Variable-pitch drive redundancy fault-tolerant method and system for high-power wind turbine generator

Abstract

The embodiment of the invention provides a variable-pitch drive redundancy fault-tolerant method and system for a high-power wind turbine generator, and the method comprises the steps: decoupling a complex electromechanical coupling system into two independent electrical and mechanical analysis dimensions at a diagnosis level through constructing an electrical state observer and a mechanical state observer which are parallel; the method comprises the following steps of: comparing and analyzing an electrical residual error and a mechanical residual error generated by two observers, and accurately tracing a root cause of a fault, thereby realizing clear identification of two faults which are similar in representation but different in root source, such as electrical driving failure and mechanical resistance increase, further evaluating the confidence coefficient of the faults, and generating a clear maintenance instruction. In this way, maintenance resource waste and power generation loss caused by fault misjudgment are avoided, and then the operation reliability and the intelligent operation and maintenance level of the variable-pitch driving system of the high-power wind turbine generator are remarkably improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of fault diagnosis technology, specifically to a redundancy fault-tolerant method and system for the pitch drive of a high-power wind turbine. Background Technology

[0002] The pitch control system of a high-power wind turbine is a core component for power control and safety protection. Its reliability and safety directly affect the stability and power generation efficiency of the entire unit. Given that wind turbines are typically deployed in harsh, remote environments, a failure in the pitch drive system can lead to prolonged downtime, significant power generation losses, and high maintenance costs. Therefore, to improve operational efficiency and reduce economic losses, it is crucial to develop a solution capable of accurate fault diagnosis and fault-tolerant control of the pitch drive system.

[0003] While existing technologies have conducted some research on fault diagnosis for pitch drive systems, a key technical challenge remains: accurately distinguishing between electrical and mechanical faults, which are fundamentally different in nature. When the system experiences an abnormal blade rotation speed lower than the commanded speed, the root cause could be either an electrical component failure such as a degraded drive performance, or an abnormal increase in mechanical resistance in the transmission system due to factors like bearing jamming. Although these two types of faults have different root causes, they exhibit highly similar external characteristics, making traditional diagnostic methods prone to confusion and misdiagnosis. Such incorrect diagnosis not only fails to solve the fundamental problem but also wastes spare parts costs and prolongs maintenance cycles, severely impacting the economic efficiency and reliable operation of wind turbine units.

[0004] Therefore, an optimized method for redundancy and fault tolerance in the pitch drive of high-power wind turbines is needed. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art, and provides a redundancy fault-tolerant method and system for the pitch drive of high-power wind turbine units.

[0006] In a first aspect, embodiments of the present invention provide a redundancy and fault-tolerant method for the pitch drive of a high-power wind turbine, comprising: Obtain the q-axis voltage command, the actual q-axis current, and the actual electric angular velocity of the motor; Electrical state observation is performed based on q-axis voltage command, actual electric angular velocity of the motor, and actual q-axis current to obtain the q-axis current and electrical residual predicted by the electrical observer; Mechanical state observation is performed based on the actual electric angular velocity of the motor and the actual q-axis current to obtain the electric angular velocity and mechanical residual predicted by the mechanical observer; Fault confidence assessment is performed based on electrical and mechanical residuals to obtain fault attribute vectors; Fault classification and maintenance instruction generation are performed based on fault attribute vectors to obtain fault diagnosis results and maintenance codes.

[0007] In some possible embodiments, electrical state observation is performed based on the q-axis voltage command, the actual electrical angular velocity of the motor, and the actual q-axis current to obtain the q-axis current and electrical residual predicted by the electrical observer, including: Based on the voltage equation of PMSM, the first Luenberger observer model is constructed; The q-axis voltage command and the actual electric angular velocity of the motor are input into the Luenberger observer model to obtain the q-axis current predicted by the electrical observer. The absolute value of the difference between the q-axis current predicted by the electrical observer and the actual q-axis current is calculated as the electrical residual.

[0008] In some possible embodiments, the first Luenberger observer model is expressed by the formula:

[0009] in, The q-axis current predicted by the electrical observer. It is a permanent magnet flux linkage. For the gain of the gas observation instrument, This represents the actual q-axis current. This is the q-axis voltage command. and The motor resistance and q-axis inductance under healthy conditions. This is the actual electrical angular velocity of the motor.

[0010] In some possible embodiments, mechanical state observation is performed based on the actual electrical angular velocity of the motor and the actual q-axis current to obtain the electrical angular velocity and mechanical residual predicted by the mechanical observer, including: Based on the motor motion equations, a second Luenberger observer model is constructed. The actual q-axis current is input into the second Luenberger observer model to obtain the electric angular velocity predicted by the mechanical observer; The absolute value of the difference between the electric angular velocity predicted by the mechanical observer and the actual electric angular velocity of the motor is taken as the mechanical residual.

[0011] In some possible embodiments, the second Luenberger observer model is expressed by the formula:

[0012] in, The torque constant is The electric angular velocity predicted by the mechanical observer. For external load torque, For the gain of the mechanical observer, and The total moment of inertia and coefficient of viscous friction under healthy conditions are given. This represents the actual q-axis current. This is the actual electrical angular velocity of the motor.

[0013] In some possible embodiments, fault confidence assessment is performed based on electrical and mechanical residuals to obtain a fault attribute vector, including: The electrical and mechanical residuals are normalized to obtain normalized electrical and mechanical residuals. The normalized electrical residual and normalized mechanical residual are integrated using a sliding window to obtain the integral values ​​of the electrical residual and the mechanical residual; The confidence level of the electrical residual integral value and the mechanical residual integral value is calculated to obtain the fault attribute vector.

[0014] In some possible embodiments, confidence levels are calculated on the integral values ​​of electrical and mechanical residuals to obtain the fault attribute vector, including: The confidence level of the integral values ​​of electrical and mechanical residuals is calculated using the following formula:

[0015]

[0016]

[0017] in, This is the integral value of the electrical residual. This is the integral value of the mechanical residual. This is a fault attribute vector.

[0018] Secondly, embodiments of the present invention provide a high-power wind turbine pitch drive redundancy fault-tolerant system, comprising: The data acquisition module is used to acquire the q-axis voltage command, the actual q-axis current, and the actual electric angular velocity of the motor. The electrical condition observation module is used to observe the electrical condition based on the q-axis voltage command, the actual electric angular velocity of the motor, and the actual q-axis current in order to obtain the q-axis current and electrical residual predicted by the electrical observer. The mechanical condition observation module is used to observe the mechanical condition based on the actual electric angular velocity of the motor and the actual q-axis current in order to obtain the electric angular velocity and mechanical residual predicted by the mechanical observer. The fault confidence assessment module is used to assess fault confidence based on electrical and mechanical residuals to obtain fault attribute vectors. The fault classification and maintenance instruction generation module is used to classify faults and generate maintenance instructions based on fault attribute vectors to obtain fault diagnosis results and maintenance codes.

[0019] In some possible embodiments, the electrical condition monitoring module is further configured to: Based on the voltage equation of PMSM, the first Luenberger observer model is constructed; The q-axis voltage command and the actual electric angular velocity of the motor are input into the Luenberger observer model to obtain the q-axis current predicted by the electrical observer. The absolute value of the difference between the q-axis current predicted by the electrical observer and the actual q-axis current is calculated as the electrical residual.

[0020] In some possible embodiments, the mechanical condition observation module is further configured to: Based on the motor motion equations, a second Luenberger observer model is constructed. The actual q-axis current is input into the second Luenberger observer model to obtain the electric angular velocity predicted by the mechanical observer; The absolute value of the difference between the electric angular velocity predicted by the mechanical observer and the actual electric angular velocity of the motor is taken as the mechanical residual.

[0021] Compared with existing technologies, this invention provides a redundancy-tolerant fault-tolerant method and system for high-power wind turbine pitch drive. By constructing parallel electrical and mechanical state observers, it decouples the complex electromechanical coupling system into two independent analytical dimensions—electrical and mechanical—at the diagnostic level. Through comparative analysis of the electrical and mechanical residuals generated by the two observers, it accurately traces the root cause of faults, thereby clearly identifying two types of faults—electrical drive failure and increased mechanical drag—that appear similar but have different root causes. Furthermore, it assesses the confidence level of the fault and generates clear maintenance instructions. This approach avoids wasted maintenance resources and power generation losses due to misdiagnosis, significantly improving the operational reliability and intelligent operation and maintenance level of high-power wind turbine pitch drive systems. Attached Figure Description

[0022] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0023] Figure 1A flowchart of a high-power wind turbine pitch drive redundancy fault-tolerant method according to an embodiment of the present invention; Figure 2 This is a data flow diagram of a high-power wind turbine pitch drive redundancy fault-tolerant method according to an embodiment of the present invention. Figure 3 This is a block diagram of a high-power wind turbine pitch drive redundancy fault-tolerant system according to an embodiment of the present invention. Detailed Implementation

[0024] To enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0025] Unless otherwise specifically stated, the technical or scientific terms used in the embodiments of this invention should be understood in their ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains. The terms "comprising" or "including," as used in the embodiments of this invention, do not limit the shapes, numbers, steps, actions, operations, components, elements, and / or groups thereof mentioned, nor do they exclude the appearance or addition of one or more other different shapes, numbers, steps, actions, operations, components, elements, and / or groups thereof, or the inclusion of these.

[0026] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of the invention. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale, and techniques, methods, and apparatus known to those skilled in the art may not be discussed in detail; however, where appropriate, the illustrated techniques, methods, and apparatus should be considered part of the specification. In all the examples shown and discussed herein, any other specific example may have different values. It should be noted that similar symbols and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0027] In the description of the embodiments of the present invention, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In the embodiments of the present invention, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in the embodiments of the present invention, as well as the features of different embodiments or examples.

[0028] Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0029] In the technical solution of this invention, a redundancy and fault-tolerant method for the pitch drive of a high-power wind turbine is proposed. Figure 1 This is a flowchart of a high-power wind turbine pitch drive redundancy fault-tolerant method according to an embodiment of the present invention. Figure 2 This is a system architecture diagram of a high-power wind turbine pitch drive redundancy fault-tolerant method according to an embodiment of the present invention. Figure 1 and Figure 2 As shown, the high-power wind turbine pitch drive redundancy fault-tolerant method according to an embodiment of the present invention includes the following steps: S1, acquiring the q-axis voltage command, the actual q-axis current, and the actual electrical angular velocity and actual q-axis current of the motor; S2, performing electrical state observation based on the q-axis voltage command, the actual electrical angular velocity of the motor, and the actual q-axis current to obtain the q-axis current and electrical residual predicted by the electrical observer; S3, performing mechanical state observation based on the actual electrical angular velocity of the motor and the actual q-axis current to obtain the electrical angular velocity and mechanical residual predicted by the mechanical observer; S4, performing fault confidence assessment based on the electrical residual and the mechanical residual to obtain a fault attribute vector; S5, performing fault classification and maintenance command generation based on the fault attribute vector to obtain fault diagnosis results and maintenance codes.

[0030] Specifically, in S1, the q-axis voltage command, actual q-axis current, and actual electric angular velocity of the motor are acquired. Here, the q-axis voltage command, actual q-axis current, and actual electric angular velocity of the motor and actual q-axis current constitute the data foundation for subsequent state observation, residual calculation, and fault assessment. By accurately and in real-time capturing the key electrical and mechanical state parameters of the pitch drive system during operation, the effectiveness of the subsequent fault diagnosis model and the reliability of the final diagnosis results are guaranteed.

[0031] The q-axis voltage command refers to the target q-axis voltage component used to drive the motor, calculated by the upper-level controller of the pitch system (e.g., a PID controller) based on the deviation between the desired blade angle or rotational speed and the actual value. It is an internal control signal representing the electromagnetic torque the system expects the motor to output. Therefore, it is not obtained through physical sensor measurement, but rather calculated and generated in real-time by the pitch controller's internal algorithm. For example, when the pitch controller outputs a control quantity to adjust the motor torque based on the deviation between the actual and target blade angles using PID or other control strategies, this control quantity, after appropriate coordinate transformation (such as inverse Park transformation), yields the corresponding q-axis voltage command. The data acquisition module directly captures this digital command value at a preset sampling frequency by reading the controller's internal registers, shared memory, or real-time communication interface (such as a high-speed digital bus).

[0032] The actual q-axis current is the real q-axis current component flowing through the motor windings, directly reflecting the magnitude of the electromagnetic torque actually generated by the motor. In acquiring the actual q-axis current, firstly, a precision current sensor (e.g., a Hall effect current sensor or a current sampling circuit based on shunt resistors) installed on the motor driver output side (typically the three-phase power lines) is used to measure the actual three-phase current of the motor with high precision. These sensors convert the current signal into a voltage signal. Subsequently, these analog voltage signals are sent to a high-speed analog-to-digital converter (ADC) for digitization. The acquired three-phase digital current values ​​are then transformed from a three-phase stationary coordinate system (abc coordinate system) to a two-phase stationary coordinate system using a Clarke transform in a digital signal processor (DSP) or microcontroller, and then from the two-phase stationary coordinate system to a two-phase rotating coordinate system using a Park transform. During the Park transform, the rotor position information of the motor is required to ensure the accuracy of the coordinate transformation. Finally, the q-axis current component in the dq coordinate system is extracted. The entire current sampling and transformation process must be performed synchronously at high frequency to capture the dynamic response of the motor.

[0033] The actual electrical angular velocity of a motor is the true rate of rotation of the motor rotor. It is typically obtained using position sensors mounted on the end of the motor shaft, such as resolvers or high-precision absolute encoders. These sensors provide high-resolution rotor position information. The data acquisition module calculates the motor's mechanical angular velocity by performing differential operations on the continuously sampled position information and dividing by the sampling time interval. Based on the motor's own pole pair parameter (p), this is converted into the actual electrical angular velocity using the following formula: .

[0034] It is worth mentioning that the acquisition of these parameters requires high precision and strict synchronization to ensure that the subsequent electrical condition observer and mechanical condition observer can make accurate predictions and residual calculations based on a consistent time base.

[0035] Specifically, S2 involves observing the electrical state based on the q-axis voltage command, the actual electrical angular velocity of the motor, and the actual q-axis current to obtain the predicted q-axis current and electrical residual by the electrical observer. It should be understood that the stability of the pitch drive system's performance is directly related to the overall safety and power generation efficiency of the wind turbine. As the core actuator, the health of the motor's electrical system is paramount among all considerations. By constructing a mathematical model that accurately simulates the physical characteristics of the motor in a healthy state—the state observer—and using this model to predict the motor's q-axis current, the model's predicted output can be compared in real time with the value actually measured by sensors. The difference between these two, the electrical residual, constitutes a quantitative indicator of the system's health status. In one example, in an ideal, fault-free operating environment, this residual value should approach zero infinitely; conversely, a sustained and significant deviation in the residual value strongly indicates a possible electrical fault in the system, such as motor winding parameter drift or inter-turn short circuits. This step is a necessary prerequisite for early warning of potential faults and for improving the reliability and safety of system operation. The electrical residuals generated are the core inputs upon which subsequent fault confidence assessment and accurate fault classification depend.

[0036] Electrical state observation, in essence, applies state estimation techniques from control theory. It constructs a mathematical model of a system and uses measurable inputs (such as voltage commands) and outputs (such as rotational speed) to infer internal state variables (such as predicted current) that are either directly measurable or prohibitively expensive to measure. The Luenberger observer is a specific and widely used implementation of a closed-loop state observer. By introducing a feedback correction mechanism for output errors, it ensures that the state estimates asymptotically converge to the true values, offering advantages such as simple structure and reliable performance. Electrical residuals, on the other hand, are a key term within the model-based diagnostic framework. They are explicitly defined as the difference between the actual measured output of the system and the model's predicted output, serving as direct evidence of unhealthy system behavior.

[0037] In practical implementation, firstly, based on the voltage equations of the PMSM, a first Luenberger observer model is constructed. Here, the Luenberger observer, a classic tool in modern control theory, is essentially a mathematical model constructed in parallel with the observed object (here, a permanent magnet synchronous motor). By introducing a correction feedback term based on the output error, it drives the estimated values ​​of the model's internal state variables to continuously and asymptotically converge to the state variables of the actual system. In this technical solution, the observer model is strictly established based on the q-axis voltage balance equations of the permanent magnet synchronous motor in the dq synchronous rotating coordinate system. Specifically, the first Luenberger observer model is expressed by the following formula:

[0038] in, The q-axis current predicted by the electrical observer. It is a permanent magnet flux linkage. For the gain of the gas observation instrument, This represents the actual q-axis current. This is the q-axis voltage command. and The motor resistance and q-axis inductance under healthy conditions. This is the actual electrical angular velocity of the motor.

[0039] Next, the q-axis voltage command and the actual electrical angular velocity of the motor are input into the Luenberger observer model to obtain the q-axis current predicted by the electrical observer. During this stage, the system's data acquisition module continuously collects three key data points from the control system and sensors: the q-axis voltage command, the actual electrical angular velocity of the motor, and the actual q-axis current. These three real-time data points, along with pre-set motor health model parameters and tuned observer gain, are substituted into the aforementioned observer differential equation. In digital control systems, numerical integration algorithms such as the Euler method or higher-order Runge-Kutta methods are typically used to discretize and solve this differential equation, thereby calculating the latest prediction of the q-axis current by the electrical observer at each control cycle.

[0040] Furthermore, the absolute value of the difference between the q-axis current predicted by the electrical observer and the actual q-axis current is calculated as the electrical residual. By calculating the absolute value of the difference between the q-axis current predicted by the electrical observer and the actual q-axis current, a signal that can directly reflect the degree of system anomaly can be generated. Specifically, after obtaining the predicted q-axis current value output by the observer, the system immediately compares it with the actual q-axis current measured at the same sampling time. The absolute value of the difference between the two is formally defined as the electrical residual, which quantifies the deviation between the dynamic behavior of the actual electrical subsystem of the motor and its expected behavior under healthy conditions (described by the model). Under normal operating conditions where the motor is fault-free and the parameters do not change, the observer can ensure that its predicted value is in high agreement with the actual value. At this time, the electrical residual will be very small and fluctuate within a very small range. However, once the motor experiences an electrical fault, such as an inter-turn short circuit in the winding causing its actual resistance or inductance to deviate from the healthy value in the model, the actual physical response of the motor will no longer follow the original mathematical model. This leads to a significant and persistent deviation between the actual current and the model-predicted current, resulting in a sharp increase in the amplitude of the electrical residual. This amplified residual signal forms a solid foundation for the subsequent in-depth analysis and judgment by the fault confidence assessment module.

[0041] Specifically, S3 involves observing the mechanical state of the motor based on its actual electrical angular velocity and q-axis current to obtain the predicted electrical angular velocity and mechanical residual from the mechanical observer. It should be understood that the motor's ultimate function is to output torque and drive the mechanical load; its mechanical dynamic characteristics directly reflect the health of the entire transmission chain. For example, bearing wear and lubrication failure can increase the coefficient of friction, while blade icing or damage can alter the system's total moment of inertia and external load. These mechanical anomalies may be difficult to distinguish accurately through purely electrical observation. Therefore, by constructing a mathematical model (i.e., a mechanical observer) specifically to describe the motor's mechanical motion in a healthy state, and using this model to predict the motor's core mechanical state variables (electrical angular velocity in this scheme), the predicted values ​​can be compared with the actual measured values. The deviation between the two, i.e., the mechanical residual, can serve as a direct indication of mechanical anomalies or faults in the system. A system with good mechanical performance should maintain a mechanical residual close to zero; a significant increase in the residual indicates a possible mechanical fault. Therefore, this step is crucial for accurate fault location (distinguishing between electrical and mechanical faults), and the generated mechanical residuals are another key input for subsequent fault confidence assessment and fault classification. In other words, the focus of diagnosis shifts from the electrical subsystem of the motor to the mechanical subsystem. By modeling and observing the dynamic behavior of the motor and its load, the aim is to identify and quantify fault symptoms caused by abnormalities in mechanical components (such as friction and changes in inertia), thereby achieving precise monitoring of the mechanical health of the pitch drive system. This complements electrical condition observation, together forming a complete fault diagnosis system.

[0042] Mechanical state observation refers to the process of estimating the internal mechanical state variables of a system by using its mechanical motion equations model, combined with measurable inputs (such as driving current) and outputs (such as actual velocity). The second Luenberger observer is applied to the second part of the system, namely the mechanical subsystem, whose model is based on rigid body dynamics equations.

[0043] In practical implementation, firstly, a second Luenberger observer model is constructed based on the motor's motion equations. Similar to the electrical observer, the Luenberger observer architecture is also used here to ensure the robustness and convergence of the predictions. Its theoretical basis is the application of Newton's second law to rotational motion, namely the balance relationship between electromagnetic torque, load torque, inertial torque, and frictional torque. Specifically, the second Luenberger observer model is expressed by the following formula:

[0044] in, The torque constant is The electric angular velocity predicted by the mechanical observer. For external load torque, For the gain of the mechanical observer, and The total moment of inertia and coefficient of viscous friction under healthy conditions are given. This represents the actual q-axis current. This is the actual electrical angular velocity of the motor.

[0045] Next, the actual q-axis current is input into the second Luenberger observer model to obtain the electric angular velocity predicted by the mechanical observer. In this stage, the system uses the actual q-axis current as the input for generating electromagnetic torque, and the actual electric angular velocity of the motor as the basis for feedback correction. These real-time acquired data, along with preset motor health mechanical parameters, estimated external load torque, and calibrated observer gain, are substituted into the aforementioned mechanical observer differential equation. Within each sampling cycle of the digital control system, the equation is solved using a numerical integration algorithm to calculate the mechanical observer's prediction of the motor's electric angular velocity at the current moment.

[0046] Furthermore, the absolute value of the difference between the electrical angular velocity predicted by the mechanical observer and the actual electrical angular velocity of the motor is used as the mechanical residual. By calculating the difference between the electrical angular velocity predicted by the mechanical observer and the actual electrical angular velocity of the motor, an index signal that can quantify the behavioral deviation of the mechanical system can be generated. Specifically, after obtaining the predicted electrical angular velocity output by the observer, the system directly compares it with the electrical angular velocity actually measured by the sensor at the same moment. The absolute value of the difference between these two values ​​is defined as the mechanical residual. This mechanical residual value quantifies the difference between the actual mechanical dynamic response of the motor and transmission system and its expected response under healthy conditions (defined by the model). When the mechanical parts of the system are working normally, because the model parameters match the actual situation, the observer's predicted value will track the actual value very accurately, thus keeping the mechanical residual at a very low level. Conversely, if the system experiences mechanical failure, such as bearing damage leading to an increase in the coefficient of viscous friction, or blade icing leading to an increase in the total moment of inertia, then the actual motion state will deviate from the predicted trajectory of the healthy model. This will result in a non-negligible and persistent deviation between the actual and predicted speeds, ultimately manifesting as a significant increase in mechanical residuals. This increased residual signal provides clear evidence of a deterioration in the health of the mechanical system for subsequent failure confidence assessments.

[0047] Specifically, in step S4, a fault confidence assessment is performed based on electrical and mechanical residuals to obtain a fault attribute vector. It should be understood that, firstly, sensor noise, abrupt changes in control commands, and minor disturbances in the external environment can all cause instantaneous spikes in the residual signal. Directly judging based on these spikes is highly likely to lead to false alarms. Secondly, when a system fault occurs, both electrical and mechanical residuals may increase simultaneously, making it difficult and unreliable to simply compare their magnitudes to determine the fault source. Therefore, in the technical solution of this invention, a series of signal processing and mathematical transformations are used to filter out noise interference, enhance the saliency of fault features, and ultimately clearly quantify the fault type attributed to the detected anomaly in the form of confidence scores. The final output fault attribute vector is a mathematical entity containing confidence score assignments, providing an ideal, decoupled input feature for subsequent classification algorithms.

[0048] In practice, the electrical and mechanical residuals are first normalized to obtain normalized electrical and mechanical residuals. It should be understood that electrical and mechanical residuals originate from different physical domains of the system (electrical and mechanical domains), and their dimensions, numerical ranges, and dynamic characteristics differ significantly. Electrical residuals reflect current deviations, while mechanical residuals reflect angular velocity deviations. If they are directly compared or merged without processing, the residuals with larger numerical ranges will dominate subsequent calculations, thus obscuring the useful fault information contained in the residuals with smaller numerical ranges, leading to biased evaluation results and even incorrect diagnostic conclusions. Normalization eliminates this adverse effect caused by differences in dimensions and numerical ranges, placing the two residuals on a unified, comparable scale, ensuring the fairness and accuracy of subsequent fault confidence assessments.

[0049] Specifically, a standardized mathematical transformation method can be used to map the instantaneous values ​​of the original electrical and mechanical residuals to a pre-defined, uniform numerical range, such as [-1, 1]. In a specific example, this can be achieved using methods such as min-max scaling. This method requires obtaining the maximum and minimum values ​​of the electrical and mechanical residuals under various operating conditions (including healthy and various fault states) through experiments or simulations beforehand, and then using these boundaries to linearly scale the residuals obtained in real time.

[0050] Next, a sliding window integration is performed on the normalized electrical and mechanical residuals to obtain the integrated values ​​of the electrical and mechanical residuals. It should be understood that sensor noise, transient responses of the control system, or slight disturbances in the external environment can all cause brief, meaningless spikes in the normalized residuals. Directly judging based on these instantaneous values ​​can easily lead to false alarms. Sliding window integration, by accumulating the normalized residuals over a certain time window, effectively smooths out these random noises and glitches, preserving the true trend of the signal. More importantly, it accumulates evidence of faults, because a persistent fault will inevitably cause the residuals to deviate from normal levels continuously over a period of time, significantly increasing their integral value, thus providing a strong criterion for distinguishing between real faults and temporary disturbances.

[0051] Sliding window integration is a signal processing technique that observes local segments of a data sequence through a moving window and performs integration (or summation in this context) on the data points within the window. This method effectively extracts local trend features of the signal and suppresses high-frequency noise. The electrical residual integral value represents the sum of all normalized electrical residual sample values ​​within the current time window, reflecting the cumulative effect of recent electrical subsystem anomalies. The mechanical residual integral value represents the sum of all normalized mechanical residual sample values ​​within the current time window, reflecting the cumulative effect of recent mechanical subsystem anomalies.

[0052] Specifically, this can be achieved by maintaining a fixed-length time window (or a fixed number of sample points), which slides forward over time. At each sampling moment, the system calculates and updates the sum of all normalized electrical residual samples within the current window to obtain the current electrical residual integral value; simultaneously, it calculates the sum of all normalized mechanical residual samples within the window in the same way to obtain the current mechanical residual integral value. This process is continuous, thereby generating two time-varying residual integral value signals.

[0053] Furthermore, confidence scores are calculated on the electrical and mechanical residual integral values ​​to obtain the fault attribute vector. It should be understood that while the electrical and mechanical residual integral values ​​obtained after integration can reflect the cumulative effect of the fault well, they are still two independent values, not intuitive enough, and fail to clearly indicate the probability of the fault occurring or its attribution. The purpose of confidence score calculation is to transform these two integral values ​​into a standardized fault attribute vector with clear probabilistic meaning. This vector not only represents the overall confidence level of the system fault but also clearly reveals whether the fault is more likely to originate from the electrical or mechanical components. In this way, it provides direct and quantitative input for subsequent automated diagnosis and decision-making.

[0054] Specifically, the confidence levels of the integral values ​​of electrical and mechanical residuals are calculated using the following formula:

[0055]

[0056]

[0057] in, This is the integral value of the electrical residual. This is the integral value of the mechanical residual. For fault attribute vectors, For electrical system fault confidence, Confidence level for mechanical system failure.

[0058] Specifically, in step S5, fault classification and maintenance instruction generation are performed based on the fault attribute vector to obtain fault diagnosis results and maintenance codes. It should be understood that the preceding steps, through state observation and confidence assessment, condense the complex system state into a structured fault attribute vector, but this set of values ​​is still not intuitive enough for maintenance personnel. Therefore, in the technical solution of this invention, fault classification and maintenance instruction generation are performed based on the fault attribute vector to transform this quantified, machine-readable fault attribute vector into a clear, human-understandable fault diagnosis result, and further map it into standardized, executable maintenance codes. This greatly simplifies the fault troubleshooting process, reduces reliance on the experience of maintenance personnel, shortens downtime for maintenance, and thus improves the availability and power generation efficiency of wind turbine units.

[0059] The fault diagnosis result is a clear and unambiguous natural language text that clearly indicates the most likely fault type or faulty component in the system, such as "abnormal electrical parameters of the pitch drive motor" or "speed sensor signal drift".

[0060] The maintenance code is a standardized alphanumeric code (e.g., E-001, M-005). This code is linked to the company's internal maintenance knowledge base, and each code uniquely corresponds to a detailed set of maintenance procedures, a list of required spare parts, and safety precautions. This standardized coding facilitates recording, retrieval, and management by the computer system, and also allows maintenance personnel to quickly locate and execute repair plans.

[0061] In practice, the first step is to determine the current operating status of the system based on the specific values ​​of the fault attribute vector. A direct and efficient approach is to use a threshold-based rule classification method. This method requires pre-setting one or more confidence thresholds, such as high-confidence thresholds and low-confidence thresholds. Subsequently, faults are categorized using a series of logical judgment rules: Healthy Status: If both the electrical system fault confidence and the mechanical system fault confidence values ​​are lower than the preset low confidence threshold, then the system is considered to be fault-free and in a healthy operating state. Electrical Fault: If the confidence level of an electrical system fault is significantly higher than that of a mechanical system fault, and the electrical system fault confidence level itself exceeds a preset high confidence threshold, while the mechanical system fault confidence level is lower than a preset low confidence threshold, then the system is determined to have experienced an electrical system fault. Mechanical Fault: In contrast to electrical fault, if the confidence level of a mechanical system fault is significantly higher than that of an electrical system fault, and the mechanical system fault confidence level itself exceeds a preset high confidence threshold, while at the same time the electrical system fault confidence level is lower than a preset low confidence threshold, then the system is judged to have experienced a mechanical system fault. Coupled / Uncertain Fault: In other cases, such as when both the electrical system fault confidence and the mechanical system fault confidence values ​​are higher than the preset high confidence threshold, or when both values ​​fall between the low confidence threshold and the high confidence threshold, the system can determine that a coupled fault has occurred where electrical and mechanical systems interact, or that it has entered an uncertain state that cannot be clearly classified. In this case, further manual inspection or activation of more complex diagnostic logic is required.

[0062] Following fault classification, the system needs to convert the classification results into specific fault diagnosis results and maintenance codes. This is typically accomplished through a pre-defined mapping table, which in software implementation can be a lookup table, hash table, or a series of conditional statements. This mapping table associates each fault category with a set of predefined diagnostic text and maintenance codes.

[0063] For example, this mapping table can be represented as: Input category: Health status Diagnostic output: "System is running normally" Output maintenance code: "H-000" Input category: Electrical fault Diagnostic output: "An electrical subsystem anomaly was detected. Please check the motor windings, converter, or current sensor." Output maintenance code: "E-001" Input category: Mechanical failure Diagnostic output: "An anomaly was detected in the mechanical subsystem. Please check the encoder, bearings, or gearbox lubrication." Output maintenance code: "M-001" Input category: Coupling failure Diagnostic output: "An electrical and mechanical coupling fault was detected. A comprehensive inspection of the pitch drive unit is recommended." Output maintenance code: "C-001" Once the fault classification sub-step determines the fault category, this sub-step immediately queries the mapping table, extracts the corresponding text and code, and uses it as the final output.

[0064] Taking the solution of this invention as an example, suppose that at a certain moment, after the calculations of the preceding steps, the fault attribute vector obtained by the system contains an electrical fault confidence score of 0.92 and a mechanical fault confidence score of 0.15. The system's preset classification thresholds are: a high confidence threshold of 0.8 and a low confidence threshold of 0.2. The fault classification process is as follows: The system compares the input confidence score value with the preset thresholds: the electrical system fault confidence score is 0.92, which is greater than the preset high confidence threshold of 0.8; the mechanical system fault confidence score is 0.15, which is less than the preset low confidence threshold of 0.2. This situation satisfies the judgment rule that "the electrical system fault confidence score is greater than the high confidence threshold, and the mechanical system fault confidence score is less than the low confidence threshold," so the system classifies the current fault as "electrical fault." The maintenance instruction generation process is as follows: The system uses the classification result of "electrical fault" as an index to query the internally stored maintenance instruction mapping table. According to the mapping relationship defined above, the system can find the corresponding entry. The final fault diagnosis result is: "An electrical subsystem anomaly detected. Please check the motor windings, converter, or current sensor." The maintenance code output is: "E-001." This result can then be pushed to the wind farm's central monitoring system or displayed directly on the on-site human-machine interface, guiding the operation and maintenance team to carry out precise and efficient maintenance work. In this way, complex dynamic system data is transformed into simple and clear operation and maintenance decisions, forming a complete redundancy and fault-tolerant closed loop.

[0065] In summary, the redundancy and fault-tolerant method for high-power wind turbine pitch drive according to embodiments of the present invention is explained. By constructing parallel electrical and mechanical state observers, the complex electromechanical coupling system is decoupled into two independent analytical dimensions—electrical and mechanical—at the diagnostic level. By comparing and analyzing the electrical and mechanical residuals generated by the two observers, the root cause of the fault can be accurately traced. This enables clear identification of two types of faults—electrical drive failure and increased mechanical resistance—that appear similar but have different root causes. Furthermore, the confidence level of the fault is assessed, and clear maintenance instructions are generated. In this way, the waste of maintenance resources and power generation loss caused by misdiagnosis of faults are avoided, thereby significantly improving the operational reliability and intelligent operation and maintenance level of the high-power wind turbine pitch drive system.

[0066] Furthermore, a redundant fault-tolerant system for the pitch drive of a high-power wind turbine is also provided.

[0067] Figure 3 This is a block diagram of a high-power wind turbine pitch drive redundancy and fault-tolerant system according to an embodiment of the present invention. Figure 3 As shown, the high-power wind turbine pitch drive redundancy fault-tolerant system 300 according to an embodiment of the present invention includes: a data acquisition module 310, used to acquire q-axis voltage command, q-axis actual current, and actual electric angular velocity of the motor; an electrical condition observation module 320, used to perform electrical condition observation based on the q-axis voltage command, actual electric angular velocity of the motor, and actual q-axis current to obtain the q-axis current and electrical residual predicted by the electrical observer; a mechanical condition observation module 330, used to perform mechanical condition observation based on the actual electric angular velocity of the motor and actual q-axis current to obtain the electric angular velocity and mechanical residual predicted by the mechanical observer; a fault confidence assessment module 340, used to perform fault confidence assessment based on the electrical residual and mechanical residual to obtain a fault attribute vector; and a fault classification and maintenance instruction generation module 350, used to perform fault classification and maintenance instruction generation based on the fault attribute vector to obtain fault diagnosis results and maintenance codes.

[0068] Furthermore, the electrical state observation module 320 is specifically used for: constructing a first Luenberger observer model based on the voltage equation of the PMSM; inputting the q-axis voltage command and the actual electric angular velocity of the motor into the Luenberger observer model to obtain the q-axis current predicted by the electrical observer; and calculating the absolute value of the difference between the q-axis current predicted by the electrical observer and the actual q-axis current as the electrical residual.

[0069] Furthermore, the mechanical state observation module 330 is specifically used for: constructing a second Luenberger observer model based on the motor motion equation; inputting the actual q-axis current into the second Luenberger observer model to obtain the electric angular velocity predicted by the mechanical observer; and calculating the absolute value of the difference between the electric angular velocity predicted by the mechanical observer and the actual electric angular velocity of the motor as the mechanical residual.

[0070] As described above, the high-power wind turbine pitch drive redundancy fault-tolerant system 300 according to embodiments of the present invention can be implemented in various wireless terminals, such as servers with high-power wind turbine pitch drive redundancy fault-tolerant algorithms. In one possible implementation, the high-power wind turbine pitch drive redundancy fault-tolerant system 300 according to embodiments of the present invention can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the high-power wind turbine pitch drive redundancy fault-tolerant system 300 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the high-power wind turbine pitch drive redundancy fault-tolerant system 300 can also be one of many hardware modules of the wireless terminal.

[0071] Alternatively, in another example, the high-power wind turbine pitch drive redundancy fault-tolerant system 300 and the wireless terminal can also be separate devices, and the high-power wind turbine pitch drive redundancy fault-tolerant system 300 can be connected to the wireless terminal via wired and / or wireless networks, and transmit interactive information in accordance with an agreed data format.

[0072] 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 redundancy fault-tolerant method for large power wind turbine variable pitch drive, characterized in that, The method comprises the following steps: Obtaining a q-axis voltage instruction, a q-axis actual current and an actual electrical angular velocity of a motor; Performing electrical state observation based on the q-axis voltage instruction, the actual electrical angular velocity of the motor and the q-axis actual current to obtain a q-axis current predicted by an electrical observer and an electrical residual error; Performing mechanical state observation based on the actual electrical angular velocity of the motor and the q-axis actual current to obtain an electrical angular velocity predicted by a mechanical observer and a mechanical residual error; Performing fault confidence evaluation based on the electrical residual error and the mechanical residual error to obtain a fault attribute vector; Performing fault classification and maintenance instruction generation based on the fault attribute vector to obtain a fault diagnosis result and a maintenance code.

2. The redundancy fault tolerant method for pitch drive of high power wind turbine generators as claimed in claim 1 wherein, The electrical state observation based on the q-axis voltage instruction, the actual electrical angular velocity of the motor and the q-axis actual current to obtain the q-axis current predicted by the electrical observer and the electrical residual error comprises the following steps: Constructing a first Luenberger observer model based on a voltage equation of the PMSM; Inputting the q-axis voltage instruction and the actual electrical angular velocity of the motor into the Luenberger observer model to obtain the q-axis current predicted by the electrical observer; Calculating an absolute value of a difference between the q-axis current predicted by the electrical observer and the q-axis actual current as the electrical residual error.

3. The high-power wind turbine variable pitch drive redundancy fault tolerant method of claim 1, wherein, The first Luenberger observer model is expressed by a formula as follows: wherein, is the q-axis current predicted by the electrical observer, is the permanent magnet flux linkage, is the gain of the electrical observer, is the q-axis actual current, is the q-axis voltage command, and is the motor resistance and q-axis inductance in healthy state, is the motor actual electrical angular velocity.

4. The high-power wind turbine variable pitch drive redundancy fault tolerant method of claim 1, wherein, The mechanical state observation based on the actual electrical angular velocity of the motor and the q-axis actual current to obtain the electrical angular velocity predicted by the mechanical observer and the mechanical residual error comprises the following steps: Constructing a second Luenberger observer model based on a motor motion equation; Inputting the q-axis actual current into the second Luenberger observer model to obtain the electrical angular velocity predicted by the mechanical observer; Calculating an absolute value of a difference between the electrical angular velocity predicted by the mechanical observer and the actual electrical angular velocity of the motor as the mechanical residual error.

5. The high-power wind turbine variable pitch drive redundancy fault tolerant method of claim 1, wherein, The second Luenberger observer model is expressed by a formula as follows: wherein, is a torque constant, is a mechanical observer predicted electrical angular velocity, is an external load torque, is a gain of the mechanical observer, and is the total moment of inertia and viscous friction coefficient in the healthy state, is a q-axis actual current, is a motor actual electrical angular velocity.

6. The high-power wind turbine variable pitch drive redundancy fault tolerant method of claim 1, wherein, The fault confidence evaluation based on the electrical residual error and the mechanical residual error to obtain the fault attribute vector comprises the following steps: Performing normalization processing on the electrical residual error and the mechanical residual error to obtain a normalized electrical residual error and a normalized mechanical residual error; Performing sliding window integration on the normalized electrical residual error and the normalized mechanical residual error to obtain an electrical residual error integral value and a mechanical residual error integral value; Performing confidence calculation on the electrical residual error integral value and the mechanical residual error integral value to obtain the fault attribute vector.

7. The redundancy fault tolerant method for pitch drive of high power wind turbine generators as claimed in claim 6 wherein, The confidence calculation on the electrical residual error integral value and the mechanical residual error integral value to obtain the fault attribute vector comprises the following steps: The confidence calculation on the electrical residual error integral value and the mechanical residual error integral value is performed by a formula as follows: wherein, is an electrical residual integral value, is a mechanical residual integral value, is a fault attribute vector.

8. A high-power wind turbine variable pitch drive redundant fault tolerant system, characterized by, The method comprises the following steps: A data acquisition module is configured to obtain a q-axis voltage instruction, a q-axis actual current and an actual electrical angular velocity of a motor; An electrical state observation module is configured to perform electrical state observation based on the q-axis voltage instruction, the actual electrical angular velocity of the motor and the q-axis actual current to obtain a q-axis current predicted by an electrical observer and an electrical residual error; A mechanical state observation module is configured to perform mechanical state observation based on the actual electrical angular velocity of the motor and the q-axis actual current to obtain an electrical angular velocity predicted by a mechanical observer and a mechanical residual error; a fault confidence assessment module, configured to perform fault confidence assessment based on the electrical residual and the mechanical residual to obtain a fault attribute vector; a fault classification and maintenance instruction generation module, configured to perform fault classification and maintenance instruction generation based on the fault attribute vector to obtain a fault diagnosis result and a maintenance code.

9. The high-power wind turbine variable pitch drive redundant fault tolerant system of claim 8, wherein, The electrical state observation module is specifically further configured to: construct a first Luenberger observer model based on a voltage equation of the PMSM; input the q-axis voltage instruction and the actual electrical angular velocity of the motor into the Luenberger observer model to obtain a q-axis current predicted by the electrical observer; calculate an absolute value of a difference between the q-axis current predicted by the electrical observer and an actual q-axis current as the electrical residual.

10. The high-power wind turbine variable pitch drive redundant fault tolerant system of claim 8, wherein, The mechanical state observation module is specifically further configured to: construct a second Luenberger observer model based on a motion equation of the motor; input the actual q-axis current into the second Luenberger observer model to obtain an electrical angular velocity predicted by the mechanical observer; calculate an absolute value of a difference between the electrical angular velocity predicted by the mechanical observer and the actual electrical angular velocity of the motor as the mechanical residual.

Images