An ESO Adaptive Bandwidth Switching Method for Suppressing Torsional Vibration in Wind Turbine Shafts

Abstract

This invention discloses an ESO adaptive bandwidth switching method for suppressing torsional vibration of wind turbine shaft systems, belonging to the field of wind power generation control technology. It constructs an extended state observer architecture with low-bandwidth and high-bandwidth parallel connections. Macroscopic error features reflecting the severity of wind conditions are extracted through low-pass filtering. A residence time locking mechanism and an asymmetric weight update strategy are combined to generate continuous adaptive fusion weights. Finally, the two state estimation vectors are weighted and fused, and disturbance compensation is superimposed to generate a bandwidth adaptive control command output to the wind turbine drivetrain. The beneficial effects of this invention are: it achieves efficient suppression of torsional vibration of the wind turbine shaft system under all operating conditions, ensuring both high-frequency noise suppression capability under stable wind conditions and rapid disturbance tracking capability under gust conditions, avoiding control jitter and switching shocks, significantly reducing fatigue damage to transmission components, and improving the operational reliability and economy of wind turbine units.

Description

Technical Field

[0001] This invention relates to the field of wind power generation control technology, and in particular to an ESO adaptive bandwidth switching method for suppressing torsional vibration of wind turbine shaft systems. Background Technology

[0002] When wind turbine generators operate under complex wind conditions, the physical flexibility of the transmission chain makes them highly susceptible to torsional vibration of the shaft system under the impact of gusts or random high-frequency noise. This vibration not only affects the quality of power output but also accelerates fatigue damage to critical components such as the gearbox and main shaft.

[0003] In existing technologies, Extended State Observers (ESOs) are widely used in torsional vibration suppression of wind turbine shaft systems. This method achieves disturbance compensation control by estimating the system state and total disturbance in real time. However, existing ESOs typically use fixed bandwidth parameters, which leads to an inherent trade-off between disturbance immunity and noise immunity, resulting in the following performance issues: (1) When the observer bandwidth is set to a large value, although the tracking speed of disturbance can be improved, the high-frequency measurement noise will be significantly amplified, and the noise will be introduced into the control loop through the observer estimation channel, resulting in high-frequency jitter of the fan shaft torsional vibration. (2) When the observer bandwidth is small, the system has good high-frequency noise suppression capability, but it is slow to respond to sudden disturbances. Under the condition of sudden wind speed change, it cannot compensate for the disturbance in time, and the wind turbine shaft system will have a large transient response deviation. (3) Existing offline parameter optimization methods are mostly based on specific working conditions. Due to the strong randomness and nonlinear uncertainty of the actual wind field environment, once the real-time working conditions deviate from the preset range, the offline parameters often fail due to the lack of real-time adaptive capability, making it difficult to balance the stability and accuracy of the wind turbine shaft system under all working conditions. Summary of the Invention

[0004] The purpose of this invention is to provide an ESO adaptive bandwidth switching method for suppressing torsional vibration of wind turbine shaft systems. By constructing a parallel structure of dual-bandwidth extended state observers and combining error feature extraction, threshold determination and weight scheduling mechanisms, the adaptive dynamic adjustment of the observer bandwidth is achieved. This solves the problems of existing fixed ESO bandwidth, which makes it difficult to balance anti-disturbance performance and noise suppression performance, offline parameter tuning methods, and frequent switching in multimodal control strategies.

[0005] The objective of this invention is achieved through the following technical solution: An ESO adaptive bandwidth switching method for suppressing torsional vibration of wind turbine shaft systems includes the following steps: S1. Construct a dual-bandwidth parallel extended state observer architecture: In the torsional vibration suppression control loop of the wind turbine shaft system, deploy a low-bandwidth extended state observer for steady-state suppression of high-frequency measurement noise and a high-bandwidth extended state observer for transient rapid tracking of sudden disturbances. The low-bandwidth extended state observer and the high-bandwidth extended state observer share the same control input signal and sensor feedback error signal, and perform state estimation independently, and output state estimation vectors respectively. To address the inherent contradiction between high-frequency measurement noise amplification and transient disturbance rejection hysteresis in wind turbine drivetrains under complex wind conditions, a problem that traditional single-bandwidth observers cannot simultaneously solve, a dual-bandwidth parallel architecture is adopted. This architecture decouples noise suppression and disturbance rejection by using two independently operating extended state observers connected in parallel in the control loop to form a dual-bandwidth backup architecture. This results in no significant physical delay, thus providing real-time state data support for subsequent adaptive dynamic switching of bandwidth based on error characteristics.

[0006] S2. Wind condition feature extraction: Low-pass filtering is performed on the sensor feedback error signal to extract macroscopic error features that reflect the severity of wind conditions; In the actual operating environment of wind turbine generators, the real-time error signal fed back by the system often exhibits complex frequency band coupling characteristics, including both high-frequency white noise and low-frequency deviations caused by sudden changes such as gusts. If the original error signal is directly used for subsequent controller bandwidth adaptive switching, the switching logic is prone to frequent false triggering due to high-frequency noise interference, which in turn exacerbates mechanical fatigue and torsional oscillations in the drivetrain. To avoid false triggering of the switching logic by conventional high-frequency noise, this application introduces a low-pass filter to filter the real-time physical feedback error, extracts the macroscopic error value reflecting wind conditions, and configures an appropriate filter weight coefficient. While filtering out microscopic high-frequency white noise to prevent false triggering, it also ensures that phase lag is eliminated when sudden strong winds occur, enabling rapid perception of error changes.

[0007] S3. Logical Judgment and Weight Allocation: Compare the macroscopic error features extracted in step S2 with the preset threshold for sudden wind changes, trigger dwell time locking based on the comparison result and output the target fusion weight; In actual operation, since gusts or aerodynamic disturbances are not ideal step signals, the feedback errors they cause often fluctuate around the critical point. If only a single amplitude threshold is used for observer state switching, frequent jumps in the target fusion weights will inevitably occur when the error signal hovers around the threshold. This high-frequency control state switching will directly result in severe high-frequency chattering of the electromagnetic torque, which not only fails to effectively resist disturbances but also easily induces low-frequency torsional resonance in the wind turbine drivetrain, causing severe mechanical fatigue damage to the gearbox and generator main shaft. Therefore, this application sets a wind condition change detection threshold. When the macroscopic error value is slightly higher than the normal background noise, it is determined to be a gust of wind, and the state switching command is immediately activated and a dwell time self-locking timer is started. During the lockout period, the system forcibly points the target fusion weights to a high-bandwidth state to avoid frequent switching caused by the fluctuation of the feedback signal around the threshold.

[0008] S4. Asymmetric adaptive bandwidth switching: Based on the target fusion weights output in step S3, an asymmetric update strategy is used to generate continuous adaptive fusion weights. Under gust conditions, the system quickly switches to the high-bandwidth extended state observer, and under stable conditions, it smoothly switches back to the low-bandwidth extended state observer based on the set recovery time constant. Establish actual adaptive fusion weights for the dual-bandwidth observer state, and transform the discrete instructions output by the front-end modal locking logic into continuous and safe control weights to avoid secondary impacts on the transmission chain caused by sudden electromagnetic torque changes.

[0009] S5. State Fusion Reconstruction and Control Output: Based on the adaptive fusion weights output in step S4, the state estimation vector output in step S1 is weighted and fused to obtain the reconstructed comprehensive state vector. Then, the comprehensive state vector is sent to the state feedback controller and disturbance compensation is superimposed to generate an adaptive bandwidth control command and output to the wind turbine drive train.

[0010] In conventional variable-gain adaptive control, directly and dynamically modifying the bandwidth parameter within a single observer can cause numerical integral jumps in the state estimate at the moment of switching, leading to abrupt changes in the control command. This application innovatively employs a state fusion and reconstruction mechanism at the control execution end. Since the two parallel-running observers at the underlying level maintain independent continuous integration operations without any parameter jumps within themselves, their parallel output state variables are directly weighted and fused. This mathematically eliminates state jumps, achieving smooth, disturbance-free switching. Furthermore, in step S2, the low-pass filter adopts a discrete-time first-order low-pass filter, and its discrete-domain update formula is:

[0011]

[0012] In the formula, For the first Real-time physical feedback error per sampling period This is the current time's filter error value. This is the filtering error value from the previous time step. These are the filter weight coefficients. The extracted macroscopic error absolute characteristic value.

[0013] Furthermore, in step S3, the logical judgment formula for comparing the macroscopic error characteristics with the preset threshold for sudden wind changes is as follows:

[0014]

[0015] In the formula, This is the alarm threshold. To lock in the time for the gusts to linger, To control the sampling step size, This is the remaining time in the countdown to the current moment. This is the remaining time in the countdown from the previous moment. Weigh the target state together; when for The time refers to the high-bandwidth extended state observer, when for The time refers to the low-bandwidth extended state observer.

[0016] Furthermore, in step S4, the update formula for the asymmetric update strategy is:

[0017] In the formula, The actual adaptive fusion weights at the current moment. The actual adaptive fusion weights from the previous time step. The weighted recovery time constant under stable wind conditions; When the target state fusion weight is less than the actual adaptive fusion weight of the previous time step, quickly switch to the high-bandwidth extended state observer; when the target state fusion weight is greater than or equal to the actual adaptive fusion weight of the previous time step, smoothly switch back to the low-bandwidth extended state observer according to the weight recovery time constant.

[0018] Furthermore, in step S5, the formula for calculating the reconstructed integrated state vector is:

[0019] In the formula, This is the three-dimensional state estimation vector output by the low-bandwidth extended state observer at the current time. This is the three-dimensional state estimation vector output by the high-bandwidth extended state observer at the current time. This is the reconstructed integrated state vector.

[0020] Furthermore, in step S5, the expanded form of the reconstructed integrated state vector is: ,in, This is the estimated vector of the basic state of the reconstructed system. This is the estimated total disturbance for the system extension; The formula for calculating the adaptive bandwidth control command is:

[0021] In the formula, K is the state feedback gain matrix calculated by pole placement. This is the control gain compensation coefficient of the system; This is an adaptive bandwidth control command.

[0022] The present invention has the following advantages: 1. By constructing a parallel structure of low-bandwidth extended state observers and high-bandwidth extended state observers, adaptive scheduling of observer bandwidth in the time domain is achieved. Under stable wind conditions, the system is dominated by the low-bandwidth extended state observer, effectively suppressing high-frequency sensor noise, reducing high-frequency vibration of the wind turbine shaft system, and helping to extend the mechanical fatigue life of the wind turbine gearbox and main shaft. Under the influence of sudden gusts and other disturbances, the system switches to high-bandwidth observer dominance, improving the ability to track disturbances and thus reducing transient response deviation. Through the above mechanism, the system's anti-disturbance performance and noise suppression performance can be balanced under different operating conditions.

[0023] 2. A dual-bandwidth observer parallel operation structure is adopted, combined with a threshold determination and weight scheduling mechanism based on error characteristics, to overcome the limitations of offline parameter optimization and achieve real-time adaptive adjustment to wind condition changes. By extracting macroscopic feature quantities from the error signal and performing mode determination and asymmetric weight updates based on these features, the system can adaptively switch control strategies according to wind condition changes. It can quickly switch under gust conditions to ensure disturbance rejection speed, and linearly recover weights according to the time constant under stable conditions, avoiding secondary mechanical shocks to the drivetrain caused by sudden changes in control commands. This improves the system's adaptability and operational stability under complex wind conditions, solving the problems of fixed parameters, poor environmental adaptability, and inability to cope with unforeseen extreme gusts inherent in conventional offline parameter optimization strategies.

[0024] 3. The bandwidth switching is achieved by using a weighted fusion method of dual observer output states. The two bottom-level observers always maintain independent and continuous integral operations, which completely blocks the jump of state estimate values ​​from a mathematical mechanism, realizes the smooth transition of control mode without disturbance, can be directly embedded into the existing wind turbine main control system, can effectively suppress shaft torsional vibration, significantly reduce fatigue damage to core transmission components such as gearbox and main shaft, extend the service life of the unit, and greatly reduce the overall operation and maintenance cost of wind farm. Attached Figure Description

[0025] Figure 1 This is a control principle diagram of the present invention.

[0026] Figure 2 This is a schematic diagram of the process of the present invention.

[0027] Figure 3 This is the time-domain response diagram of the present invention under high-frequency random noise input in noisy conditions.

[0028] Figure 4 This is a timing waveform diagram of the asymmetric fusion weight dynamic scheduling under the sudden gust of wind, according to the present invention.

[0029] Figure 5 The waveform diagram shows the comparative verification of the transmission chain disturbance rejection performance of the adaptive switching method and the single bandwidth observer control method of this invention when facing extreme step gust impacts. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0031] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0032] It should be noted that, unless otherwise specified, the embodiments and features described in this invention can be combined with each other.

[0033] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0034] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of this invention is in use, or the orientation or positional relationship commonly understood by those skilled in the art. They are only used for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention. In addition, the terms "first," "second," etc., are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0035] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0036] refer to Figure 1 As shown in Figure 2, one embodiment of the present invention is as follows: An ESO adaptive bandwidth switching method for suppressing torsional vibration of wind turbine shaft systems includes the following steps: S1. Construct a dual-bandwidth parallel extended state observer architecture: In the torsional vibration suppression control loop of the wind turbine shaft system, deploy a low-bandwidth extended state observer for steady-state suppression of high-frequency measurement noise and a high-bandwidth extended state observer for transient rapid tracking of sudden disturbances. The low-bandwidth extended state observer and the high-bandwidth extended state observer share the same control input signal and sensor feedback error signal, and perform state estimation independently, and output state estimation vectors respectively.

[0037] S2. Wind condition feature extraction: Low-pass filtering is performed on the sensor feedback error signal to extract macroscopic error features that reflect the severity of wind conditions; The low-pass filter uses a discrete-time first-order low-pass filter, and its discrete-domain update formula is as follows:

[0038]

[0039] In the formula, For the first Real-time physical feedback error per sampling period This is the current time's filter error value. This is the filtering error value from the previous time step. Here, 'a' represents the filter weighting coefficient; in this embodiment, 'a' is 0.05 to achieve the optimal balance between noise immunity and response sensitivity. The extracted macroscopic error absolute characteristic value.

[0040] Macro error index Eliminating invalid disturbances provides a reliable basis for accurately determining whether the system needs to switch to high-bandwidth anti-disturbance modes, achieving synergy between the agility of error perception and the stability of control decisions.

[0041] S3. Logical Judgment and Weight Allocation: The macroscopic error features extracted in step S2 are compared with a preset threshold for sudden wind changes. Based on the comparison result, the dwell time is locked, and the target fusion weight is output. The logical judgment formula for comparing the macroscopic error features with the preset threshold for sudden wind changes is as follows: ,

[0042]

[0043] In the formula, The alarm threshold value is set to 0.008 in this embodiment. The gust dwell time is set to 4.0 seconds in this embodiment. In this embodiment, to control the sampling step size: Second, This is the remaining time in the countdown to the current moment. This is the remaining time in the countdown from the previous moment. Weigh the target state together; when for The time refers to the high-bandwidth extended state observer, when it is The time refers to the low-bandwidth extended state observer.

[0044] S4. Asymmetric Adaptive Bandwidth Switching: Based on the target fusion weights output in step S3, an asymmetric update strategy is used to generate continuous adaptive fusion weights. Under gust conditions, the system quickly switches to the high-bandwidth extended state observer; under stable conditions, it smoothly switches back to the low-bandwidth extended state observer based on a set recovery time constant. The update formula for the asymmetric update strategy is:

[0045] In the formula, The actual adaptive fusion weights at the current moment. The actual adaptive fusion weights from the previous time step. The weighted recovery time constant under stable wind conditions is set to 1 second in this embodiment; When the target state fusion weight is less than the actual adaptive fusion weight of the previous moment, a rapid switch to the high-bandwidth extended state observer is implemented. When the target state fusion weight is greater than or equal to the actual adaptive fusion weight of the previous moment, a smooth switch back to the low-bandwidth extended state observer is implemented according to the weight recovery time constant. Specifically, a rapid switching strategy is implemented under gusty wind conditions. When a strong wind alarm is triggered, a rapid switch is completed in a very short time, transferring control to the high-bandwidth observer to achieve rapid disturbance rejection and reduce system transient overshoot. Under stable wind conditions, a smooth recovery is implemented. When the dwell time ends and the macroscopic error falls back to be determined as a stable wind condition, the fusion weight is linearly reduced based on the set recovery time constant to achieve a smooth transition to the low-bandwidth observer and avoid secondary impact on the drivetrain caused by sudden electromagnetic torque changes.

[0046] S5. State Fusion Reconstruction and Control Output: Based on the adaptive fusion weights output in step S4, the state estimation vector output in step S1 is weighted and fused to obtain the reconstructed comprehensive state vector. This comprehensive state vector is then fed into the state feedback controller and disturbance compensation is superimposed to generate an adaptive bandwidth control command, which is output to the wind turbine drivetrain. Through deep physical collaboration with the aforementioned wind condition perception and weight switching mechanism, the state variables independently output by the low-bandwidth and high-bandwidth observers are directly weighted and fused using the actual adaptive fusion weights at the current moment. The reconstructed fused state is fed into the state feedback controller to separate the basic state vector and the extended total disturbance estimate. The system's basic state estimation vector is multiplied by the state feedback gain matrix to form the basic feedback control law, and the system's extended total disturbance estimate is transformed into a feedforward compensation law. Combined with the control gain compensation coefficient, the final bandwidth adaptive control command is generated. Under complex alternating wind conditions, it can adaptively output reverse compensation torque, achieving dual optimization of transient disturbance immunity and steady-state noise reduction in the drivetrain.

[0047] The formula for calculating the reconstructed integrated state vector is as follows:

[0048] In the formula, This is the three-dimensional state estimation vector output by the low-bandwidth extended state observer at the current time. This is the three-dimensional state estimation vector output by the high-bandwidth extended state observer at the current time. The reconstructed integrated state vector has the following expanded form: ,in, This is the estimated vector of the basic state of the reconstructed system. This is the estimated total disturbance for the system extension.

[0049] The formula for calculating the adaptive bandwidth control command is:

[0050] In the formula, K is the state feedback gain matrix calculated by pole placement. This is the control gain compensation coefficient of the system; This is an adaptive bandwidth control command. For the state feedback gain matrix K, firstly, based on the target damping ratio required for torsional vibration attenuation of the wind turbine drive shaft system and the desired closed-loop response bandwidth, the desired dominant pole is selected in the left half-plane of the complex plane. Then, combined with the dynamic state equation of the drive train, the state feedback gain matrix that can place the closed-loop system poles to the target positions is solved by the pole placement method.

[0051] The method of this invention was compared with methods using a single high-bandwidth observer and a single low-bandwidth observer through simulation. A nonlinear dynamic model of the wind turbine generator drivetrain was built in the Simulink simulation platform, and high-frequency white noise was continuously injected into the sensor feedback loop to simulate a real industrial measurement environment. Subsequently, to test the robustness of the system under extreme conditions, an extreme step gust of wind impact disturbance signal was introduced as the core test scenario at a set time point in the simulation, and the torsional angle of the wind turbine shaft system was used as the evaluation index.

[0052] To realistically simulate the measurement interference of the transmission chain sensors during actual operation, a measurement noise simulation environment is set up, with the system measurement noise source being... Figure 3 The band-limited white noise shown is precisely set to its noise power. The sampling time was configured to 0.001 seconds, and the random number seed was fixed at 23341. This parameter configuration aims to ensure that experiments are conducted under a uniform high-frequency observation noise benchmark, guaranteeing the repeatability and comparability of the simulation experiments.

[0053] like Figure 4 , 5 As shown, the deep noise suppression operation mode under stable wind conditions: the 28-30 seconds of the simulation run is the stable wind period, and the macroscopic error extracted by the system is... It consistently fluctuates below the 0.008 threshold, countdown timer. ,Depend on Figure 4 It can be seen that the actual adaptive fusion weights Maintaining a value of 1. At this point, the system is entirely controlled by the low-bandwidth extended state observer, combined with... Figure 5 As can be seen, the torsional angle of the wind turbine shaft is relatively smooth at this time, indicating that the present invention effectively suppresses high-frequency sensor noise under stable wind conditions, thereby reducing gearbox wear. Simultaneously, the high-bandwidth extended state observer maintains synchronous operation to estimate system disturbances in real time, thus ensuring the continuity and consistency of disturbance estimation during observer switching.

[0054] From the perspective of steady-state control accuracy, during the simulated 28-30 second period of stable wind conditions, the single high-bandwidth observer is affected by the high-frequency measurement noise of the sensor, resulting in obvious high-frequency chattering of the torsion angle. However, the method of this invention extracts the macroscopic error feature value through a discrete-time first-order low-pass filter and determines that its value is lower than the set wind condition change detection threshold. Therefore, the asymmetric weighting logic of the system strictly locks the adaptive fusion weight to 1, so that the control loop is completely dominated by the low-bandwidth observer. Relying on its excellent low-pass filtering characteristics, it effectively filters out high-frequency sensor noise, making the output control command smoother. Its steady-state noise suppression effect is basically the same as that of the single low-bandwidth observer.

[0055] Rapid switching of operating modes under gust disturbance: A step gust disturbance is introduced at the 30-second mark of the simulation. The system error increases rapidly, and the filtering error... The threshold of 0.008 is immediately exceeded. At this point, the system triggers a switching mechanism, resetting the dwell time to 4.0 seconds. Figure 4 Actual adaptive fusion weights The system rapidly decreases from 1 to 0, thereby switching control to the high-bandwidth extended state observer. Due to the high-bandwidth extended state observer's faster dynamic response capability, the system's response speed to disturbances is significantly improved. Through... Figure 5 It can be observed that the transient overshoot value obtained by the method of the present invention is about 0.83 deg, which is significantly lower than about 0.90 deg of a single low-bandwidth extended state observer, effectively reducing the transient overshoot value of the transmission chain torsion angle.

[0056] From the perspective of transient immunity limits, when an extreme step gust of wind is introduced at the 30th second, a single low-bandwidth observer, due to its slow response speed, experiences an overshoot peak of approximately 0.90 degrees in the shaft torsion angle. However, this invention, at the instant the gust enters, rapidly exceeds the wind condition change detection threshold by extracting macroscopic errors. At this moment, the system's adaptive fusion weights instantly abruptly change to 0, and a dwell time self-locking timer is activated. During the lockout period, the system forcibly directs the target fusion weights to a high-bandwidth state. This method enables the system to switch to high-bandwidth observer dominance in a timely manner, controlling the overshoot amplitude to approximately 0.83 degrees, achieving an immunity performance close to that of a single high-bandwidth observer.

[0057] Smooth switching mode after shaft torsion angle reduction: when the dwell time ends and When the value falls below the 0.008 threshold again, the system triggers the weight recovery mechanism. Figure 4 Actual adaptive fusion weights According to the set time constant The control variable is gradually restored from 0 to 1, ensuring continuity during state transitions and enabling a smooth transition of control from the high-bandwidth extended state observer to the low-bandwidth extended state observer. Figure 5It can be seen that at 44s, the method of the present invention has the same noise suppression performance as the low-bandwidth observer, which reduces the frequency of the torsional angular vibration of the wind turbine shaft system and realizes a smooth transition from the high-bandwidth observer to the low-bandwidth observer. This process ensures the continuity of control commands, and the system re-enters the steady-state operation period of noise suppression.

[0058] From the perspective of the smooth transition from interference immunity to noise suppression, observe Figure 5 Within the magnified region of 38–50 seconds, it can be observed that around the 44th second, as the shaft torsion angle gradually decreases, the system enters a new steady-state process. During this process, this method introduces a dwell time constraint to prevent switching from being triggered just below the threshold, thereby suppressing misjudgments and high-frequency chattering caused by measurement noise near the threshold. Simultaneously, a linear ramp function based on the time constant is used to continuously recover the fused weights, avoiding control input shocks caused by abrupt weight changes. This ensures the continuity and smoothness of the active damping control quantity during the handover process, enabling the system to transition smoothly to the low-bandwidth observer-dominated state without significant disturbance after completing transient disturbance rejection. The controller's output waveform smoothly converges from a state containing significant high-frequency measurement noise to a noise-free, smooth state. This intuitively verifies that the system can smoothly switch to the low-bandwidth observer-dominated mode after completing transient disturbance rejection, achieving the recovery of steady-state noise suppression performance; moreover, the transition process is continuous, without significant oscillations or shocks.

[0059] The following table analyzes the advantages of improved transmission chain reliability and full-scenario operation and maintenance:

[0060] As can be seen from the table above: In terms of steady-state noise suppression and fatigue mitigation: a single high-bandwidth observer cannot effectively filter out high-frequency measurement noise, resulting in a noise suppression index as high as 9.8491 × 10⁻⁶. -4 This results in extremely high steady-state fatigue wear of the transmission chain; however, the method of this invention, by adaptively switching to a low-bandwidth mode under stable wind conditions, successfully suppresses the noise reduction index to 4.4722×10. -4 This is comparable to that of a single low-bandwidth controller. This data directly demonstrates that the invention effectively suppresses high-frequency noise, fundamentally ensuring extremely low steady-state fatigue wear of the transmission chain.

[0061] In terms of transient disturbance immunity and fault prevention: When faced with sudden gusts, the single low-bandwidth observer suffers a significant degradation in its disturbance immunity index to 4.5198×10 due to phase lag and sluggish response. -2 This significantly increases the risk of transient gearbox fracture. In contrast, the dynamic switching mechanism of this invention can rapidly tangentially contact the high-bandwidth observer at the moment of impact, optimizing the disturbance rejection index by several times to 1.7568×10⁻⁶. -2The disturbance rejection performance is similar to that of a high-bandwidth observer. This significant improvement demonstrates the system's strong and rapid suppression capability under extreme conditions, successfully reducing the risk of transient gearbox failure.

[0062] Overall Operation and Maintenance and Economic Benefits: After introducing the adaptive dynamic switching strategy, the system shows improvements in terms of steady-state fatigue wear of the transmission chain, risk of transient gearbox fracture, and expected overall operation and maintenance costs. This performance improvement mainly stems from the system's adaptive dynamic switching of control over high-bandwidth and low-bandwidth observers. Compared to a single-bandwidth observer, the method of this invention improves the system's response capability under disturbances while ensuring steady-state noise suppression performance, thereby reducing the risk of long-term fatigue damage to the mechanical structure to a certain extent and contributing to improving the operational reliability and economy of the wind power generation system.

[0063] The above simulation comparison shows that the present invention achieves coordinated optimization of disturbance suppression performance and noise suppression performance through adaptive dynamic allocation of control power, thereby effectively alleviating the trade-off between transient disturbance rejection performance and steady-state noise suppression performance in traditional linear control methods. Compared with a single low-bandwidth observer and a single high-bandwidth observer, it has better overall performance and improves the stability and reliability of wind turbine operation.

[0064] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An ESO adaptive bandwidth switching method for suppressing torsional vibration of wind turbine shaft systems, characterized in that: Includes the following steps: S1. Construct a dual-bandwidth parallel extended state observer architecture: In the torsional vibration suppression control loop of the wind turbine shaft system, deploy a low-bandwidth extended state observer for steady-state suppression of high-frequency measurement noise and a high-bandwidth extended state observer for transient rapid tracking of sudden disturbances. The low-bandwidth extended state observer and the high-bandwidth extended state observer share the same control input signal and sensor feedback error signal, and perform state estimation independently, and output state estimation vectors respectively. S2. Wind condition feature extraction: Low-pass filtering is performed on the sensor feedback error signal to extract macroscopic error features that reflect the severity of wind conditions; S3. Logical Judgment and Weight Allocation: Compare the macroscopic error features extracted in step S2 with the preset threshold for sudden wind changes, trigger dwell time locking based on the comparison result and output the target fusion weight; S4. Asymmetric adaptive bandwidth switching: Based on the target fusion weights output in step S3, an asymmetric update strategy is used to generate continuous adaptive fusion weights. Under gust conditions, the system quickly switches to the high-bandwidth extended state observer, and under stable conditions, it smoothly switches back to the low-bandwidth extended state observer based on the set recovery time constant. S5. State Fusion Reconstruction and Control Output: Based on the adaptive fusion weights output in step S4, the state estimation vector output in step S1 is weighted and fused to obtain the reconstructed comprehensive state vector. Then, the comprehensive state vector is sent to the state feedback controller and disturbance compensation is superimposed to generate an adaptive bandwidth control command and output to the wind turbine drive train.

2. The ESO adaptive bandwidth switching method for suppressing torsional vibration of a wind turbine shaft system according to claim 1, characterized in that: In step S2, the low-pass filter adopts a discrete-time first-order low-pass filter, and its discrete-domain update formula is as follows: , In the formula, For the first Real-time physical feedback error per sampling period This is the current time's filter error value. This represents the filtering error value from the previous time step. These are the filter weight coefficients. The extracted macroscopic error absolute characteristic value.

3. The ESO adaptive bandwidth switching method for suppressing torsional vibration of a wind turbine shaft system according to claim 2, characterized in that: In step S3, the logical judgment formula for comparing the macroscopic error characteristics with the preset threshold for sudden wind changes is as follows: , In the formula, This is the alarm threshold. To lock in the time for the gusts to linger, To control the sampling step size, This is the remaining time in the countdown to the current moment. This is the remaining time in the countdown from the previous moment. Weigh the target state together; when for The time refers to the high-bandwidth extended state observer, when for The time refers to the low-bandwidth extended state observer.

4. The ESO adaptive bandwidth switching method for suppressing torsional vibration of a wind turbine shaft system according to claim 3, characterized in that: In step S4, the update formula for the asymmetric update strategy is: In the formula, The actual adaptive fusion weights at the current moment. The actual adaptive fusion weights from the previous time step. The weighted recovery time constant under stable wind conditions; When the target state fusion weight is less than the actual adaptive fusion weight of the previous time step, quickly switch to the high-bandwidth extended state observer; when the target state fusion weight is greater than or equal to the actual adaptive fusion weight of the previous time step, smoothly switch back to the low-bandwidth extended state observer according to the weight recovery time constant.

5. The ESO adaptive bandwidth switching method for suppressing torsional vibration of a wind turbine shaft system according to claim 4, characterized in that: In step S5, the formula for calculating the reconstructed integrated state vector is: In the formula, This is the three-dimensional state estimation vector output by the low-bandwidth extended state observer at the current time. This is the three-dimensional state estimation vector output by the high-bandwidth extended state observer at the current time. This is the reconstructed integrated state vector.

6. The ESO adaptive bandwidth switching method for suppressing torsional vibration of a wind turbine shaft system according to claim 5, characterized in that: In step S5, the expanded form of the reconstructed comprehensive state vector is as follows: ,in, This is the estimated vector of the basic state of the reconstructed system. This is the estimated total disturbance for the system extension; The formula for calculating the adaptive bandwidth control command is: In the formula, K is the state feedback gain matrix calculated by pole placement. This is the control gain compensation coefficient of the system; This is an adaptive bandwidth control command.

Images