time, it was determined that the chaos index was slowly rising.

[0048] If the chaos index is determined to show a rapid upward trend, i.e., v≥ If the fault type is determined to be a structural fault such as broken rotor bars or uneven air gap, this type of fault is a sudden, progressive mechanical or electrical structural abnormality that directly disrupts the steady-state characteristics of the stator current, causing a significant increase in the chaotic characteristics of the current signal in a short period of time. Therefore, the value of the chaos index K will rise rapidly. However, if the chaos index is determined to be rising slowly, i.e., v < 0, then the fault type is determined to be structural fault. If the fault type is found to be stator winding insulation aging, the insulation aging is a gradual deterioration process. Its impact on the stator current is gradual and will not significantly change the chaotic characteristics of the current in a short period of time. It will only cause the chaotic index K to increase slowly at a low rate.

[0049] S4. Based on the aforementioned results, output the diagnostic results, which include the health index, fault type, and graded warning signals.

[0050] In this embodiment, different graded early warning signals are output based on different health status classification results. Specifically, an operation and maintenance prompt signal is output when a minor abnormality is determined, and an emergency shutdown suggestion signal is output when a serious fault is determined.

[0051] Specifically, before outputting diagnostic results, consistency verification of all prerequisite results is required to ensure logical matching among the health index, health status classification, and fault type. For example, if the health index H is in the range of 70% ≤ H < 85% and the chaos index rises slowly, the fault type corresponds to stator winding insulation aging. In this case, the classification warning signal must match the maintenance prompt requirements for minor anomalies. If the health index H < 50% and the chaos index rises rapidly, the fault type corresponds to rotor bar breakage or uneven air gap. The classification warning signal must trigger an emergency shutdown recommendation to avoid misjudgment of maintenance due to contradictory results. After verification, all results are packaged in a preset format. The health index must retain two decimal places to accurately present the quantitative value of stator health status. The fault type must clearly indicate the specific cause. If there is no clear fault type, it must be marked as "no abnormal fault characteristics".

[0052] The tiered early warning signal output adopts a combination of signal type, priority, and prompt content to adapt to the operation and maintenance needs under different operating conditions. When the health status is determined to be healthy (H≥85%), a normal operation prompt signal is output. This signal has the lowest priority and only displays a green status icon and health index value on the remote operation and maintenance platform. The local main control system only records data and does not trigger audible and visual alarms, indicating that the stator of the unit is in good operating condition and no additional operation and maintenance is required. When a minor abnormality is determined (70%≤H<85%), an operation and maintenance prompt signal is output. This signal has a medium priority. A yellow warning window pops up on the remote operation and maintenance platform, indicating the health index, the corresponding fault type (such as early aging of stator winding insulation), and the prompt content "It is recommended to carry out a special inspection within 72 hours". The field control box issues an intermittent yellow audible and visual alarm, which reminds operation and maintenance personnel to pay attention to potential hazards without affecting the normal operation of the unit.

[0053] When the health status is determined to be moderately abnormal (50%≤H<70%), a fault aggravation warning signal is output, with priority raised to the second highest level. The remote operation and maintenance platform issues a red flashing warning and simultaneously pushes the warning information to the mobile terminals of operation and maintenance personnel. A continuous yellow audible and visual alarm is activated on-site, with the message clearly stating "Fault characteristics continue to intensify; shutdown and maintenance recommended within 24 hours." A chaotic index trend curve is also attached to the diagnostic report to assist operation and maintenance personnel in judging the rate of fault development. When a serious fault is determined (H<50%), an emergency shutdown recommendation signal is output. This signal has the highest priority. The remote operation and maintenance platform immediately locks the warning interface and issues a strong alert sound effect. The on-site control box activates a strong red audible and visual alarm and simultaneously sends a shutdown trigger signal to the wind turbine main control system. For example, a 3-minute manual confirmation delay can be set to allow time for emergency handling. The diagnostic report clearly indicates the fault type, health index, and the warning message "Immediate shutdown; forced operation is prohibited" to prevent the fault from escalating and causing serious equipment accidents such as stator winding burnout and rotor damage.

[0054] After the diagnostic results are output, local storage and cloud backup are completed simultaneously. The stored content includes the complete diagnostic report, all raw data (stator current signal, chaotic index sequence, operating parameters), and warning trigger timestamps, retaining at least 90 days of historical data for subsequent traceability and analysis. Simultaneously, the remote operation and maintenance platform supports exporting and printing the diagnostic results, facilitating maintenance personnel in creating maintenance records. For recurring minor or moderate anomalies, fault thresholds can be optimized based on historical output data. Including early warning trigger conditions, continuously improve the adaptability and reliability of the diagnostic system.

[0055] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A weak fault diagnosis method based on chaotic signals of wind turbine stator current, characterized in that, The aforementioned diagnostic methods include: The current signal of the wind turbine stator is acquired and the current signal is preprocessed, including row sequence component transformation, adaptive fundamental suppression and detrending processing. Calculate the chaos index based on the preprocessed signal; A health index is calculated based on the chaos index, and a health status classification is determined based on the health index. Based on the aforementioned results, the diagnostic results are output, including health index, fault type, and graded warning signals.

2. The weak fault diagnosis method based on chaotic signals of wind turbine stator current according to claim 1, characterized in that, The preprocessing process specifically includes: converting the current signal into positive-sequence, negative-sequence, and zero-sequence components through Clarke-Park transform; extracting the negative-sequence component and then using an adaptive filter to filter out the residual fundamental wave and its main harmonic interference from the negative-sequence current signal; and performing detrending processing on the filtered signal to obtain the final signal sequence to be analyzed.

3. The weak fault diagnosis method based on chaotic signals of wind turbine stator current according to claim 2, characterized in that, The chaos index is calculated using the following formula: In the formula, The mean square shift is obtained by transforming the signal sequence to be analyzed, where n is the step size. This indicates taking the limit with respect to the step size.

4. The weak fault diagnosis method based on chaotic signals of wind turbine stator current according to claim 3, characterized in that, The health index is calculated using the following formula. : In the formula, This is the fault threshold.

5. The weak fault diagnosis method based on chaotic signals of wind turbine stator current according to claim 4, characterized in that, The range of fault threshold values ​​is determined by manual judgment based on real-time collected unit speed and output power.

6. The weak fault diagnosis method based on chaotic signals of wind turbine stator current according to claim 5, characterized in that, The mean square shift is obtained by transforming the signal sequence to be analyzed. Specifically, this includes performing a 0-1 test on the preprocessed signal sequence to obtain the transformed variable. and ,based on and Calculate the mean square displacement.

7. The weak fault diagnosis method based on chaotic signals of wind turbine stator current according to claim 6, characterized in that, According to the health index The health status is classified and determined as follows: when H ≥ 85%, it is considered to be in a healthy state; when 70% ≤ H < 85%, it is considered to be slightly abnormal; when 50% ≤ H < 70%, it is considered to be moderately abnormal; and when H < 50%, it is considered to be seriously faulty.

8. The weak fault diagnosis method based on chaotic signals of wind turbine stator current according to claim 4, characterized in that, The method also includes determining the fault type based on the rate of increase of the chaotic index. If the chaotic index rises rapidly, the fault type is diagnosed as equipment faults such as rotor bar breakage and uneven air gap; if the chaotic index rises slowly, the fault type is diagnosed as insulation aging.

9. A weak fault diagnosis method based on chaotic signals of wind turbine stator current according to claim 7, characterized in that, Based on different health status classification results, different classification warning signals are output, specifically including: when it is determined to be a minor abnormality, an operation and maintenance prompt signal is output, and when it is determined to be a serious fault, an emergency shutdown suggestion signal is output.

Images