Single-phase fan harmonic monitoring method based on overmodulation algorithm and related equipment

Abstract

The invention relates to a single-phase fan harmonic monitoring method based on an overmodulation algorithm and related equipment, and the method comprises the steps: dynamically dividing a control mode of an inverter into at least a conventional SPWM region, an overmodulation region and a square wave region based on a value range of a modulation coefficient M; when the inverter is in the conventional SPWM area, sine waves are compared with triangular carriers to generate SPWM waves which are used as PWM signals for driving the inverter; when the inverter is in the square wave area, each switching tube of the inverter is controlled to be conducted once in each fundamental wave period; when the inverter is in the over-modulation region, closed-loop harmonic management is carried out: real-time sampling and harmonic analysis are carried out on the output voltage of the inverter to obtain an analysis result; on the basis of the real-time operation parameters of the fan and the analysis result, the harmonic wave change trend in the future control period is predicted; and dynamically adjusting the conduction time of a switching tube in the inverter based on the harmonic analysis result and the harmonic change trend. The single-phase fan has the effect that the requirements for high rotating speed and low-noise operation of the single-phase fan are met at the same time.

Description

Technical Field

[0001] This application relates to the field of wind turbine drive control technology, and in particular to a single-phase wind turbine harmonic monitoring method and related equipment based on an overmodulation algorithm. Background Technology

[0002] Single-phase fans, as a key power and ventilation device, are widely used in household appliances, industrial equipment, automotive electronics, and communication base stations. The core of their drive control technology lies in adjusting the voltage and frequency applied to the motor windings via an inverter to achieve speed control.

[0003] In existing technologies, sinusoidal pulse-width modulation (SPWM) control is the mainstream technology for unidirectional fan drives. This technology outputs a sinusoidal current through the inverter, forming a stable rotating magnetic field within the motor stator, driving the fan rotor to rotate smoothly. It has the advantages of low operating noise and low vibration. However, the performance of conventional sinusoidal pulse-width modulation is constrained by both the voltage boundary circle determined by the inverter's DC bus voltage and the current boundary circle determined by the rated current of the power devices and the heating limit of the motor windings. Its maximum output fundamental voltage is limited, making it difficult to further increase the maximum speed of the fan, which cannot meet the urgent demand for high speed in emerging application scenarios such as data center cooling and supercharging equipment.

[0004] To overcome speed limitations, a square wave control scheme exists in the industry. This scheme outputs a square wave voltage from the inverter, which can significantly increase the fundamental component of the output voltage, thereby driving the fan to achieve higher speeds. However, the square wave voltage contains a large number of low-order harmonics, which can cause motor current distortion and increased torque ripple, ultimately resulting in a significant increase in fan operating noise, severely impacting user experience and equipment quietness.

[0005] Therefore, the existing technology presents a clear technical contradiction: while sinusoidal wave control can ensure low-noise operation, it cannot meet the high-speed requirements; while square wave control can achieve high speed, it must sacrifice low-noise characteristics, that is, it cannot simultaneously meet the high-speed and low-noise operation requirements of single-phase fans, so it needs to be improved. Summary of the Invention

[0006] In order to simultaneously meet the requirements of high speed and low noise operation of single-phase fans, this application provides a harmonic monitoring method and related equipment for single-phase fans based on an overmodulation algorithm.

[0007] Firstly, this application provides a single-phase wind turbine harmonic monitoring method based on an overmodulation algorithm, employing the following technical solution: Based on the range of values ​​for the modulation coefficient M, the control mode of the inverter is dynamically divided into at least the conventional SPWM region, the overmodulation region, and the square wave region. Specifically, when the modulation coefficient M satisfies: 0 < M ≤ 1, it is in the conventional SPWM region; when the modulation coefficient M satisfies: 1 < M < Mmax, it is in the overmodulation region; and when the modulation coefficient M = Mmax, it is in the square wave region. Whenever the inverter is in the normal SPWM zone, a sine wave is compared with a triangular carrier wave to generate an SPWM wave, which is used as the PWM signal to drive the inverter. Whenever the inverter is in the square wave region, each switch of the control inverter is turned on once per fundamental wave cycle; Whenever the inverter is in the overmodulation zone, closed-loop harmonic management is performed: the inverter's output voltage is sampled and harmonic analyzed in real time to obtain the analysis results; based on the real-time operating parameters of the wind turbine and the analysis results, the harmonic variation trend within the future control cycle is predicted using a preset prediction model; wherein, the real-time operating data of the wind turbine includes at least the motor stator current, wind turbine speed, and load data; based on the harmonic analysis results and the harmonic variation trend, the on-time of the switching transistors in the inverter is dynamically adjusted to increase the fundamental voltage amplitude while suppressing waveform distortion within a preset threshold.

[0008] By adopting the above technical solution, the speed range is first expanded through partitioning. Addressing the specific problem of high noise due to large harmonics in the overmodulation region in the background technology, a "detection-prediction-control" solution is provided. This solution is applied within the overmodulation region, specifically using harmonic analysis and prediction to sense the system state, and precise control to balance the conflict between speed and noise. Its effect is to fundamentally solve the technical challenge of balancing "high speed" and "low noise" in the background technology, providing an intelligent and adaptive control scheme.

[0009] Optionally, the real-time sampling and harmonic analysis of the inverter's output voltage to obtain analysis results includes: The sliding window fast Fourier transform algorithm is used to perform high-frequency sampling and real-time spectrum analysis on the output voltage signal, and the amplitude and phase information of the specified low-order harmonic components are extracted as the analysis results.

[0010] By adopting the above technical solution and using sliding window FFT, the time delay caused by data caching and calculation is greatly reduced compared with traditional FFT, realizing the "real-time" or "near real-time" capture of harmonic components, providing a timely and accurate data foundation for subsequent dynamic control, and avoiding control failure caused by information lag.

[0011] Optionally, the dynamic adjustment of the on-time of the switching transistors in the inverter to suppress harmonic distortion within a preset threshold while increasing the fundamental voltage amplitude includes: Based on the current modulation coefficient, an initial modulation wave in the form of a quasi-square wave is generated. The quasi-square wave refers to a waveform with its top clipped, which is between a sine wave and a pure square wave. With the total harmonic distortion not exceeding a preset threshold as a constraint, calculate the balance coefficient between the harmonic amplitude and the fundamental amplitude; When the harmonic amplitude is predicted to exceed a preset threshold based on the harmonic variation trend, the initial modulation waveform is optimized based on the balance coefficient to reduce the width of the flat-top segment of the initial modulation waveform, thereby suppressing the growth of low-order harmonics.

[0012] By adopting the above technical solution, a specific means of fine-tuning within the overmodulation region is provided. It does not simply sacrifice voltage for harmonics, but rather, by fine-tuning the switching timing, it "precisely" eliminates harmful harmonics while ensuring the fundamental voltage (i.e., rotational speed) is increased as much as possible, thus achieving the optimal balance of performance.

[0013] Optionally, based on the real-time operating parameters of the wind turbine and the analysis results, the harmonic variation trend within the future control cycle is predicted using a preset prediction model, including: Based on the real-time operating parameters of the wind turbine and the analysis results, the amplitude of a specific harmonic, the switching loss, and the torque pulsation prediction value are predicted in the future control cycle using a preset prediction model. The dynamic adjustment of the on-time of the switching transistors in the inverter, to increase the fundamental voltage amplitude while suppressing waveform distortion within a preset threshold, includes: Multiple candidate modulation wave correction values ​​are generated, and corresponding candidate modulation waves are generated based on the correction values; wherein each modulation wave correction value corresponds to the adjustment of the width of the quasi-square wave flat top segment. Based on a pre-stored empirical model describing the relationship between the modulation wave correction and the changes in various performance indicators, and the predicted values ​​of the specific harmonic amplitude, switching loss, and torque ripple in the future control cycle, the specific harmonic amplitude, switching loss, and torque ripple corresponding to each candidate modulation wave are estimated respectively. Based on the estimated amplitude of the specific subharmonic, the switching loss value, and the torque ripple value, calculate the multi-objective cost of each candidate scheme; The candidate modulation wave that minimizes the multi-objective cost is selected, and the candidate modulation wave is compared with a triangular carrier wave to generate the final PWM signal used to drive the inverter switching transistor, thereby realizing the dynamic adjustment of the switching transistor's on-time.

[0014] By adopting the above technical solution, the single harmonic control problem in the overmodulation region is creatively transformed into a multi-objective collaborative optimization problem. Through intelligent prediction and search, it dynamically adjusts the switching sequence (conduction time) to ensure that while the wind turbine achieves high speed (high fundamental voltage), its harmonic distortion, switching losses, and torque ripple are also systematically suppressed to optimal levels. This achieves a balance between high speed, low noise, high efficiency, and high stability, solving the technical challenge that existing technologies cannot simultaneously address.

[0015] Optionally, the multi-objective cost = w1 × the estimated amplitude of the specific subharmonic + w2 × the estimated switching loss + w3 × the torque ripple; where w1, w2 and w3 are weighting coefficients; The calculation of the multi-objective cost value for each candidate solution includes: Calculate the dissatisfaction with the three performance indicators of the current specific harmonic amplitude, switching loss and torque ripple, where the dissatisfaction with any performance indicator is U=max(0, the current value of the performance indicator − its preset expected value); sum the dissatisfaction with the performance indicators to obtain the total dissatisfaction. The ratio of dissatisfaction with each performance indicator to the total dissatisfaction is used as its corresponding weight coefficients w1, w2, and w3, respectively. The multi-objective cost value of each candidate solution is calculated using the final weight coefficients w1, w2, and w3.

[0016] By adopting the above technical solution, this approach completely eliminates subjective judgment and ambiguity in weight setting. The magnitude of the weight directly and proportionally reflects the severity of the performance indicator's "failure to meet the standard." System control resources will automatically and precisely tilt towards the area with the "most prominent problem," achieving optimal dynamic allocation of control resources and significantly improving the system's overall performance ceiling and adaptive capability.

[0017] Optionally, the generation of the initial modulation wave in quasi-square wave form includes: Based on a preset modulation waveform library, the current modulation coefficient M, and the operating status of the wind turbine, the optimal initial modulation waveform type is selected from the modulation waveform library for the current control cycle; wherein, the modulation waveform library stores multiple different modulation waveforms for the same modulation coefficient M; based on the selected initial modulation waveform type, an initial modulation waveform in the form of a quasi-square wave is generated.

[0018] By adopting the above technical solution, compared to using a single overmodulation algorithm, this solution intelligently selects the initial waveform that best matches the current operating conditions from the modulation waveform library. This provides a starting point with better harmonic characteristics and greater control potential for subsequent precise control, reducing the difficulty of harmonic suppression from the source. For different operating states (such as load changes or different target speeds), the system can automatically call the most suitable waveform type, effectively addressing the inherent defect of harmonic performance degradation under certain specific operating conditions with a single waveform, thus improving the overall performance of the wind turbine under various operating conditions. This solution, together with the subsequent real-time fine-tuning steps, forms a perfect division of labor between "coarse tuning" and "fine tuning." Intelligent pre-selection is responsible for significantly improving the operating point (coarse tuning), while harmonic prediction and precise control are responsible for the final refined compensation (fine tuning). This division of labor maximizes the efficiency and effectiveness of the entire control system.

[0019] Optionally, when the modulation coefficient M satisfies: 0 < M ≤ 1, it is in the conventional SPWM region; when the modulation coefficient M satisfies: 1 < M < Mmax, it is in the overmodulation region; when the modulation coefficient M = Mmax, it is in the square wave region; including: When 0 < M < Mlow1, it is in the normal SPWM region; When Mlow1≤M≤Mhigh1, it is in the first transition zone; where Mlow1<1<Mhigh1; When Mhigh1 < M < Mlow2, it is in the overmodulation region; where Mlow2 < Mmax; When Mlow2≤M≤Mhigh2, it is in the second transition region; where Mhigh2=Mmax; The method further includes: When the current modulation coefficient M is within the transition region, a smooth transition operation is performed: standard modulation waveforms of two standard partitions adjacent to the transition region are generated in parallel; based on the relative position of the current modulation coefficient M within the current transition region, a mixing weighting coefficient ρ is calculated; the two parallel-generated standard modulation waveforms are weighted and fused using the mixing weighting coefficient, and the fused waveform is used as the modulation wave to generate the PWM signal driving the inverter; wherein, the transition region is within the first transition region or the second transition region; the standard partitions include the conventional SPWM region, the overmodulation region, and the square wave region.

[0020] By adopting the above technical solution, through continuous weighted mixing of waveforms, electromagnetic torque pulsation and current surges caused by modulation waveform abrupt changes at the boundary of the partition are fundamentally avoided, making the fan speed switching process extremely smooth and stable, and improving the user experience of high-end applications; throughout the entire transition zone, the fundamental voltage rises smoothly with the increase of the modulation coefficient M and the mixing weight coefficient ρ, ensuring the linearity and naturalness of the fan acceleration process and avoiding the feeling of "step" or "jerk" in speed; this solution is not sensitive to small changes in system parameters (such as bus voltage and motor parameters), and can ensure the smoothness of boundary switching under any circumstances, improving the overall reliability of the product.

[0021] Secondly, this application provides a single-phase wind turbine harmonic monitoring system based on an overmodulation algorithm, employing the following technical solution: The dynamic partitioning control module is used to dynamically divide the inverter's control mode into at least a conventional SPWM region, an overmodulation region, and a square wave region based on the range of values ​​of the modulation coefficient M. Specifically, when the modulation coefficient M satisfies: 0 < M ≤ 1, it is in the conventional SPWM region; when the modulation coefficient M = Mmax, it is in the square wave region; and when the modulation coefficient M satisfies: 1 < M < Mmax, it is in the overmodulation region. The real-time harmonic detection module is used to perform closed-loop harmonic management whenever the inverter is in the overmodulation zone: it samples and analyzes the inverter's output voltage in real time to obtain the analysis results. The harmonic trend prediction module is used to predict the harmonic change trend within a future control cycle based on the real-time operating parameters of the wind turbine and the analysis results, using a preset prediction model; wherein, the real-time operating data of the wind turbine includes at least the motor stator current, wind turbine speed and load data. The wind turbine precision control module is used to generate an SPWM wave by comparing a sine wave with a triangular carrier wave whenever it is in the normal SPWM zone, and use this as the PWM signal to drive the inverter; it is also used to control each switch of the inverter to conduct once per fundamental cycle whenever it is in the square wave zone; and it is also used to dynamically adjust the conduction time of the switch in the inverter based on the harmonic analysis results and the harmonic change trend whenever it is in the overmodulation zone, so as to increase the voltage fundamental amplitude while suppressing waveform distortion within a preset threshold.

[0022] Thirdly, this application provides a single-phase wind turbine harmonic monitoring device based on an overmodulation algorithm, including a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executed as described in any of the first aspects.

[0023] Fourthly, this application provides a computer-readable storage medium storing a computer program that can be loaded by a processor and executed as described in any of the first aspects.

[0024] In summary, the present application includes at least one of the following beneficial technical effects: The present application breaks through the speed limit of conventional sine wave control, significantly increasing the maximum speed of the fan, and can meet the requirements of high-speed application scenarios; compared with traditional square wave control, since the core waveform characteristics of the sine wave are retained, the harmonic content during the operation of the fan is effectively controlled, and the noise is reduced to a certain extent compared with square wave control, taking into account both high-speed and low-noise operation; Furthermore, by establishing a "prediction - adjustment - feedback" closed loop, the conduction time parameter is corrected according to the actual harmonic detection result every cycle, ensuring that during the dynamic change of the modulation coefficient M (1 < M ≤ Mmax), the harmonic ratio is always controlled within the preset range, thereby taking into account the improvement of voltage utilization rate and the stability of motor operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0026] Figure 1 It is a schematic flowchart of a single-phase fan harmonic supervision method based on an overmodulation algorithm disclosed in an embodiment of the present application.

[0027] Figure 2 It is a structural block diagram of a single-phase fan harmonic supervision system based on an overmodulation algorithm disclosed in an embodiment of the present application.

[0028] Explanation of reference numerals: 201, dynamic partition control module; 202, harmonic real-time detection module; 203, harmonic trend prediction module; 204, fan precise control module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0029] The following will further elaborate on the present application in conjunction with the attached Figure 1-2 for a more detailed description.

[0030] This application discloses a single-phase wind turbine harmonic monitoring method based on an overmodulation algorithm (hereinafter referred to as the wind turbine harmonic monitoring method). The main executing component is a wind turbine harmonic monitoring system, which is communicatively connected to an inverter power circuit (composed of a DC bus power supply, filter capacitors, and a full-bridge circuit consisting of four switching transistors), a sampling circuit (including a bus voltage sensor and an output phase current sensor), and a controlled unidirectional wind turbine (such as a single-phase permanent magnet synchronous wind turbine). The wind turbine harmonic monitoring system first calculates the required modulation coefficient M based on the target speed, performs dynamic partitioning, selects the corresponding control mode (SPWM, overmodulation, or square wave), and generates a corresponding modulation wave signal internally according to the corresponding control mode. This signal is then compared with a carrier wave (usually a triangular wave) to generate four complementary PWM signals with dead time. These PWM signals are amplified in voltage and current by a gate driver chip and then directly applied to the gates of the four switching transistors in the inverter power circuit to control the inverter's on / off state. After the wind turbine harmonic monitoring system outputs a PWM wave to drive the inverter, the sampling circuit captures the inverter's output voltage and current in real time. The system analyzes the sampled voltage signal and extracts the current harmonic components. Then, combining current motor speed, load current, and other operating parameters with historical harmonic data, the system predicts harmonic changes in the near future. Finally, the system uses the combined current harmonic values ​​and future predictions to dynamically adjust the PWM generation strategy (mainly the switching transistor on-time) for the next control cycle, thereby achieving a balance between speed increase and noise suppression. The following section will combine... Figure 1 The specific implementation principles and steps are explained in detail.

[0031] S101, based on the range of values ​​of the modulation coefficient M, dynamically divides the control mode of the inverter into at least a conventional SPWM region, an overmodulation region, and a square wave region; wherein, when the modulation coefficient M satisfies: 0 < M ≤ 1, it is in the conventional SPWM region; when the modulation coefficient M satisfies: 1 < M < Mmax, it is in the overmodulation region; when the modulation coefficient M = Mmax, it is in the square wave region.

[0032] S102, whenever in the normal SPWM zone, uses a sine wave to compare with a triangular carrier wave to generate an SPWM wave, which is used as the PWM signal to drive the inverter.

[0033] S103 controls each switch of the inverter to conduct once per fundamental cycle whenever it is in the square wave region.

[0034] S104, whenever the inverter is in the overmodulation zone, closed-loop harmonic management is performed: the inverter's output voltage is sampled and harmonic analyzed in real time to obtain the analysis results; based on the real-time operating parameters of the wind turbine and the analysis results, the harmonic change trend in the future control cycle is predicted through a preset prediction model; the real-time operating data of the wind turbine includes at least the motor stator current, wind turbine speed and load data; based on the harmonic analysis results and harmonic change trends, the on-time of the switching transistors in the inverter is dynamically adjusted to increase the voltage fundamental amplitude while suppressing waveform distortion within a preset threshold.

[0035] The step S1O4, "Real-time sampling and harmonic analysis of the inverter's output voltage to obtain analysis results," specifically includes the following sub-steps: S1041 uses a sliding window fast Fourier transform algorithm to perform high-frequency sampling and real-time spectrum analysis on the output voltage signal, and extracts the amplitude and phase information of the specified low-order harmonic components as the analysis results. The "dynamically adjusting the on-time of the switching transistors in the inverter to suppress waveform distortion within a preset threshold while increasing the fundamental voltage amplitude" in S1O4 specifically includes the following steps: S1042, Based on the current modulation coefficient, generate an initial modulation wave in the form of a quasi-square wave. A quasi-square wave is a waveform with its top clipped, which is between a sine wave and a pure square wave. With the total harmonic distortion not exceeding a preset threshold as a constraint, calculate the balance coefficient between the harmonic amplitude and the fundamental amplitude; When the harmonic amplitude is predicted to exceed a preset threshold based on the harmonic variation trend, the initial modulation waveform is optimized based on the balance coefficient to reduce the width of the flat-top segment of the initial modulation waveform, thereby suppressing the growth of low-order harmonics.

[0036] In implementation, the wind turbine harmonic monitoring system reads the target speed ω_ref of the wind turbine (given by the user or issued by the upper-level system) and compares it with the actual speed ω_actual estimated by the encoder or back EMF. The difference is then output as a q-axis current command Iq_ref to control the motor torque via a speed PI regulator. It should be noted that for single-phase wind turbines, control is typically simplified to controlling the direct and quadrature axis currents in a synchronous rotating coordinate system.

[0037] The wind turbine harmonic system compares the q-axis current command Iq_ref with the sampled actual current Iq. Its output is decoupled and calculated to finally generate voltage commands Vq and Vd in a two-phase rotating coordinate system. Then, through inverse Park transformation, the voltage commands Vq and Vd in the rotating coordinate system are converted into voltage vectors Uα and Uβ in a two-phase stationary coordinate system.

[0038] The amplitude of the voltage vector is the required amplitude of the fundamental voltage, U_ref = sqrt(Uα² + Uβ²); the phase angle of the voltage vector is θ = atan2(Uβ, Uα). For SPWM, the modulation coefficient M is calculated as: M = U_ref / (U_dc / 2), where U_dc is the DC bus voltage. This calculated value of M is the modulation coefficient required to achieve the current target speed.

[0039] The wind turbine harmonic monitoring system automatically enters the corresponding control range based on the real-time calculated M value: 1. Standard SPWM Region (0 < M ≤ 1, control mode is SPWM): This region is a linear modulation region, executing the standard SPWM algorithm. Upon entering the SPWM region, the wind turbine harmonic monitoring system calculates the instantaneous value U_ref[k] of the sinusoidal modulated wave every control cycle K. Its discrete-time expression is: U_ref[k] = M × sin(2 × π × f × k × T_s + θ); where M is the modulation coefficient mentioned above, f is the target fundamental frequency, proportional to the target rotational speed; T_s is the control cycle; k is the counting index of the current control cycle; and θ is the current phase angle (in radians) of the voltage vector.

[0040] Then, the wind turbine harmonic monitoring system internally generates a high-frequency triangular carrier U_tri[k]. Its discrete-time expression can be expressed as: U_tri[k] = (2 / π) * arcsin(sin(2 * π * f_c * k * T_s)); or, more commonly, it is implemented in the DSP through an incrementing and decrementing counter whose value varies linearly between -1 and +1, with a period of T_c = 1 / f_c (carrier period). Therefore, the triangular carrier value can be regarded as a variable that changes periodically in the range [-1, +1].

[0041] The wind turbine harmonic monitoring system compares U_ref[k] with U_tri[k]. If U_ref[k] > U_tri[k], the generated PWM signal for driving the upper tube (S1) of the first bridge arm is at an active level (high level), while the generated PWM signal for driving the lower tube (S4) of the first bridge arm is at an inactive level (low level). When U_ref[k] < U_tri[k], the generated PWM signal for driving S1 is at an inactive level, while the generated PWM signal for driving S4 is at an active level.

[0042] Simultaneously, a PWM signal is generated to drive the upper transistor (S3) of the second bridge arm, ensuring that it is always in the opposite logic state to the PWM signal driving S1. A PWM signal is also generated to drive the lower transistor (S2) of the second bridge arm, ensuring that it is always in the opposite logic state to the PWM signal driving S3.

[0043] A preset delay time (i.e., dead time) is inserted before the rising edge (the transition edge from invalid to valid) of all four PWM signals. During this dead time, it is ensured that the two complementary signals (such as S1 and S4) are both at invalid levels.

[0044] Finally, the four PWM signals generated above, with dead time inserted, are output to the gate drive circuit, ultimately driving the four switching transistors to operate.

[0045] 2. Overmodulation Zone (1 < M < Mmax, e.g., Mmax = 1.27, control mode is overmodulation): When the required speed of the wind turbine increases, and the required voltage exceeds the maximum sinusoidal fundamental voltage that the bus voltage can provide, this zone is entered. At this time, the amplitude of the modulation wave will exceed the peak value of the triangular carrier wave. At the trough and peak, the switching transistor will enter saturation conduction or cutoff state, forming a "flat peak" for a period of time. The output PWM wave changes from an SPWM wave to a "quasi-square wave". According to Fourier analysis, this waveform contains a larger fundamental voltage component than SPWM (thus increasing the speed), but also introduces significant low-order (e.g., 5th, 7th) harmonics. Specifically, after entering this zone, the wind turbine harmonic monitoring system no longer directly uses the standard sine wave, but instead calculates an overmodulated waveform U_mod[k]. An exemplary calculation scheme is as follows: Define A = M * π / 4; calculate the angle α = arcsin((1-2 / π*A) / (1-4 / π²*A)); for intervals such as 0 ≤ ωt < α and π-α ≤ ωt < π + α, the modulated wave is clamped: U_mod[k] = 1 (when 0 ≤ ωt < α); U_mod[k] = -1 (when π-α ≤ ωt < π + α). For other angle intervals, a modified sine segment is used: U_mod[k] = M * sin(ωt) / (cos(ωt) - cos(α)). This is an example of a known overmodulation algorithm, and its key parameters, such as the angle α_initial characterizing the width of the flat top, are documented. In this invention, the U_mod[k] generated in this step is the "initial modulation wave", that is, a quasi-square wave. Based on this "initial modulation wave", the wind turbine harmonic monitoring system will execute the following steps in sequence: "real-time harmonic detection based on improved FFT", "harmonic change trend prediction" and "precise optimization of switch conduction time based on prediction".

[0046] 3. Square Wave Region (M=Mmax, control mode is square wave): This is the modulation limit, outputting a square wave with a fixed duty cycle of 50%. In this region, the wind turbine harmonic monitoring system abandons the mechanism of comparing the modulated wave with the carrier wave, and instead adopts a direct lookup table method based on the voltage vector angle. Specifically, the wind turbine harmonic monitoring system obtains the current voltage vector angle θ (usually expressed as electrical angle, ranging from 0° to 360°) based on rotor position sensor information or through integration calculation, and then determines the 60° sector it belongs to based on the angle θ. For example: Sector I: 0°≤θ<60°, Sector II: 60°≤θ<120°, and so on; the wind turbine harmonic monitoring system directly reads the corresponding switch drive logic from a fixed switch state table based on the determined sector, and maintains this state until entering the next sector.

[0047] Example of switch status representation: Sector I: Drive S1=1, S4=1, S2=0, S3=0; Sector II: Drive S1=1, S2=1, S3=0, S4=0; Sector III: Drive S3=1, S2=1, S1=0, S4=0; Sector IV: drive S3=1, S4=1, S1=0, S2=0; Sector V: Drive S2=1, S4=1, S1=0, S3=0; Sector VI: Drive S2=1, S1=1, S3=0, S4=0; "1" represents a forced active level (on), and "0" represents a forced inactive level (off).

[0048] In addition, the wind turbine harmonic monitoring system also requires a dead time to be inserted when switching sectors, i.e., when the switching state is about to change. The controller first invalidates all drive signals of the switching transistors that are about to change, and after a dead time delay, outputs the drive signal corresponding to the new sector. Finally, the generated PWM signal is output to the gate drive circuit.

[0049] Specifically, the implementation principle of the above-mentioned "real-time harmonic detection based on improved FFT" steps is as follows: The wind turbine harmonic monitoring system synchronously samples the inverter output voltage U_ab(t) at a frequency f_sample during each PWM carrier cycle, obtaining a discrete sequence U_ab[n]. A pre-defined array of length N (N is typically a power of 2, such as 128, covering two fundamental cycles) is used as a sliding window. At the beginning of each new control cycle, the oldest set of data (e.g., 64 points) is removed from the window, and the latest 64 sampled points are added to the end of the window. This ensures that the data within the window is always up-to-date and achieves seamless data updates. A Hanning window is applied to the N-point data within the sliding window to reduce spectral leakage. An N-point FFT operation is then performed on the windowed data to obtain its spectrum F[k]. Then, based on the known fundamental frequency f1 and sampling frequency fs, the position k1 of the spectral line corresponding to the fundamental is calculated. The positions k5, k7, ... of the spectral lines of key lower harmonics such as the 5th and 7th harmonics are located. The amplitude of each harmonic is accurately calculated using an interpolation algorithm: H5 = 2 * |F[k5]| / N (multiplying by 2 recovers the single-sided spectral amplitude, dividing by N normalizes it). Then, the phase angle of the corresponding spectral line in the spectrum F[k] is extracted: φ5 = arg(F[k5]). This yields a clear and quantified analysis result: (H5, φ5). This analysis result is a data structure, for example: Harmonic_Data = {H1, H5, φ5, H7, φ7...}, which contains the real-time amplitude and phase information of the fundamental and key harmonics.

[0050] Next, the implementation principle of the "harmonic variation trend prediction" step is as follows: The wind turbine harmonic monitoring system is used to predict harmonic trends based on the aforementioned analysis results and a pre-set neural network algorithm. Specifically: In each control cycle, the wind turbine harmonic monitoring system inputs the current feature vector (X[k], including H5, φ5, and Iq[k], Id[k], ω[k], M[k]) into the deployed LSTM model. The LSTM model outputs the predicted harmonic amplitude H5_predicted, which indicates the possible future trend of harmonics. Specifically, this application pre-operates on the wind turbine under various speeds and load conditions in a laboratory environment, collecting a large amount of sample data. Each sample includes: input features: instantaneous / effective value of stator current, speed command / feedback, and estimated load torque; output label: harmonic amplitude / phase detected by the improved FFT for the next 1-2 cycles. This massive amount of data is used to train the LSTM model, enabling it to infer the future harmonic variation patterns from historical operating states. During the actual operation of wind turbines, the wind turbine harmonic monitoring system is used to input the operating parameters and harmonic data sequences of the most recent period (e.g., 10 control cycles before the current time) into the LSTM model so that the LSTM model can output the predicted values ​​of the key harmonic component amplitudes in the next 1-2 control cycles (i.e., harmonic amplitude H5_predicted). This allows the wind turbine harmonic monitoring system to "predict" the upcoming harmonic spikes.

[0051] Finally, the specific implementation principle of the "precise optimization of switch conduction time based on prediction" step is as follows: The wind turbine harmonic monitoring system has a pre-set total harmonic distortion threshold (H5_threshold, i.e., the preset threshold). The harmonic amplitude H5_predicted is compared with H5_threshold. When the LSTM model predicts that H5_predicted in a future cycle will exceed the total harmonic distortion threshold (H5_threshold), a dynamic adjustment operation is triggered: if H5_predicted > H5_threshold, a balance coefficient K is calculated as K = min(1, (H5_predicted - H5_threshold) / H5_threshold); K is between 0 and 1.

[0052] Then, a modulation wave correction operation is performed (this is the core of achieving "precise control," used to fine-tune the quasi-square wave U_mod[k] generated earlier, i.e., the specific operation of adjusting the switch conduction time). The specific operation is as follows: After generating the initial modulation wave U_mod, in the flat-top segment of U_mod[k] (i.e., the part where its value is equal to +1 or -1), a correction term based on K, Δ=K * β * sign(U_mod[k]), is introduced to optimize U_mod[k], resulting in U_mod_optimized[k]. Here, U_mod_optimized[k]=U_mod[k]-K * β * sign(U_mod[k]); β is an adjustable gain coefficient (e.g., 0.05), and sign is the sign function (sign function or signum function). Finally, the optimized U_mod_optimized[k] is compared with the triangular carrier U_tri[k] to generate the final PWM signal driving the switch. The specific driving operation is as follows: When U_mod_optimized[k] > U_tri[k], the upper bridge arm switch (S1) of the first bridge arm of the inverter is turned on, and the lower bridge arm switch (S4) of the same bridge arm is turned off. When U_mod_optimized[k] < U_tri[k], switch (S1) is turned off, and switch (S4) is turned on. Simultaneously, the switch states of the second bridge arm are reversed compared to the first bridge arm: when switch (S1) is turned on, the upper bridge arm switch (S3) of the second bridge arm is turned off, and its lower bridge arm switch (S2) is turned on. When switch (S1) is turned off, switch (S3) is turned on, and switch (S2) is turned off. A preset dead time is inserted before any switch transitions from the off state to the on state; during this dead time, the switch and its complementary switch are both turned off to prevent bridge arm shoot-through short circuits. Finally, the PWM signal processed by the above steps is sent to the gate drive circuit, which amplifies it and drives the corresponding switching transistor to operate.

[0053] Optionally, the step S104, "based on the real-time operating parameters and analysis results of the wind turbine, predicting the harmonic variation trend within the future control cycle using a preset prediction model," includes the following steps: S1043, based on the real-time operating parameters and analysis results of the wind turbine, predicts the amplitude of a specific harmonic, the value of switching loss, and the predicted value of torque pulsation within the future control cycle through a preset prediction model. The step S104, "Dynamically adjusting the on-time of the switching transistors in the inverter to increase the fundamental voltage amplitude while suppressing waveform distortion within a preset threshold," also includes the following steps: S1044, generate multiple candidate modulation wave correction values, and generate corresponding candidate modulation waves based on the correction values; wherein, each modulation wave correction value corresponds to the adjustment of the width of the quasi-square wave flat top segment. Based on a pre-stored empirical model describing the relationship between the modulation wave correction and the changes in various performance indicators, and the predicted values ​​of the specific harmonic amplitude, switching loss, and torque ripple in the future control cycle, the specific harmonic amplitude, switching loss, and torque ripple corresponding to each candidate modulation wave are estimated respectively. Based on the estimated specific harmonic amplitude, switching loss value, and torque ripple value, calculate the multi-objective cost of each candidate scheme; The candidate modulation wave that minimizes the multi-objective cost is selected and compared with the triangular carrier wave to generate the final PWM signal used to drive the inverter switching transistor, thereby realizing the dynamic adjustment of the switching transistor's on-time.

[0054] The multi-objective cost value is calculated as follows: w1 × estimated specific harmonic amplitude + w2 × estimated switching loss + w3 × torque ripple; where w1, w2, and w3 are weighting coefficients. The step "calculate the multi-objective cost value of each candidate scheme" specifically includes: calculating the dissatisfaction of the three performance indicators—current specific harmonic amplitude, switching loss, and torque ripple—where the dissatisfaction of any performance indicator is U = max(0, current value of the performance indicator − its preset expected value); summing the dissatisfaction of the performance indicators to obtain the total dissatisfaction; using the ratio of the dissatisfaction of each performance indicator to the total dissatisfaction as its corresponding weighting coefficients w1, w2, and w3, respectively, and using the finally obtained weighting coefficients w1, w2, and w3, calculating the multi-objective cost value of each candidate scheme.

[0055] In implementation, this application pre-establishes two-dimensional lookup tables for the turn-on energy Eon and turn-off energy Eoff within a switching cycle in a laboratory environment, based on datasheets or measured data. These two tables use "collector current Ic" and "DC bus voltage Udc" as inputs.

[0056] The wind turbine harmonic monitoring system samples Ic and Udc in real time during each PWM cycle; it then looks up Eon and Eoff under the current operating conditions using a table; and calculates the average switching loss Psw = (Eon + Eoff) × fsw for each switch in one PWM cycle, where fsw is a preset, fixed switching frequency. Simultaneously, it measures the switch loss Ptotal (conduction loss Pcond + switching loss Psw) and estimates the junction temperature Tj = Tc + Ptotal × Rth(j−c), where Rth(j−c) is the thermal resistance value obtained from the switch's (e.g., IGBT or MOSFET) datasheet; Tc is the case temperature measured by a temperature sensor (e.g., a thermistor) pre-installed on the inverter module's heatsink, close to the switch mounting point; and Ptotal is the total switching loss. The power loss is calculated as follows: Ptotal = conduction loss Pcond + average switching loss Psw; Pcond = Ic² + Rds(on); Rds(on) is the on-resistance. Rds(on) is obtained through a pre-stored two-dimensional lookup table. This lookup table uses junction temperature Tj and gate drive voltage Vgs as input variables. During development, typical Rds(on) values ​​at different junction temperatures Tj and gate drive voltages Vgs are obtained from the switch datasheet. This lookup table is then constructed and burned into the wind turbine harmonic monitoring system. In real-time control, the wind turbine harmonic monitoring system obtains accurate Rds(on) values ​​through the following steps: obtaining the actual or setpoint value of the current gate drive voltage Vgs; estimating the current junction temperature Tj using the method described above; using (Tj, Vgs) as an index to look up the two-dimensional table; and obtaining the accurate Rds(on) value through bilinear interpolation.

[0057] Furthermore, as mentioned above, the preset prediction model (LSTM model) will use the input feature vector X[k] as input to predict the output harmonic amplitude H5_predicted. Based on this, this application proposes that the original feature vector X[k] (including H5, φ5, and Iq[k], Id[k], ω[k], M[k]) further include the DC bus voltage Udc, the estimated junction temperature of the switching transistor Tj, and the estimated historical switching loss Psw.

[0058] Accordingly, in conjunction with the above, a large amount of sample data with the aforementioned feature vectors were collected in advance in a laboratory environment. The input of the sample data is the aforementioned feature vector X[k], and the output labels include not only the harmonic amplitude H5_predicted mentioned above, but also the switching loss value Psw_predicted and the torque ripple prediction value Torque_predicted. These sample data are used to train the LSTM model, enabling it to predict three output harmonic amplitude values: H5_predicted, switching loss value Psw_predicted, and torque ripple prediction value Torque_predicted. This minimizes the combined error Loss (such as the weighted root mean square of the three output errors) between the predicted and actual values. For example, Loss = α × MSE(H5_predicted, H5_actual) + β * MSE(Psw_predicted, Psw_actual) + γ * MSE(Torque_predicted, Torque_actual), where α, β, and γ refer to preset weights, and MSE(A, B) is the difference between A and B. Finally, the training of the LSTM model is completed.

[0059] Then, the wind turbine harmonic monitoring system is used to perform the following optimization process in each control cycle: 1. Obtaining Predicted Values: Input the currently obtained input feature vector X[k] into the pre-trained LSTM model so that the LSTM model predicts and outputs the harmonic amplitude H5_predicted, switching loss Psw_predicted, and torque ripple prediction value Torque_predicted for several future control cycles.

[0060] 2. Define the modulation wave correction Δ' as a decision variable. Its physical meaning is the adjustment amount of the angle α_initial of the flat top width of the initial modulation wave U_mod[k]. Within a reasonable range of values ​​[Δ'min, Δ'max], N candidate correction amounts are generated by uniform sampling method, forming a set {Δ'1, Δ'2, ..., Δ'N}.

[0061] 3. For each candidate correction Δ'i, 1≤i≤N, calculate its corresponding flat-top angle α_i = α_initial + Δ'i. Use this Δ'i as a new parameter and re-execute the standard overmodulation algorithm to generate a new, corrected modulation wave (i.e., candidate modulation wave) U_mod[k]_i.

[0062] 4. The wind turbine harmonic monitoring system pre-stores an empirical coefficient model that describes the relative impact of a unit change in the correction amount Δ' on three performance indicators (harmonic amplitude, switching loss, and torque ripple). Let the coefficients be a, c, and d, respectively. For each candidate correction amount Δ'i, its performance indicator is estimated using the following formula: H5_predicted_i= H5_predicted×(1+a×Δ'i); Psw_predicted_i=Psw_predicted×(1+c×Δ'i); Torque_ predicted_i=Torque_ predicted×(1+d×Δ'i).

[0063] Finally, for each candidate correction amount Δ'i, the estimated performance index is substituted into the preset objective cost function to calculate the cost value J_i: J_i = w1 × H5_predicted_i + w2 × Psw_predicted_i + w3 × Torque_predicted_i; where w1, w2, and w3 are weight coefficients ≥ 0.

[0064] 5. Compare all cost values ​​{J_1, J_2, ..., J_N}, and select the candidate modulation wave corresponding to the minimum cost value J as the final modulation wave of this control cycle. Finally, the modulation wave correction operation mentioned above (i.e., the operation of introducing a K-based correction term Δ=K*β*sign(U_mod[k]) to optimize U_mod[k] to obtain U_mod_optimized[k]) is omitted. Instead, the final modulation wave obtained here is directly compared with the triangular carrier wave to generate the final PWM signal driving the switch. The specific driving operation has been disclosed above.

[0065] Regarding the weight coefficients w1, w2, and w3 mentioned in the target cost function when calculating the target cost value, in other embodiments, w1, w2, and w3 can be preset values; however, the embodiments of this application propose that before calculating the target cost value, the optimal values ​​of the weight coefficients w1, w2, and w3 are first obtained as the current weight coefficient combination, and then the current weight coefficient combination is substituted into the multi-objective cost function to solve for the cost value of each candidate correction amount.

[0066] Specifically, the wind turbine harmonic monitoring system has three pre-stored expected values ​​for performance indicators (harmonic amplitude, switching loss value, and torque ripple) as the benchmark targets for optimization: Hdesired: the upper limit of the expected amplitude of a specific harmonic (such as the 5th harmonic), Psw_desired: the upper limit of the expected switching loss, and Trip_desired: the upper limit of the expected torque ripple.

[0067] In each control cycle, the wind turbine harmonic monitoring system obtains the current values ​​of three performance indicators by sampling sensors or calling the prediction model, and calculates the dissatisfaction of each indicator relative to its expected value, U=max(0, current performance value − expected value).

[0068] Specifically: Harmonic amplitude dissatisfaction UH=max(0,Hcurrent−Hdesired); Switching loss dissatisfaction: UP=max(0,Psw_current−Psw_desired); Torque ripple dissatisfaction: UT=max(0,Trip_current−Trip_desired).

[0069] The above formula ensures that when a certain performance is better than or equal to the expected value, the dissatisfaction is zero; only when the performance is worse than the expected value will positive dissatisfaction occur, and the larger the difference, the higher the dissatisfaction.

[0070] Deterministic weights are assigned through normalization: the wind turbine harmonic monitoring system calculates the sum of three dissatisfactions: Utotal = UH + UP + UT; subsequently, through normalization, the proportion of each dissatisfaction in the total dissatisfaction is directly determined as the weight coefficient of that performance indicator in the cost function. w1 = UH / Utotal; w2 = UP / Utotal; w3 = UT / Utotal. If Utotal = 0 (i.e. all performance has met expectations), the wind turbine harmonic monitoring system automatically adopts a set of default balance weights, for example (w1, w2, w3) = (1 / 3, 1 / 3, 1 / 3).

[0071] Optionally, the "generating an initial modulation wave in quasi-square wave form" in S1042 includes the following steps; Based on the preset modulation waveform library, the current modulation coefficient M, and the operating status of the wind turbine, the optimal initial modulation waveform type is selected from the modulation waveform library for the current control cycle. The modulation waveform library stores multiple different modulation waveforms for the same modulation coefficient M. Based on the selected initial modulation waveform type, an initial modulation waveform in the form of a quasi-square wave is generated.

[0072] During implementation, in the development phase, a modulation waveform library was pre-built and burned into the non-volatile memory of the wind turbine harmonic monitoring system through simulation and experimentation. The database construction process is as follows: For a series of discrete modulation coefficients M (e.g., from 1.05 to Mmax, with step sizes of 0.01 or 0.05), two quasi-square waves with different properties are generated respectively: Type A: Overmodulation algorithm waveform. This type of waveform uses, for example, a single-mode or dual-mode overmodulation algorithm to clip the sine wave by calculating the clipping angle α, thus forming a quasi-square wave. The harmonic components of this type of waveform are predominantly low-order (e.g., 5th, 7th). Type B: Third Harmonic Injection SPWM Waveform. By injecting a specific proportion of the third harmonic into a standard sine wave, the mid-voltage is raised, creating a flat-top effect. The harmonic components of this type of waveform are predominantly multiples of the third order (e.g., 3rd, 9th).

[0073] Then, the key generation parameters corresponding to each type of waveform are stored. For example, for type A, the clipping angle α (an angle value between 0 and π / 2 radians) is stored, and for type B, the third harmonic injection rate η (i.e., the ratio of the third harmonic amplitude to the fundamental amplitude, a decimal between 0 and 1 / 6) is stored. Each data entry is labeled with the corresponding modulation coefficient M, waveform type identifier, and its theoretical total harmonic distortion (THD) under ideal conditions.

[0074] In actual operation, the wind turbine harmonic monitoring system performs the following closed-loop operation in each control cycle K: 1. Based on the currently calculated required modulation coefficient M[k], retrieve all waveform data entries applicable to this M[k] value from the modulation waveform library to form a candidate list; 2. Based on the real-time operating status of the wind turbine (such as the magnitude of the load current and the growth trend of specific harmonics in historical harmonic data), an optimal waveform type is pre-selected from the candidate list. The pre-selection logic can be as follows: if the wind turbine harmonic monitoring system detects a persistently high level of the 5th harmonic recently, then waveform type B (3rd harmonic injection) is preferred. This is because its harmonic spectrum is characterized by multiples of the 3rd order, which can effectively avoid the 5th harmonic problem and improve the spectral structure from the root. Conversely, type A is preferred. The result of the selection is a waveform type identifier and its corresponding key generation parameter (α or η).

[0075] 3. The wind turbine harmonic monitoring system calculates and generates the initial modulation wave U_mod[k] in quasi-square wave form for the current control cycle in real time based on the selected waveform type and key generation parameters. Specifically: For type A, the wind turbine harmonic monitoring system obtains the electrical angle θ = ωt (ω is the electrical angular velocity, t is time) of the current voltage vector, and calculates the instantaneous value U_mod(θ) of the modulated wave according to the following rules based on the electrical angle θ and the clipping angle α: When θ belongs to the interval [0, α) or [π-α, π+α) or [2π-α, 2π), the modulated wave is clamped to the maximum value (1 after normalization) or the minimum value (-1); U_mod(θ) = sign(sin(θ)); When θ belongs to other intervals, the modulated wave enters the sinusoidal correction segment, and its value is calculated by the following formula: U_mod (θ)= M * sin(θ) / (M * cos(θ)- cos(α)); (This formula is a known formula of the standard overmodulation algorithm, used for smooth connection in the non-clamped region); Finally, traverse all discrete angle points of a fundamental period and repeat the above process to generate the initial modulated wave U_mod[k] for the entire period.

[0076] For type B, the wind turbine harmonic monitoring system is used to obtain the current electrical angle θ, and then calculate the fundamental component U_fundamental = sin(θ) and the third harmonic component U_third_harmonic = η*sin(3*θ); Then, the fundamental component and the third harmonic component are added to obtain the unnormalized modulation wave U_unnormalized = U_fundamental + U_third_harmonic; this unnormalized modulation wave is then normalized to limit its peak value to the range of [-1, +1] to ensure correct comparison with the triangular carrier wave. The normalization formula is: U_mod(θ) = U_unnormalized / (1+η); it should be noted that since the peak value of the waveform changes from 1 to (1+η) after the third harmonic is injected, it is divided by (1+η) to scale it back to the standard range. This operation is the key to improving the voltage utilization rate of this method; finally, the above process is repeated for all discrete angle points of a fundamental cycle to generate the initial modulation wave U_mod[k] with flat-top characteristics for the entire cycle.

[0077] Optionally, the statement in S101 that "when the modulation coefficient M satisfies: 0 < M ≤ 1, it is in the normal SPWM region; when the modulation coefficient M satisfies: 1 < M < Mmax, it is in the overmodulation region; when the modulation coefficient M = Mmax, it is in the square wave region" includes: When 0 < M < Mlow1, it is in the normal SPWM region; When Mlow1≤M≤Mhigh1, it is in the first transition zone; where Mlow1<1<Mhigh1; When Mhigh1 < M < Mlow2, it is in the overmodulation region; where Mlow2 < Mmax; When Mlow2≤M≤Mhigh2, it is in the second transition region; where Mhigh2=Mmax; The wind turbine harmonic monitoring method also includes the following steps: When the current modulation coefficient M is within the transition region, a smooth transition operation is performed: standard modulation waveforms of two standard partitions adjacent to the transition region are generated in parallel; based on the relative position of the current modulation coefficient M within the current transition region, the mixing weight coefficient ρ is calculated; the two parallel-generated standard modulation waveforms are weighted and fused using the mixing weight coefficient, and the fused waveform is used as the modulation wave to generate the PWM signal driving the inverter; wherein, the transition region is within the first transition region or the second transition region; the standard partitions include the conventional SPWM region, the overmodulation region, and the square wave region.

[0078] In implementation, a first transition region (SPWM-overmodulation transition region) and a second transition region (overmodulation-square wave transition region) are predefined. The first transition region (SPWM overmodulation transition region) is defined in the interval Mlow1≤M≤Mhigh1, which covers the boundary of the standard SPWM region (M=1); for example: Mlow1=0.95, Mhigh1=1.05.

[0079] The second transition region (overmodulation-square wave transition region) is defined in the interval Mlow2≤M≤Mhigh2, which covers Mmax, i.e., the boundary between the overmodulation region and the square wave region; for example: Mlow2=Mmax-0.1, Mhigh2= Mmax.

[0080] The following section takes the first transition zone (SPWM-overmodulation transition zone) as an example to explain in detail the implementation steps of its smooth transition. The implementation steps of the smooth transition in the second transition zone are completely similar. Simply replace the SPWM waveform with the overmodulation waveform and the square wave with the pure square wave.

[0081] Step 1: Calculate the hybrid weighting coefficient ρ: After calculating the current modulation coefficient M, the wind turbine harmonic monitoring system first determines whether it is in the transition zone. If M is in the first transition zone (Mlow1≤M≤Mhigh1), then a hybrid weighting coefficient ρ=(M-Mlow1) / (Mhigh1-Mlow1) is calculated; if M is in the second transition zone (Mlow2≤M≤Mhigh2), the corresponding hybrid weighting coefficient ρ=(M-Mlow2) / (Mhigh2-Mlow2).

[0082] Step 2: Synchronously generate two reference waveforms: The wind turbine harmonic monitoring system is used to generate two reference modulation waves in parallel when the current M is in the first transition zone: 1. Standard SPWM modulation wave (U_SPWM): using the parameter M=1, a standard sine modulation wave with an amplitude of 1 is generated; 2. Overmodulation modulation wave (U_mod): using the current actual M value, the corresponding overmodulation quasi-square wave is generated through the standard overmodulation algorithm (such as the method of calculating the clipping angle α mentioned above).

[0083] The wind turbine harmonic monitoring system generates two reference modulation waves in parallel when the current M is in the second transition zone: 1. Overmodulated modulation wave (U_mod): Using the current actual M value, a corresponding overmodulated quasi-square wave (U_mod) is generated through a standard overmodulation algorithm (such as the method for calculating the clipping angle α described above). 2. Pure square wave: A standard pure square wave (U_Square) with an amplitude of ±1 is generated, and its phase is synchronized with the current voltage vector angle.

[0084] Step 3: The weighted coefficient ρ obtained from the practical calculation is used to weight and fuse the two reference modulation waves obtained in Step 2 to generate the final modulation wave U_Final. For example, when M is in the first transition zone, U_Final = (1-ρ)×U_SPWM +ρ×U_mod; when M is in the second transition zone, U_Final = (1-ρ)×U_mod +ρ×U_Square.

[0085] Step 4: Send the synthesized final modulated wave U_Final to the PWM comparator, compare it with the triangular carrier wave, and then generate the PWM signal to drive the switching transistor based on the comparison result and the specific driving operation disclosed above.

[0086] This application also discloses a single-phase wind turbine harmonic monitoring system based on an overmodulation algorithm. (Refer to...) Figure 2 ,include: The dynamic partition control module 201 is used to dynamically divide the control mode of the inverter into at least a conventional SPWM region, an overmodulation region, and a square wave region based on the range of values ​​of the modulation coefficient M. Specifically, when the modulation coefficient M satisfies: 0 < M ≤ 1, it is in the conventional SPWM region; when the modulation coefficient M = Mmax, it is in the square wave region; and when the modulation coefficient M satisfies: 1 < M < Mmax, it is in the overmodulation region. The real-time harmonic detection module 202 is used to perform closed-loop harmonic management whenever the inverter is in the overmodulation zone: to sample and analyze the inverter's output voltage in real time and obtain the analysis results; The harmonic trend prediction module 203 is used to predict the harmonic change trend within the future control cycle based on the real-time operating parameters and analysis results of the wind turbine and through a preset prediction model; wherein, the real-time operating data of the wind turbine includes at least the motor stator current, wind turbine speed and load data. The wind turbine precision control module 204 is used to generate an SPWM wave by comparing a sine wave with a triangular carrier wave whenever it is in the normal SPWM zone, and use this as the PWM signal to drive the inverter; it is also used to control each switch of the inverter to conduct once per fundamental cycle whenever it is in the square wave zone; and it is also used to dynamically adjust the conduction time of the switch in the inverter based on the harmonic analysis results and harmonic variation trend whenever it is in the overmodulation zone, so as to increase the voltage fundamental amplitude while suppressing waveform distortion within a preset threshold.

[0087] Optionally, the harmonic real-time detection module 202 is also used to perform high-frequency sampling and real-time spectrum analysis on the output voltage signal using a sliding window fast Fourier transform algorithm, and extract the amplitude and phase information of the specified low-order harmonic components as the analysis results.

[0088] Optionally, the wind turbine precision control module 204 is also used to generate an initial modulation wave in the form of a quasi-square wave based on the current modulation coefficient. A quasi-square wave refers to a waveform with a flattened top, which is between a sine wave and a pure square wave. With the total harmonic distortion not exceeding a preset threshold as a constraint, the balance coefficient between the harmonic amplitude and the fundamental amplitude is calculated. When it is predicted that the harmonic amplitude will exceed the preset threshold based on the harmonic variation trend, the initial modulation wave waveform is optimized based on the balance coefficient to reduce the width of the flat-top segment of the initial modulation wave waveform, thereby suppressing the growth of low-order harmonics.

[0089] Optionally, the harmonic trend prediction module 203 is also used to predict the amplitude of a specific harmonic, the switching loss value, and the torque pulsation prediction value within the future control cycle based on the real-time operating parameters and analysis results of the wind turbine and through a preset prediction model. The wind turbine precision control module 204 is also used to generate multiple candidate modulation wave correction values ​​and generate corresponding candidate modulation waves based on the correction values. Each modulation wave correction value corresponds to the adjustment of the width of the quasi-square wave flat top segment. Based on a pre-stored empirical model describing the relationship between the modulation wave correction value and the changes in various performance indicators, as well as the predicted amplitude of the specific harmonic, the switching loss value, and the torque ripple prediction value in the future control cycle, the specific harmonic amplitude, the switching loss value, and the torque ripple value corresponding to each candidate modulation wave are estimated respectively. Based on the estimated specific harmonic amplitude, the switching loss value, and the torque ripple value, the multi-objective cost value of each candidate scheme is calculated. The candidate modulation wave that minimizes the multi-objective cost value is selected, and the candidate modulation wave is compared with the triangular carrier wave to generate the final PWM signal used to drive the inverter switching transistor, thereby realizing the dynamic adjustment of the switching transistor's on-time.

[0090] Optionally, the wind turbine precision control module 204 is also used to calculate the dissatisfaction of three performance indicators: current specific harmonic amplitude, switching loss, and torque pulsation. The dissatisfaction of any performance indicator is U=max(0, the current value of the performance indicator − its preset expected value). The dissatisfaction of the performance indicators is summed to obtain the total dissatisfaction. The ratio of the dissatisfaction of each performance indicator to the total dissatisfaction is used as its corresponding weight coefficients w1, w2, and w3, respectively. The multi-objective cost value of each candidate scheme is calculated using the finally obtained weight coefficients w1, w2, and w3.

[0091] Optionally, the wind turbine precision control module 204 is also used to select the optimal initial modulation waveform type for the current control cycle from the modulation waveform type library based on the preset modulation waveform type library, the current modulation coefficient M, and the operating status of the wind turbine; wherein, the modulation waveform type library stores a variety of different modulation waveforms for the same modulation coefficient M; based on the selected initial modulation waveform type, an initial modulation waveform in the form of a quasi-square wave is generated.

[0092] Optionally, a smooth transition module is also included, which performs a smooth transition operation when the current modulation coefficient M is in the transition region: generating standard modulation waveforms of two standard partitions adjacent to the transition region in parallel; calculating the mixing weight coefficient ρ based on the relative position of the current modulation coefficient M in the current transition region; using the mixing weight coefficient to perform weighted fusion of the two parallel-generated standard modulation waveforms, and using the fused waveform as the modulation wave to generate the PWM signal driving the inverter; wherein, when the transition region is in the first transition region or the second transition region; the standard partitions include the conventional SPWM region, the overmodulation region, and the square wave region.

[0093] This application also discloses a single-phase wind turbine harmonic monitoring device based on an overmodulation algorithm. The single-phase wind turbine harmonic monitoring device based on the overmodulation algorithm includes a memory and a processor. The memory stores a computer program that can be loaded by the processor and executed as described above for the single-phase wind turbine harmonic monitoring method based on the overmodulation algorithm.

[0094] This application also discloses a computer-readable storage medium that stores a computer program that can be loaded by a processor and executed as described above regarding the single-phase wind turbine harmonic control method based on an overmodulation algorithm. The computer-readable storage medium includes, for example, various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0095] It should be noted that in this paper, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0096] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit the scope of protection of the application. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on these embodiments, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

Claims

1. A harmonic control method for single-phase wind turbines based on an overmodulation algorithm, characterized in that, include: Based on the range of values ​​for the modulation coefficient M, the control mode of the inverter is dynamically divided into at least the conventional SPWM region, the overmodulation region, and the square wave region. Specifically, when the modulation coefficient M satisfies: 0 < M ≤ 1, it is in the conventional SPWM region; when the modulation coefficient M satisfies: 1 < M < Mmax, it is in the overmodulation region; and when the modulation coefficient M = Mmax, it is in the square wave region. Whenever the inverter is in the normal SPWM zone, a sine wave is compared with a triangular carrier wave to generate an SPWM wave, which is used as the PWM signal to drive the inverter. Whenever the inverter is in the square wave region, each switch of the control inverter is turned on once per fundamental wave cycle; Whenever the inverter is in the overmodulation zone, closed-loop harmonic management is performed: the inverter's output voltage is sampled and harmonic analyzed in real time to obtain the analysis results; based on the real-time operating parameters of the wind turbine and the analysis results, the harmonic variation trend within the future control cycle is predicted using a preset prediction model; wherein, the real-time operating data of the wind turbine includes at least the motor stator current, wind turbine speed, and load data; based on the harmonic analysis results and the harmonic variation trend, the on-time of the switching transistors in the inverter is dynamically adjusted to increase the fundamental voltage amplitude while suppressing waveform distortion within a preset threshold.

2. The single-phase wind turbine harmonic monitoring method based on overmodulation algorithm according to claim 1, characterized in that, The real-time sampling and harmonic analysis of the inverter's output voltage yields the following analysis results: The sliding window fast Fourier transform algorithm is used to perform high-frequency sampling and real-time spectrum analysis on the output voltage signal, and the amplitude and phase information of the specified low-order harmonic components are extracted as the analysis results.

3. The single-phase wind turbine harmonic monitoring method based on overmodulation algorithm according to claim 1, characterized in that, The dynamic adjustment of the on-time of the switching transistors in the inverter, to increase the fundamental voltage amplitude while suppressing harmonic distortion within a preset threshold, includes: Based on the current modulation coefficient, an initial modulation wave in the form of a quasi-square wave is generated. The quasi-square wave refers to a waveform with its top clipped, which is between a sine wave and a pure square wave. With the total harmonic distortion not exceeding a preset threshold as a constraint, calculate the balance coefficient between the harmonic amplitude and the fundamental amplitude; When the harmonic amplitude is predicted to exceed a preset threshold based on the harmonic variation trend, the initial modulation waveform is optimized based on the balance coefficient to reduce the width of the flat-top segment of the initial modulation waveform, thereby suppressing the growth of low-order harmonics.

4. The single-phase wind turbine harmonic monitoring method based on overmodulation algorithm according to claim 1, characterized in that, Based on the real-time operating parameters of the wind turbine and the analysis results, a preset prediction model is used to predict the harmonic variation trend within the future control cycle, including: Based on the real-time operating parameters of the wind turbine and the analysis results, the amplitude of a specific harmonic, the switching loss, and the torque pulsation prediction value are predicted in the future control cycle using a preset prediction model. The dynamic adjustment of the on-time of the switching transistors in the inverter, to increase the fundamental voltage amplitude while suppressing waveform distortion within a preset threshold, includes: Multiple candidate modulation wave correction values ​​are generated, and corresponding candidate modulation waves are generated based on the correction values; wherein each modulation wave correction value corresponds to the adjustment of the width of the quasi-square wave flat top segment. Based on a pre-stored empirical model describing the relationship between the modulation wave correction and the changes in various performance indicators, and the predicted values ​​of the specific harmonic amplitude, switching loss, and torque ripple in the future control cycle, the specific harmonic amplitude, switching loss, and torque ripple corresponding to each candidate modulation wave are estimated respectively. Based on the estimated amplitude of the specific subharmonic, the switching loss value, and the torque ripple value, calculate the multi-objective cost of each candidate scheme; The candidate modulation wave that minimizes the multi-objective cost is selected, and the candidate modulation wave is compared with a triangular carrier wave to generate the final PWM signal used to drive the inverter switching transistor, thereby realizing the dynamic adjustment of the switching transistor's on-time.

5. The single-phase wind turbine harmonic monitoring method based on overmodulation algorithm according to claim 4, characterized in that, The multi-objective cost = w1 × estimated amplitude of the specific subharmonic + w2 × estimated switching loss + w3 × torque ripple; where w1, w2 and w3 are weighting coefficients; The calculation of the multi-objective cost value for each candidate solution includes: Calculate the dissatisfaction with the three performance indicators of the current specific harmonic amplitude, switching loss and torque ripple, where the dissatisfaction with any performance indicator is U=max(0, the current value of the performance indicator − its preset expected value); sum the dissatisfaction with the performance indicators to obtain the total dissatisfaction. The ratio of dissatisfaction with each performance indicator to the total dissatisfaction is used as its corresponding weight coefficients w1, w2, and w3, respectively. The multi-objective cost value of each candidate solution is calculated using the final weight coefficients w1, w2, and w3.

6. The single-phase wind turbine harmonic monitoring method based on overmodulation algorithm according to claim 3, characterized in that, The initial modulation wave in the form of a quasi-square wave includes: Based on a preset modulation waveform library, the current modulation coefficient M, and the operating status of the wind turbine, the optimal initial modulation waveform type is selected from the modulation waveform library for the current control cycle; wherein, the modulation waveform library stores multiple different modulation waveforms for the same modulation coefficient M; based on the selected initial modulation waveform type, an initial modulation waveform in the form of a quasi-square wave is generated.

7. The single-phase wind turbine harmonic monitoring method based on overmodulation algorithm according to claim 1, characterized in that, When the modulation coefficient M satisfies: 0 < M ≤ 1, it is within the conventional SPWM region; When the modulation coefficient M satisfies: 1 < M < Mmax, it is in the overmodulation region; When the modulation coefficient M = Mmax, it is in the square wave region; including: When 0 < M < Mlow1, it is in the normal SPWM region; When Mlow1≤M≤Mhigh1, it is in the first transition zone; where Mlow1<1<Mhigh1; When Mhigh1 < M < Mlow2, it is in the overmodulation region; where Mlow2 < Mmax; When Mlow2≤M≤Mhigh2, it is in the second transition region; where Mhigh2=Mmax; The method further includes: When the current modulation coefficient M is within the transition region, a smooth transition operation is performed: standard modulation waveforms of two standard partitions adjacent to the transition region are generated in parallel; based on the relative position of the current modulation coefficient M within the current transition region, a mixing weighting coefficient ρ is calculated; the two parallel-generated standard modulation waveforms are weighted and fused using the mixing weighting coefficient, and the fused waveform is used as the modulation wave to generate the PWM signal driving the inverter; wherein, the transition region is within the first transition region or the second transition region; the standard partitions include the conventional SPWM region, the overmodulation region, and the square wave region.

8. A single-phase wind turbine harmonic monitoring system based on an overmodulation algorithm, characterized in that, include, The dynamic partition control module (201) is used to dynamically divide the control mode of the inverter into at least a conventional SPWM region, an overmodulation region, and a square wave region based on the range of values ​​of the modulation coefficient M. When the modulation coefficient M satisfies: 0 < M ≤ 1, it is in the conventional SPWM region; when the modulation coefficient M = Mmax, it is in the square wave region; when the modulation coefficient M satisfies: 1 < M < Mmax, it is in the overmodulation region. The real-time harmonic detection module (202) is used to perform closed-loop harmonic management whenever the inverter is in the overmodulation zone: to sample and analyze the inverter's output voltage in real time and obtain the analysis results; The harmonic trend prediction module (203) is used to predict the harmonic change trend in the future control cycle based on the real-time operating parameters of the wind turbine and the analysis results, through a preset prediction model; wherein, the real-time operating data of the wind turbine includes at least the motor stator current, wind turbine speed and load data. The wind turbine precision control module (204) is used to generate an SPWM wave by comparing a sine wave with a triangular carrier wave whenever it is in the normal SPWM zone, and use it as the PWM signal to drive the inverter; it is also used to control each switch of the inverter to conduct once per fundamental cycle whenever it is in the square wave zone; it is also used to dynamically adjust the conduction time of the switch in the inverter based on the harmonic analysis results and the harmonic change trend whenever it is in the overmodulation zone, so as to increase the voltage fundamental amplitude while suppressing waveform distortion within a preset threshold.

9. A single-phase wind turbine harmonic monitoring device based on an overmodulation algorithm, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executed as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer program is stored that can be loaded by a processor and executed as described in any one of claims 1 to 7.

Images