Real-time detection method for health state of direct-current support capacitor of wind power converter

Abstract

The invention provides a wind power converter DC support capacitor health state real-time detection method, and belongs to the technical field of wind power converter DC support capacitor health state detection. Comprising the following steps: collecting a ripple voltage signal and a temperature signal of a DC bus capacitor in real time; variational mode decomposition is carried out on the ripple voltage signals, and time domain features, frequency domain features and nonlinear features of multi-mode components are extracted; meanwhile, mean value features and dynamic change rate features are extracted from the temperature signals; on the basis of robust principal component analysis, dimension reduction and feature enhancement are carried out on the extracted features, and a state feature matrix is generated; inputting the state feature matrix into a pre-trained intelligent learning model, and outputting a health state classification result of the DC support capacitor; according to the method, the health state of the capacitor is accurately reflected in real time, degradation and faults are found in time, and the method can be widely applied to real-time monitoring of the wind power converter.

Description

Technical Field

[0001] The present application relates to the technical field of wind power converter DC support capacitor health status detection, and in particular to a real-time detection method for the health status of a wind power converter DC support capacitor. Background Art

[0002] Wind turbine converters are key components in wind power systems. Their primary function is to convert and match energy between wind power generation and the power grid, ensuring efficient and safe operation of wind turbines and stable power output. Research shows that over 27% of faults in wind power systems are caused by faults in the converter. Therefore, its reliability directly determines the safe and stable operation of the entire wind power system. Wind turbine converters consist of components such as IGBT modules, DC bus capacitor modules, driver modules, and protection circuits. The components most prone to failure in wind turbine converters are the IGBT modules and DC bus capacitor modules, with a combined failure rate exceeding 50%. Failures in these components can cause converter shutdown, impacting the reliable operation of the entire wind power system.

[0003] As a key component in wind turbine converters, IGBT modules are responsible for converting and transmitting electrical energy between the generator and grid, making them essential core components. The DC bus capacitors, acting as energy storage devices, primarily balance the power between the generator and grid converters, stabilize the DC bus voltage, and absorb the AC current in both converter stages. Failure of either or both components can trigger a chain reaction, ultimately paralyzing the entire wind turbine system.

[0004] Traditional monitoring of the DC link capacitors of wind turbine converters relies on offline testing or simple threshold determination, which is difficult to meet real-time monitoring requirements and cannot accurately reflect capacitor degradation trends and complex fault characteristics. Therefore, there is an urgent need to develop a detection method that can accurately monitor the health of wind turbine converter DC link capacitors in real time, detect degradation and faults promptly, and improve the reliability and stability of wind power systems. Summary of the Invention

[0005] In view of the above-mentioned defects or deficiencies in the prior art, the present application aims to provide a real-time detection method for the health status of the DC support capacitor of a wind power converter, thereby realizing the status monitoring of key components of the wind power converter, mainly the IGBT module and the DC bus aluminum electrolytic capacitor; The detection method comprises the following steps: Real-time acquisition of ripple voltage signal and temperature signal of DC bus capacitor; Performing variational modal decomposition on the ripple voltage signal to obtain multimodal components, extracting time domain features, frequency domain features, and nonlinear features of the multimodal components; and extracting mean features and dynamic change rate features of the temperature signal; Based on robust principal component analysis, the extracted features are subjected to dimensionality reduction and feature enhancement to generate a state feature matrix; the extracted features include the time domain features, the frequency domain features, the nonlinear features, the mean features, and the dynamic change rate features; Inputting the state feature matrix into a pre-trained intelligent learning model to output a health state classification result of the DC link capacitor, wherein the health state classification result includes a normal state, a slightly degraded state, or a severely degraded state; A real-time alarm is triggered based on the health status classification results, and abnormal parameters are recorded in the monitoring system.

[0006] According to the technical solution provided in the embodiment of the present application, after the real-time acquisition of the ripple voltage signal and the temperature signal of the DC bus capacitor, the following steps are included: determining a working state of the converter according to the ripple voltage signal and the temperature signal; The performing variational modal decomposition on the ripple voltage signal comprises the following steps: If the working state of the converter is normal, variational modal decomposition is performed on the ripple voltage signal.

[0007] According to the technical solution provided in the embodiment of the present application, performing variational modal decomposition on the ripple voltage signal includes the following steps: The ripple voltage signal is decomposed into 10×n modal components, and the sample entropy, mean value and weighted permutation entropy of each modal component are extracted as nonlinear features.

[0008] According to the technical solution provided in the embodiment of the present application, the dynamic change rate characteristics of the temperature signal include a maximum temperature value and a temperature change amplitude.

[0009] According to the technical solution provided in the embodiment of the present application, the intelligent learning model is a long short-term memory network, and its training process includes data partitioning, cross-validation optimization and classification label definition, wherein the label of the normal state is 0, the label of the slightly degraded state is 1, and the label of the severely degraded state is 2.

[0010] According to the technical solution provided in the embodiment of the present application, triggering a real-time alarm based on the health status classification result includes the following steps: If the health status classification result is the severe degradation state, an alarm signal is triggered and abnormal parameters are recorded; If the temperature change amplitude in the dynamic change rate feature exceeds a first preset threshold, a heat dissipation abnormality alarm is triggered synchronously.

[0011] According to the technical solution provided in the embodiment of the present application, the ripple voltage signal of the DC bus capacitor is collected in real time by a voltage sensor, the voltage sensor is installed at the positive and negative poles of the capacitor, and the sampling frequency is not less than 1kHz; the temperature signal of the DC bus capacitor is collected in real time by an ambient temperature sensor and a capacitor surface temperature sensor, the ambient temperature sensor and the capacitor surface temperature sensor are respectively installed at the radiator and the surface of the capacitor casing.

[0012] According to the technical solution provided in the embodiment of the present application, before obtaining the health status classification result of the DC link capacitor based on the state characteristic matrix, the following steps are included: Get the equivalent series resistance value of the capacitor; Obtaining a health status classification result of the DC link capacitor based on the state characteristic matrix includes the following steps: If the equivalent series resistance value is less than a second preset threshold, obtaining a health status classification result of the DC link capacitor based on the state characteristic matrix; After obtaining the equivalent series resistance value of the capacitor, the following steps are further included: If the equivalent series resistance value is greater than or equal to the second preset threshold, the health status classification result is forced to be a slightly degraded state; if the equivalent series resistance value is greater than or equal to the third preset threshold, the health status classification result is forced to be a severely degraded state; The third preset threshold is greater than the second preset threshold.

[0013] According to the technical solution provided in the embodiment of the present application, the method further includes the following steps: Get the accumulated running time of the capacitor in real time; If the temperature change amplitude in the dynamic change rate feature exceeds a first preset threshold, a heat dissipation abnormality alarm is synchronously triggered, including the following steps: If the accumulated operating time of the capacitor is less than a first preset time, and if the temperature change amplitude in the dynamic change rate characteristic exceeds a first preset threshold, a heat dissipation abnormality alarm is triggered synchronously.

[0014] According to the technical solution provided in the embodiment of the present application, after obtaining the accumulated running time of the capacitor in real time, the following steps are also included: If the cumulative operating time of the capacitor is greater than or equal to the first preset time, then if the temperature change amplitude in the dynamic change rate characteristic exceeds a fourth preset threshold, a heat dissipation abnormality alarm is triggered synchronously; the fourth preset threshold is less than the first preset threshold.

[0015] It can be seen from the above technical solution that this application has at least the following beneficial effects: This method uses variational mode decomposition (VMD) to perform multimodal decomposition on the ripple voltage signal, extracting time-domain (peak value, RMS value), frequency-domain (main frequency, harmonic distortion), and nonlinear features (sample entropy, weighted permutation entropy) to comprehensively characterize the capacitor state. Mean and dynamic rate of change features (such as temperature rise rate and temperature variation amplitude) are extracted from the temperature signal to comprehensively reflect the capacitor's thermal characteristics. Robust principal component analysis (RPCA) is used to reduce and enhance the multi-source features, separating useful signals from noise and improving feature quality. Through multi-source signal fusion and intelligent feature processing, this method overcomes the limitations of traditional single-parameter threshold methods, which are susceptible to noise interference (such as transient ripple fluctuations caused by sudden load changes), significantly reducing noise interference and improving classification accuracy. Furthermore, the state feature matrix is ​​input into a pre-trained intelligent learning model to output a health status classification result for the DC link capacitor. This model accurately reflects the capacitor's health status in real time and promptly detects degradation and faults. The system is simple in structure and can be widely used for real-time monitoring of wind turbine converters. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A flowchart of the steps of the method for real-time detection of the health status of the DC support capacitor of a wind power converter provided in an embodiment of the present application; Figure 2 A topology diagram of a doubly-fed wind turbine converter provided in an embodiment of the present application; Figure 3 The converter circuit topology provided in the embodiment of the present application; Figure 4 A feature extraction flow chart provided for an embodiment of the present application; Figure 5 A schematic diagram of the entire system structure provided in an embodiment of the present application. DETAILED DESCRIPTION

[0017] The terms "first", "second" and "third" in this application specification and the accompanying drawings are used to distinguish different objects rather than to limit a specific order.

[0018] In the embodiments of this application, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0019] To make the description of the following embodiments clear and concise, a brief introduction to the related technologies is first given: The status monitoring of DC support capacitors in traditional wind turbine converters relies on offline detection or simple threshold judgment, which is difficult to meet real-time monitoring needs and cannot accurately reflect the degradation trend and complex fault characteristics of the capacitors.

[0020] In view of this, an embodiment of the present application provides a real-time detection method for the health status of the DC support capacitor of a wind power converter. In order to make the technical solution of the present application clearer and easier to understand, the following introduces a real-time detection method for the health status of the DC support capacitor of a wind power converter provided in an embodiment of the present application.

[0021] like Figure 1 As shown, the following steps are included: S1, real-time acquisition of the ripple voltage signal and temperature signal of the DC bus capacitor; Furthermore, the ripple voltage signal of the DC bus capacitor is collected in real time through a voltage sensor, which is installed at the positive and negative poles of the capacitor, and the sampling frequency is not less than 1kHz; the temperature signal of the DC bus capacitor is collected in real time through an ambient temperature sensor and a capacitor surface temperature sensor, which are installed at the radiator and the surface of the capacitor casing respectively.

[0022] Specifically, real-time ripple voltage and temperature signals are collected from the sensor. The specific process is as follows: Figure 5 As shown in the figure, sensors and other external devices are installed on the converter, and a data acquisition system is used to extract the capacitor's ripple voltage and temperature signals. The voltage sensor is mounted on both sides of the capacitor, and two temperature sensors measure the ambient temperature and the capacitor's surface temperature, respectively. The signal conditioning circuit transmits the collected signals to the data acquisition device, and ultimately transmits them to the host computer via USB to obtain the converter's busbar capacitor ripple voltage and temperature signals.

[0023] S2. performing variational modal decomposition on the ripple voltage signal to obtain multimodal components, and extracting time domain features, frequency domain features, and nonlinear features of the multimodal components; and simultaneously extracting mean features and dynamic change rate features of the temperature signal; Furthermore, after the real-time acquisition of the ripple voltage signal and the temperature signal of the DC bus capacitor, the following steps are included: determining a working state of the converter according to the ripple voltage signal and the temperature signal; The performing variational modal decomposition on the ripple voltage signal comprises the following steps: If the working state of the converter is normal, variational modal decomposition is performed on the ripple voltage signal.

[0024] Specifically, taking a doubly fed wind turbine generator set as an example, the doubly fed wind turbine converter topology is as follows: Figure 2 As shown, the converter topology is as follows Figure 3 As shown. Capacitors can be divided into two categories according to their application location. One category is used as filter capacitors on the AC side, and the other category is used on the DC side to support DC and absorb bus ripple. The capacitors of wind power converters are aluminum electrolytic capacitors supported by the DC bus. In the capacitor health status monitoring of wind power converters, ripple voltage is an important characteristic indicator, which is usually used to evaluate the health status of capacitors. Voltage ripple refers to the periodic fluctuation of the voltage across the capacitor, which is usually caused by the switching frequency of the converter or the operation of power electronic equipment. The standard of voltage ripple is defined by the following aspects: (1) Ripple Amplitude: The amplitude of voltage ripple is the difference between the maximum and minimum values ​​of the capacitor voltage fluctuation. Voltage ripple is usually expressed as peak-to-peak value (Vpp), and its calculation formula is: V pp =V max -V min Among them, V max and V min Represent the maximum and minimum values ​​of the capacitor voltage respectively.

[0025] (2) The RMS value of the ripple voltage is calculated as follows:

[0026] This formula is used to measure the intensity of voltage fluctuations. is the RMS value of the voltage; T is one cycle of the ripple voltage (cycle length), in seconds (s); v ( t ) is the instantaneous value of the ripple voltage signal changing with time, which is a function of time.

[0027] (3) Ripple factor: This is a commonly used indicator of voltage ripple, defined as the ratio of the effective value of the AC component to the DC component. It indicates the intensity of the ripple. The larger the value, the more severe the ripple. Its calculation formula is:

[0028] Among them, V AC is the effective value of the AC part (ripple part), V DC is the voltage of the DC part.

[0029] (4) Ripple frequency: This is the frequency of the voltage ripple cycle, which is usually determined by the operating frequency of the converter. The higher the frequency, the finer the ripple. The voltage ripple frequency should be related to the converter operating frequency, switching frequency, etc. Its calculation formula is:

[0030] Among them, T ripple is the ripple period, is the ripple frequency.

[0031] (5) Total harmonic distortion: The quality of the ripple is evaluated by calculating the harmonic components of the voltage ripple signal. The higher the total harmonic distortion, the more high-frequency harmonic components in the voltage ripple, which usually means that the capacitor is aging or the converter is unstable. The calculation formula is:

[0032] Among them, V1 is the amplitude of the fundamental wave (DC voltage), V2, V3, ···V n is the amplitude of each harmonic order.

[0033] IEC standards generally stipulate that during normal operation of equipment, the amplitude of voltage fluctuations and ripples must be within specified ranges, typically not exceeding 10% of the rated voltage. In the wind power industry, the peak (peak-to-peak) value of voltage ripple is generally required to be no more than 2%-5% of the rated voltage. Total harmonic distortion (THD) must typically be controlled below 5% to ensure system efficiency and stability. Furthermore, in the wind power sector, temperature is a key parameter for equipment such as wind turbine converters and capacitors. As a core component of wind turbine converters, capacitors generally should not operate at temperatures exceeding 85°C. Temperatures above this limit can affect the lifespan and stability of capacitors. The IEC 60721-3-3 standard generally specifies an operating temperature range of 0°C to 40°C.

[0034] In summary, when the voltage ripple amplitude is less than 5% and the temperature is between 0°C and 40°C, the converter is judged to be operating normally; when the voltage ripple amplitude is between 5% and 10% and the temperature is between 40°C and 85°C, the converter is judged to be operating degraded; and when the voltage ripple amplitude is greater than 10% and the temperature exceeds 85°C, the converter is judged to be operating faulty.

[0035] Furthermore, performing variational modal decomposition on the ripple voltage signal comprises the following steps: The ripple voltage signal is decomposed into 10×n modal components, and the sample entropy, mean value and weighted permutation entropy of each modal component are extracted as nonlinear features.

[0036] Furthermore, the dynamic change rate characteristics of the temperature signal include a maximum temperature value and a temperature change amplitude.

[0037] Furthermore, the intelligent learning model is a long short-term memory network, and its training process includes data partitioning, cross-validation optimization and classification label definition, wherein the label of the normal state is 0, the label of the slightly degraded state is 1, and the label of the severely degraded state is 2.

[0038] S3. Based on robust principal component analysis, perform dimensionality reduction and feature enhancement on the extracted features to generate a state feature matrix; the extracted features include the time domain features, the frequency domain features, the nonlinear features, the mean features, and the dynamic change rate features; Specifically, if Figure 4 As shown in the figure, the ripple voltage and temperature signals are first collected respectively, the VMD mode number a=10 is set, and the VMD algorithm is executed to decompose the original signal X(t) into 10×n modal components, where Mi(t) represents the i-th modal component.

[0039]

[0040] Get 10×n modal components M1(t), M2(t),…, M an (t), these components represent the information of the signal in different frequency bands.

[0041] Then, key features are extracted from the modal components. The sample entropy, mean, and weighted permutation entropy of the ripple voltage modal component and the mean temperature, maximum temperature, and temperature variation of the temperature modal component are calculated. A feature vector is generated for each modal component:

[0042] Where b is the number of features extracted.

[0043] Integrate the features of all modal components to form a complete initial feature matrix.

[0044]

[0045] The shape of the initial feature matrix F is a × b. Each column of the initial feature matrix is ​​normalized to ensure that the dimensions of different features are consistent.

[0046] Finally, dimensionality reduction is used to reduce redundant features and extract the most discriminative features. The covariance of the standardized initial feature matrix F is calculated using the following formula:

[0047] Where n is the number of samples, represents the i-th sample vector in the feature matrix F, is the sample mean vector Decompose the covariance matrix to obtain eigenvalues ​​and eigenvectors. According to the contribution rate of the eigenvalues, select the first k principal components. The characteristic matrix after dimensionality reduction is:

[0048] Where W k Is a matrix composed of the first k eigenvectors. Output the reduced eigenvector matrix FPCA , the shape is a×k.

[0049] The regularization parameter λ is set for the feature matrix F obtained from the feature extraction stage, where a and b are the number of rows and columns of the feature matrix.

[0050]

[0051] RPCA is used to extract enhanced features L and S, where is the nuclear norm of the matrix L, is the L1 norm of the matrix S.

[0052]

[0053] After RPCA enhancement, the feature matrix may still contain features that are unrelated to the target variable. The low-rank feature matrix L after RPCA enhancement is input and the normalized mutual information (NMI) is used to further filter out the features that are most correlated with the target variable.

[0054]

[0055] Where NMI(X;Y) is the normalized information, IMI(X;Y) is the mutual information between X and Y, H(X) represents the information entropy of X, and H(Y) represents the information entropy of Y. p (x,y) is the joint probability distribution of X and Y, p (x) represents the marginal probability distribution of X, p (y) represents the marginal probability distribution of Y.

[0056] Sort the features by NMI value and select the k features with the highest correlation: (k is the number of features after dimensionality reduction), and finally the state feature matrix is ​​obtained.

[0057] S4. Input the state feature matrix into a pre-trained intelligent learning model, and output a health status classification result of the DC link capacitor, where the health status classification result includes a normal state, a slightly degraded state, or a severely degraded state; Specifically, the health status of capacitors is classified into three levels according to the capacitance value: Level 1: Normal state Capacitance value range: 100% - 90% of rated capacitance.

[0058] This defines a capacitor as performing well, showing no noticeable signs of aging or degradation. The ripple voltage and equivalent series resistance are both within the design range. The voltage ripple is low (less than 5% of the rated value), the ESR increase is low (close to the initial value), and the temperature change is stable.

[0059] Level 2: Slightly degraded state Capacitance value range: 89% - 70% of rated capacitance.

[0060] This is defined as a capacitor experiencing a certain degree of performance degradation, possibly due to long-term use or high ambient temperature, resulting in a decrease in filtering capability. This occurs when the voltage ripple increases slightly (between 5% and 10% of the rated value), the ESR increases significantly (greater than 50% of the initial value), and the temperature changes exceed the normal range but do not reach the danger threshold.

[0061] Level 3: Severely degraded state Capacitance value range: less than 70% of the rated capacitance value.

[0062] This condition is defined as severe capacitor degradation, with a sharp drop in filtering capability, leading to unstable inverter output and the risk of failure. Characteristics include a significant increase in voltage ripple (exceeding 10% of the rated value), a significant increase in ESR (reaching 2-3 times the initial value), and a rapid temperature rise, approaching or exceeding the rated operating temperature of the capacitor.

[0063] Based on the above capacitor health status definition, different levels of capacitor degradation are simulated by adjusting the power input under laboratory conditions.

[0064] Ripple voltage V ripple (t) is the high-frequency component in the DC bus voltage, which is related to the load power P and the degradation degree ΔC of the capacitor C.

[0065]

[0066] in is the load current, operating power P and DC voltage V dc Related; is the equivalent capacitance value, taking into account the degradation of the capacitance; is the rated capacitance value; is the capacitance decay factor.

[0067] The range of setting the operating power P is divided into three cases: low load, medium load and high load.

[0068] Low load: P∈[10%,40%], ; Medium load: P∈[41%,80%], ; High load: P∈[81%,100%], .

[0069] Temperature T(t) is the operating temperature of a capacitor and is related to the load power P and the internal resistance of the capacitor. High load power and capacitor degradation will increase the temperature rise.

[0070]

[0071] is the ambient temperature (constant); is the thermal resistance coefficient (related to heat dissipation conditions); is the capacitor power loss; , ESR is the equivalent series resistance; Where ESRnom is the nominal equivalent series resistance, is the change in equivalent series resistance.

[0072] Low load: P∈[10%,40%], ; Medium load: P∈[41%,80%], ; High load: P∈[81%,100%], .

[0073] The training process of the intelligent learning model (Long Short-Term Memory Network, LSTM) mainly includes the following three steps: Step 1: Data Partitioning The collected capacitor status data (including ripple voltage characteristics, temperature characteristics, and equivalent series resistance, etc.) is divided into training, validation, and test sets. This application uses a ratio of 70%-80% training set, 10%-15% validation set, and 10%-15% test set. When dividing, pay attention to stratified sampling to ensure that each health status category (normal, slightly degraded, severely degraded) is evenly distributed in each data set to avoid model misjudgment due to data skew. For example, if the normal state samples account for 70% of the original data, the proportion of normal samples in the training set, validation set, and test set should also be maintained at approximately 70%.

[0074] Step 2: Cross-validation optimization During the training process, the LSTM hyperparameters are optimized through K-fold cross-validation (e.g., 5-fold or 10-fold). The specific steps are as follows: Parameter grid definition: Set the range of hyperparameters to be optimized, such as learning rate (0.001-0.1), number of layers (1-3), number of neurons (64-256), dropout rate (0.2-0.5), etc.

[0075] Cross-validation iterations: Divide the training set into K subsets, select K-1 subsets each time to train the model, and use the remaining subset as the validation set. Repeat this K times to evaluate the performance of different parameter combinations. For example, when using 5-fold cross-validation, the model is trained five times, each time using 4 / 5 of the data for training and 1 / 5 for validation.

[0076] Performance evaluation and parameter selection: The optimal parameter combination is screened using metrics such as validation set accuracy and F1 score. For example, if a parameter combination has the highest average accuracy across five validation runs, it is selected as the final hyperparameter.

[0077] Step 3: Classification label definition Health states are mapped to numerical labels: 0 for normal, 1 for slightly degraded, and 2 for severely degraded. This definition directly influences the design of the LSTM output layer—it must contain three neurons (corresponding to the three categories) and use a softmax activation function to convert the output values ​​into a probability distribution. During training, a cross-entropy loss function is used, minimizing the difference between the predicted probabilities and the true labels through an optimizer (such as Adam).

[0078] S5. Trigger a real-time alarm based on the health status classification result, and record abnormal parameters to the monitoring system.

[0079] Furthermore, triggering a real-time alarm according to the health status classification result includes the following steps: If the health status classification result is the severe degradation state, an alarm signal is triggered and abnormal parameters are recorded; If the temperature change amplitude in the dynamic change rate feature exceeds a first preset threshold, a heat dissipation abnormality alarm is triggered synchronously.

[0080] Specifically, based on the health status classification results, the system outputs the capacitor status and triggers an alarm or maintenance instruction. If the status is "severely degraded" or "failed," the system triggers an alarm signal, records the abnormal time and key parameters, and outputs the health status classification results to the monitoring system for display and archiving.

[0081] In a preferred embodiment, before obtaining the health status classification result of the DC link capacitor based on the state characteristic matrix, the following steps are included: Get the equivalent series resistance value of the capacitor; Specifically, an LCR meter (such as the Keysight E4980A) is used to measure the impedance response of the DC bus capacitor by injecting a 1kHz sinusoidal excitation signal across the DC bus capacitor while the converter is operating. The equivalent series resistance (ESR) is calculated based on the total impedance, capacitive reactance, and real-time capacitance. The ESR measurement is transmitted to the control board via a signal conditioning circuit, with the sampling frequency synchronized with the voltage and temperature signals.

[0082] Obtaining a health status classification result of the DC link capacitor based on the state characteristic matrix includes the following steps: If the equivalent series resistance value is less than a second preset threshold, obtaining a health status classification result of the DC link capacitor based on the state characteristic matrix; After obtaining the equivalent series resistance value of the capacitor, the following steps are further included: If the equivalent series resistance value is greater than or equal to the second preset threshold, the health status classification result is forced to be a slightly degraded state; if the equivalent series resistance value is greater than or equal to the third preset threshold, the health status classification result is forced to be a severely degraded state; The third preset threshold is greater than the second preset threshold.

[0083] Specifically, the second preset threshold is set to 1.5 times the initial ESR value. For example, if the capacitor's initial ESR value is 0.01Ω, the second preset threshold is 0.015Ω. The third preset threshold is set to 2.0 times the initial ESR value, which is 0.02Ω. The threshold data is stored in the non-volatile memory of the monitoring system.

[0084] This embodiment takes into account the correlation between ESR and capacitor degradation. When the capacitor ages, electrolyte drying or electrode corrosion causes a significant increase in ESR. This change occurs earlier than capacitance decay, making it a sensitive indicator of early degradation. Therefore, under sudden load changes or environmental interference, the ripple voltage and temperature characteristics may show temporary abnormalities, while the ESR changes are continuous. Forced classification through thresholds can avoid misjudgment.

[0085] In a preferred embodiment, the method further comprises the following steps: Get the accumulated running time of the capacitor in real time; If the temperature change amplitude in the dynamic change rate feature exceeds a first preset threshold, a heat dissipation abnormality alarm is synchronously triggered, including the following steps: If the accumulated operating time of the capacitor is less than a first preset time, and if the temperature change amplitude in the dynamic change rate characteristic exceeds a first preset threshold, a heat dissipation abnormality alarm is triggered synchronously.

[0086] Specifically, a real-time clock module is built into the monitoring system to record the total operating time of the capacitor since its initial commissioning. This data is stored in EEPROM to prevent loss during power outages. Time-to-threshold mapping rules: The first preset duration is 10,000 hours (approximately 1.14 years), corresponding to the typical maintenance cycle of a wind turbine converter. The first preset threshold: The temperature change amplitude threshold for new capacitors (operating time <10,000 hours) is 8°C / min.

[0087] Furthermore, after obtaining the cumulative running time of the capacitor in real time, the following steps are also included: If the cumulative operating time of the capacitor is greater than or equal to the first preset time, then if the temperature change amplitude in the dynamic change rate characteristic exceeds a fourth preset threshold, a heat dissipation abnormality alarm is triggered synchronously; the fourth preset threshold is less than the first preset threshold.

[0088] Specifically, the fourth preset threshold for older capacitors (operating time ≥ 10,000 hours) is adjusted to 5°C / min, and the threshold is reduced by 1°C / min for every 5,000 hours. For example: 10,000-15,000 hours: threshold = 5°C / min; 15,000-20,000 hours: threshold = 4°C / min; ≥ 20,000 hours: threshold = 3°C / min.

[0089] This implementation accounts for changes in temperature sensitivity caused by the decline in heat dissipation performance of aging capacitors. New capacitors may briefly overheat under high load, but aged capacitors heat up more quickly under the same operating conditions. Fixed thresholds can easily lead to early missed alarms or late false alarms. Dynamically adjusting the temperature alarm threshold based on the capacitor's cumulative operating time creates a differentiated alarm strategy with a looser threshold for new capacitors and a stricter threshold for older capacitors, improving alarm accuracy.

[0090] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. The above is only the preferred implementation method of this application. It should be pointed out that due to the limitations of textual expression, there are objectively infinite specific structures. For ordinary technicians in this technical field, without departing from the principles of the present invention, they can also make several improvements, modifications or changes, and can also combine the above technical features in an appropriate manner; these improvements, modifications, changes or combinations, or the direct application of the inventive concept and technical solution to other occasions without improvement, should be regarded as the scope of protection of this application.

Claims

1. A real-time detection method for the health status of a DC support capacitor of a wind power converter, characterized in that: The following steps are involved: Real-time acquisition of ripple voltage signal and temperature signal of DC bus capacitor; performing variational modal decomposition on the ripple voltage signal to obtain multimodal components, and extracting time domain features, frequency domain features, and nonlinear features of the multimodal components; Extracting mean value features and dynamic change rate features from the temperature signal; Based on robust principal component analysis, the extracted features are subjected to dimensionality reduction and feature enhancement to generate a state feature matrix; the extracted features include the time domain features, the frequency domain features, the nonlinear features, the mean features, and the dynamic change rate features; Inputting the state feature matrix into a pre-trained intelligent learning model to output a health state classification result of the DC link capacitor, wherein the health state classification result includes a normal state, a slightly degraded state, or a severely degraded state; A real-time alarm is triggered based on the health status classification results, and abnormal parameters are recorded in the monitoring system.

2. The method for real-time detection of the health status of a DC link capacitor of a wind power converter according to claim 1, characterized in that: After the real-time acquisition of the ripple voltage signal and the temperature signal of the DC bus capacitor, the following steps are included: determining a working state of the converter according to the ripple voltage signal and the temperature signal; The performing variational modal decomposition on the ripple voltage signal comprises the following steps: If the working state of the converter is normal, variational modal decomposition is performed on the ripple voltage signal.

3. The method for real-time detection of the health status of a DC link capacitor of a wind power converter according to claim 1, characterized in that: The performing variational modal decomposition on the ripple voltage signal comprises the following steps: The ripple voltage signal is decomposed into 10×n modal components, and the sample entropy, mean value and weighted permutation entropy of each modal component are extracted as nonlinear features.

4. The method for real-time detection of the health status of a DC link capacitor of a wind power converter according to claim 1, characterized in that: The dynamic change rate characteristics of the temperature signal include a maximum temperature value and a temperature change amplitude.

5. The method for real-time detection of the health status of a DC link capacitor of a wind power converter according to claim 1, characterized in that: The intelligent learning model is a long short-term memory network, and its training process includes data partitioning, cross-validation optimization and classification label definition, wherein the label of the normal state is 0, the label of the slightly degraded state is 1, and the label of the severely degraded state is 2.

6. The method for real-time detection of the health status of a DC link capacitor of a wind power converter according to claim 4, characterized in that: The triggering of a real-time alarm according to the health status classification result comprises the following steps: If the health status classification result is the severe degradation state, an alarm signal is triggered and abnormal parameters are recorded; If the temperature change amplitude in the dynamic change rate feature exceeds a first preset threshold, a heat dissipation abnormality alarm is triggered synchronously.

7. The method for real-time detection of the health status of a DC link capacitor of a wind power converter according to claim 1, characterized in that: The ripple voltage signal of the DC bus capacitor is collected in real time through a voltage sensor. The voltage sensor is installed at the positive and negative poles of the capacitor, and the sampling frequency is not less than 1kHz. The temperature signal of the DC bus capacitor is collected in real time through an ambient temperature sensor and a capacitor surface temperature sensor. The ambient temperature sensor and the capacitor surface temperature sensor are installed at the radiator and the surface of the capacitor casing respectively.

8. The method for real-time detection of the health status of a DC link capacitor of a wind power converter according to claim 1, characterized in that: Before obtaining the health status classification result of the DC link capacitor based on the state characteristic matrix, the following steps are included: Get the equivalent series resistance value of the capacitor; Obtaining a health status classification result of the DC link capacitor based on the state characteristic matrix includes the following steps: If the equivalent series resistance value is less than a second preset threshold, obtaining a health status classification result of the DC link capacitor based on the state characteristic matrix; After obtaining the equivalent series resistance value of the capacitor, the following steps are further included: If the equivalent series resistance value is greater than or equal to the second preset threshold, the health status classification result is forced to be a slightly degraded state; if the equivalent series resistance value is greater than or equal to the third preset threshold, the health status classification result is forced to be a severely degraded state; The third preset threshold is greater than the second preset threshold.

9. The method for real-time detection of the health status of a DC link capacitor of a wind power converter according to claim 6, characterized in that: The method further comprises the following steps: Get the accumulated running time of the capacitor in real time; If the temperature change amplitude in the dynamic change rate feature exceeds a first preset threshold, a heat dissipation abnormality alarm is synchronously triggered, including the following steps: If the accumulated operating time of the capacitor is less than a first preset time, and if the temperature change amplitude in the dynamic change rate characteristic exceeds a first preset threshold, a heat dissipation abnormality alarm is triggered synchronously.

10. The method for real-time detection of the health status of a DC link capacitor of a wind power converter according to claim 9, characterized in that: After obtaining the accumulated running time of the capacitor in real time, the following steps are also included: If the cumulative operating time of the capacitor is greater than or equal to the first preset time, then if the temperature change amplitude in the dynamic change rate characteristic exceeds a fourth preset threshold, a heat dissipation abnormality alarm is triggered synchronously; the fourth preset threshold is less than the first preset threshold.

Images