A double-fed wind turbine parasitic capacitance closed-loop control method

Abstract

The application provides a kind of double-fed wind turbine parasitic capacitance closed-loop control method, belong to wind generating set electrical parameter identification and shaft current suppression technical field, can significantly alleviate or solve the existing parasitic capacitance, it is difficult to adapt to oil film thickness, PWM high frequency component, shell vibration and aging change, and the calculation result is difficult to be used for active suppression of shaft current Problem, the application collects electrical quantity, mechanical quantity, environment and service state quantity and inverter control parameter, establishes the dynamic parasitic capacitance model of introducing lubricating oil film dynamic thickness, conductor skin effect and air gap modulation quantity, and obtains nonlinear mapping relationship based on L-M iteration of finite element prior constraint;According to real-time parasitic capacitance prediction shaft voltage and shaft current, execute PWM parameter correction, bearing insulation early warning and shaft voltage active compensation.The method can improve the stability of capacitance identification and shaft current suppression effect.

Description

Technical Field

[0001] This invention belongs to the field of electrical parameter identification and shaft current suppression technology for wind turbine generator sets, specifically relating to a closed-loop control method for parasitic capacitance of doubly fed wind turbines. Background Technology

[0002] With the continuous development of wind power generation technology, doubly-fed induction generators (DFIGs) have been widely used in onshore wind farms, offshore wind farms, and large-scale renewable energy grid-connected systems due to their characteristics such as variable speed and constant frequency operation, relatively small converter capacity, and flexible power regulation. During the operation of a DFIG, the rotor side is typically controlled by a PWM converter to achieve coordinated regulation of active power, reactive power, and speed. However, the PWM converter generates high-frequency pulse components with high voltage change rates during switching operations. These high-frequency components easily form complex common-mode voltage coupling paths between the motor windings, rotor core, shaft, bearings, and casing. When the high-frequency common-mode voltage is coupled to the shaft and bearings via internal parasitic capacitance, it can induce shaft voltage and shaft current, leading to problems such as bearing oil film breakdown, raceway erosion, deterioration of lubrication, and reduced bearing life. Therefore, accurately identifying the internal parasitic capacitance parameters of the DFIG and further utilizing these parameters to suppress shaft current has become a crucial technical issue for the safe and stable operation of DFIGs.

[0003] From the perspective of electromagnetic field distribution, a doubly-fed induction generator (DFIG) exhibits parasitic capacitances at least between the rotor windings and the casing, between the rotor windings and the shaft, and between the casing and the shaft. Simultaneously, there are equivalent impedances between the casing and the shaft related to bearing oil film, grounding status, and insulation status. In traditional analysis, these parasitic parameters are often treated as fixed parameters or estimated using a parallel-plate capacitor approximation model. While these methods are relatively simple to calculate, they fail to accurately reflect the dynamic characteristics of DFIGs under conditions of startup, speed-up, constant-speed operation, variable-load operation, high-temperature and high-humidity operation, and long-term service aging. Especially in actual operation, the casing-shaft parasitic capacitance is affected not only by temperature, humidity, and rotational speed, but also by the combined effects of bearing lubricating oil film thickness, PWM high-frequency carrier frequency, casing vibration, rotor eccentricity, air gap changes, and insulation material aging. Its variation exhibits significant nonlinearity and time-varying characteristics.

[0004] While some existing technologies can estimate parasitic capacitance by collecting data such as voltage, current, speed, temperature, and humidity, or fit the relationship between parasitic capacitance and operating parameters using nonlinear iterative algorithms, these methods typically remain at the level of "numerical capacitance calculation." On one hand, their iterative solution process lacks prior constraints from the electromagnetic field finite element method, making them prone to parameter oscillations, model drift, or calculation results that do not conform to physical laws when sensor data is missing, sampling noise increases, or operating conditions change abruptly. On the other hand, they do not fully consider the long-term deviations in capacitance reference values ​​caused by factors such as insulation material aging, bearing wear, and changes in the dielectric properties of lubricating oil after long-term wind turbine operation, resulting in insufficient adaptability of the model under long-term service scenarios.

[0005] Furthermore, existing parasitic capacitance calculations typically fail to form an effective closed loop with the wind turbine control system. Even with relatively accurate parasitic capacitance parameters, they are difficult to use for proactive adjustment of the inverter's PWM dead time or switching frequency, and cannot proactively suppress high-frequency common-mode voltage coupling paths based on real-time shaft current risks. Simultaneously, existing bearing condition monitoring relies heavily on temperature, vibration, or shaft current threshold alarms, lacking early warning criteria for insulation performance based on abnormal drift of the casing-shaft parasitic capacitance, capacitance fluctuation rate, and current prediction residuals. This makes it difficult to promptly identify early risks such as unstable bearing insulating oil film, localized breakdown, or insulation degradation. For shaft voltage compensation, existing solutions often employ fixed impedance or fixed grounding branches, failing to dynamically adjust the impedance matching relationship of the compensation branch based on real-time parasitic capacitance and the PWM high-frequency spectrum. Consequently, under complex operating conditions, insufficient compensation or excessive compensation losses are prone to occur.

[0006] To address this, a closed-loop control method for the parasitic capacitance of a doubly fed fan is proposed. Summary of the Invention

[0007] The present invention aims to solve at least one of the technical problems existing in the prior art, and provides a closed-loop control method for parasitic capacitance of doubly fed wind turbines.

[0008] This invention provides a closed-loop control method for the parasitic capacitance of a doubly-fed wind turbine, comprising the following steps: S1: Collect electrical, mechanical, environmental, and service status parameters of the doubly-fed induction generator (DFIG) during operation, as well as inverter control parameters; and construct an equivalent model of the parasitic capacitance of the DFIG to determine the parasitic capacitance between the rotor windings and the casing. Parasitic capacitance between rotor windings and shaft Parasitic capacitance between the housing and the shaft and the equivalent impedance between the housing and the shaft Voltage-current constraint relationship; S2: Construct operating condition characteristic variables based on the electrical quantities, mechanical quantities, environmental and service status quantities and inverter control parameters, and establish a dynamic parasitic capacitance calculation model; wherein, the dynamic parasitic capacitance calculation model shall at least incorporate the dynamic thickness of the bearing lubricating oil film, the conductor skin effect corresponding to the PWM carrier frequency and its harmonic components, and the air gap modulation amount caused by the housing vibration; S3. Based on the prior constraints of electromagnetic field finite element method, establish the objective function of LM iteration, and determine the model parameters of the dynamic parasitic capacitance calculation model through LM iteration, so as to obtain the nonlinear mapping relationship between dynamic parasitic capacitance and the operating condition characteristic variables. S4: Determine the common-mode voltage based on the rotor-side three-phase voltage or the inverter output three-phase voltage, and determine the common-mode voltage change rate based on differential or multi-scale sampling analysis of adjacent sampling points; and calculate the dynamic parasitic capacitance in real time based on the nonlinear mapping relationship, and predict the shaft voltage and shaft current based on the dynamic parasitic capacitance, common-mode voltage and common-mode voltage change rate, and construct a shaft current risk index; S5: Based on the shaft current risk index, perform at least one of the following control or diagnostic operations: inverter PWM parameter correction, bearing insulation performance warning, and active shaft voltage compensation. Use the shaft voltage, shaft current, and / or compensation response results after performing the control or diagnostic operation as the input for the next sampling period to update the shaft current risk index.

[0009] Furthermore, in step S2, the parasitic capacitance between the housing and the rotating shaft... The channel is determined based on the series medium channels formed by the air gap layer and the lubricating oil film layer, and is calculated according to the following relationship:

[0010] in, The radius of the pivot is 1. For effective coupling length, For air gaps, For the dynamic thickness of the lubricating oil film, The complex permittivity of air is 1. is the complex dielectric constant of the lubricating oil.

[0011] Specifically, the dynamic thickness of the lubricating oil film is determined based on the dynamic viscosity of the lubricating oil, the shaft entrainment speed, the equivalent elastic modulus, the dynamic load of the bearing, the equivalent radius of curvature, and the surface roughness correction term, and is calculated according to the following relationship:

[0012]

[0013] in, This is the correction factor for the spatial distribution of the oil film. , , , For lubrication model parameters, The equivalent radius of curvature, For the dynamic viscosity of lubricating oil, The speed of the rotating shaft is the suction speed. For the equivalent elastic modulus, The pressure viscosity coefficient, For bearing dynamic load, This is a correction item for the bearing contact type. This is a surface roughness correction term.

[0014] Specifically, the conductor skin effect corresponding to the PWM carrier frequency and its harmonic components is calculated by introducing the dynamic parasitic capacitance through the skin depth and surface impedance corresponding to each order harmonic component of the PWM carrier frequency; for the PWM carrier frequency... The first harmonic component has the following angular frequency:

[0015] The skin depths in the shaft and housing materials are as follows:

[0016] The high-frequency surface impedance is determined based on the skin depth, and the high-frequency surface impedance is coupled with the dielectric capacitance channel to obtain the frequency-dependent equivalent parasitic capacitance.

[0017] Preferably, the air gap modulation caused by the casing vibration is introduced through a dynamic air gap. The calculation process for the dynamic air gap is determined according to the following relationship:

[0018] in, For static air gaps, , This represents the rotor eccentric displacement component. For the first The amplitude of the vibration modes of the casing. For the corresponding mode shape function, The vibration frequency, For phase; Furthermore, the coupling current between the housing and the shaft is corrected based on the rate of change of the dynamic capacitance, as shown in the following formula:

[0019] in, This is the voltage between the casing and the shaft.

[0020] Specifically, in step S3, the objective function of the LM iteration includes a data fitting term, a finite element prior constraint term, a finite element sensitivity constraint term, a physical boundary constraint term, and a parameter smoothing constraint term, and is established according to the following relationship:

[0021] in, For data fitting terms, These are the prior constraints for the finite element method. For finite element sensitivity constraints, For physical boundary constraints, For parameter smoothing constraint terms, , , , These are the corresponding weighting coefficients; When the data missing rate within the sampling window exceeds a preset missing rate threshold, or the sampling noise variance exceeds a preset noise threshold, the weight of the finite element prior constraint term is increased, and the data fitting weight corresponding to the abnormal sampling data is decreased.

[0022] Furthermore, the method also includes an online aging adaptive correction step: The long-term offset trend of parasitic capacitance is determined based on the current prediction residual within the stable operating window, and the parasitic capacitance is aging corrected according to the following relationship:

[0023] in, This refers to the parasitic capacitance after aging correction. The parasitic capacitance output by the physical constraint model. This is the multiplicative aging correction factor. This is the additive reference offset. express , or ; When the rate of change of rotational speed, the rate of change of temperature, the rate of change of humidity, and the vibration amplitude all meet the stable operating conditions, and the current prediction residual continuously shifts in the same direction within multiple sampling windows, the multiplicative aging correction coefficient and the additive reference offset are updated.

[0024] Further, in step S4, multi-scale sampling analysis is used to determine the common-mode voltage change rate and / or the shaft voltage change rate. The multi-scale sampling analysis includes: The short-scale voltage change rate is determined based on adjacent sampling points; The mesoscale voltage change rate is determined based on the first preset sampling interval; The long-scale voltage change rate is determined based on the second preset sampling interval; The transient intensity factor is determined based on the difference between the short-scale and long-scale voltage change rates; and Based on the transient intensity factor, weights are assigned to short-scale, medium-scale, and long-scale voltage change rates, and the voltage change rates at different scales are weighted and fused to obtain multi-scale voltage change rates.

[0025] Furthermore, the inverter PWM parameter correction includes: Based on the dynamic parasitic capacitance calculated in real time, the common-mode voltage determined by the electrical quantities, and the common-mode voltage change rate determined by differential or multi-scale sampling analysis of adjacent sampling points, the shaft current and shaft voltage are predicted. A shaft current risk index is constructed based on predicted shaft current, predicted shaft voltage, and the rate of change of dynamic parasitic capacitance. When the shaft current risk index exceeds the preset safety threshold, the PWM dead time is corrected within the allowable dead time range, and / or the PWM switching frequency is corrected within the allowable switching frequency range. The correction targets for the PWM dead time and / or PWM switching frequency include reducing the predicted shaft current, reducing the predicted shaft voltage, limiting torque ripple, and limiting output current harmonic distortion.

[0026] Specifically, the bearing insulation performance early warning includes: determining the reference housing-shaft parasitic capacitance under the same operating conditions based on the current temperature, humidity, rotational speed, dynamic thickness of the lubricating oil film, and housing vibration amplitude; determining the normalized capacitance deviation based on the deviation between the real-time housing-shaft parasitic capacitance and the reference housing-shaft parasitic capacitance under the same operating conditions; determining the capacitance fluctuation rate based on the change in housing-shaft parasitic capacitance within adjacent sampling windows; and determining the bearing insulation performance early warning level based on the normalized capacitance deviation, capacitance fluctuation rate, and current prediction residual. The active shaft voltage compensation includes: when the shaft voltage or shaft current risk exceeds the compensation trigger threshold, determining the equivalent admittance of the high-frequency coupling channel between the housing and the shaft based on the real-time housing-shaft parasitic capacitance and the PWM carrier frequency and its harmonic components, and adjusting at least one of the equivalent resistance, equivalent inductance, and equivalent capacitance of the compensation branch according to the equivalent admittance, so that the compensation branch forms impedance matching with the high-frequency coupling channel within the target frequency range to suppress the shaft voltage and bearing current.

[0027] The beneficial effects of this invention are as follows: This invention incorporates the dynamic thickness of the bearing lubricating oil film, the conductor skin effect corresponding to the PWM carrier frequency and its harmonic components, and the air gap modulation caused by casing vibration into the dynamic parasitic capacitance calculation process, making the parasitic capacitance results more consistent with the actual operating state of the doubly-fed induction generator (DFIG). Furthermore, this invention introduces electromagnetic field finite element prior constraints, sensitivity constraints, and physical boundary constraints during the LM iteration process, which can suppress parameter oscillations and improve the reliability of the dynamic parasitic capacitance identification results when data is missing, noise increases, or operating conditions change abruptly. This invention also compensates for capacitance reference deviations caused by factors such as insulation material aging, changes in lubricating oil dielectric properties, and bearing wear through an online aging adaptive correction mechanism, enabling the model to adapt to parameter drift after long-term wind turbine service. Additionally, it can predict shaft voltage and shaft current based on real-time dynamic parasitic capacitance and adjust the PWM dead time or switching frequency accordingly, thereby reducing the risk of high-frequency common-mode voltage coupling to form shaft current through parasitic capacitance. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating the steps of a closed-loop control method for parasitic capacitance of a doubly-fed wind turbine according to a specific embodiment of the present invention. Figure 2 The flowchart illustrates the algorithm of a closed-loop control method for parasitic capacitance of a doubly fed wind turbine according to a specific embodiment of the present invention. Detailed Implementation

[0029] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0030] like Figure 1 As shown in the figure, a closed-loop control method for parasitic capacitance of a doubly-fed wind turbine provided by a specific embodiment of the present invention includes the following steps: S1: Collect electrical, mechanical, environmental, and service status parameters, as well as inverter control parameters, of the doubly-fed induction generator (DFIG) during operation. The electrical parameters include at least two of the following: rotor-side voltage, rotor-side current, common-mode voltage, or three-phase output voltage used to calculate the common-mode voltage, leakage current, and shaft voltage. The mechanical parameters include at least two of the following: generator speed, casing vibration amplitude, bearing load, and rotor eccentricity. The environmental and service status parameters include the motor's internal temperature, humidity, and service life. The inverter control parameters include at least one of the following: PWM carrier frequency and switching status. Furthermore, construct an equivalent model of the DFIG's parasitic capacitance to determine the parasitic capacitance between the rotor windings and the casing. Parasitic capacitance between rotor windings and shaft Parasitic capacitance between the housing and the shaft and the equivalent impedance between the housing and the shaft Voltage-current constraint relationship; S2: Construct operating condition characteristic variables based on the electrical quantities, mechanical quantities, environmental and service status quantities and inverter control parameters, and establish a dynamic parasitic capacitance calculation model; wherein, the dynamic parasitic capacitance calculation model shall at least incorporate the dynamic thickness of the bearing lubricating oil film, the conductor skin effect corresponding to the PWM carrier frequency and its harmonic components, and the air gap modulation amount caused by the housing vibration; S3. Based on the prior constraints of electromagnetic field finite element method, establish the objective function of LM iteration, and determine the model parameters of the dynamic parasitic capacitance calculation model through LM iteration, so as to obtain the nonlinear mapping relationship between dynamic parasitic capacitance and the operating condition characteristic variables. S4: Determine the common-mode voltage based on the rotor-side three-phase voltage or the inverter output three-phase voltage, and determine the common-mode voltage change rate based on differential or multi-scale sampling analysis of adjacent sampling points; and calculate the dynamic parasitic capacitance in real time based on the nonlinear mapping relationship, and predict the shaft voltage and shaft current based on the dynamic parasitic capacitance, common-mode voltage and common-mode voltage change rate, and construct a shaft current risk index; S5: Based on the shaft current risk index, perform at least one of the following control or diagnostic operations: inverter PWM parameter correction, bearing insulation performance warning, and active shaft voltage compensation. Use the shaft voltage, shaft current, and / or compensation response results after performing the control or diagnostic operation as the input for the next sampling period to update the shaft current risk index.

[0031] In one implementation, the electrical quantities, mechanical quantities, environmental and service status quantities, and inverter control parameters are collected using a unified sampling clock and undergo time alignment, outlier removal, and normalization before being entered into the dynamic parasitic capacitance calculation model. For data with different sampling frequencies, interpolation resampling or sliding window statistics are used to convert them to the same calculation period, ensuring that the electrical quantities, mechanical quantities, environmental and service status quantities, and inverter control parameters within the same sampling period can collectively characterize the current operating status of the doubly-fed wind turbine.

[0032] Preferably, the operating condition characteristic variables include at least one combination of voltage amplitude, current amplitude, common-mode voltage change rate, rotational speed, vibration amplitude, bearing load, rotor eccentricity, motor internal temperature, humidity, service life, PWM carrier frequency, and switching state. The operating condition characteristic variables can be directly obtained from the acquired data or derived quantities calculated from the acquired data, such as rotational speed change rate, temperature change rate, humidity change rate, vibration energy, voltage spectral components, or common-mode voltage harmonic amplitude.

[0033] Furthermore, the equivalent model of parasitic capacitance of the doubly-fed wind turbine is used to characterize the high-frequency coupling path between the rotor winding, shaft, bearing, and housing. The parasitic capacitance between the rotor winding and housing, the parasitic capacitance between the rotor winding and shaft, the parasitic capacitance between the housing and shaft, and the equivalent impedance between the housing and shaft together constitute the basis for predicting shaft voltage and shaft current. After the controller performs PWM parameter correction, bearing insulation performance warning, or active shaft voltage compensation based on the prediction results, it feeds back the corrected shaft voltage, shaft current, or compensation response results to the next sampling period, thereby forming a closed-loop control process for dynamic parasitic capacitance identification and shaft current suppression.

[0034] Based on the above basic implementation method, in step S2, the parasitic capacitance between the housing and the rotating shaft... The channel is determined based on the series medium channels formed by the air gap layer and the lubricating oil film layer, and is calculated according to the following relationship:

[0035] in, The radius of the pivot is 1. For effective coupling length, For air gaps, For the dynamic thickness of the lubricating oil film, The complex permittivity of air is 1. The complex dielectric constant of the lubricating oil; Indicates time and angular frequency The frequency-dependent dynamic parasitic capacitance of the chassis-shaft is considered as the chassis-shaft parasitic capacitance. The dynamic calculated values ​​are used in the prediction of shaft voltage and shaft current.

[0036] Furthermore, the air gap layer and the lubricating oil film layer participate in electric field coupling as different dielectric layers between the housing and the shaft. When the thickness of the lubricating oil film or the air gap increases, the equivalent dielectric distance between the housing and the shaft increases, and the parasitic capacitance between the housing and the shaft tends to decrease; when the thickness of the lubricating oil film decreases, the air gap decreases, or the complex dielectric constant of the lubricating oil increases, the parasitic capacitance between the housing and the shaft tends to increase. Thus, the parasitic capacitance between the housing and the shaft can reflect the dynamic changes in the bearing lubrication state and the housing-shaft clearance state.

[0037] In one embodiment, the complex dielectric constant of the lubricating oil is determined based on the real part of the dielectric constant of the lubricating oil and the equivalent conductivity of the lubricating oil, and its expression is as follows:

[0038] in, Let be the real part of the dielectric constant of the lubricating oil. The equivalent conductivity of the lubricating oil. Angular frequency, The dielectric constant is the vacuum dielectric constant. The real part of the dielectric constant and the equivalent conductivity of the lubricating oil can be obtained from the lubricating oil material parameters, temperature and humidity calibration data, or historical operating data.

[0039] In one specific embodiment, the dynamic thickness of the lubricating oil film is determined based on the dynamic viscosity of the lubricating oil, the shaft entrainment speed, the equivalent elastic modulus, the bearing dynamic load, the equivalent radius of curvature, and the surface roughness correction term, and is calculated according to the following relationship:

[0040]

[0041] in, This is the correction factor for the spatial distribution of the oil film. , , , For lubrication model parameters, The equivalent radius of curvature, For the dynamic viscosity of lubricating oil, The speed of the rotating shaft is the suction speed. For the equivalent elastic modulus, The pressure viscosity coefficient, The bearing dynamic load is determined based on values ​​collected by a bearing load sensor, or calculated based on generator speed, casing vibration amplitude, and rotor eccentricity. This is a correction item for the bearing contact type. For surface roughness correction; the lubrication model parameters The bearing contact form correction term can be determined based on the bearing type, lubricant type, and test calibration data, or it can be obtained by fitting the results of elastohydrodynamic lubrication simulation; The surface roughness correction term is determined based on the bearing type, rolling element geometry parameters, and contact ellipticity; Determined based on the surface roughness parameters of the bearing raceway and rolling elements.

[0042] In this embodiment, the oil film spatial distribution correction coefficient This is used to characterize the non-uniform distribution of oil film thickness along the circumferential and axial directions within the bearing contact area. The oil film spatial distribution correction coefficient can be determined based on bearing structural parameters, rotor eccentricity, bearing load direction, housing vibration amplitude, and surface roughness parameters. In simplified calculations, the oil film spatial distribution correction coefficient can be taken as a preset calibration value or a distribution function obtained by fitting the finite element lubrication simulation results.

[0043] Furthermore, the shaft winding speed is determined based on the generator speed and bearing contact pair geometric parameters; the lubricating oil dynamic viscosity is determined based on the motor internal temperature and lubricating oil viscosity-temperature curve; and the bearing dynamic load is determined based on the values ​​collected by the bearing load sensor, or calculated based on the generator speed, rotor eccentricity, housing vibration amplitude, and rotor imbalance parameters. Therefore, under high temperature, low speed, heavy load, or enhanced vibration conditions, the change in the dynamic thickness of the lubricating oil film can be reflected in the calculation results of the housing-shaft parasitic capacitance.

[0044] In another specific embodiment, the conductor skin effect corresponding to the PWM carrier frequency and its harmonic components is introduced into the dynamic parasitic capacitance calculation process through the skin depth and surface impedance corresponding to each order harmonic component of the PWM carrier frequency; for the PWM carrier frequency... The first harmonic component has the following angular frequency:

[0045] The skin depths in the shaft and housing materials are as follows:

[0046] The high-frequency surface impedance is determined based on the skin depth, and then coupled with the high-frequency surface impedance to the dielectric capacitance channel to obtain the frequency-dependent equivalent parasitic capacitance. The high-frequency surface impedance can be determined based on the shaft radius, the equivalent radius of the housing, the effective coupling length, the material conductivity, and the skin depth. In one example, the high-frequency surface impedance can be expressed as:

[0047] in, Let be the equivalent radius of the casing.

[0048] In this embodiment, the conductor skin effect does not directly change the geometric capacitance between the housing and the shaft, but rather alters the current distribution path on the conductor surfaces of the shaft and housing under the influence of the PWM carrier frequency and its harmonic components, causing the equivalent parasitic capacitance obtained from the voltage-current relationship to exhibit frequency-dependent characteristics. Therefore, when calculating the dynamic parasitic capacitance, the skin depth and surface impedance are introduced as frequency-dependent correction quantities into the dielectric capacitance channel.

[0049] Specifically, the magnetic permeability and electrical conductivity of the shaft material and housing material are determined by a preset material parameter library, factory material parameters, or on-site calibration data. For the PWM carrier frequency... The first harmonic component is used to calculate the skin depth in the shaft and housing materials based on the corresponding angular frequency, and the high-frequency surface impedance is determined based on the skin depth. The high-frequency surface impedance, together with the dielectric capacitance channel formed by the air gap layer and the lubricating oil film layer, interacts to obtain the equivalent admittance at the corresponding frequency.

[0050] Furthermore, if the common-mode voltage contains multiple PWM harmonic components, the frequency weight is determined based on the amplitude or energy proportion of each harmonic component, and the equivalent parasitic capacitance at different frequencies is weighted and fused to obtain a comprehensive dynamic parasitic capacitance for shaft current prediction. Through this processing, the model can adapt to the shaft current prediction requirements under different PWM carrier frequencies, different switching states, and different common-mode voltage spectrum distributions.

[0051] In another specific embodiment, the air gap modulation caused by the housing vibration is introduced through a dynamic air gap. The calculation process for the dynamic air gap is determined according to the following relationship:

[0052] in, For static air gaps, , This represents the rotor eccentric displacement component. For the first The amplitude of the vibration modes of the casing. The corresponding mode shape function is determined based on casing modal testing, finite element modal analysis, or a preset mode shape parameter library. The vibration frequency, For phase; Furthermore, the coupling current between the housing and the shaft is corrected based on the rate of change of the dynamic capacitance, as shown in the following formula:

[0053] in, This is the voltage between the casing and the shaft.

[0054] Furthermore, the dynamic air gap is used to characterize the change in the housing-shaft clearance under the combined effects of static air gap, rotor eccentricity, and housing vibration modal displacement. The rotor eccentricity displacement component can be determined based on displacement sensors, bearing vibration signals, or rotor dynamics models; the housing vibration modal amplitude can be obtained by modal decomposition of vibration signals collected by housing vibration sensors, or it can be calculated based on a preset modal parameter library and the current vibration amplitude.

[0055] Specifically, when the housing vibration causes periodic changes in the air gap, the parasitic capacitance between the housing and the shaft modulates over time, and the dynamic capacitance change rate is used in the calculation of the coupling current between the housing and the shaft. Compared to calculating the coupling current using only a fixed capacitance, this method introduces... The study can reflect the effects of rapid changes in air gap, sudden changes in capacitance, and vibration modulation on shaft current, thereby improving the accuracy of predicting shaft current risk under conditions of strong vibration, rotor eccentricity, and bearing wear.

[0056] In another specific embodiment, in step S3, the LM iteration objective function includes a data fitting term, a finite element prior constraint term, a finite element sensitivity constraint term, a physical boundary constraint term, and a parameter smoothing constraint term, and is established according to the following relationship:

[0057] in, For data fitting terms, These are the prior constraints for the finite element method. For finite element sensitivity constraints, For physical boundary constraints, For parameter smoothing constraint terms, , , , These are the corresponding weighting coefficients; When the data missing rate within the sampling window exceeds a preset missing rate threshold, or the sampling noise variance exceeds a preset noise threshold, the weight of the finite element prior constraint term is increased, and the data fitting weight corresponding to abnormal sampling data is decreased. The finite element prior constraint term is determined based on the prior parasitic capacitance value output by the pre-established electromagnetic field finite element model. The finite element sensitivity constraint term is determined based on the sensitivity matrix of the prior parasitic capacitance value relative to the operating condition characteristic variable. The upper and lower limits of capacitance in the physical boundary constraint term can be determined based on the motor structure dimensions, dielectric parameters of the insulating material, finite element simulation results, and historical stable operating data. The parameter smoothing constraint term is used to limit the change in model parameters within adjacent sampling windows. The weighting coefficients... Preset calibration values ​​can be used, or adaptive adjustments can be made based on data missing rate, noise variance, and operating condition variation.

[0058] Furthermore, the data fitting term is determined based on the residual between the measured current and the predicted current. The predicted current is determined based on the dynamic parasitic capacitance, the common-mode voltage change rate, and the dynamic parasitic capacitance change rate. The finite element prior constraint term is used to limit the calculated result of the dynamic parasitic capacitance from deviating from the prior parasitic capacitance value output by the electromagnetic field finite element model. The finite element sensitivity constraint term is used to limit the variation trend of the dynamic parasitic capacitance relative to temperature, humidity, rotational speed, dynamic thickness of the lubricating oil film, PWM carrier frequency, and air gap modulation from deviating from the finite element sensitivity matrix. The physical boundary constraint term is used to limit the parasitic capacitance from falling within a preset physical reasonable range. The parameter smoothing constraint term is used to suppress abnormal abrupt changes in model parameters between adjacent sampling windows.

[0059] Specifically, during the LM iteration process, the predicted current and comprehensive residual are first calculated based on the current model parameters, and then the model parameters are updated based on the comprehensive residual. When the updated comprehensive objective function decreases, the LM damping factor is reduced to improve the convergence speed; when the updated comprehensive objective function increases, the LM damping factor is increased to enhance the iteration stability. When the data missing rate or noise variance within the sampling window exceeds the corresponding threshold, the weights of the finite element prior constraint terms and physical boundary constraint terms are increased, so that the model remains within the search range that conforms to the laws of electromagnetic fields even when the data is incomplete or the noise is strong.

[0060] In another specific embodiment, the method further includes an online aging adaptive correction step: The long-term offset trend of parasitic capacitance is determined based on the current prediction residual within the stable operating window, and the parasitic capacitance is aging corrected according to the following relationship:

[0061] in, This refers to the parasitic capacitance after aging correction. The parasitic capacitance output by the physical constraint model. This is the multiplicative aging correction factor. This is the additive reference offset. express , or ; When the rate of change of rotational speed, the rate of change of temperature, the rate of change of humidity, and the vibration amplitude all meet the stable operating conditions, and the current prediction residual continuously shifts in the same direction within multiple sampling windows, the multiplicative aging correction coefficient and the additive reference offset are updated; the rate of change of rotational speed, the rate of change of temperature, and the rate of change of humidity are determined based on the generator speed, the internal temperature of the motor, and the humidity within adjacent sampling windows, respectively; the current prediction residual is the difference between the measured casing-shaft coupling current and the coupling current predicted based on the current dynamic parasitic capacitance, or the difference between the measured shaft current and the predicted shaft current.

[0062] Furthermore, the current prediction residual is the difference between the measured casing-shaft coupling current and the coupling current predicted based on the current dynamic parasitic capacitance, or the difference between the measured shaft current and the predicted shaft current. To avoid misjudging short-term operating condition fluctuations as aging offsets, the online aging adaptive correction is only performed within a stable operating window. The stable operating window is a sampling window where the rate of change of speed, temperature, humidity, and vibration amplitude all meet the corresponding stability thresholds; the multiple continuous stable operating windows can be three or more continuous sampling windows. The stability thresholds can be determined based on the fan's rated speed, normal temperature rise rate, normal humidity change rate, and allowable vibration amplitude. When short-term operating condition changes, start-up / shutdown processes, or fault impacts cause sudden changes in the residual, aging correction is not performed to avoid misjudging transient disturbances as aging offsets.

[0063] Furthermore, when the current prediction residual maintains a consistent directional shift across multiple consecutive stable operating windows, it is determined that the dynamic parasitic capacitance model has a long-term reference deviation, and the multiplicative aging correction coefficient and additive reference offset are updated. The multiplicative aging correction coefficient is used to compensate for proportional capacitance deviations caused by insulation material aging, changes in lubricating oil dielectric properties, and bearing wear; the additive reference offset is used to compensate for overall deviations caused by changes in casing grounding status, installation status, or sensor zero-point drift after long-term operation. In this way, the model can maintain the stability and reliability of the dynamic parasitic capacitance identification results during long-term wind turbine operation.

[0064] In one specific implementation, step S4 involves using multi-scale sampling analysis to determine the common-mode voltage change rate and / or the shaft voltage change rate. The multi-scale sampling analysis includes: The short-scale voltage change rate is determined based on adjacent sampling points; the medium-scale voltage change rate is determined based on a first preset sampling interval; the long-scale voltage change rate is determined based on a second preset sampling interval; a transient intensity factor is determined based on the difference between the short-scale and long-scale voltage change rates; and weights are assigned to the short-scale, medium-scale, and long-scale voltage change rates based on the transient intensity factor, and the voltage change rates at different scales are weighted and fused to obtain a multi-scale voltage change rate; wherein, when the transient intensity factor increases, the weight of the short-scale voltage change rate is increased; when the transient intensity factor decreases, the weight of the long-scale voltage change rate is increased; the first preset sampling interval is greater than the interval between adjacent sampling points, and the second preset sampling interval is greater than the first preset sampling interval; in one example, the transient intensity factor is determined according to the following relationship:

[0065] in, For short-scale voltage change rate, For long-scale voltage change rate, To prevent constants with zero denominators, the weights for different scales can be determined by looking up the transient intensity factor in a table, or by using a normalization function, ensuring that the sum of the weights for each scale is 1.

[0066] Specifically, the short-scale voltage change rate is determined based on the voltage difference between adjacent sampling points and is used to capture PWM switching actions, common-mode voltage spikes, and shaft voltage abrupt changes; the medium-scale voltage change rate is determined based on the voltage difference within a first preset sampling interval and is used to characterize transient changes during acceleration, load changes, or compensation control processes; the long-scale voltage change rate is determined based on the voltage difference within a second preset sampling interval and is used to reduce the impact of sampling noise on the voltage change rate calculation results during steady-state operation.

[0067] Furthermore, the transient intensity factor is determined based on the difference between the short-scale voltage change rate and the long-scale voltage change rate. When the transient intensity factor increases, it indicates the presence of strong high-frequency disturbances or transient abrupt changes, thus increasing the weight of the short-scale voltage change rate; conversely, when the transient intensity factor decreases, it indicates a relatively stable operating state, thus increasing the weight of the long-scale voltage change rate. Therefore, multi-scale sampling analysis can balance steady-state noise immunity and transient response capability.

[0068] In one specific implementation, the inverter PWM parameter correction includes: Based on the dynamic parasitic capacitance calculated in real time, the common-mode voltage determined by the electrical quantities, and the common-mode voltage change rate determined by differential or multi-scale sampling analysis of adjacent sampling points, the shaft current and shaft voltage are predicted; a shaft current risk index is constructed based on the predicted shaft current, predicted shaft voltage, and dynamic parasitic capacitance change rate; when the shaft current risk index exceeds a preset safety threshold, the PWM dead time is corrected within the allowable dead time range, and / or the PWM switching frequency is corrected within the allowable switching frequency range; wherein, the correction objectives of the PWM dead time and / or PWM switching frequency include reducing the predicted shaft current, reducing the predicted shaft voltage, limiting torque ripple, and limiting output current harmonic distortion.

[0069] In this embodiment, the shaft current risk index is determined jointly based on the predicted shaft current, predicted shaft voltage, and dynamic parasitic capacitance change rate. The closer the predicted shaft current is to or exceeds the shaft current safety threshold, the closer the predicted shaft voltage is to or exceeds the shaft voltage safety threshold, or the greater the dynamic parasitic capacitance change rate, the higher the shaft current risk index. When the shaft current risk index exceeds a preset safety threshold, the controller performs PWM parameter correction to reduce the risk of common-mode voltage coupling to the shaft and bearing via parasitic capacitance. In one example, within the allowable dead time and switching frequency ranges, the controller calculates the predicted shaft current, predicted shaft voltage, torque ripple, and output current harmonic distortion corresponding to candidate PWM parameters, and selects the candidate PWM parameter with the lowest comprehensive evaluation value as the corrected PWM parameter. The comprehensive evaluation value is obtained by weighting the predicted shaft current, predicted shaft voltage, torque ripple, and harmonic distortion terms.

[0070] Specifically, the PWM dead time correction includes increasing or decreasing the dead time within the allowable dead time range to reduce the combined evaluation values ​​of the common-mode voltage change rate, predicted shaft voltage, and predicted shaft current. The PWM switching frequency correction includes adjusting the PWM carrier frequency within the allowable switching frequency range to ensure that the PWM carrier frequency and its harmonic components avoid the shaft current sensitive frequency band. The PWM parameter correction also simultaneously constrains torque ripple and output current harmonic distortion to avoid the shaft current suppression process adversely affecting the grid connection performance and electromagnetic torque stability of the doubly-fed wind turbine.

[0071] In one specific implementation, the bearing insulation performance early warning includes: determining the reference housing-shaft parasitic capacitance under the same operating conditions based on the current temperature, humidity, rotational speed, dynamic thickness of the lubricating oil film, and housing vibration amplitude; determining the normalized capacitance deviation based on the deviation between the real-time housing-shaft parasitic capacitance and the reference housing-shaft parasitic capacitance under the same operating conditions; determining the capacitance fluctuation rate based on the change in housing-shaft parasitic capacitance within adjacent sampling windows; and determining the bearing insulation performance early warning level based on the normalized capacitance deviation, capacitance fluctuation rate, and current prediction residual. The active shaft voltage compensation includes: when the shaft voltage or shaft current risk exceeds the compensation trigger threshold, determining the equivalent admittance of the high-frequency coupling channel between the housing and the shaft based on the real-time housing-shaft parasitic capacitance and the PWM carrier frequency and its harmonic components, and adjusting at least one of the equivalent resistance, equivalent inductance, and equivalent capacitance of the compensation branch according to the equivalent admittance, so that the compensation branch forms impedance matching with the high-frequency coupling channel within the target frequency range to suppress the shaft voltage and bearing current.

[0072] In this embodiment, the reference chassis-shaft parasitic capacitance under the same operating conditions is determined based on historical stable samples, finite element prior samples, or calibration samples corresponding to temperature, humidity, rotational speed, dynamic thickness of lubricating oil film, and chassis vibration amplitude. This capacitance is used to characterize the normal level of the chassis-shaft parasitic capacitance under the current operating conditions. Comparing the real-time chassis-shaft parasitic capacitance with the reference chassis-shaft parasitic capacitance under the same operating conditions eliminates the influence of normal temperature, humidity, rotational speed, and vibration changes on the capacitance, thereby improving the accuracy of insulation performance early warning. The reference chassis-shaft parasitic capacitance under the same operating conditions can be determined based on historical stable operating samples, finite element prior samples, or factory calibration samples. When the current operating conditions are not completely consistent with the sample operating conditions, the corresponding reference chassis-shaft parasitic capacitance under the same operating conditions can be obtained through interpolation, regression models, or table lookup methods.

[0073] When at least two of the normalized capacitance deviation, capacitance volatility, and current prediction residual exceed the corresponding preset threshold, the bearing insulation performance warning level is output; when the normalized capacitance deviation, capacitance volatility, current prediction residual, and predicted shaft current all exceed the corresponding high-risk threshold, the bearing insulation breakdown risk warning is output.

[0074] The compensation branch includes at least one of an adjustable resistor branch, an adjustable inductor branch, an adjustable capacitor branch, or an active compensation branch. The equivalent admittance of the chassis-shaft high-frequency coupling channel can be determined based on the real-time chassis-shaft parasitic capacitance, chassis-shaft equivalent impedance, and PWM carrier frequency and its harmonic components.

[0075] Specifically, when at least two of the normalized capacitance deviation, capacitance volatility, and current prediction residual exceed their corresponding preset thresholds, a bearing insulation performance warning level is output; when the normalized capacitance deviation, capacitance volatility, current prediction residual, and predicted shaft current all exceed their corresponding high-risk thresholds, a bearing insulation breakdown risk warning is output. The bearing insulation performance warning levels include at least two of the following: slight degradation level, abnormal fluctuation level, and high-risk level.

[0076] In one specific implementation, such as Figure 2 As shown, taking a doubly-fed induction wind turbine with a rated power of 2MW as an example, the doubly-fed induction wind turbine includes a stator, rotor windings, shaft, bearings, housing, rotor-side PWM inverter, shaft voltage acquisition unit, leakage current acquisition unit, vibration acquisition unit, temperature and humidity acquisition unit, and controller. The controller is used to execute the doubly-fed wind turbine parasitic capacitance closed-loop control method of this embodiment.

[0077] In this embodiment, the controller collects the operating data of the doubly-fed induction generator (DFIG) with a sampling period of 1ms. The electrical quantities include rotor-side three-phase voltage, rotor-side three-phase current, leakage current, and shaft voltage; the mechanical quantities include generator speed, casing vibration amplitude, bearing load, and rotor eccentricity; the environmental and service status quantities include motor internal temperature, humidity, and service life; and the inverter control parameters include PWM carrier frequency, switching status, and DC bus voltage. For example, within a certain sampling window, the generator speed is 1520 r / min, the motor internal temperature is 72℃, the humidity is 68%RH, the casing vibration amplitude is 0.12 mm, the bearing load is 18 kN, the rotor eccentricity is 0.035 mm, the PWM carrier frequency is 2.5 kHz, the DC bus voltage is 1050 V, and the service life is 6 years.

[0078] The controller calculates the common-mode voltage based on the rotor-side three-phase voltage or the inverter output three-phase voltage:

[0079] The common-mode voltage change rate is determined based on differential or multi-scale sampling analysis of adjacent sampling points. Specifically, the controller calculates the short-scale, medium-scale, and long-scale voltage change rates respectively, and determines the transient intensity factor based on the difference between the short-scale and long-scale voltage change rates. When a common-mode voltage spike caused by PWM switching is detected, the weight of the short-scale voltage change rate is increased; when the fan is in a constant-speed stable operation state, the weight of the long-scale voltage change rate is increased to reduce the impact of sampling noise on shaft current prediction.

[0080] The controller constructs an equivalent model of the parasitic capacitance of the doubly-fed wind turbine, denoting the parasitic capacitance between the rotor winding and the casing as... The parasitic capacitance between the rotor winding and the shaft is denoted as... The parasitic capacitance between the housing and the shaft is denoted as... And the equivalent impedance between the housing and the shaft is denoted as .in, As key parameters in the prediction of shaft voltage and shaft current, they are determined based on the series medium channels formed by the air gap layer and the lubricating oil film layer.

[0081] In this embodiment, the controller first determines the dynamic thickness of the bearing lubricating oil film based on temperature, rotational speed, and bearing load. The dynamic viscosity of the lubricating oil is obtained from the current temperature and the lubricating oil viscosity-temperature curve. The shaft entrainment speed is obtained from the generator speed and the geometric parameters of the bearing contact pair. The dynamic load of the bearing is corrected based on the values ​​collected by the bearing load sensor and the vibration amplitude of the housing. The controller calculates the equivalent lubricating oil film thickness within the current sampling window to be 0.85 μm, and determines the spatial distribution correction coefficient of the oil film based on the rotor eccentricity and the bearing load direction, thereby obtaining the spatially distributed lubricating oil film thickness. .

[0082] Simultaneously, the controller calculates the skin depth in the shaft and housing materials based on the PWM carrier frequency and its harmonic components. For example, for the first-order PWM carrier frequency component, the angular frequency is:

[0083] For the second and third harmonic components, the angular frequencies are respectively:

[0084] The controller calculates the skin depth at different frequencies based on the conductivity and permeability of the shaft and housing materials, and further determines the corresponding high-frequency surface impedance. The high-frequency surface impedance, together with the dielectric capacitance channel formed by the air gap layer and the lubricating oil film layer, yields the frequency-dependent equivalent housing-shaft parasitic capacitance.

[0085] Specifically, the controller determines the dynamic air gap based on the casing vibration signal and the rotor eccentricity. For example, the static air gap is 1.2 mm, the current rotor eccentricity components in the horizontal and vertical directions are 0.025 mm and 0.024 mm, respectively, the first-order vibration mode frequency of the casing is 45 Hz, and the vibration mode amplitude is 0.12 mm. The controller calculates the dynamic air gap based on the static air gap, the rotor eccentricity displacement components, and the casing vibration mode displacement. And substitute this dynamic air gap into In the computational model.

[0086] Based on the above parameters, the controller establishes a dynamic parasitic capacitance calculation model and calls the prior parasitic capacitance values ​​and sensitivity matrix output from the pre-established electromagnetic field finite element model. The finite element prior samples include parasitic capacitance calibration values ​​under different temperatures, humidity levels, rotational speeds, oil film thicknesses, air gaps, and PWM frequencies. The controller incorporates finite element prior constraints, finite element sensitivity constraints, physical boundary constraints, and parameter smoothing constraints into the LM iteration objective function, and determines the model parameters of the dynamic parasitic capacitance calculation model through LM iteration.

[0087] In this embodiment, when the sensor data within the sampling window is complete and the noise variance is lower than the preset noise threshold, the controller increases the weight of the data fitting term, so that the model is corrected more based on the current measured data. When the temperature and humidity sensor experiences short-term packet loss or the leakage current sampling noise variance increases, the controller increases the weight of the finite element prior constraint term and decreases the data fitting weight corresponding to the abnormal sampling data, so that the model output remains within the physically reasonable range determined by the finite element prior and avoids abnormal oscillations in the iteration results.

[0088] Through the above calculations, the controller obtains the dynamic parasitic capacitance within the current sampling window, for example:

[0089] The above values ​​are merely illustrative and do not constitute a limitation on the scope of protection of this application.

[0090] Subsequently, the controller predicts the shaft voltage and shaft current based on real-time dynamic parasitic capacitance, common-mode voltage, and common-mode voltage change rate. For example, within the current sampling window, the predicted peak shaft voltage is 18V, the predicted peak shaft current is 0.92A, and the dynamic parasitic capacitance change rate is higher than the baseline change rate under the same operating condition. The controller constructs a shaft current risk index based on the predicted shaft voltage, predicted shaft current, and dynamic parasitic capacitance change rate. When the shaft current risk index exceeds a preset safety threshold, the controller triggers PWM parameter correction.

[0091] During PWM parameter correction, the controller adjusts the PWM dead time within the allowable dead time range and the PWM carrier frequency within the allowable switching frequency range. For example, the controller adjusts the PWM carrier frequency from 2.5kHz to 2.2kHz to ensure that the main harmonic components of the PWM avoid the current shaft current sensitive frequency band; simultaneously, it corrects the PWM dead time from 3.0μs to 2.6μs to reduce the common-mode voltage change rate. During PWM parameter correction, the controller simultaneously monitors output current harmonic distortion and torque ripple to ensure that shaft current suppression does not lead to a significant decrease in grid-connected current quality and electromagnetic torque stability.

[0092] In another sampling window, if the controller detects real-time... Compared to the benchmark under the same operating conditions If the normalization deviation continuously exceeds a preset first threshold and the capacitance fluctuation rate exceeds a preset second threshold, a bearing insulation performance warning will be output. For example, the current reference housing-shaft parasitic capacitance under the same operating conditions is 0.36nF, while the real-time... The value is 0.44 nF, with a normalized deviation of approximately 22.2%. Simultaneously, the capacitance fluctuation rate increases across multiple consecutive sampling windows, and the current prediction residual between the measured and predicted coupling currents continuously shifts in the same direction. Based on this, the controller determines that there is a risk of abnormal fluctuations in the bearing insulation performance and outputs a secondary warning message. If the subsequent predicted shaft current further exceeds the high-risk threshold, a bearing insulation breakdown risk warning is output.

[0093] Furthermore, when the shaft voltage or shaft current risk exceeds the compensation trigger threshold, the controller performs active shaft voltage compensation. The compensation branch includes at least one of an adjustable resistor branch, an adjustable capacitor branch, and an active compensation branch. The controller adjusts the compensation based on real-time... The equivalent admittance of the high-frequency coupling channel between the chassis and the shaft is determined by the PWM carrier frequency and its harmonic components. Based on this, the equivalent impedance of the compensation branch is adjusted to achieve impedance matching between the compensation branch and the high-frequency coupling channel between the chassis and the shaft within the main harmonic frequency band of the PWM. For example, when... When the shaft voltage increases and causes the peak value of the shaft voltage to rise, the controller reduces the equivalent impedance of the compensation branch in the target high-frequency band to enhance the high-frequency discharge capability; when the shaft voltage returns to a safe range, the controller gradually increases the equivalent impedance of the compensation branch to reduce compensation loss.

[0094] After performing PWM parameter correction and / or active shaft voltage compensation, the controller continues to collect shaft voltage, shaft current, and compensation response results, using them as input for the next sampling cycle. If the shaft current risk index drops to a safe range within the next sampling cycle, the current control parameters are maintained or gradually restored to economic operating parameters according to preset recovery rules; if the shaft current risk index remains above the safe threshold, the PWM dead time, PWM switching frequency, or compensation branch impedance are adjusted until the shaft current risk index drops to a safe range.

[0095] In this embodiment, the calculation results of dynamic parasitic capacitance are no longer merely used as offline parameters, but are directly involved in shaft voltage prediction, shaft current suppression, bearing insulation performance early warning, and active shaft voltage compensation control. Therefore, this embodiment can improve the bearing insulation status identification capability and active shaft current suppression capability of doubly-fed induction generator (DFIG) fans under complex operating conditions such as high temperature, high humidity, strong vibration, rotor eccentricity, PWM high-frequency impact, and long-term aging.

[0096] In summary, this embodiment has at least the following technical effects: The dynamic parasitic capacitance calculation model incorporates the dynamic thickness of the bearing lubricating oil film, the conductor skin effect corresponding to the PWM carrier frequency and its harmonic components, and the air gap modulation caused by the housing vibration. This enables the housing-shaft parasitic capacitance to reflect the coupling effect of lubrication state, high-frequency electromagnetic boundary and mechanical gap changes on the capacitance, thus improving the accuracy of the dynamic parasitic capacitance calculation results under complex working conditions. By introducing finite element prior constraints, finite element sensitivity constraints, physical boundary constraints, and parameter smoothing constraints into the LM iteration objective function, the model can still remain within the search range that conforms to the laws of electromagnetic fields even when sampling data is missing or noise increases, thus reducing the probability of parameter oscillations and non-physical solutions. The online aging adaptive correction mechanism compensates for the long-term shift in parasitic capacitance caused by aging of insulation materials, changes in the dielectric properties of lubricating oil, bearing wear, and changes in grounding status, thereby improving the model's adaptability during the long-term service of the wind turbine. Determining the common-mode voltage change rate or shaft voltage change rate through multi-scale sampling analysis can improve noise immunity accuracy during steady-state operation and enhance the ability to capture sudden changes under transient conditions such as PWM switching action, speed increase and load change. Environmental and service status parameters, including internal motor temperature, humidity, and service life, are used to calculate dynamic parasitic capacitance. The results are then used for PWM parameter correction, bearing insulation performance early warning, and active shaft voltage compensation. This allows the parasitic capacitance identification results to participate in the shaft current suppression closed loop, improving the bearing insulation reliability and operational safety of the doubly fed wind turbine.

[0097] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A closed-loop control method for the parasitic capacitance of a doubly-fed wind turbine, characterized in that, Includes the following steps: S1: Collect electrical, mechanical, environmental, and service status parameters of the doubly-fed induction generator (DFIG) during operation, as well as inverter control parameters; and construct an equivalent model of the parasitic capacitance of the DFIG to determine the parasitic capacitance between the rotor windings and the casing. Parasitic capacitance between rotor windings and shaft Parasitic capacitance between the housing and the shaft and the equivalent impedance between the housing and the shaft Voltage-current constraint relationship; S2: Construct operating condition characteristic variables based on the electrical quantities, mechanical quantities, environmental and service status quantities and inverter control parameters, and establish a dynamic parasitic capacitance calculation model; wherein, the dynamic parasitic capacitance calculation model shall at least incorporate the dynamic thickness of the bearing lubricating oil film, the conductor skin effect corresponding to the PWM carrier frequency and its harmonic components, and the air gap modulation amount caused by the housing vibration; S3. Based on the prior constraints of electromagnetic field finite element method, establish the objective function of LM iteration, and determine the model parameters of the dynamic parasitic capacitance calculation model through LM iteration, so as to obtain the nonlinear mapping relationship between dynamic parasitic capacitance and the operating condition characteristic variables. S4: Determine the common-mode voltage based on the rotor-side three-phase voltage or the inverter output three-phase voltage, and determine the common-mode voltage change rate based on differential or multi-scale sampling analysis of adjacent sampling points; and calculate the dynamic parasitic capacitance in real time based on the nonlinear mapping relationship, and predict the shaft voltage and shaft current based on the dynamic parasitic capacitance, common-mode voltage and common-mode voltage change rate, and construct a shaft current risk index; S5: Based on the shaft current risk index, perform at least one of the following control or diagnostic operations: inverter PWM parameter correction, bearing insulation performance warning, and active shaft voltage compensation. Use the shaft voltage, shaft current, and / or compensation response results after performing the control or diagnostic operation as the input for the next sampling period to update the shaft current risk index.

2. The closed-loop control method for parasitic capacitance of a doubly-fed wind turbine according to claim 1, characterized in that, In step S2, the parasitic capacitance between the housing and the rotating shaft The channel is determined based on the series medium channels formed by the air gap layer and the lubricating oil film layer, and is calculated according to the following relationship: in, The radius of the pivot is 1. For effective coupling length, For air gaps, For the dynamic thickness of the lubricating oil film, The complex permittivity of air is 1. is the complex dielectric constant of the lubricating oil.

3. The closed-loop control method for parasitic capacitance of a doubly-fed wind turbine according to claim 2, characterized in that, The dynamic thickness of the lubricating oil film is determined based on the dynamic viscosity of the lubricating oil, the shaft entrainment speed, the equivalent elastic modulus, the dynamic load of the bearing, the equivalent radius of curvature, and the surface roughness correction term, and is calculated according to the following relationship: in, This is the correction factor for the spatial distribution of the oil film. , , , For lubrication model parameters, The equivalent radius of curvature, For the dynamic viscosity of lubricating oil, The speed of the rotating shaft is the suction speed. For the equivalent elastic modulus, The pressure viscosity coefficient, For bearing dynamic load, This is a correction item for the bearing contact type. This is a surface roughness correction term.

4. The closed-loop control method for parasitic capacitance of a doubly-fed wind turbine according to claim 1, characterized in that, The conductor skin effect corresponding to the PWM carrier frequency and its harmonic components is introduced into the dynamic parasitic capacitance calculation process through the skin depth and surface impedance corresponding to each order harmonic component of the PWM carrier frequency; for the PWM carrier frequency... The first harmonic component has the following angular frequency: The skin depths in the shaft and housing materials are as follows: The high-frequency surface impedance is determined based on the skin depth, and the high-frequency surface impedance is coupled with the dielectric capacitance channel to obtain the frequency-dependent equivalent parasitic capacitance.

5. The closed-loop control method for parasitic capacitance of a doubly-fed wind turbine according to claim 1, characterized in that, The air gap modulation caused by the casing vibration is introduced through a dynamic air gap. The calculation process for the dynamic air gap is determined according to the following relationship: in, For static air gaps, , This represents the rotor eccentric displacement component. For the first The amplitude of the vibration modes of the casing. For the corresponding mode shape function, The vibration frequency, For phase; Furthermore, the coupling current between the housing and the shaft is corrected based on the rate of change of the dynamic capacitance, as shown in the following formula: in, This is the voltage between the casing and the shaft.

6. The closed-loop control method for parasitic capacitance of a doubly-fed wind turbine according to claim 1, characterized in that, In step S3, the objective function of the LM iteration includes data fitting terms, finite element prior constraint terms, finite element sensitivity constraint terms, physical boundary constraint terms, and parameter smoothing constraint terms, and is established according to the following relationship: in, For data fitting terms, These are the prior constraints for the finite element method. For finite element sensitivity constraints, For physical boundary constraints, For parameter smoothing constraint terms, , , , These are the corresponding weighting coefficients; When the data missing rate within the sampling window exceeds a preset missing rate threshold, or the sampling noise variance exceeds a preset noise threshold, the weight of the finite element prior constraint term is increased, and the data fitting weight corresponding to the abnormal sampling data is decreased.

7. The closed-loop control method for parasitic capacitance of a doubly-fed wind turbine according to claim 1, characterized in that, The method also includes an online aging adaptive correction step: The long-term offset trend of parasitic capacitance is determined based on the current prediction residual within the stable operating window, and the parasitic capacitance is aging corrected according to the following relationship: in, This refers to the parasitic capacitance after aging correction. The parasitic capacitance output by the physical constraint model. This is the multiplicative aging correction factor. This is the additive reference offset. express , or ; When the rate of change of rotational speed, the rate of change of temperature, the rate of change of humidity, and the vibration amplitude all meet the stable operating conditions, and the current prediction residual continuously shifts in the same direction within multiple sampling windows, the multiplicative aging correction coefficient and the additive reference offset are updated.

8. The closed-loop control method for parasitic capacitance of a doubly-fed wind turbine according to claim 1, characterized in that, In step S4, multi-scale sampling analysis is used to determine the common-mode voltage change rate and / or the shaft voltage change rate. The multi-scale sampling analysis includes: The short-scale voltage change rate is determined based on adjacent sampling points; The mesoscale voltage change rate is determined based on the first preset sampling interval; The long-scale voltage change rate is determined based on the second preset sampling interval; The transient intensity factor is determined based on the difference between the short-scale and long-scale voltage change rates; and Based on the transient intensity factor, weights are assigned to short-scale, medium-scale, and long-scale voltage change rates, and the voltage change rates at different scales are weighted and fused to obtain multi-scale voltage change rates.

9. The closed-loop control method for parasitic capacitance of a doubly-fed wind turbine according to claim 1, characterized in that, The inverter PWM parameter correction includes: Based on the dynamic parasitic capacitance calculated in real time, the common-mode voltage determined by the electrical quantities, and the common-mode voltage change rate determined by differential or multi-scale sampling analysis of adjacent sampling points, the shaft current and shaft voltage are predicted. A shaft current risk index is constructed based on predicted shaft current, predicted shaft voltage, and the rate of change of dynamic parasitic capacitance; and When the shaft current risk index exceeds the preset safety threshold, the PWM dead time is corrected within the allowable dead time range, and / or the PWM switching frequency is corrected within the allowable switching frequency range. The correction targets for the PWM dead time and / or PWM switching frequency include reducing the predicted shaft current, reducing the predicted shaft voltage, limiting torque ripple, and limiting output current harmonic distortion.

10. The closed-loop control method for parasitic capacitance of a doubly-fed wind turbine according to any one of claims 1 to 9, characterized in that, The bearing insulation performance early warning includes: determining the reference housing-shaft parasitic capacitance under the same operating conditions based on the current temperature, humidity, rotational speed, dynamic thickness of the lubricating oil film, and housing vibration amplitude; determining the normalized capacitance deviation based on the deviation between the real-time housing-shaft parasitic capacitance and the reference housing-shaft parasitic capacitance under the same operating conditions; determining the capacitance fluctuation rate based on the change in housing-shaft parasitic capacitance within adjacent sampling windows; and determining the bearing insulation performance early warning level based on the normalized capacitance deviation, capacitance fluctuation rate, and current prediction residual. The active shaft voltage compensation includes: when the shaft voltage or shaft current risk exceeds the compensation trigger threshold, determining the equivalent admittance of the high-frequency coupling channel between the housing and the shaft based on the real-time housing-shaft parasitic capacitance and the PWM carrier frequency and its harmonic components, and adjusting at least one of the equivalent resistance, equivalent inductance, and equivalent capacitance of the compensation branch according to the equivalent admittance, so that the compensation branch forms impedance matching with the high-frequency coupling channel within the target frequency range to suppress the shaft voltage and bearing current.

Images