Wind power converter open-circuit fault identification method

Abstract

The invention discloses a wind power converter open-circuit fault identification method, and the method comprises the steps: constructing a dynamic health state baseline in real time through employing q-axis and d-axis current instructions capable of accurately reflecting the dynamic intention of a system, and carrying out the feed-forward compensation of Park vector module length, thereby actively stripping the interference of the change of a normal working condition on fault features, and improving the fault identification precision. Therefore, the compensated fault features can be kept stable in a healthy state, obvious deviation is generated only when a real fault occurs, effective decoupling of the fault features and system transient response is achieved, the signal-to-noise ratio of diagnosis is obviously improved, and safe and stable operation of a wind power generation system is effectively guaranteed.

Description

Technical Field

[0001] This application relates to the field of fault identification, specifically to a method for identifying open-circuit faults in wind power converters. Background Technology

[0002] As an important clean energy source, the stable operation of wind power is crucial for the transformation of the energy structure. Wind power converters are the core equipment of wind power generation systems, responsible for the conversion and control of electrical energy. However, the failure rate of their internal power electronic components is relatively high, with open-circuit faults being a common and serious type of fault. If open-circuit faults are not identified and handled promptly and accurately, they may cause problems such as overcurrent and torque pulsation, and in severe cases, even damage the converter and the entire wind turbine, threatening the stable operation of the power grid. Therefore, developing an efficient and reliable open-circuit fault identification solution for wind power converters has become an urgent need in the industry.

[0003] Currently, methods for identifying open-circuit faults in wind power converters mainly rely on the analysis of electrical signals such as current and voltage. These methods typically achieve good results under steady-state conditions, but face significant challenges in practical applications. The fundamental problem lies in the fact that wind power systems operate in a continuously dynamic environment, with transient processes such as random fluctuations in wind speed and voltage drops in the grid being commonplace. During these dynamic processes, the current and voltage distortions caused by normal transient responses exhibit characteristics very similar to those of open-circuit faults, making them difficult to distinguish and resulting in the so-called feature aliasing problem. Most existing diagnostic methods employ judgment logic based on static health baselines and fixed thresholds; this rigid strategy cannot adapt to complex, variable operating conditions. This problem of fault feature distortion and aliasing under variable operating conditions severely restricts the reliability and accuracy of existing technologies in practical engineering applications.

[0004] Therefore, an optimized method for identifying open-circuit faults in wind power converters is needed. Summary of the Invention

[0005] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a method for identifying open-circuit faults in wind power converters. This method utilizes q-axis and d-axis current commands, which accurately reflect the dynamic intent of the system, to construct a dynamic health baseline in real time. This baseline is then used to feedforward compensate the Parker vector magnitude, thereby actively eliminating the interference of normal operating condition changes on fault characteristics. In this way, the compensated fault characteristics remain stable in a healthy state, only deviating significantly when a real fault occurs. This effectively decouples fault characteristics from the system's transient response, significantly improving the signal-to-noise ratio of the diagnosis and effectively ensuring the safe and stable operation of the wind power generation system.

[0006] According to one aspect of this application, a method for identifying open-circuit faults in a wind power converter is provided, comprising: Obtain the three-phase output current, q-axis current command, and d-axis current command; The three-phase output current, q-axis current command, and d-axis current command are preprocessed to obtain the filtered three-phase current, synchronized q-axis command, and synchronized d-axis command. The filtered three-phase current is subjected to Clarke transform and magnitude calculation to obtain shaft current, Axial current and Parker vector magnitude; Based on synchronized q-axis and d-axis commands, dynamic baseline feedforward compensation is performed on the Parker vector magnitude to obtain the compensated Parker vector magnitude. right shaft current, Fault feature extraction and decision-making are performed using shaft current and compensated Parker vector magnitude to obtain fault signs and fault locations.

[0007] Compared with existing technologies, the open-circuit fault identification method for wind power converters provided in this application constructs a dynamic health baseline in real time by utilizing q-axis and d-axis current commands that can accurately reflect the dynamic intentions of the system. This baseline is then used to feedforward compensate the Parker vector magnitude, thereby actively eliminating the interference of normal operating condition changes on fault characteristics. As a result, the compensated fault characteristics remain stable in a healthy state and only deviate significantly when a real fault occurs. This effectively decouples fault characteristics from the transient response of the system, significantly improves the signal-to-noise ratio of diagnosis, and effectively ensures the safe and stable operation of the wind power generation system. Attached Figure Description

[0008] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0009] Figure 1 This is a flowchart of a wind power converter open-circuit fault identification method according to an embodiment of this application; Figure 2 This is a schematic diagram of data flow in the wind power converter open-circuit fault identification method according to an embodiment of this application. Detailed Implementation

[0010] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0011] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not specifically singular and may include plural forms. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0012] While this application makes various references to certain modules of the systems according to embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The modules described are merely illustrative, and different aspects of the systems and methods may use different modules.

[0013] Flowcharts are used in this application to illustrate the operations performed by the system according to embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

[0014] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0015] The technical solution of this application proposes a method for identifying open-circuit faults in wind power converters. Figure 1 This is a flowchart of a wind power converter open-circuit fault identification method according to an embodiment of this application. Figure 2 This is a schematic diagram of the data flow in the wind power converter open-circuit fault identification method according to an embodiment of this application. Figure 1 and Figure 2 As shown, the wind power converter open-circuit fault identification method according to an embodiment of this application includes the following steps: S1, acquiring three-phase output current, q-axis current command, and d-axis current command; S2, preprocessing the three-phase output current, q-axis current command, and d-axis current command to obtain filtered three-phase current, synchronized q-axis command, and synchronized d-axis command; S3, performing Clarke transform and modulus calculation on the filtered three-phase current to obtain... shaft current, S4, Based on synchronized q-axis and d-axis commands, perform dynamic baseline feedforward compensation on the Parker vector magnitude to obtain the compensated Parker vector magnitude; S5, for... shaft current, Fault feature extraction and decision-making are performed using shaft current and compensated Parker vector magnitude to obtain fault signs and fault locations.

[0016] Specifically, S1 acquires the three-phase output current, q-axis current command, and d-axis current command. Traditional methods rely solely on external measurement signals such as three-phase current. During transient processes such as severe wind speed fluctuations or grid disturbances, normal transient responses can cause current distortion, whose characteristics are very similar to open-circuit faults, leading to numerous false alarms or missed alarms. This application introduces internal variables of the converter control system (such as dq-axis current commands or torque commands) as proxy variables for the operating conditions. These internal command signals are purer and respond faster than externally measured physical quantities (such as wind speed), and can more directly and accurately reflect the dynamic intent of the control system, thereby achieving more precise compensation.

[0017] In practice, the three-phase output current is acquired by installing high-precision current sensors on the converter output lines. The sensors convert the analog current signal into a voltage signal, which is then digitized by an analog-to-digital converter (ADC) at a high sampling rate (e.g., 20kHz). The q-axis and d-axis current commands are acquired by directly reading digital variables from the controller's memory or registers. The key is that these two data acquisitions must be precisely synchronized. This is achieved, for example, by aligning the ADC sampling with the controller's main program's timer interrupts or PWM update events, ensuring that all acquired data points correspond to the same instant in time.

[0018] Specifically, S2 preprocesses the three-phase output current, q-axis current command, and d-axis current command to obtain filtered three-phase current, synchronized q-axis command, and synchronized d-axis command. It should be understood that the originally acquired signals often contain noise and unnecessary interference components, which, if not processed, will seriously affect the accuracy of fault identification. Specifically, the three-phase output current collected from the sensors is inevitably affected by high-frequency harmonics generated by converter switching operations, electromagnetic interference, and measurement circuit noise. These high-frequency components and noise may mask the true fault characteristics or be misjudged as faults in a healthy state, leading to false alarms from the diagnostic system. Although the q-axis current command and d-axis current command are digital signals within the control system and are generally relatively pure, synchronization and necessary smoothing are still crucial to ensure their precise time alignment with the filtered three-phase current and to eliminate any potential subtle jitter or non-ideals within the control system. By preprocessing the three-phase output current, q-axis current command, and d-axis current command, noise in the signal can be removed, fault characteristics can be highlighted, and all input data can be made highly consistent on the time axis. This provides a high-quality and reliable data foundation for subsequent Clarke transform, Parker vector magnitude calculation, and dynamic baseline feedforward compensation.

[0019] The filtered three-phase current refers to the original three-phase output current that has been low-pass filtered to remove high-frequency noise and switching harmonics, retaining a pure current signal that reflects the system's power frequency and low-frequency dynamic characteristics. This processing helps improve the signal-to-noise ratio and makes fault-induced characteristic changes more apparent. The synchronized q-axis command and synchronized d-axis command refer to the control commands that are precisely time-aligned with the filtered three-phase current after the original q-axis current command and d-axis current command have undergone fine time calibration and possible smoothing.

[0020] In practice, the three-phase output current is first filtered. This process is typically implemented using digital filters. For example, low-pass filters, such as Butterworth filters or moving average filters, can be used. By designing an appropriate cutoff frequency, digital low-pass filters can effectively filter out high-frequency harmonics caused by the converter switching frequency and its harmonics, as well as random noise introduced by sensors, while preserving the fundamental frequency component and any possible fault-related low-frequency characteristics to the maximum extent possible. Simultaneously, the q-axis and d-axis current commands are synchronized. Since these command signals are generated internally by the controller, they are typically already in digital form and synchronized with the control cycle. Here, it is necessary to ensure that these command signals are perfectly aligned with the filtered three-phase currents in terms of timestamps. If there is a difference between the sampling rate of the three-phase currents and the update frequency of the commands, interpolation or resampling is required to ensure that both are processed on the same time base. Furthermore, to eliminate potential minor jitter or transient jumps within the control system, the command signals are lightly smoothed, for example, using a first-order inertial element or a simple moving average.

[0021] Specifically, S3 involves performing Clarke transform and magnitude calculation on the filtered three-phase current to obtain... shaft current, Shaft current and Parker vector magnitude. It should be understood that the original three-phase current signals are mutually coupled and periodically changing, making direct analysis of their fault characteristics quite complex. The Clarke transform can convert the three-phase AC signal into a stationary α-β two-phase orthogonal coordinate system, thus reducing the three-dimensional (three-phase) signal to a two-dimensional (two-phase) signal. This dimensionality reduction not only reduces data complexity, but more importantly, when an open-circuit fault occurs in the converter, the symmetry of the three-phase current is disrupted. Reflected in the α-β coordinate system, the trajectory or magnitude of the Parker vector will undergo significant and regular changes, making fault characteristics easier to identify and quantify. Calculating the Parker vector magnitude further transforms the two-dimensional vector into a single scalar characteristic. This scalar value should remain relatively stable during normal operation, but will exhibit significant fluctuations or deviations during a fault, thus becoming a sensitive fault indicator.

[0022] The Parker vector magnitude refers to the length of the Parker vector in the α-β plane. It is a scalar representing the instantaneous amplitude of the current space vector. Under healthy operating conditions, if the three-phase currents are balanced and sinusoidal, the Parker vector magnitude should be a constant. Once an open-circuit fault occurs, the symmetry of the current is disrupted, and its magnitude will fluctuate or decrease. Therefore, the Parker vector magnitude is an important characteristic quantity for judging open-circuit faults.

[0023] In practice, firstly, the filtered three-phase current is subjected to a Clarke transform to obtain... shaft current and Axis currents. The Clarke transform is a linear transformation that projects balanced three-phase currents onto mutually orthogonal α-β stationary coordinate axes; furthermore, based on... shaft current and Calculate the Parker vector magnitude using the shaft current. Specifically, the Parker vector magnitude is calculated using the following formula:

[0024] in, for shaft current, for Axial current. This formula intuitively represents the current in the α-β plane caused by... and The Euclidean distance from the origin to the endpoint of the constructed vector is the magnitude of the vector. At each sampling time, the value calculated in the previous step is... and Substituting the value into this formula yields the current Parker vector magnitude. .

[0025] Specifically, S4, based on synchronized q-axis and d-axis commands, performs dynamic baseline feedforward compensation on the Parker vector magnitude to obtain the compensated Parker vector magnitude. It should be understood that the inherent technical weakness of the feedforward compensation stage in the original diagnostic mechanism lies in the static nature of the normalization processing method. This method only considers the amplitude of the control command, completely ignoring a crucial physical relationship in the actual operation of wind power converters: the negative correlation between the dynamic intensity of the system control command and the reliability of fault characteristics. When the wind power system encounters severe transient conditions such as grid voltage drops or drastic wind speed changes, the control system will issue instantaneous and drastically changing current commands to maintain grid stability. Due to the inherent inertia of the physical system and the non-ideal nature of the control model, the Parker vector magnitude of the actual output current cannot instantaneously and perfectly track this command, resulting in a transient deviation that is numerically large in a healthy state. Traditional static normalization methods will incorrectly identify this normal transient deviation as a fault characteristic, causing the diagnostic system to output a large number of false alarms under severe transient conditions, seriously affecting the reliability of the diagnostic results. To address the transient false alarm problem caused by the aforementioned technical reasons, an adaptive tolerance normalization technique based on the dynamic fluctuations of control commands is proposed. Specifically, by introducing dynamic information from control commands, adaptive compensation for fault characteristic quantities is achieved, effectively distinguishing between real faults and transient responses under normal operating conditions, thus fundamentally solving the transient false alarm problem.

[0026] Among them, the compensated Parker vector magnitude is the final Parker vector magnitude after dynamic baseline feedforward compensation. It can more accurately reflect whether there is an open circuit fault in the converter and has higher robustness to changes in normal operating conditions.

[0027] In practice, firstly, the transient volatility of the synchronized q-axis and d-axis commands is quantified to obtain a transient volatility metric. This step aims to quantify the dynamic severity of the system control commands. Specifically, by acquiring the q-axis current command related to the dynamics of active or torque control and the d-axis current command related to the dynamics of reactive or voltage control, the time derivatives of these two control command components are calculated, and then the Euclidean norms of these two rates of change are calculated, thus obtaining a transient volatility metric that comprehensively reflects the drastic changes in the overall control command vector. More specifically, this process is expressed by the formula:

[0028] in, It is a transient volatility measure, a real-time changing scalar used to describe the severity of system transients; It is a synchronous q-axis current command; It is a synchronous d-axis current command. This step creates an indicator that is highly sensitive to dynamic disturbances in the system. Its value is close to zero during steady-state system operation, but it will show a significant spike instantaneously when encountering transient events such as sudden changes in wind speed or power grid failures. In this way, a precise and reliable quantitative input is established for subsequent adaptive adjustment of tolerance, realizing real-time perception of the system's excitation level; Next, a dynamic tolerance factor is generated based on the transient volatility measure. That is, the transient volatility measure, obtained in the pure physical quantity form from the previous steps, is transformed into a regulating factor with clear control significance that can be directly used to modify the normalization formula. Specifically, the transient volatility measure... As input, it is mapped to a dimensionless dynamic tolerance factor through a modified nonlinear sigmoid function. The design of this function makes it possible when... When I was very young, Maintain at the baseline value; while when When the preset threshold is exceeded, This will then increase smoothly and non-linearly, eventually saturating at a preset maximum value, thus achieving a smooth and bounded adjustment of the tolerance. More specifically, this process can be expressed by the formula:

[0029] in, It is a dynamic tolerance factor; It is a measure of the transient volatility of the input; It is the maximum tolerance increment, which defines the upper limit of tolerance amplification; It is the slope of the transition zone, which controls how quickly the tolerance changes; This is the volatility threshold, which defines the threshold of transient severity at which tolerance adjustments are initiated. In this way, an intelligent adjustment mechanism is constructed. It can maintain high diagnostic sensitivity in steady state and improve diagnostic tolerance in transient state according to the actual operating state of the system, thereby generating a control factor that can directly reflect the strength of transient suppression, enabling the diagnostic system to dynamically adjust its judgment criteria according to the operating conditions. Furthermore, the dynamic tolerance factor, dynamic baseline, and Parker vector magnitude are subjected to volatility-adaptive tolerance normalization to obtain the compensated Parker vector magnitude. That is, the dynamic tolerance factor generated in the preceding steps is applied to the final fault characteristic calculation to fundamentally improve the original normalization mechanism. Specifically, in the denominator of the original normalization formula, the dynamic baseline is... With dynamic tolerance factor Multiplication allows the normalized benchmark to dynamically scale according to the severity of the transient. This step adaptively adjusts the denominator so that even if the numerator increases due to normal dynamic mismatch during a severe transient, the denominator will remain unchanged. The value of the fault indicator is amplified by an equal or even greater magnitude as the value increases dramatically, thereby suppressing the final output fault indicator value below the safety threshold. More specifically, this process is expressed by the formula:

[0030] in, It is an improved and compensated Parker vector magnitude, which serves as the final fault indicator; It is the instantaneous magnitude of the Parker vector; It is a dynamic baseline related to the current command; It is a dynamic tolerance factor; This is a small constant set to prevent the denominator from being zero. Specifically, during severe transients, even if the numerator increases due to normal dynamic mismatch, the denominator will be amplified by an equal or even greater magnitude due to the sharp increase in alpha(t), thus suppressing the final output fault indicator value below the safe threshold. In this way, a fault indicator with high robustness to transient conditions can be generated, effectively distinguishing between genuine fault signals and normal system responses caused by drastic changes in operating conditions, fundamentally solving the false alarm problem.

[0031] This approach significantly improves the robustness and reliability of open-circuit fault diagnosis methods for wind turbine converters under dynamic operating conditions. Specifically, the improved mechanism can sense the transient fluctuations of system control commands in real time and dynamically adjust the tolerance boundary for fault judgment accordingly. This effectively suppresses false alarms caused by the normal dynamic response of the system under strong transient disturbances such as grid voltage drops and drastic wind speed changes. This enables the diagnostic system to accurately distinguish between true fault characteristics and normal electrical quantity distortions under transient operating conditions, achieving a high signal-to-noise ratio for fault identification. Ultimately, this not only significantly reduces the false alarm rate of the diagnostic system and enhances its practical application value in the complex and variable environment of real wind farms, but also increases the trust of maintenance personnel in the diagnostic results, providing solid technical support for achieving more accurate and reliable wind turbine condition monitoring and predictive maintenance.

[0032] Specifically, S5, for shaft current, Fault feature extraction and decision-making are performed using shaft current and the compensated Parker vector magnitude to obtain fault indicators and fault locations. Specifically, the compensated Parker vector magnitude, obtained through the aforementioned dynamic baseline feedforward compensation and exhibiting high robustness to transient conditions, reliably determines the occurrence of a fault and further identifies its specific location. Here, it should be understood that the previous step successfully separated dynamic fluctuations under normal operating conditions from fault features, ensuring that the compensated Parker vector magnitude remains stable in a healthy state, while exhibiting significant and distinguishable changes when a real open-circuit fault occurs. Therefore, in the technical solution of this application, by further analyzing this highly refined fault indicator and combining it with auxiliary information from the shaft current, the true fault signal is accurately identified amidst noise and interference, and the specific faulty component is identified. This provides clear guidance for timely maintenance and fault isolation of wind power converters, avoiding false alarms and missed alarms, and ensuring the safe and stable operation of the wind power system.

[0033] The fault flag is a binary output indicating whether an open-circuit fault exists at the current moment (e.g., 1 indicates a fault, 0 indicates no fault); the fault location further specifies which phase or which switching device of the converter has an open-circuit fault. For example, for a three-phase full-bridge converter, it may be necessary to identify whether the upper or lower arm of the IGBT in phase A, phase B, or phase C is open-circuited.

[0034] In practice, firstly, a moving average is performed on the time series of the compensated Parker vector magnitude within the sliding window to obtain the real-time fault indicator. It should be understood that although the aforementioned dynamic baseline feedforward compensation has greatly improved the robustness of the compensated Parker vector magnitude under varying operating conditions and effectively suppressed false alarms caused by normal transient fluctuations, the signal may still be affected by residual random noise, measurement errors, or subtle jitter in the control system during actual acquisition and processing. If these instantaneous high-frequency fluctuations are directly used for fault diagnosis, they may lead to frequent jumps or false triggers of fault indicators, thereby reducing the stability and reliability of the diagnostic system. By performing a moving average on the compensated Parker vector magnitude, these residual instantaneous fluctuations can be further smoothed, highlighting the continuous changing trend of the compensated Parker vector magnitude when a fault occurs, giving it a higher signal-to-noise ratio and stronger anti-interference capability when used as a fault indicator, thus improving the accuracy and stability of fault detection.

[0035] In summary, the moving average calculation is a commonly used digital signal processing method that smooths time series data by calculating the average of all data points within a sliding window to eliminate short-term fluctuations and reveal long-term trends. The real-time fault index is the output of the moving average calculation. It is a smoothed and more stable value that is directly compared with a preset threshold to determine whether a fault exists.

[0036] During this process, for each new compensated Parker vector magnitude, the system adds it to a sliding window storing recently compensated Parker vector magnitude values; simultaneously, the oldest data point in the window is removed to ensure that the window always contains the latest and a fixed number of data points; subsequently, the system calculates the arithmetic mean of all compensated Parker vector magnitude values ​​within the sliding window, using this average as the real-time fault indicator for the current moment. This process is repeated in each sampling period. It is worth noting that the length of the sliding window is a critical parameter in the design, and its selection requires a trade-off between smoothing effect and response speed. Next, the real-time fault indicator is compared with a preset fixed threshold to obtain the fault flag. It should be understood that all the aforementioned complex signal processing and compensation steps, including dynamic baseline feedforward compensation and moving average calculation, ultimately aim to generate a real-time fault indicator that is stable and close to zero in a healthy state, but significantly increases in value when a real open-circuit fault occurs. Therefore, by setting an appropriate fixed threshold, a binary decision mechanism that effectively divides the system into normal operation and fault occurrence states can be established, which is the foundation for automated fault diagnosis. It can filter out minor disturbances that may exist under complex operating conditions, ensuring that an alarm is triggered only when the fault characteristics are sufficiently obvious and persistent, thereby minimizing false alarms and improving the reliability and practicality of the diagnostic system. The introduction of the threshold enables the system to make objective judgments based on quantified data.

[0037] The preset fixed threshold is a pre-determined constant value. Setting this threshold is crucial, as it serves as the boundary between normal and fault states. Its selection requires comprehensive consideration of diagnostic sensitivity, anti-interference capability, and false alarm rate. It is typically determined through extensive experimental testing, historical data analysis, or simulation to ensure timely fault detection while minimizing false alarms.

[0038] During this process, the system acquires the real-time fault indicator at each sampling period and compares it with a fixed threshold pre-stored in the controller's memory. The comparison logic determines whether the real-time fault indicator is greater than the preset fixed threshold. Specifically, the fault flag generation logic can be expressed as follows: if the real-time fault indicator is greater than the preset fixed threshold, the fault flag is set to 1, indicating that an open-circuit fault has been detected; conversely, if the real-time fault indicator is less than or equal to the preset fixed threshold, the fault flag is set to 0, indicating that the system is operating normally. This logical judgment is performed continuously, updating the fault flag status at each sampling moment. Once the real-time fault indicator first exceeds the preset fixed threshold, the fault flag is set to 1, indicating that the system has identified an open-circuit fault. Subsequently, even if the real-time fault indicator fluctuates slightly, as long as its value remains higher than the preset fixed threshold, the fault flag will remain at 1. Only when the fault is eliminated and the real-time fault indicator continuously falls below the preset fixed threshold will the fault flag be reset to 0. Furthermore, in response to a fault flag of one, based on shaft current, The fault location was determined by analyzing the time series of shaft current and the compensated Parker vector magnitude. It should be understood that although the aforementioned steps have successfully and accurately identified open-circuit faults under complex operating conditions, i.e., obtained fault indicators, they are insufficient to guide actual maintenance work. Converters typically consist of multiple power electronic switching devices (such as IGBTs), and an open-circuit fault in each switching device will have different effects on the converter's operation. To quickly and accurately isolate and repair faults, minimize downtime, and reduce maintenance costs, it is necessary to identify which phase (phase A, phase B, phase C) and which switching device (upper or lower bridge arm) the fault occurred in. Therefore, by analyzing the time series of the fault occurrence and subsequent data... shaft current, By conducting in-depth analysis of the time series of shaft current and compensated Parker vector magnitude, the unique fingerprint features formed in these signals by different open-circuit fault modes can be used to accurately determine the specific location of the fault.

[0039] In this process, firstly, analysis shaft current and Trajectory characteristics of shaft current: Under healthy conditions, the balanced three-phase currents in... - A circular trajectory is typically depicted on a plane. When a single-phase open-circuit fault occurs, the symmetry of the three-phase current is broken, which will lead to... shaft current - The shaft current trajectory changes from a circle to a specific ellipse, an open curve, or a distorted trajectory containing straight line segments. Different open-circuit fault types (e.g., open circuit in the upper arm of phase A, open circuit in the lower arm of phase B, etc.) will... shaft current - Differential and identifiable trajectory deformation patterns are generated on the axial current plane. By comparing the currently observed trajectory with pre-stored trajectory templates of various typical open-circuit fault modes, the phase and type of the fault can be preliminarily determined. Next, the fluctuation pattern of the time series of the compensated Parker vector magnitude is analyzed: Besides shaft current - The trajectory of the shaft current changes, and the compensated Parker vector magnitude also exhibits specific periodic fluctuations or instantaneous decrease patterns after an open-circuit fault. For example, some open-circuit faults may cause the compensated Parker vector magnitude to show one or two significant decreases within each fundamental cycle, and the magnitude and duration of these decreases are also related to the fault type. These dynamic characteristics of the compensated Parker vector magnitude are then compared with... shaft current and Combining the analysis of shaft current trajectory deformation can enhance the accuracy of fault location; Furthermore, the bridge arm position is determined based on the polarity of the fault phase current: for an open-circuit fault in a certain phase, its impact is often related to the polarity of the phase current. For example, if the fault occurs in the upper bridge arm IGBT of a certain phase, the current in that phase will be blocked when it attempts to flow in the positive direction, causing the current waveform to be clamped or disappear during the positive half-cycle; conversely, if the fault occurs in the lower bridge arm IGBT, the current in that phase will be blocked when it attempts to flow in the negative direction. Further analysis of the fault occurrence time... shaft current and The instantaneous value of the shaft current can be used to infer the current polarity of the faulty phase, thereby distinguishing whether it is an open circuit fault in the upper or lower bridge arm. Subsequently, in practical applications, fault location is often achieved by establishing a fault mode library. This library stores the fault modes corresponding to various known open-circuit fault types. shaft current and The system collects the trajectory characteristics of the shaft current, the fluctuation pattern of the time series of the compensated Parker vector magnitude, and other auxiliary information. When a fault flag of one is detected, the system will update the currently collected data. shaft current, The time series characteristics of shaft current and compensated Parker vector magnitude are compared with all patterns in the pattern library, and the most matching fault type and location are determined by pattern matching algorithms (such as distance-based, correlation-based, or neural network-based).

[0040] In summary, the wind power converter open-circuit fault identification method according to the embodiments of this application is explained. It constructs a dynamic health state baseline in real time by utilizing q-axis and d-axis current commands that can accurately reflect the dynamic intention of the system, and uses this baseline to feedforward compensate the Parker vector magnitude, thereby actively eliminating the interference of normal operating condition changes on fault characteristics. In this way, the compensated fault characteristics can remain stable in the healthy state and only deviate significantly when a real fault occurs. This achieves effective decoupling of fault characteristics and system transient response, significantly improves the signal-to-noise ratio of diagnosis, and effectively ensures the safe and stable operation of the wind power generation system.

[0041] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for open circuit fault identification of a wind power converter, characterized in that, The method comprises: acquiring three-phase output currents, a q-axis current instruction and a d-axis current instruction; preprocessing the three-phase output currents, the q-axis current instruction and the d-axis current instruction to obtain filtered three-phase currents, synchronized q-axis instructions and synchronized d-axis instructions; The filtered three-phase currents are subjected to a Clarke transformation and a module length calculation to obtain the shaft current, the shaft current and the Parker vector module length; performing dynamic baseline feedforward compensation on a Park vector module length based on the synchronized q-axis instructions and the synchronized d-axis instructions to obtain a compensated Park vector module length; To Shaft current, Fault feature extraction and decision are made on shaft current and compensated park vector module to get fault flag and fault location.

2. The wind power converter open circuit fault identification method of claim 1, wherein, The filtered three-phase currents are subjected to a Clarke transformation and a module length calculation to obtain the shaft current, the shaft current and the Parker vector module length, comprising: The filtered three-phase currents are subjected to a Clarke transformation to obtain the axis currents and the axis currents; based on axial current and axial current, the poynting vector modulus is calculated.

3. The wind power converter open circuit fault identification method of claim 2, wherein, based on axial current and axial current, calculating the poynting vector modulus, comprising: calculating the poynting vector modulus with the following formula: wherein is shaft current, is shaft current.

4. The wind power converter open circuit fault identification method of claim 1, wherein, performing dynamic baseline feedforward compensation on a Park vector module length based on the synchronized q-axis instructions and the synchronized d-axis instructions to obtain a compensated Park vector module length, comprising: quantifying transient fluctuation of control instructions based on the synchronized q-axis instructions and the synchronized d-axis instructions to obtain a transient fluctuation metric; generating a dynamic tolerance factor based on the transient fluctuation metric; performing fluctuation adaptive tolerance normalization on the dynamic tolerance factor, a dynamic baseline and the Park vector module length to obtain the compensated Park vector module length.

5. The wind power converter open circuit fault identification method of claim 4, wherein, quantifying transient fluctuation of control instructions based on the synchronized q-axis instructions and the synchronized d-axis instructions to obtain a transient fluctuation metric, comprising: wherein, is a transient fluctuation metric; is a synchronized q-axis current command, and is a synchronized d-axis current command.

6. The wind power converter open circuit fault identification method of claim 4, wherein, generating a dynamic tolerance factor based on the transient fluctuation metric, comprising: wherein, is a dynamic tolerance factor; is a transient volatility measure; is a maximum tolerance increment; is a transition zone slope, and is a volatility threshold.

7. The wind power converter open circuit fault identification method of claim 4, wherein, performing fluctuation adaptive tolerance normalization on the dynamic tolerance factor, a dynamic baseline and the Park vector module length to obtain the compensated Park vector module length, comprising: wherein is the compensated park vector magnitude; is the park vector magnitude; is the dynamic baseline associated with the current command; is the dynamic tolerance factor, and is a small constant set to prevent the denominator from being zero.

8. The wind power converter open circuit fault identification method of claim 1, wherein, For shaft current, fault feature extraction and decision are made on the shaft current and the compensated park vector module length to obtain the fault flag and the fault location, including: performing a moving average calculation on a time series of the compensated Park vector module length within a sliding window to obtain a real-time fault indicator; comparing the real-time fault indicator with a preset fixed threshold to obtain a fault flag; in response to the fault flag being one, based on the shaft current, the shaft current and the time series of compensated park vector magnitudes, determining the fault location.

Images