A wind turbine generator control method for large-scale synchronous power grid

Abstract

This invention discloses a wind turbine control method for large-scale synchronous power grids, relating to the field of power system renewable energy grid-connected control technology. By constructing a dual-input fuzzy inference unit, it achieves full-condition disturbance identification based on the magnitude and direction of frequency deviation and frequency change rate, distinguishing between falling disturbances, rising disturbances, and recovery states, and determining three levels of disturbance intensity: weak, medium, and strong. It employs a virtual inertia time constant that dynamically adapts to the frequency change rate to achieve inertia support matching with disturbance intensity; it suppresses power surges through leading and trailing edge smoothing filtering; and it automatically locks inertia and strengthens primary frequency regulation during the frequency recovery phase, while simultaneously constructing multi-level safety constraints in conjunction with speed margin. This application can adaptively match the disturbance characteristics of large power grids, solving problems such as unreasonable VSG inertia input, mismatched frequency regulation response, and easy deterioration during frequency recovery, thus balancing frequency support effectiveness and unit operation safety.

Description

Technical Field

[0001] This invention relates to the field of power system new energy grid connection control technology, and in particular to a wind turbine control method for large-scale synchronous power grids. Background Technology

[0002] With the deepening of the global energy structure transformation, the installed capacity and penetration rate of new energy power generation such as wind power in the power system continue to rise. In large-scale synchronous grids, the large-scale integration of wind power is mainly achieved through power electronic converters. This grid connection method has led to a relative decrease in the proportion of traditional synchronous generator sets, while the widespread integration of power electronic interface units has resulted in a significant reduction in the system's equivalent rotational inertia. The reduction in system inertia directly weakens the grid's ability to resist power disturbances. When disturbance events such as UHVDC blocking, sudden load changes, or large fluctuations in new energy power occur, the dynamic characteristics of the grid frequency change change significantly, the frequency change rate increases, the stability domain narrows, and the risk of frequency instability increases significantly, seriously threatening the safe and stable operation of the power system.

[0003] Virtual synchronous machine (VSM) technology, as an effective means to solve the grid connection stability problem of new energy sources, simulates the inertial response and primary frequency regulation characteristics of synchronous generators in the control strategy of wind turbine units. This enables power electronic interface units to actively participate in grid frequency regulation, providing necessary frequency support for the system. The core of VSM control lies in the organic combination of virtual inertia control and primary frequency regulation control. Virtual inertia mainly suppresses rapid frequency changes in the early stages of disturbances by releasing stored rotational kinetic energy, buying time for primary frequency regulation response. Primary frequency regulation, on the other hand, continuously provides power support through the droop relationship between active power and frequency deviation, gradually restoring system power balance. The two control mechanisms work together to maintain grid frequency stability.

[0004] However, existing virtual synchronous machine control methods for wind turbines have significant limitations in practical applications. Because the parameters of virtual inertia and primary frequency regulation are typically fixed based on typical operating conditions or empirical values, the control strategy cannot adaptively adjust to the actual intensity and type of grid disturbances, leading to a severe mismatch between control performance and disturbance characteristics. Specifically, in weak disturbance scenarios, excessively large inertia and frequency regulation parameters can generate unnecessary power surges, causing frequency overshoot and increased mechanical stress on the equipment. In strong disturbance scenarios, conservative parameter settings result in insufficient inertia support and slow frequency regulation response, failing to effectively suppress drastic frequency fluctuations. More importantly, existing technologies provide a rather coarse identification of disturbance states, typically relying solely on a single threshold for frequency deviation. This fails to fully consider the rich state information inherent in the magnitude and direction of frequency deviation and rate of change, making it difficult to accurately distinguish between different operating stages such as frequency drop, rise, bottoming out and recovery, and fallback recovery. This inaccurate disturbance identification further exacerbates the discrepancy between the control strategy and actual needs, resulting in a large inertia output being maintained during the frequency recovery phase, leading to over-support. This, in turn, causes a secondary frequency drop or overshoot, delaying the frequency recovery process. Furthermore, existing virtual synchronous machine controls generally lack safety constraint mechanisms linked to the wind turbine's own operating state, failing to fully consider the limiting effect of speed margin on inertia release and frequency regulation output. When the wind turbine is operating in the speed boundary region, forcibly engaging inertia or frequency regulation support may lead to safety accidents such as turbine stall, overcurrent, or grid disconnection.

[0005] In summary, the core problem with current virtual synchronous machine (VRM) control technology for wind turbines lies in the insufficient adaptive coordination between inertia and frequency regulation, making it impossible to achieve precise dynamic control based on grid disturbance characteristics and turbine operating status. This technical bottleneck severely restricts the effectiveness of VRMs in frequency stability control of large-scale synchronous grids. There is an urgent need to propose a VRM control method that can adaptively match disturbance characteristics under all operating conditions, achieve coordinated coordination between inertia and frequency regulation, and ensure turbine operating safety, in order to meet the higher requirements for frequency stability control in large-scale synchronous grids with high wind power integration. Summary of the Invention

[0006] This invention addresses the shortcomings of existing wind turbine virtual synchronous machines (VSGs), such as fixed virtual inertia and frequency regulation parameters, unreasonable timing of inertia input, low disturbance identification accuracy, and imperfect safety constraint mechanisms. It provides a wind turbine VSG virtual inertia-primary frequency regulation adaptive cooperative control method for large-scale synchronous power grids. By constructing a full-condition adaptive disturbance identification mechanism, a dynamic virtual inertia adjustment mechanism, a refined cooperative allocation mechanism for inertia and frequency regulation, and a multi-level safety constraint mechanism based on speed margin, it achieves adaptive cooperative control of virtual inertia and primary frequency regulation. This effectively solves technical problems in existing technologies, such as mismatch between control parameters and disturbance intensity, easy deterioration during frequency recovery, severe power surges, and insufficient protection for safe operation of the turbine.

[0007] The wind turbine VSG virtual inertia-primary frequency regulation adaptive cooperative control method for large-scale synchronous power grids provided by this invention includes the following technical features:

[0008] Signal acquisition and feature calculation steps: Real-time acquisition of the grid instantaneous frequency f, wind turbine speed ω, and current active power output P; calculation of the frequency deviation Δf = f - f0, where f0 is the rated frequency; calculation of the frequency change rate df / dt; calculation of the speed margin Δω = ω - ω_min, where ω_min is the wind turbine's safe minimum speed. These signal acquisition and feature calculation steps provide the necessary raw signals and feature quantities for subsequent fuzzy inference and control decision-making.

[0009] The dual-input fuzzy inference discrimination steps are as follows: Construct a fuzzy inference unit with frequency deviation Δf and frequency change rate df / dt as dual inputs. Based on the magnitude of |Δf|, it is divided into three levels: small, medium, and large. The small level corresponds to |Δf| < 0.1Hz, the medium level corresponds to 0.1Hz ≤ |Δf| < 0.2Hz, and the large level corresponds to |Δf| ≥ 0.2Hz. Based on the magnitude of |df / dt|, it is divided into three levels: slow change, fast change, and abrupt change. The slow change level corresponds to |df / dt| < 0.5Hz / s, the fast change level corresponds to 0.5Hz / s ≤ |df / dt| < 1.0Hz / s, and the abrupt change level corresponds to |df / dt| ≥ 1.0Hz / s. The fuzzy inference unit comprehensively judges the frequency state based on the sign direction of Δf and df / dt, distinguishing the following four operating conditions: when Δf < 0 and df / dt < 0, it is determined to be a downward disturbance operating condition, indicating that the frequency is lower than the rated value and continues to decrease; when Δf < 0 and df / dt ≥ 0, it is determined to be a bottoming out or recovery operating condition, indicating that the frequency is lower than the rated value but stops decreasing and begins to rise; when Δf > 0 and df / dt > 0, it is determined to be an upward disturbance operating condition, indicating that the frequency is higher than the rated value and continues to rise; when Δf > 0 and df / dt ≤ 0, it is determined to be an upward fall or recovery operating condition, indicating that the frequency is higher than the rated value but stops rising and begins to fall. The fuzzy inference unit determines three levels of disturbance: weak disturbance, medium disturbance, and strong disturbance according to the disturbance intensity judgment rule. Weak disturbance corresponds to a combination where |Δf| is small and |df / dt| changes slowly. Medium disturbance corresponds to all operating conditions except weak and strong disturbance. Strong disturbance corresponds to a combination where |Δf| is large or |df / dt| changes drastically. The fuzzy inference unit outputs the inertia allowable output coefficient Kv and the frequency modulation gain correction coefficient Kf_adapt according to the preset fuzzy inference rules. In the recovery state, i.e., the condition of falling to the bottom / recovery or rising and falling / recovery, Kv=0 is set to lock the inertia output, and Kf_adapt is taken as [1.1,1.2] to strengthen the primary frequency modulation support. In the disturbance state, i.e., the condition of falling disturbance or rising disturbance, the output is graded according to the intensity: when the disturbance is weak, Kv∈[0,0.2] and Kf_adapt∈[1.0,1.1]; when the disturbance is medium, Kv∈[0.4,0.7] and Kf_adapt∈[1.1,1.2]; when the disturbance is strong, Kv∈[0.8,1.0] and Kf_adapt∈[1.2,1.3].

[0010] The variable virtual inertia control steps are as follows: First, the virtual inertia time constant is adaptively adjusted based on the magnitude of df / dt, calculated using the formula TJ = TJ0·(1+λ·|df / dt|), where TJ0 is the base virtual inertia time constant, ranging from 2s to 4s, λ is the inertia enhancement coefficient, ranging from 0.5 to 1.0, and |df / dt| is the absolute value of the frequency change rate. Second, inertia support power is generated based on the adaptive virtual inertia time constant, calculated using the formula Pv = -TJ·(df / dt)·PN / f0, where PN is the VSG rated power and f0 is the rated frequency. This step increases the virtual inertia time constant with increasing frequency change rate, ensuring stronger inertia support during more severe disturbances and maintaining a smaller base inertia during weak disturbances to avoid unnecessary power surges.

[0011] The adaptive primary frequency modulation step involves generating a base frequency modulation power Pf0 and correcting it using a frequency modulation gain correction coefficient Kf_adapt to obtain the adaptive frequency modulation power Pf = Kf_adapt·Pf0. This allows the frequency modulation gain to adaptively adjust with the disturbance strength. This step increases the frequency modulation gain to strengthen frequency support during strong disturbances and maintains moderate frequency modulation to avoid over-modulation during weak disturbances.

[0012] Leading and trailing edge smoothing filtering steps: The inertial support power Pv is subjected to leading and trailing edge smoothing filtering. During the inertial input phase, a first-order inertial filter is used, with the filtering equation being Pv_limit = Pv / (1 + Ts·s), where Ts is the filtering time constant, ranging from 0.05s to 0.2s. During the inertial output phase, a second-order damping filter is used to smoothly reduce Pv_limit to zero. This combined smoothing strategy ensures that the inertial output increases slowly during input and smoothly returns to zero during output, avoiding sudden power surges that could impact the power grid and the wind turbine's mechanical system.

[0013] Total Additional Power Synthesis Step: The total additional power is synthesized by weighting the coefficients output by fuzzy inference. The calculation formula is ΔP=Kv·Pv_limit+Pf, where Kv is the inertia allowable deployment coefficient, Pv_limit is the smoothed inertia support power, and Pf is the adaptive frequency modulation power. This step weights and synthesizes the inertia support and primary frequency modulation according to the cooperative allocation coefficient to form a complete frequency support additional power command.

[0014] Safety constraints and VSG execution steps: A multi-level safety constraint domain is constructed based on the turbine speed margin Δω. The total additional power ΔP is then constrained and corrected before being sent to the VSG active power control loop for execution. Specific constraint rules include: when the speed margin Δω < 0.1 pu, the inertia support output is forcibly locked; when Δω < 0.05 pu, the primary frequency regulation output is limited to 50% of the rated regulation; the rate of change of total additional power |dP / dt| is limited to ≤ 0.2 pu / s; and ΔP is limited and corrected before being sent to the VSG active power control loop. This safety constraint mechanism determines the turbine's operating status through the speed margin and automatically locks the inertia or reduces the frequency regulation under low-speed conditions to prevent turbine stall, overcurrent, or grid disconnection.

[0015] Furthermore, the present invention also provides a wind turbine VSG control system, including a frequency acquisition module, a speed acquisition module, a fuzzy inference module, a variable inertia module, an adaptive frequency modulation module, a safety constraint module, and a VSG main control module. The frequency acquisition module and the speed acquisition module acquire the grid frequency signal and the wind turbine speed signal, respectively, and send them to the fuzzy inference module and the variable inertia module. The fuzzy inference module outputs the inertia projection coefficient and the frequency modulation gain correction coefficient based on the frequency deviation and the frequency change rate. The variable inertia module calculates the virtual inertia time constant and generates the inertia support power based on the frequency change rate. The adaptive frequency modulation module generates the adaptive frequency modulation power based on the frequency modulation gain correction coefficient. The safety constraint module performs constraint correction on the total additional power in combination with the speed margin. The VSG main control module receives the corrected total additional power and sends it to the VSG active power control loop for execution.

[0016] The beneficial effects of this invention are:

[0017] This invention, through a dual-input fuzzy inference unit, comprehensively considers the magnitude and direction of frequency deviation and frequency change rate, achieving complete differentiation of four operating conditions: frequency drop disturbance, frequency rise disturbance, frequency drop bottoming-out recovery, and frequency rise-falling-out recovery. It also provides refined discrimination of weak, medium, and strong disturbance intensities, significantly improving control accuracy. By using a virtual inertia time constant that adapts in real-time to the frequency change rate, it achieves dynamic matching between inertia support and disturbance intensity. This ensures that stronger disturbances result in larger inertia to quickly smooth frequency fluctuations, while maintaining smaller inertia during weak disturbances to avoid power surges. Furthermore, by automatically setting the inertia deployment coefficient to zero during the recovery phase and retaining only the enhanced primary frequency modulation, it addresses the underlying mechanism... This avoids the problems of secondary frequency drop, overshoot, and slow recovery caused by excessive inertia output; the combined smoothing strategy of first-order inertial filtering and second-order damping filtering effectively suppresses power surges during the inertia input and output phases, reducing the mechanical stress of the wind turbine and the output oscillation of the converter; the multi-level safety constraint mechanism combined with speed margin achieves dual protection of grid frequency support and wind turbine safety; by forming a coordinated control mechanism with frequency regulation as the main focus and inertia as the auxiliary focus, it conforms to the operating characteristics of large synchronous grids with abundant inertia and more critical primary frequency regulation capability, reducing the ineffective action of virtual inertia and reducing the energy consumption and equipment burden of the wind turbine. Attached Figure Description

[0018] Figure 1 This is the overall control flowchart of the present invention;

[0019] Figure 2 This is a schematic diagram of the dual-input fuzzy inference unit structure of the present invention;

[0020] Figure 3 This is a safety constraint structure diagram of the present invention;

[0021] Figure 4 This is a comparison diagram of the frequency control of the present invention;

[0022] Figure 5 This is a diagram showing the fan speed fluctuation of the method of the present invention. Detailed Implementation

[0023] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. The illustrative embodiments and descriptions herein are used to explain the present invention, but are not intended to limit the present invention.

[0024] like Figure 1As shown, the embodiments of the present invention mainly include the following steps: signal acquisition and feature calculation step S1, dual-input fuzzy inference and discrimination step S2, variable virtual inertia control step S3, adaptive primary frequency regulation step S4, leading and trailing edge smoothing filtering step S5, total additional power synthesis step S6, and safety constraints and VSG execution step S7. The following describes the complete implementation process of the method of the present invention in detail, taking the VSG retrofit of a 1.5MW doubly-fed wind turbine as an example, combined with specific parameters. The system rated frequency f0 is set to 50Hz, and the wind turbine VSG rated power PN is set to 1.5MW. In the dynamic virtual inertia parameters, the basic virtual inertia time constant TJ0 is set to 3s, and the inertia enhancement coefficient λ is set to 0.7. In the smoothing filtering parameters, the filtering time constant Ts is set to 0.1s, and the inertia exit smoothing time is set to 0.2s. In the primary frequency regulation parameters, the rated frequency regulation coefficient Kf is set to 10, the frequency regulation dead zone is set to |Δf|<0.03Hz, and the frequency regulation output limit is set to ±10%PN, i.e., ±0.15MW. In the safety constraint parameters, the minimum safe speed of the wind turbine, ω_min, is set to 0.85 pu. When the speed margin Δω < 0.1 pu, the inertia is locked; when Δω < 0.05 pu, the frequency regulation is dated to 50%; the total additional power change rate |dP / dt| ≤ 0.2 pu / s; and the total additional power is limited to ±15%PN, i.e., ±0.225MW. The sampling frequency is set to 100Hz, i.e., the sampling period is 0.01s.

[0025] The specific implementation process of signal acquisition and feature calculation step S1 is as follows: The instantaneous frequency f of the power grid is acquired in real time through a frequency sensor. The frequency sensor can be a phase-locked loop circuit or a digital frequency measurement device, and the sampling frequency is 100Hz, that is, the power grid frequency data is acquired once every 10ms; The wind turbine rotation speed ω is acquired in real time through a speed sensor. The speed sensor can be a rotary transformer or a photoelectric encoder to directly obtain the wind turbine rotor mechanical speed signal; The current active power output P is acquired in real time through a power sensor. The power sensor can be a voltage and current transformer in conjunction with a power calculation module to directly obtain the current output active power value of the wind turbine. The acquired raw signals are sent to the calculation unit for feature quantity calculation: the frequency deviation Δf is calculated using the formula Δf=f-f0, where f0 is the system rated frequency of 50Hz; the frequency change rate df / dt is calculated using the formula df / dt=(f(k)-f(k-1)) / 0.01s, which is the frequency difference between two adjacent sampling times divided by the sampling period of 0.01s, used to characterize the dynamic change trend of the grid frequency; the speed margin Δω is calculated using the formula Δω=ω-ω_min, where ω_min is the minimum safe speed of the wind turbine of 0.85pu, and the speed margin characterizes the current speed reserve of the wind turbine, used for subsequent safety constraint judgment. The signal acquisition and feature calculation step S1 provides the required raw signals and feature quantities for subsequent steps and is the input basis of the entire control method.

[0026] The specific implementation process of the dual-input fuzzy inference discrimination step S2 is as follows: Figure 2 The dual-input fuzzy inference unit structure is shown. Figure 2 A fuzzy inference structure with the absolute value of frequency deviation |Δf| and the absolute value of frequency change rate |df / dt| as dual inputs is illustrated. |Δf| is divided into three fuzzy levels: small, medium, and large, while |df / dt| is divided into three fuzzy levels: slowly changing, rapidly changing, and drastically changing. The inference logic of the fuzzy inference unit, which comprehensively judges the frequency state and disturbance intensity level based on the sign direction of Δf and df / dt, and outputs the inertia allowable projection coefficient Kv and the frequency modulation gain correction coefficient Kf_adapt, is also shown. A fuzzy inference unit with frequency deviation Δf and frequency change rate df / dt as dual inputs is constructed, where the absolute value of frequency deviation |Δf| is the first input variable, and the absolute value of frequency change rate |df / dt| is the second input variable. The absolute value of frequency deviation |Δf| is fuzzy-divided into three levels: small, medium, and large. Small corresponds to |Δf| < 0.1 Hz, indicating a small frequency deviation; medium corresponds to 0.1 Hz ≤ |Δf| < 0.2 Hz, indicating a medium frequency deviation; and large corresponds to |Δf| ≥ 0.2 Hz, indicating a large frequency deviation. The absolute value of the rate of change of frequency |df / dt| is fuzzy-divided into three levels: slow, fast, and abrupt. Slow corresponds to |df / dt| < 0.5 Hz / s, indicating a relatively slow frequency change; fast corresponds to 0.5 Hz / s ≤ |df / dt| < 1.0 Hz / s, indicating a medium frequency change; and abrupt corresponds to |df / dt| ≥ 1.0 Hz / s, indicating a very drastic frequency change.

[0027] The fuzzy inference unit comprehensively judges the frequency state based on the sign direction of Δf and df / dt, distinguishing the following four typical operating conditions: The first is the frequency drop disturbance condition, determined when Δf < 0 and df / dt < 0, indicating that the frequency is below the rated value and in a continuous decreasing state, corresponding to power grid power deficit disturbances such as UHVDC blocking and generator disconnection; the second is the frequency drop bottoming out or recovery condition, determined when Δf < 0 and df / dt ≥ 0, indicating that the frequency is below the rated value but the rate of decrease has slowed down or begun to recover, corresponding to... The first type is the recovery phase after the grid frequency bottoms out; the second type is the rising disturbance condition, which is determined when Δf > 0 and df / dt > 0, indicating that the frequency is higher than the rated value and is in a continuous rising state, corresponding to power surplus disturbances in the grid, such as a sudden increase in large-scale new energy power generation or a sudden decrease in load; the third type is the rising and falling or recovery condition, which is determined when Δf > 0 and df / dt ≤ 0, indicating that the frequency is higher than the rated value but the rising rate has slowed down or begun to fall back, corresponding to the recovery phase after the grid frequency reaches its peak.

[0028] The fuzzy inference unit classifies disturbance intensity into three levels—weak, medium, and strong—based on the disturbance intensity judgment rules: weak disturbance corresponds to a combination where |Δf| is small and |df / dt| changes slowly, indicating that the grid frequency is subject to a slight disturbance; medium disturbance corresponds to all operating condition combinations other than weak and strong disturbances, indicating that the grid frequency is subject to a moderate level of disturbance; and strong disturbance corresponds to a combination where |Δf| is large or |df / dt| changes drastically, indicating that the grid frequency is subject to a severe disturbance requiring strong support.

[0029] The fuzzy inference unit outputs the inertia allowable output coefficient Kv and the frequency modulation gain correction coefficient Kf_adapt according to the preset fuzzy inference rules. In the recovery state, i.e., the condition of bottoming out / recovery or rising and falling back / recovery, Kv is set to 0 to lock the inertia output, so as to avoid excessive release of inertia during the recovery phase. At the same time, Kf_adapt is set to [1.1,1.2] to strengthen the primary frequency modulation support and accelerate the frequency recovery process. Under disturbance conditions, i.e., falling or rising disturbances, control coefficients are output according to intensity levels: for weak disturbances, Kv∈[0,0.2] and Kf_adapt∈[1.0,1.1], a small proportion of inertia support is provided while maintaining a moderate frequency modulation gain; for medium disturbances, Kv∈[0.4,0.7] and Kf_adapt∈[1.1,1.2], a medium proportion of inertia support is provided while moderately increasing the frequency modulation gain; for strong disturbances, Kv∈[0.8,1.0] and Kf_adapt∈[1.2,1.3], a near-full proportion of inertia support is provided while significantly increasing the frequency modulation gain, thus achieving a graded adjustment mechanism of "less support for weak disturbances, moderate support for medium disturbances, and strong support for strong disturbances".

[0030] The specific implementation process of the variable virtual inertia control step S3 is as follows: The virtual inertia time constant TJ is adaptively adjusted according to the magnitude of df / dt. The dynamic calculation formula for the virtual inertia time constant is TJ = TJ0·(1+λ·|df / dt|), where TJ0 is the basic virtual inertia time constant, ranging from 2s to 4s (3s in this embodiment), λ is the inertia reinforcement coefficient, ranging from 0.5 to 1.0 (0.7 in this embodiment), and |df / dt| is the absolute value of the frequency change rate. This formula ensures that the virtual inertia time constant increases linearly with the increase of the frequency change rate, guaranteeing stronger inertia support when the disturbance is more severe. When |df / dt|=0, TJ=TJ0=3s, maintaining the basic inertia level; when |df / dt|=1.0Hz / s, TJ=3×(1+0.7×1.0)=5.1s, the inertia time constant increases by 70%; when |df / dt|=1.2Hz / s, TJ=3×(1+0.7×1.2)=5.52s, the inertia time constant increases by 84%. The inertia support power is generated based on the adaptive virtual inertia time constant. The formula for calculating the inertia support power is Pv=-TJ·(df / dt)·PN / f0, where PN is the VSG rated power of 1.5MW and f0 is the rated frequency of 50Hz. The negative sign indicates that the direction of the inertia-supported power is opposite to the direction of the frequency change rate: when df / dt < 0 (i.e., the frequency decreases), Pv > 0 indicates that the output of positive active power supports the frequency; when df / dt > 0 (i.e., the frequency increases), Pv < 0 indicates that the absorption of active power suppresses the frequency increase. This step increases the virtual inertia time constant with the increase of the frequency change rate, ensuring that the inertia support is stronger when the disturbance is more severe, which can quickly smooth out violent frequency fluctuations and buy valuable time for primary frequency regulation; under weak disturbances, it maintains a small base inertia to avoid unnecessary power surges and reduce the mechanical stress of the wind turbine and the output oscillation of the converter.

[0031] The specific implementation process of adaptive primary frequency modulation step S4 is as follows: A base frequency modulation power Pf0 is generated. Primary frequency modulation adopts a droop control method, the basic principle of which is to adjust the active power output proportionally according to the frequency deviation. The formula for calculating the base frequency modulation power is Pf0 = -Kf·Δf, where Kf is the rated frequency modulation coefficient of 10, and Δf is the frequency deviation. To avoid frequent actions of the frequency modulation system to small frequency fluctuations, a frequency modulation dead zone is set to |Δf| < 0.03Hz. When the frequency deviation is within the dead zone, the primary frequency modulation power output is zero; when the frequency deviation exceeds the dead zone, the primary frequency modulation power is calculated based on the droop characteristics. The base frequency modulation power is corrected by the frequency modulation gain correction coefficient Kf_adapt to obtain the adaptive frequency modulation power Pf = Kf_adapt·Pf0, allowing the frequency modulation gain to adaptively adjust with the disturbance intensity. This step increases the frequency modulation gain to strengthen frequency support during strong disturbances and maintains moderate frequency modulation to avoid over-modulation during weak disturbances. The frequency regulation power is also set to a power output limit of ±0.15MW, or ±10%PN, to prevent excessive primary frequency regulation power from adversely affecting the wind turbine and the power grid.

[0032] The specific implementation process of the leading and trailing edge smoothing filtering step S5 is as follows: Leading and trailing edge smoothing filtering is applied to the inertia support power Pv to avoid power surges during the inertia input and output phases that could impact the power grid and wind turbine mechanical systems. During the inertia input phase, a first-order inertial filter is used, with the filtering equation Pv_limit=Pv / (1+Ts·s), where Ts is the filtering time constant, ranging from 0.05s to 0.2s, and is set to 0.1s in this embodiment. This first-order inertial filter ensures that the inertia support power increases slowly according to an exponential law during input, rather than instantaneously reaching the target value, thus effectively suppressing power surges. During the inertia output phase, a second-order damped filter is used. The transfer function of the second-order damped filter can be expressed as G(s)=ωn² / (s²+2ζωns+ωn²), where ζ is the damping ratio, taken as greater than 1 to achieve over-damping and ensure smooth output, and ωn is the natural frequency, tuned according to the inertia output smoothing time of 0.2s. This second-order damping filter ensures that Pv_limit smoothly decreases to zero according to the second-order system response law when the inertia is deactivated, avoiding the sudden removal of inertia-supported power. The combined smoothing strategy enables the inertia output to rise slowly when it is activated and smoothly return to zero when it is deactivated, achieving a soft switch of inertia-supported power and significantly reducing the mechanical stress of the wind turbine and the oscillation of the converter output.

[0033] The specific implementation process of the total additional power synthesis step S6 is as follows: The total additional power is synthesized by weighting the coefficients output by fuzzy inference. The calculation formula is ΔP=Kv·Pv_limit+Pf, where Kv is the inertia allowable deployment coefficient, Pv_limit is the smoothed inertia support power, and Pf is the adaptive frequency modulation power. This step weights and synthesizes the inertia support and primary frequency modulation according to the cooperative allocation coefficient to form a complete frequency support additional power command. The inertia support power is proportionally adjusted through Kv. When Kv=0, the inertia support is completely locked, and when Kv=1.0, the inertia support is fully deployed. The primary frequency modulation power is gain-corrected through Kf_adapt to match the frequency modulation output with the disturbance intensity.

[0034] For the specific implementation process of safety constraints and VSG execution step S7, please refer to Figure 3 The safety constraint structure shown. Figure 3 This paper illustrates a multi-level safety constraint domain structure constructed based on the wind turbine speed margin Δω, including inertia blocking constraints, frequency regulation derating constraints, power change rate constraints, and total power limiting constraints. The total additional power ΔP is then constrained and corrected before being sent to the VSG active power control loop for execution. The specific constraint rules include the following levels: The first level constraint is an inertia blocking constraint. When the speed margin Δω < 0.1pu, the inertia support output is forcibly blocked, i.e., the inertia allowable input coefficient Kv is forcibly set to zero to prevent the wind turbine from further decelerating under low-speed conditions, thus avoiding stalling or grid disconnection risks. The second level constraint is a frequency regulation derating constraint. When Δω < 0.05pu, the primary frequency regulation output is limited to 50% of the rated regulation, i.e., the frequency regulation output limit is reduced from ±0.15MW to ±0.075MW to prevent stalling or grid disconnection under low-speed conditions. Furthermore, excessive output during frequency regulation leads to a further decrease in wind turbine speed. The third constraint is a power change rate constraint, limiting the total additional power change rate |dP / dt| ≤ 0.2pu / s, meaning the change in active power per second does not exceed 20% of the rated power, preventing rapid power command jumps from impacting the converter and grid. The fourth constraint is a total power limiting constraint, limiting ΔP to not exceed ±15%PN, or ±0.225MW, preventing excessive frequency support power from exceeding the wind turbine's regulation capacity and the grid's acceptance capacity. The corrected total additional power ΔP is sent to the VSG main control module, where it is superimposed on the original active power output command of the wind turbine in the VSG active power control loop to generate the final active power reference value for the converter. By controlling the converter to output actual active power, effective support for the grid frequency is achieved.

[0035] In terms of specific application scenario analysis, the control method of this invention can be adapted to various typical disturbance scenarios. Taking the frequency drop disturbance scenario as an example, when the UHVDC blocking causes a large-scale power deficit, the grid frequency drops sharply. Assuming that the power deficit occurs at t=0, the grid frequency drops continuously from 50Hz. The frequency sensor collects the frequency signal in real time and sends it to the control unit. When Δf=-0.25Hz, df / dt=-1.2Hz / s, ω=0.95pu, and speed margin Δω=0.1pu are detected, the fuzzy inference unit determines it as a falling disturbance and a strong disturbance, and outputs Kv=0.9 and Kf_adapt=1.25. Since the speed margin Δω=0.1pu is exactly at the critical point of the inertia blocking threshold, there is no need for blocking inertia support. The variable virtual inertia control step calculates TJ = 3 × (1 + 0.7 × 1.2) = 5.52 s, and the inertia support power Pv = -5.52 × (-1.2) × 1.5 / 50 ≈ 0.1987 MW. The adaptive primary frequency regulation step calculates the basic frequency regulation power Pf0 = -10 × (-0.25) = 2.5 MW, which, after limiting and correction, yields Pf = 0.15 MW. The inertia support power is smoothed by first-order inertial filtering. The total additional power is synthesized according to the formula ΔP = Kv·Pv_limit + Pf ≈ 0.36 MW. After safety constraint correction, ΔP is limited to 0.225 MW. This power command is sent to the VSG active power control loop for execution, controlling the converter to output positive active power to support the grid frequency. Throughout the process, dual-input fuzzy inference accurately identified strong disturbance states, dynamic inertia control provided sufficient inertia support, adaptive frequency modulation provided continuous power support, smoothing filtering suppressed power surges, and safety constraints ensured the safety of downstream equipment. The entire collaborative control process was coordinated and orderly, and the control effect was significantly better than the traditional fixed-parameter VSG method.

[0036] Taking the frequency recovery phase as an example, when the grid frequency begins to rise after bottoming out, the frequency sensor detects that Δf is still negative, but df / dt has become positive. Assuming Δf = -0.15Hz and df / dt = 0.3Hz / s are detected, the fuzzy inference unit determines the bottoming-out / recovery state based on the sign direction of Δf and df / dt, outputting Kv = 0 and Kf_adapt = 1.15. At this time, inertia support is completely locked out, retaining only enhanced primary frequency regulation. Although the variable virtual inertia control step calculates a large virtual inertia time constant and inertia support power, since Kv = 0, the inertia support power does not participate in the synthesis. The primary frequency regulation step outputs normal frequency regulation power, which is directly sent to the VSG for execution if the speed margin is sufficient. Because the inertia is automatically locked out during the recovery phase, the frequency recovery process is entirely dominated by primary frequency regulation, avoiding problems such as secondary frequency drops, overshoot, and slow recovery caused by excessive release of inertia during the recovery phase, making the frequency recovery process more stable and faster.

[0037] Taking a frequency rise disturbance scenario as an example, when a sudden increase in large-scale new energy power generation leads to power surplus, the grid frequency continues to rise. Assuming a detected Δf = 0.3Hz and df / dt = 1.5Hz / s, the fuzzy inference unit determines it as a rising disturbance and a strong disturbance, outputting Kv = 0.9 and Kf_adapt = 1.25. The variable virtual inertia control step calculates a large virtual inertia time constant and a negative inertia support power Pv, indicating the absorption of active power from the grid to suppress frequency rise. The adaptive primary frequency regulation step calculates a negative frequency regulation power Pf, indicating a reduction in active power output to cooperate with the inertia support in suppressing frequency rise. The synthesized total additional power ΔP is negative. After being sent to the VSG active power control loop for execution, it controls the converter to reduce active power output or even absorb active power, effectively suppressing grid frequency rise.

[0038] For simulation verification, to verify the effectiveness of the control method of this invention, a complete simulation model was built using Simulink, and a 1.5MW doubly-fed induction generator (DFIG) was connected to simulate frequency drop and frequency rise disturbance scenarios, respectively, comparing the control effects of the method of this invention with the traditional fixed-inertia VSG control method. Figure 4 The frequency control comparison chart shown illustrates the frequency response curves of the two methods under a frequency drop disturbance scenario. Simulation comparison curves demonstrate the control performance of the traditional fixed inertia VSG control method and the adaptive cooperative control method of this invention under a frequency drop disturbance scenario, including comparisons of key indicators such as maximum frequency deviation, frequency recovery time, power surge amplitude, and speed fluctuation amplitude. Simulation results show that the traditional method has a maximum frequency deviation of -0.36Hz and a recovery time of 14s; the method of this invention has a maximum frequency deviation of -0.24Hz and a recovery time of 10s, representing a 33.33% reduction in maximum frequency deviation and a 28.6% reduction in recovery time. Regarding equipment operation, the traditional method exhibits significant power surges during the inertia input and output phases, with a maximum surge amplitude reaching 0.3MW and a wind turbine speed fluctuation amplitude reaching 0.08pu, potentially adversely affecting the wind turbine mechanical system and converter. Figure 5 The wind turbine speed fluctuation diagram shown in the present invention illustrates the response characteristics of the wind turbine speed during the inertia input and inertia de-initiation stages after adopting the leading and trailing edge smoothing filtering strategy. The power surge amplitude of the present invention is controlled within 0.05MW, and the wind turbine speed fluctuation amplitude is controlled within 0.03pu, effectively avoiding power surges and abnormal speeds, making the equipment operation more stable and reliable.

[0039] Simulation results show that the control method of the present invention, through the combination of full-condition adaptive disturbance identification, dynamic virtual inertia adjustment, refined collaborative allocation and comprehensive safety constraints, can effectively suppress grid frequency fluctuations, significantly shorten frequency recovery time, and fully ensure the safe operation of wind turbine units. Compared with the existing technology, it has significant technical advantages, is fully adaptable to the frequency stability control requirements of large synchronous grids, and can be widely applied to large synchronous grids with a high proportion of wind power integration.

[0040] In terms of control system structure, this invention also provides a wind turbine VSG control system, including a frequency acquisition module, a speed acquisition module, a fuzzy inference module, a variable inertia module, an adaptive frequency regulation module, a safety constraint module, and a VSG main control module. The frequency acquisition module and the speed acquisition module acquire the grid frequency signal and the wind turbine speed signal, respectively, and send them to the fuzzy inference module and the variable inertia module. The fuzzy inference module outputs the inertia projection coefficient and the frequency regulation gain correction coefficient based on the frequency deviation and the frequency change rate. The variable inertia module calculates the virtual inertia time constant based on the frequency change rate and generates the inertia-supported power. The adaptive frequency regulation module generates the adaptive frequency regulation power based on the frequency regulation gain correction coefficient. The safety constraint module performs constraint correction on the total additional power in conjunction with the speed margin. The VSG main control module receives the corrected total additional power and sends it to the VSG active power control loop for execution. All modules coordinate and cooperate to achieve the adaptive collaborative control function of virtual inertia and primary frequency regulation.

[0041] In summary, this invention acquires grid frequency and wind turbine operating status information through signal acquisition and feature calculation step S1; achieves full-condition disturbance identification and intensity classification through dual-input fuzzy inference discrimination step S2; achieves dynamic matching of inertia support and disturbance intensity through variable virtual inertia control step S3; achieves follow-up adjustment of frequency regulation gain through adaptive primary frequency regulation step S4; achieves smooth switching between inertia input and output through leading and trailing edge smoothing filtering step S5; achieves coordinated allocation of inertia and frequency regulation through total additional power synthesis step S6; and achieves dual protection of frequency support and unit safety through safety constraints and VSG execution step S7. This invention comprehensively covers four typical operating conditions: frequency drop disturbance, rising disturbance, bottoming out recovery, and falling back recovery. It realizes adaptive coordinated control of virtual inertia and primary frequency regulation, balancing grid frequency support effectiveness and wind turbine operating safety, and has broad engineering application value.

[0042] The technical solutions of the present invention are not limited to the specific embodiments described above. Any technical modifications made in accordance with the technical solutions of the present invention fall within the protection scope of the present invention.

Claims

1. A wind turbine control method for large-scale synchronous power grids, characterized in that, Includes the following steps: S1: Signal Acquisition and Feature Calculation: Real-time acquisition of grid instantaneous frequency f, wind turbine speed ω, and current active power output P; calculation of frequency deviation Δf=f-f0, frequency change rate df / dt, and speed margin Δω=ω-ω_min; S2: Dual-input fuzzy inference and discrimination: Fuzzy division is performed based on the absolute value of frequency deviation |Δf| and the absolute value of frequency change rate |df / dt|. The frequency state is determined and the disturbance intensity level is distinguished according to the sign direction of Δf and df / dt. The inertia allowable projection coefficient Kv and the frequency modulation gain correction coefficient Kf_adapt are output. S3: Variable virtual inertia control: Calculate the virtual inertia time constant according to the formula TJ=TJ0·(1+λ·|df / dt|), and generate the inertia support power according to the formula Pv=-TJ·(df / dt)·PN / f0; S4: Adaptive primary frequency modulation: Generate the basic frequency modulation power Pf0 and correct it using the frequency modulation gain correction coefficient Kf_adapt to obtain the adaptive frequency modulation power Pf; S5: Leading and trailing edge smoothing filtering: Perform first-order inertial filtering and second-order damping filtering on the inertial support power Pv to obtain the smoothed inertial support power Pv_limit; S6: Total Additional Power Synthesis: Total additional power is synthesized according to the formula ΔP=Kv·Pv_limit+Pf; S7: Safety constraints and VSG execution: After constraining and correcting the total additional power ΔP based on the speed margin Δω, it is sent to the VSG active control loop for execution.

2. The wind turbine control method for large-scale synchronous power grids according to claim 1, characterized in that, In the dual-input fuzzy inference and discrimination step, the absolute value of frequency deviation |Δf| is divided into three levels: small, medium, and large. The small level corresponds to |Δf| < 0.1Hz, the medium level corresponds to 0.1Hz ≤ |Δf| < 0.2Hz, and the large level corresponds to |Δf| ≥ 0.2Hz. The absolute value of frequency change rate |df / dt| is divided into three levels: slow change, fast change, and abrupt change. The slow change level corresponds to |df / dt| < 0.5Hz / s, the fast change level corresponds to 0.5Hz / s ≤ |df / dt| < 1.0Hz / s, and the abrupt change level corresponds to |df / dt| ≥ 1.0Hz / s.

3. The wind turbine control method for large-scale synchronous power grids according to claim 2, characterized in that, The dual-input fuzzy inference discrimination step distinguishes four operating conditions based on the sign direction of Δf and df / dt: when Δf < 0 and df / dt < 0, it is determined to be a downward disturbance operating condition; when Δf < 0 and df / dt ≥ 0, it is determined to be a downward bottoming or recovery operating condition; when Δf > 0 and df / dt > 0, it is determined to be an upward disturbance operating condition; when Δf > 0 and df / dt ≤ 0, it is determined to be an upward pullback or recovery operating condition.

4. The wind turbine control method for large-scale synchronous power grids according to claim 3, characterized in that, In the recovery state (i.e., when the price has bottomed out or is recovering, or when it is rising and falling back or recovering), Kv=0 is set to lock the inertia output, and Kf_adapt is set to 1.1 to 1.

2. In the disturbance state (i.e., when the price is falling or rising), the output is graded according to the intensity: when |Δf| is small and |df / dt| changes slowly, Kv is set to 0 to 0.2 and Kf_adapt is set to 1.0 to 1.1; when |Δf| is large or |df / dt| changes drastically, Kv is set to 0.8 to 1.0 and Kf_adapt is set to 1.2 to 1.

3.

5. The wind turbine control method for large-scale synchronous power grids according to claim 1, characterized in that, In the variable virtual inertia control step, the value range of the basic virtual inertia time constant TJ0 is 2s to 4s, and the value range of the inertia enhancement coefficient λ is 0.5 to 1.

0.

6. The wind turbine control method for large-scale synchronous power grids according to claim 1, characterized in that, In the adaptive primary frequency modulation step, the basic frequency modulation power Pf0 = -Kf·Δf, where Kf is the rated frequency modulation coefficient; when |Δf| < 0.03Hz, a frequency modulation dead zone is set, and the primary frequency modulation power output is zero; the frequency modulation power output limit is set to ±10%PN.

7. The wind turbine control method for large-scale synchronous power grids according to claim 1, characterized in that, In the aforementioned leading and trailing edge smoothing filtering steps, the inertia input stage adopts a first-order inertial filter, with the filter equation being Pv_limit=Pv / (1+Ts·s), where Ts is the filtering time constant, ranging from 0.05s to 0.2s; the inertia exit stage adopts a second-order damped filter, with the transfer function being G(s)=ωn² / (s²+2ζωns+ωn²), where ζ is the damping ratio, with a value greater than 1.

8. The wind turbine control method for large-scale synchronous power grids according to claim 1, characterized in that, The safety constraints and VSG execution steps include the following constraint rules: when the speed margin Δω < 0.1 pu, the inertia support output is forcibly locked; when Δω < 0.05 pu, the primary frequency regulation output is limited to 50% of the rated regulation; the total additional power change rate |dP / dt| is limited to ≤ 0.2 pu / s; and ΔP is limited to not exceed ±15%PN.

9. A VSG control system for a wind turbine, comprising a frequency acquisition module, a speed acquisition module, a fuzzy inference module, a variable inertia module, an adaptive frequency modulation module, a safety constraint module, and a VSG main control module, characterized in that, The frequency acquisition module and the speed acquisition module acquire the power grid frequency signal and the wind turbine speed signal, respectively, and send them to the fuzzy inference module and the variable inertia module. The fuzzy inference module outputs the inertia projection coefficient and the frequency modulation gain correction coefficient based on the frequency deviation and the frequency change rate. The variable inertia module calculates the virtual inertia time constant based on the frequency change rate and generates the inertia support power. The adaptive frequency modulation module generates adaptive frequency modulation power based on the frequency modulation gain correction coefficient; the safety constraint module performs constraint correction on the total additional power in combination with the speed margin; the VSG main control module receives the corrected total additional power and sends it to the VSG active power control loop for execution.

10. The wind turbine VSG control system according to claim 9, characterized in that, The fuzzy inference module includes a signal acquisition and feature calculation unit, a dual-input fuzzy inference unit, and a coefficient output unit. The signal acquisition and feature calculation unit calculates the frequency deviation, frequency change rate, and speed margin in real time. The dual-input fuzzy inference unit performs fuzzy division based on the absolute value of the frequency deviation and the absolute value of the frequency change rate, and comprehensively judges the frequency state and distinguishes the disturbance intensity level based on the sign of the deviation and the sign of the change rate. The coefficient output unit outputs the corresponding allowable inertia projection coefficient and frequency modulation gain correction coefficient according to the disturbance intensity level.

Images