A wind turbine pitch control method based on load prediction feedforward compensation

Abstract

This invention belongs to the field of wind power generation control technology and relates to a wind turbine pitch control method based on load prediction and feedforward compensation. It first analyzes the spectrum of wind speed and load signals, and distinguishes between periodic load interference and instantaneous load interference based on discrete peak detection, cross-power spectrum, and energy ratio. Then, it analyzes the lateral energy ratio of the wind turbine shaft and the multi-band energy distribution to determine the direction of periodic load interference or the amplitude of the intensity change of instantaneous load interference. Based on this, corresponding feedforward compensation signals are generated: for periodic load interference, a synchronously reversed sinusoidal compensation signal is generated; for instantaneous load interference, a filtered compensation signal proportional to its intensity change amplitude is generated. Finally, the feedforward signal, after amplitude and rate limiting processing, is fused with the power feedback signal to output an integrated pitch angle command. This invention improves the active suppression capability of the pitch system for complex loads by distinguishing and compensating for different load types.

Description

Technical Field

[0001] This invention belongs to the field of wind power generation control technology, specifically a wind turbine pitch control method based on load prediction feedforward compensation. Background Technology

[0002] As an important clean energy source, wind power generation has seen its units become larger, which has led to prominent dynamic mechanical load issues and seriously threatened the fatigue life of components. Variable pitch control is the key to achieving power regulation and load suppression. By adjusting the blade pitch angle to change the aerodynamic angle of attack, it is possible not only to control the generator output power to be maintained near the rated value, but also to effectively reduce the impact of extreme wind conditions on the structure.

[0003] In the existing technology, there are also some related solutions involving pitch control methods for wind turbine generator sets. For example, a pitch control method for a doubly fed wind turbine generator set, published in Chinese Patent Publication No. CN103527405A, provides a suitable feedforward pitch angle through a feedforward control strategy based on speed deviation. This angle is then added to the output of the feedback PI controller as the set point for the pitch angle, which simplifies the feedforward control system and reduces the requirements for feedforward control accuracy.

[0004] Another Chinese patent publication number CN116292080A discloses a method and apparatus for variable pitch control of a wind turbine generator set. It identifies the current wind conditions, obtains the actual pitch angle and the target minimum pitch angle of the wind turbine generator set, and controls the operation of the wind turbine generator set based on the minimum pitch angle, thereby improving the effect of variable pitch control.

[0005] However, the existing technology still has the following limitations: it fails to effectively distinguish between periodic load interference and instantaneous load interference, resulting in lag in control response and insufficient compensation accuracy. Furthermore, if the corresponding characteristics appear in the wind speed spectrum and load spectrum, and cannot be accurately identified and compensated accordingly, it will affect the operating efficiency and structural safety of the wind turbine. Summary of the Invention

[0006] To overcome the shortcomings of the prior art, this invention provides a wind turbine pitch control method based on load prediction feedforward compensation, which can effectively solve the problems mentioned in the prior art.

[0007] The objective of this invention can be achieved through the following technical solution: a wind turbine pitch control method based on load prediction feedforward compensation, comprising: if there is a discrete peak in the wind speed spectrum that is higher than the background, and at the discrete peak frequency, the load spectrum and the wind speed spectrum satisfy the periodic correlation condition, then the type of disturbance currently experienced by the wind turbine is determined to be periodic load disturbance; otherwise, it is determined to be instantaneous load disturbance.

[0008] The lateral energy ratio and multi-frequency band energy distribution of the wind turbine shaft are analyzed to confirm the specific characteristics of the identified interference types. The specific characteristics of the periodic load interference are the interference direction and the specific characteristics of the instantaneous load interference are the amplitude of the interference intensity change.

[0009] Based on the type of interference and its specific characteristics, a corresponding feedforward compensation signal is generated. The feedforward compensation process includes: if it is periodic load interference, a synchronous reverse compensation signal is generated; if it is instantaneous load interference, a filtered compensation signal proportional to the amplitude of the interference intensity change is generated.

[0010] After the feedforward compensation signal is processed by amplitude and rate limiting, it is fused with the power feedback control signal to output an integrated pitch angle command.

[0011] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) The present invention can accurately distinguish between periodic load interference and instantaneous load interference by combining the joint analysis of wind speed spectrum and load spectrum, combined with discrete peak detection, cross power spectrum and energy ratio determination. For periodic interference, a synchronous reverse compensation signal is generated to achieve phase cancellation; for instantaneous interference, a filter compensation signal proportional to the amplitude of the interference intensity change is generated.

[0012] (2) The present invention integrates the feedforward compensation signal with the power feedback control signal after the feedforward compensation signal is processed by limiting the amplitude and limiting the rate, thereby avoiding over-adjustment and oscillation, ensuring the smoothness and feasibility of the pitch angle command, and thus improving the stability of power generation. Attached Figure Description

[0013] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0014] Figure 1 This is a flowchart of the method of the present invention.

[0015] Figure 2 This is the logic diagram for axial and lateral judgment of periodic load interference in this invention.

[0016] Figure 3 This is a flowchart of the load spectrum partitioning process of the present invention. Detailed Implementation

[0017] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Reference Figure 1 As shown, the present invention provides a wind turbine pitch control method based on load prediction feedforward compensation, including: S1. If there is a discrete peak in the wind speed spectrum that is higher than the background, and at the discrete peak frequency, the load spectrum and the wind speed spectrum satisfy the periodic correlation condition, then the type of disturbance currently experienced by the wind turbine is determined to be periodic load disturbance; otherwise, it is determined to be instantaneous load disturbance.

[0019] Considering that in variable pitch control, periodic disturbances manifest as specific discrete frequency peaks in the wind speed spectrum and can excite the load to oscillate at the corresponding frequency; while instantaneous disturbances manifest as broadband energy impacts with no significant discrete peaks in the wind speed spectrum, it is difficult to achieve corresponding suppression of the load if a single control strategy is adopted.

[0020] Wind speed and wind turbine load signals are acquired by an anemometer installed on the top of the wind turbine nacelle and a load sensor installed in the front flange connection area of ​​the wind turbine main shaft or near the main shaft bearing. Within a preset time window, the time domain signal sequences of the acquired wind speed and load signals are preprocessed by detrending and windowing, respectively.

[0021] The detrending and windowing preprocessing specifically involves: using the first-order polynomial least squares method to fit the linear equations of the wind speed signal and the load signal respectively; subtracting the value of the fitted linear equation at the corresponding time from each data point in the corresponding time-domain signal sequence to obtain the signal sequence after removing the linear trend component.

[0022] The signal sequences after removing the linear trend components are multiplied by the Hanning window to achieve windowing preprocessing.

[0023] The determination of the Hanning window is achieved through the following steps: obtaining the preset time window duration and the system sampling frequency, and taking the integer part of their product as the total number of sampling points of the signal sequence after removing the linear trend component.

[0024] Based on the total number of sampling points, construct a length of A sequence of discrete window functions, wherein the window function is a Hanning window, and its construction method is as follows: according to the order from the beginning to the end of the sequence, through... The cosine operation is performed on each sequence index point in turn. Calculate the coefficient value of a window function and substitute the result into the standard mathematical formula of the Hanning window function to obtain the Hanning window, which exhibits a symmetrical, continuous and smooth decreasing distribution from the center of the sequence to both ends, and finally approaches zero at the beginning and end of the sequence.

[0025] Using the Hanning window for windowing preprocessing can effectively reduce spurious high-frequency components caused by time-domain signal truncation, thus enabling the subsequently calculated spectrum to more accurately reflect the frequency composition of the signal itself. This provides a basis for accurately identifying discrete peaks in wind speed and load signals.

[0026] Fast Fourier Transform is performed on each time-domain signal after windowing preprocessing to obtain the corresponding complex spectrum.

[0027] Calculate the square of the spectral modulus at each frequency point in the complex spectrum, and then divide it by the length of the signal sequence to obtain the corresponding wind speed spectrum and load spectrum.

[0028] Calculate the mean and standard deviation of the wind speed spectrum amplitude, and identify spectral peaks with amplitudes greater than the sum of the mean and three times the standard deviation as discrete peaks significantly higher than the background noise.

[0029] For each identified discrete peak, the ratio of the vibration energy in the narrow band centered on the discrete peak frequency to the total broadband vibration energy is calculated in the load spectrum and denoted as the peak energy ratio.

[0030] The specific process for determining the vibration energy within the narrow band is as follows: the frequency range of the narrow band is determined by centering on each significant discrete peak frequency value identified in the wind speed spectrum, and the upper and lower boundary frequencies of the narrow band are obtained by adding and subtracting a preset half-width value from the discrete peak frequency value.

[0031] When multiple significant discrete peaks are identified, a narrow band with a different center frequency is defined for each peak. The preset half-width value needs to meet the frequency resolution obtained by the spectrum analysis. For example, the preset half-width value can be set to 2 to 5 times the frequency resolution.

[0032] For each narrowband, extract all spectral lines whose frequency values ​​fall within the frequency range of that narrowband from the load spectrum, and sum the power spectral density values ​​corresponding to each of these spectral lines. The sum obtained from this summation is the vibrational energy within the narrowband centered on the discrete peak frequency.

[0033] The broadband total vibration energy is obtained by summing the power spectral density values ​​corresponding to all discrete frequency points from the lowest to the highest analysis frequency in the load spectrum.

[0034] Multiply the conjugate complex numbers of the fast Fourier transform results of the wind speed signal sequence and the load signal sequence at each corresponding frequency point, and then divide by the length of the signal sequence to obtain the cross power spectral density complex number sequence describing the frequency domain correlation between the wind speed and the load signal.

[0035] From the calculated complex sequence of cross-power spectral density, locate the frequency point that is the same as the identified discrete peak frequency of wind speed, extract the complex result of cross-power spectral density at that frequency point, and calculate the modulus of the complex number to obtain the required cross-power spectral amplitude.

[0036] If both the peak energy ratio and the cross power spectrum amplitude exceed the corresponding discrimination threshold, then the periodic correlation condition is satisfied, and the current disturbance type of the wind turbine is periodic load disturbance. The discrimination threshold is the mean value obtained based on historical operating data plus three times the standard deviation.

[0037] If no significant discrete peak is identified, or if the identified significant discrete peak does not meet the periodic correlation condition, then the current type of disturbance to the wind turbine is determined to be instantaneous load disturbance.

[0038] It should be noted that the preset time window meets the minimum requirements of frequency resolution for spectrum analysis. Specifically, based on the dynamic characteristics of the wind turbine generator set, the rotational frequency of the wind turbine and its main low-order harmonic frequencies are obtained. In order to effectively separate the above-mentioned characteristic frequencies in spectrum analysis, the minimum frequency resolution is set to be no greater than the minimum interval between these characteristic frequencies, for example, no greater than 0.5 times the rotational frequency of the wind turbine.

[0039] Based on the principle of Discrete Fourier Transform, frequency resolution is inversely related to the length of the time window. Therefore, the theoretical time window length should be at least the reciprocal of the minimum frequency resolution.

[0040] To facilitate digital signal processing and ensure the integrity of data blocks, the preset time window must be an integer multiple of the fixed sampling period of the control system. The method for determining this is to round up the theoretical time window length to the nearest integer multiple of the sampling period, and use this as the preset time window.

[0041] The preset time window determined by the above steps ensures that the spectrum analysis has the ability to distinguish key frequency components, while also meeting the implementation constraints of the digital control system, thus ensuring the effectiveness and feasibility of subsequent signal processing.

[0042] It is understandable that spectral peaks with amplitudes greater than the sum of the mean and three times the standard deviation are identified as discrete peaks significantly higher than the background noise because the probability of such peaks being caused by random fluctuations in background noise is extremely low, and therefore they can be statistically identified as significant outliers or true signal components.

[0043] It should also be noted that the specific process of establishing the discrimination benchmark based on historical data is as follows: First, during the normal operation of the wind turbine generator, under various typical operating conditions, such as different power output ranges near the rated wind speed and common turbulence intensity ranges, wind speed and load signal samples are collected over a period of time, such as the past 30 days.

[0044] For each sample, perform the corresponding spectrum analysis and calculation process to obtain a historical peak energy ratio dataset and a historical cross power spectrum amplitude dataset containing a large number of data points.

[0045] Then, the statistical mean and standard deviation of the historical peak energy ratio dataset and the historical cross-power spectrum amplitude dataset are calculated respectively.

[0046] Finally, the statistical mean of the corresponding dataset is added to a value equal to a certain number of standard deviations, which is then set as the discrimination criterion. For example, a certain number of standard deviations can be set to three standard deviations. Implementers can adjust this adaptively according to actual requirements.

[0047] By using the statistical mean plus a certain number of standard deviations as the discrimination benchmark, false triggers caused by normal fluctuations due to random turbulence and measurement noise can be effectively filtered out, thus improving the stability of system decision-making.

[0048] S2. Analyze the lateral energy ratio and multi-band energy distribution of the wind turbine shaft to confirm the specific characteristics of the identified interference type. The specific characteristic of the periodic load interference is the interference direction, and the specific characteristic of the instantaneous load interference is the amplitude of the interference intensity change.

[0049] Considering that periodic disturbances have stable excitation directionality, mainly manifested as axial or lateral periodic bending moments, while transient disturbances are manifested as broadband, non-fixed-direction energy impacts with rapidly changing intensity over time, specific feature analysis of the identified disturbance types is conducted to provide a basis for subsequent implementation of precise feedforward compensation in a classified, directional, and quantitative manner.

[0050] Reference Figure 2 As shown, the process of analyzing the lateral energy ratio of the wind turbine shaft is as follows: the measurement signals of multiple load sensors installed on the wind turbine main shaft are calculated through geometric and mechanical relationships, a right-handed Cartesian coordinate system centered on the wind turbine main shaft is established, and independent axial bending moment time-domain signals and lateral bending moment time-domain signals are separated and output in real time under the right-handed Cartesian coordinate system; wherein, the axial bending moment refers to the bending moment component along the axis of the wind turbine main shaft, and the lateral bending moment refers to the bending moment component perpendicular to the axis of the main shaft and in the horizontal plane.

[0051] The axial bending moment time-domain signal and the lateral bending moment time-domain signal are subjected to spectral analysis consistent with the above logic. Specifically, the same detrending and windowing preprocessing is performed on their respective time-domain signal sequences, and then a fast Fourier transform is performed to obtain the axial bending moment spectrum and the lateral bending moment spectrum, both of which are power spectral density sequences.

[0052] The power spectral density values ​​of the axial bending moment component and the lateral bending moment component are obtained by summing and integrating the power spectral density values ​​at all effective frequency points within a preset time period.

[0053] The lateral energy ratio of the wind turbine shaft is obtained by dividing the frequency domain energy of the axial bending moment component by the frequency domain energy of the lateral bending moment component.

[0054] If the lateral energy ratio of the wind turbine shaft is greater than or equal to the preset reference value, it is determined to be the axial direction of the periodic load interference; otherwise, it is determined to be the lateral direction of the periodic load interference.

[0055] It should be noted that the preset duration is preferably the same preset time window length as that used in the spectrum analysis in step S1. Using the same time length eliminates the need to extract time-domain signals of different lengths for re-transformation, thus ensuring the simplicity and efficiency of the processing flow and the strict synchronization of the data time reference.

[0056] The preset reference value is based on structural symmetry and initial aerodynamic settings. Since the structural stiffness and aerodynamic excitation characteristics of the wind turbine in an ideal uniform flow field are symmetrical in the axial (thrust direction) and lateral (shear direction) directions, the preset reference value can be directly set to the value 1. If the axial-lateral energy ratio is greater than or equal to 1, the axial component is considered to be dominant, and vice versa.

[0057] It should also be noted that all effective frequency points refer to all discrete frequency points starting from a certain lowest analysis frequency greater than zero and ending at the Nyquist frequency, with intervals equal to the frequency resolution.

[0058] Reference Figure 3 As shown, the analysis of the multi-band energy distribution of the wind turbine is specifically as follows: the load spectrum is smoothed using the moving average method to obtain its envelope. Specifically, the frequency resolution is calculated using the preset time window determined in the aforementioned steps. Based on the requirement for the smoothness of the spectrum curve, an odd value is selected as the smoothness coefficient within a typical range of 3 to 7. For example, the smoothness coefficient is set to 5.

[0059] Based on the calculated frequency resolution and the target smoothness, set a sliding window with a width of 5 frequency points.

[0060] Starting from the beginning of the load spectrum, each frequency point except the two at the very beginning and the two at the very end is processed in sequence. For the frequency point currently being processed, the power spectral density values ​​corresponding to the five frequency points, including the frequency point itself, the two frequency points immediately before it, and the two frequency points immediately after it, are summed and then divided by five to calculate the arithmetic mean of these five values.

[0061] The arithmetic mean is used to replace the original center point value, thus generating a smoothed new spectrum curve.

[0062] The algorithm iterates through every frequency point on the new spectrum curve, excluding the first and last points. If the power spectral density value of a point is greater than the values ​​of its preceding and following adjacent points, then that point is identified as a local maximum. The frequency and amplitude corresponding to all identified local maxima are recorded; the amplitude is the power spectral density value.

[0063] Using frequency as the horizontal axis, connect all the identified local maxima points in sequence to form a preliminary broken line. Then, sort all the identified local maxima points in ascending order of their frequency values ​​to form a frequency-amplitude point set.

[0064] A straight line is used to connect two adjacent ordered local maxima.

[0065] Calculate the difference between the frequency and amplitude corresponding to the next local maximum point and the previous local maximum point to obtain the change in amplitude and the change in frequency.

[0066] Dividing the amplitude change by the frequency change yields the rate of amplitude change caused by the unit frequency change represented by the straight line segment, i.e., the slope of the linear relationship.

[0067] Calculate the difference between the frequency of the point to be determined and the frequency of the previous local maximum point.

[0068] The amplitude adjustment amount is obtained by multiplying the rate of change of amplitude caused by a unit frequency change by the frequency difference at the desired frequency point.

[0069] The amplitude of the previous local maximum point is added to the amplitude adjustment amount to obtain the envelope amplitude at the frequency point to be determined.

[0070] For each discrete frequency point in the original load spectrum, repeat the above calculation to obtain the envelope amplitude corresponding to each frequency point, thus forming a complete envelope.

[0071] The frequency corresponding to the local maximum point with the lowest frequency is taken as the first dividing frequency, and the frequency corresponding to the local maximum point with the highest amplitude is taken as the second dividing frequency.

[0072] The frequency band below the first boundary frequency is designated as the low-frequency band, the frequency band between the first and second boundary frequencies is designated as the mid-frequency band, and the frequency band above the second boundary frequency is designated as the high-frequency band.

[0073] The power spectral density in the low-frequency band, mid-frequency band, and high-frequency band is integrated to obtain the frequency domain energy of each sub-band.

[0074] Calculate the percentage of high-frequency band energy relative to the total energy of all sub-bands, and use this percentage as the current high-frequency band energy percentage.

[0075] Obtain the historical high-frequency band energy percentage sequence corresponding to the preset time window duration, and calculate the statistical mean and standard deviation of the high-frequency band energy percentage in the high-frequency band energy percentage sequence.

[0076] Calculate the deviation between the current high-frequency band energy percentage and the statistical mean, and divide the deviation by the standard deviation to obtain the amplitude of the interference intensity change that characterizes the instantaneous load interference.

[0077] It should be noted that the low-frequency band is the frequency band below the wind turbine rotation frequency. Its energy mainly comes from slow dynamic processes such as average load and wind speed trend changes. The load changes slowly and mainly affects the quasi-static stress of the components.

[0078] The mid-frequency band covers the frequency band centered on the wind turbine rotation frequency and its main harmonics. Its energy is a typical characteristic of periodic load interference. It is mainly excited by periodic aerodynamic and mechanical excitations such as wind shear and tower shadow effect, and is a key frequency band that leads to structural fatigue damage.

[0079] The high-frequency band refers to the frequency band with frequencies higher than the main rotating harmonics. Its energy mainly comes from sudden gusts in turbulence and the high-order modal response of the structure. It changes rapidly and has a relatively short duration, and is the main manifestation of transient load interference in the frequency domain.

[0080] Based on the division method of the first and second boundary frequencies, the lowest peak distinguishes the low-frequency gradual component from the main energy concentration, and the highest peak distinguishes the main periodic component from the higher-frequency random pulsation component. It adapts to the specific shape of the current spectrum without the need to preset a fixed frequency boundary, thus adapting to the dynamic changes of the load spectrum structure under different units and different operating conditions.

[0081] Understandably, using a historical high-frequency band energy percentage sequence corresponding to the preset time window duration is to ensure that the current calculated value is directly comparable to the historical benchmark value, and to avoid systematic biases introduced by different analysis durations.

[0082] Dividing the deviation by the standard deviation converts the absolute deviation into a quasi-fraction measured by its own historical volatility. This makes the calculated amplitude of the change in interference intensity a dimensionless, normalized statistic that can indicate the degree to which the current interference intensity deviates from the historical norm by several standard deviations. This enables a unified and comparable assessment of the interference intensity at different times, for different units, and under different wind conditions.

[0083] S3. Generate a corresponding feedforward compensation signal based on the type of interference and its specific characteristics. The feedforward compensation process includes: generating a synchronous reverse compensation signal if it is periodic load interference, and generating a filtered compensation signal that is proportional to the amplitude of the change in interference intensity if it is instantaneous load interference.

[0084] Considering that traditional feedback control has a lag, and that a single feedforward strategy is difficult to adapt to both periodic and instantaneous disturbances at the same time, in order to achieve accurate and active suppression of the load, it is necessary to generate a highly targeted feedforward compensation signal based on the disturbance type and its specific characteristics identified in the aforementioned steps, so as to make up for the shortcomings of feedback control.

[0085] The specific method for generating a synchronous reverse compensation signal for periodic load interference is as follows: when multiple discrete peaks that satisfy the periodic correlation condition are identified, the frequency corresponding to the discrete peak with the largest spectral aggregation factor is taken as the frequency of the sinusoidal compensation signal. The spectral aggregation factor is the product of the peak energy ratio corresponding to the discrete peak and the cross power spectrum amplitude.

[0086] The ratio of the vibration energy at the discrete peak frequency to the vibration energy in the corresponding narrow band of the load spectrum is calculated. A gain coefficient is determined based on this ratio, and this gain coefficient is used as the amplitude of the sinusoidal compensation signal.

[0087] Calculate the cross-power spectral density of the wind speed signal and the load signal at the discrete peak frequencies, and obtain their complex form.

[0088] The phase angle, i.e., the phase of the cross power spectrum, is extracted from the complex cross power spectral density.

[0089] The phase of the extracted cross-power spectrum is increased by 180 degrees, and the result is the initial phase of the sinusoidal compensation signal.

[0090] Based on the frequency, amplitude, and initial phase of the sinusoidal compensated signal, a time-domain signal expression in the form of a standard sinusoidal function is constructed.

[0091] Within a discrete time period equal to the preset time window duration, a series of equally spaced discrete time point sequences are generated with the system's inherent fixed sampling period as the interval.

[0092] For each moment in the discrete time sequence, it is substituted into the standard sine function expression for calculation. Specifically, according to the standard physical relationship between circular motion and simple harmonic motion, the frequency is multiplied by a constant 2π to obtain the corresponding angular frequency, and the phase value is multiplied by a constant π / 180 to obtain the initial phase in radians.

[0093] Multiply the current time value by the angular frequency of the compensation signal, and add the initial phase in radians to calculate the sine function phase value at that time. Then, take the sine function value of this phase value. Finally, multiply the obtained sine function value by the amplitude of the sine compensation signal to obtain the initial synchronous reverse compensation signal corresponding to that discrete time.

[0094] The generated initial synchronous reverse compensation signal is input to a bandpass filter with a center frequency of discrete peak frequency. The filter performs recursive operations on each input data point according to its transfer function, filtering out components other than the center frequency, and outputting the final synchronous reverse compensation signal.

[0095] The process of generating a filter compensation signal proportional to the interference intensity for instantaneous load interference is as follows: extract all discrete frequency points belonging to the high-frequency band in the load spectrum and their corresponding power spectral density values, multiply each discrete frequency point by the corresponding power spectral density value to obtain the centroid frequency of each discrete frequency point, and then sum the centroid frequencies of all discrete frequency points to calculate the mean, and obtain the final centroid frequency.

[0096] Calculate the difference between the frequency value of each discrete frequency point and the centroid frequency, and then square the difference. Next, multiply the squared value by the power spectral density value corresponding to the frequency point to obtain the weighting value for that point. Then, sum the weighted contributions of all frequency points in the high-frequency band. Finally, divide this sum by the sum of the power spectral density values ​​of all frequency points in the high-frequency band to obtain the second-order central moment.

[0097] The equivalent bandwidth is obtained by taking the square root of the calculated second-order central moments.

[0098] Multiplying both the centroid frequency and the equivalent bandwidth by a constant 2π yields the corresponding center angular frequency and bandwidth angular frequency.

[0099] Using the center angular frequency and bandwidth angular frequency obtained from the above calculation as parameters, a standard second-order bandpass transfer function is constructed. The form of the second-order bandpass transfer function is determined by the center angular frequency, the bandwidth angular frequency, and the quality factor defined by the ratio of the two. In the continuous time domain, the frequency response characteristics of the second-order bandpass transfer function are bandpass characteristics centered on the center angular frequency and with the bandwidth angular frequency as the width.

[0100] Then, the second-order bandpass transfer function is converted into a digital filter transfer function using the bilinear transform method, which is well-known in the field. The general process of the bilinear transform method is as follows: based on the fixed sampling period of the system, the differential operator in the analog transfer function is replaced with a rational fraction about the discrete delay operator using the mapping relationship specified by the bilinear transform. By rearranging the rational fraction, a standard second-order digital bandpass filter transfer function is finally obtained in the form of a negative power polynomial of the discrete delay operator.

[0101] Obtain a discrete-time sequence with the same length as the signal sequence used in the spectrum analysis, wherein a pulse with an amplitude of 1 is set at the position corresponding to the first sampling time in the discrete-time sequence, and the amplitude is 0 at all other sampling time positions, to construct a unit impulse signal.

[0102] The unit impulse signal is input into the transfer function of a second-order digital bandpass filter. Based on the difference equation defined by the transfer function, the input unit impulse signal is recursively digitally filtered to obtain the primary compensation signal.

[0103] The overall gain coefficient is obtained by multiplying the amplitude of the interference intensity change with a preset reference amplitude that has the dimension of a pitch angle command.

[0104] The primary compensation signal is multiplied by the comprehensive gain coefficient to obtain the final filtered compensation signal.

[0105] The preset reference amplitude is obtained through historical experimental calibration. Specifically, under standard conditions, multiple sets of standard instantaneous disturbances of known intensity are applied to the unit, and the load response without compensation is recorded. Subsequently, feedforward compensation is applied tentatively with different reference amplitudes, the load suppression effect is observed, and the reference amplitude that makes the load suppression effect reach the required range is selected as the preset reference amplitude.

[0106] By extracting specific discrete peak frequencies with strong correlation between wind speed and load spectrum as compensation frequencies, and setting an inverse initial phase based on cross-power spectrum phase information, the generated compensation signal can achieve frequency synchronization and phase reversal with the actual periodic mechanical load interference in the time domain. Furthermore, through a bandpass filter with adjustable center frequency, it is ensured that the feedforward compensation energy can be highly concentrated on the target frequency band that needs to be suppressed.

[0107] By quantizing the centroid frequency and equivalent bandwidth, a primary compensation signal that matches its spectral shape is dynamically generated. The primary compensation signal is multiplied by the comprehensive gain coefficient, so that the compensation strength can follow the changes in interference intensity in real time and adaptively, achieving precise matching of strong interference-strong compensation and weak interference-weak compensation.

[0108] This invention, through joint analysis of wind speed spectrum and load spectrum, combined with discrete peak detection, cross power spectrum and energy ratio determination, can accurately distinguish between periodic load interference and instantaneous load interference. For periodic interference, a synchronous reverse compensation signal is generated to achieve phase cancellation; for instantaneous interference, a filtered compensation signal proportional to the amplitude of the interference intensity change is generated.

[0109] S4. After the feedforward compensation signal is processed by amplitude limiting and rate limiting, it is fused with the power feedback control signal and the integrated pitch angle command is output.

[0110] Considering that the feedforward compensation signal is generated based on ideal conditions, while the pitch actuator of an actual wind turbine has clear physical action boundaries and rate limits, the amplitude and rate of change of the feedforward compensation signal must be constrained to ensure the physical feasibility of control commands and avoid damage to the actuator or instability. Simultaneously, periodic and transient disturbances, due to their different dynamic characteristics, require different fusion strategies with the feedback control signal to achieve optimal overall control performance.

[0111] The specific steps for limiting the amplitude and rate of the feedforward compensation signal are as follows: obtain the instantaneous amplitude of the feedforward compensation signal at the current moment, compare the instantaneous amplitude with the upper and lower limits of the physical range of the pitch angle, respectively. If the instantaneous amplitude is greater than or equal to the upper limit, the upper limit is used as the output value; if the instantaneous amplitude is less than or equal to the lower limit, the lower limit is used as the output value; otherwise, the instantaneous amplitude is used as the output value.

[0112] Calculate the difference between the output value at the current sampling time and the output value at the previous sampling time.

[0113] If the difference exceeds the preset allowable change amount, the output value of the previous moment plus the preset allowable change amount will be used as the current output value.

[0114] If the difference is less than the opposite of the preset permissible change amount, then the output value of the previous moment minus the preset permissible change amount is used as the current output value.

[0115] Otherwise, maintain the output value at the current sampling time.

[0116] The integrated pitch angle command specifically refers to the following: when the interference type is periodic load interference, the feedforward compensation signal and the power feedback control signal are phase-synchronized and superimposed.

[0117] When the interference type is instantaneous load interference, the power feedback control signal is superimposed with the feedforward compensation signal after being filtered by the first-order inertial element.

[0118] The superimposed signal is output as an integrated pitch angle command.

[0119] It should be noted that the upper and lower limits of the physical range of the pitch angle are determined based on the aerodynamic design of the wind turbine generator set. Specifically: the upper limit is the final position for aerodynamic braking and overspeed protection, i.e., the feathering position, which is usually between 85 and 90 degrees; the lower limit corresponds to the minimum pitch angle at which the blades achieve optimal aerodynamic efficiency or rated power control, which is usually between -2 and 5 degrees. The specific values ​​of the upper and lower limits can be adaptively set by the implementer according to actual requirements.

[0120] The preset permissible variation amount is determined directly through physical testing and data statistics. Specifically, on a full-size test platform for wind turbine generators, the step change amplitude of the pitch angle command is gradually increased, while the load response of key structural components such as blades and main shafts is monitored in real time. The maximum actual rate of change that the pitch system can achieve when an acceptable load fluctuation amplitude is induced is directly observed and recorded. The maximum actual rate of change is multiplied by the system sampling period to obtain the preset permissible variation amount.

[0121] It should also be noted that when the interference type changes, the original compensation signal is linearly attenuated to zero within a preset transition period, such as 0.5-2 seconds, while the new filter compensation signal is linearly increased from zero to the calculated value.

[0122] Understandably, power feedback control can effectively maintain generator speed, but it lags behind in responding to rapid load disturbances; feedforward compensation signals, on the other hand, provide predictive compensation specifically for particular disturbances. By superimposing the two into a single integrated command, it is ensured that every pitch control action of the unit is the result of the coordinated action of basic stability regulation and corresponding load suppression, thus guaranteeing system stability.

[0123] This invention integrates the feedforward compensation signal with the power feedback control signal after amplitude and rate limiting, avoiding over-adjustment and oscillation, ensuring the smoothness and feasibility of the pitch angle command, and thus improving the stability of power generation.

[0124] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined by the present invention, they should all fall within the protection scope of the present invention.

Claims

1. A wind turbine pitch control method based on load prediction feedforward compensation, characterized in that, include: If there are discrete peaks in the wind speed spectrum that are higher than the background, and the load spectrum and the wind speed spectrum satisfy the periodic correlation condition at the discrete peak frequency, then the type of interference currently experienced by the wind turbine is determined to be periodic load interference; otherwise, it is determined to be transient load interference. The lateral energy ratio and multi-frequency band energy distribution of the wind turbine shaft are analyzed to confirm the specific characteristics of the interference type. The specific characteristics of the periodic load interference are the interference direction and the specific characteristics of the instantaneous load interference are the amplitude of the interference intensity change. Based on the type of interference and its specific characteristics, a corresponding feedforward compensation signal is generated. The feedforward compensation process includes: if it is periodic load interference, a synchronous reverse compensation signal is generated; if it is instantaneous load interference, a filtered compensation signal proportional to the amplitude of the interference intensity change is generated. After the feedforward compensation signal is processed by amplitude limiting and rate limiting, it is fused with the power feedback control signal to output an integrated pitch angle command. The specific steps for determining the current type of interference experienced by the wind turbine are as follows: Perform spectral analysis on the wind speed signal and the wind turbine load signal within a preset time window to obtain the wind speed spectrum and the load spectrum; calculate the mean and standard deviation of the wind speed spectrum amplitude; identify spectral peaks with amplitudes greater than the sum of the mean and three times the standard deviation as discrete peaks significantly higher than the background noise; for each identified discrete peak, calculate the ratio of vibration energy in the narrow band centered on the discrete peak frequency to the total broadband vibration energy in the load spectrum, and mark it as the peak energy ratio; calculate the cross-power spectrum amplitude of the wind speed spectrum and the load spectrum at the discrete peak frequency; if both the peak energy ratio and the cross-power spectrum amplitude exceed the corresponding discrimination threshold, it is determined that the periodic correlation condition is met, and the current type of interference experienced by the wind turbine is periodic load interference; if no significant discrete peak is identified, or the identified significant discrete peak does not meet the periodic correlation condition, it is determined that the current type of interference experienced by the wind turbine is instantaneous load interference. The amplitude and rate limiting processing of the feedforward compensation signal is specifically as follows: the instantaneous amplitude of the feedforward compensation signal at the current moment is obtained, and the instantaneous amplitude is compared with the upper and lower limits of the physical range of the pitch angle, respectively, and the comparison result is used as the output value after amplitude limiting processing; the difference between the output value at the current sampling moment and the output value at the previous sampling moment is calculated; if the difference exceeds the preset permissible change amount, the output value at the previous moment is added to the preset permissible change amount as the current output value; if the difference is less than the negative of the preset permissible change amount, the output value at the previous moment is subtracted from the preset permissible change amount as the current output value; otherwise, the output value at the current sampling moment is maintained.

2. The wind turbine pitch control method based on load prediction feedforward compensation according to claim 1, characterized in that: The spectrum analysis specifically refers to: The time-domain signal sequences of the collected wind speed and load signals are preprocessed by detrending and windowing, respectively. Perform a Fast Fourier Transform on each preprocessed time-domain signal to obtain the corresponding complex spectrum; Calculate the square of the spectral modulus at each frequency point in the complex spectrum, and then divide it by the length of the signal sequence to obtain the corresponding wind speed spectrum and load spectrum.

3. The wind turbine pitch control method based on load prediction feedforward compensation according to claim 1, characterized in that: The process of analyzing the lateral energy ratio of the wind turbine shaft is as follows: Calculate the frequency domain energy of the axial bending moment component and the lateral bending moment component in the load spectrum within a preset time period; Divide the frequency domain energy of the axial bending moment component by the frequency domain energy of the lateral bending moment component to obtain the lateral energy ratio of the wind turbine shaft; If the lateral energy ratio of the wind turbine shaft is greater than or equal to the preset reference value, it is determined to be the axial direction of the periodic load interference; otherwise, it is determined to be the lateral direction of the periodic load interference.

4. The wind turbine pitch control method based on load prediction feedforward compensation according to claim 1, characterized in that: The analysis of the multi-frequency band energy distribution of the wind turbine is specifically as follows: The load spectrum is divided into multiple sub-bands, including low-frequency, mid-frequency, and high-frequency bands; The power spectral density in the low-frequency band, mid-frequency band, and high-frequency band is integrated to obtain the frequency domain energy of each sub-band; Calculate the percentage of high-frequency band energy to the total energy of all sub-bands, and use this percentage as the current high-frequency band energy percentage; Obtain the historical high-frequency band energy percentage sequence corresponding to the preset time window duration, and calculate the statistical mean and standard deviation of the high-frequency band energy percentage in the high-frequency band energy percentage sequence; Calculate the deviation between the current high-frequency band energy percentage and the statistical mean, and divide the deviation by the standard deviation to obtain the amplitude of the interference intensity change that characterizes the instantaneous load interference.

5. The wind turbine pitch control method based on load prediction feedforward compensation according to claim 4, characterized in that: The load spectrum division is specifically as follows: The load spectrum is smoothed to obtain its envelope, local maxima on the envelope are identified, and their frequencies and amplitudes are recorded. The frequency corresponding to the local maximum point with the lowest frequency is taken as the first boundary frequency, and the frequency corresponding to the local maximum point with the highest amplitude is taken as the second boundary frequency. The frequency band below the first boundary frequency is designated as the low-frequency band, the frequency band between the first and second boundary frequencies is designated as the mid-frequency band, and the frequency band above the second boundary frequency is designated as the high-frequency band.

6. The wind turbine pitch control method based on load prediction feedforward compensation according to claim 1, characterized in that: The specific steps for generating a synchronous reverse compensation signal to address periodic load interference are as follows: When multiple discrete peaks that satisfy the periodic correlation condition are identified, the frequency corresponding to the discrete peak with the largest spectral aggregation factor is taken as the frequency of the sinusoidal compensation signal. The gain coefficient, determined based on the ratio of vibration energy at discrete peak frequencies to vibration energy within the narrow band of the load spectrum, is used as the amplitude of the sinusoidal compensation signal. The initial phase of the sinusoidal compensation signal is calculated based on the cross-power spectrum phase information of the load spectrum and the wind speed spectrum at discrete peak frequencies. By combining the frequency, amplitude, and initial phase of the sinusoidal compensation signal, an initial synchronous reverse compensation signal is generated. The generated initial synchronous reverse compensation signal is input into a bandpass filter with a center frequency of discrete peak frequency to filter out components other than the center frequency, and the final synchronous reverse compensation signal is output.

7. The wind turbine pitch control method based on load prediction feedforward compensation according to claim 5, characterized in that: The process of generating a filtered compensation signal proportional to the interference intensity for instantaneous load interference is as follows: Calculate the centroid frequency and equivalent bandwidth of the high-frequency power spectral density distribution, wherein the centroid frequency is the frequency-weighted average value with power spectral density as the weight, and the equivalent bandwidth is the square root of the second central moment of the power spectral density distribution. Based on the centroid frequency and equivalent bandwidth, the transfer function of a second-order bandpass filter is constructed. Obtain a discrete-time sequence with the same length as the signal sequence used in the spectrum analysis, wherein a pulse with an amplitude of 1 is set at the position corresponding to the first sampling time in the discrete-time sequence, and the amplitude is 0 at all other sampling time positions, to construct a unit impulse signal; The unit impulse signal is input into the transfer function to calculate the primary compensation signal; The overall gain coefficient is obtained by multiplying the amplitude of the interference intensity change by a preset reference amplitude with the dimension of the pitch angle command. The primary compensation signal is multiplied by the comprehensive gain coefficient to obtain the final filtered compensation signal.

8. The wind turbine pitch control method based on load prediction feedforward compensation according to claim 1, characterized in that: The integrated pitch angle command is specifically as follows: When the interference type is periodic load interference, the feedforward compensation signal and the power feedback control signal are phase-synchronized and superimposed. When the interference type is instantaneous load interference, the power feedback control signal is superimposed with the feedforward compensation signal after being filtered by the first-order inertial element. The superimposed signal is output as an integrated pitch angle command.

Images