Wind power grid-connected switching control method based on stability margin and security constraint

Abstract

The application discloses a wind power grid-following and grid-forming switching control method based on stable margin and safety constraint, which comprises collecting wind turbine state variables, online evaluating grid strength, energy and current margin; extrapolating the grid state in the prediction time domain, combining with the estimated uncertainty to calculate the robust worst stable margin of the grid-following and grid-forming mode; constructing a safe feasible set with the equipment safety constraint, calculating the prediction performance cost and dynamic hysteresis in the feasible mode, combining with the minimum residence time to determine the target mode; calculating the adaptive transition time based on the state mismatch quantity, completing the disturbance-free switching through state mapping and cubic smoothing function, and coordinating the grid-forming parameter scheduling. The application solves the existing hysteresis, easy mis-switching and switching disturbance problems of the existing static threshold switching, and realizes the high-reliable smooth switching considering the foresight, robustness and equipment safety.

Description

Technical Field

[0001] This invention relates to the field of new energy grid-connected power generation and transmission control technology, specifically to a wind power grid-connected switching control method based on stability margin and safety constraints. Background Technology

[0002] As the penetration rate of new energy sources in the power system continues to increase, the system's short-circuit capacity and synchronous inertia are declining. The grid environment at the wind turbine connection point exhibits a wide range of intensity variations, gradually evolving from a traditional strong grid to a weak grid or even an extremely weak grid. Under strong grid conditions, wind turbines typically employ grid-following control based on phase-locked loops (PLLs). This control strategy relies on the PLL to accurately track the grid phase, achieving high active and reactive power tracking accuracy, and its control structure is simple and easy to implement. However, under weak grid conditions, there is a strong coupling effect between the dynamic response of the PLL and the grid impedance. This coupling can easily trigger interactive oscillations between the converter current loop and the PLL, leading to system dynamic instability, and in severe cases, even triggering the turbine to disconnect from the grid.

[0003] To address the stability issues in weak power grids, grid-based control technology has emerged. Grid-based control eliminates the need for phase-locked loops (PLLs) and autonomously establishes voltage amplitude and frequency by internally simulating the rotor motion and excitation equations of a synchronous generator. This provides inertia and voltage support for weak power grids, significantly improving system stability. However, grid-based control exhibits relatively low power point tracking accuracy in strong power grids. Furthermore, during short-circuit faults in the grid, due to its controlled voltage source characteristics, it is highly susceptible to severe overcurrent problems caused by excessive voltage drops without sophisticated current-limiting design.

[0004] To balance efficient power generation under strong grid conditions with stable support under weak grid conditions, adaptive switching control in both grid-connected and grid-connected modes has become a current research hotspot. Existing switching control strategies mostly employ static threshold-based judgment mechanisms, i.e., real-time calculation of the system's short-circuit ratio or impedance angle, triggering mode switching when this indicator exceeds a certain fixed set value. However, existing technologies still have the following significant problems in practical engineering applications:

[0005] First, the switching process suffers from severe lag. Static thresholds can only reflect the current instantaneous grid state. When the grid strength deteriorates rapidly (such as a sudden drop in local short-circuit capacity due to a main grid fault), judgments based on the current state cannot predict the risk of instability in the short term. This often leads to the switching being triggered only when the system has already approached or crossed the stability boundary, missing the best opportunity for support.

[0006] Second, it has weak anti-interference capability and is prone to erroneous switching and frequent chattering. Online identification of grid strength inevitably involves estimation errors and measurement noise. The fixed switching threshold is extremely sensitive to noise. Near the boundary, it is very easy for identification fluctuations to cause high-frequency back-and-forth switching between grid-connected and grid-connected modes, which causes serious mechanical and electrical shocks to the converter hardware.

[0007] Third, there is a lack of systematic constraints on equipment safety boundaries. Existing switching decisions typically use stability or short-circuit ratio as a single criterion, without fully considering the unit's energy margin, current margin, DC bus voltage safety range, and rotor speed limits before and after switching. This may result in the unit switching to grid-connected mode, where although the voltage stabilizes, insufficient rapid support energy could lead to DC voltage collapse or speed exceeding limits.

[0008] Fourth, the handover process suffers from significant transient impacts. The internal integral states and synchronization mechanisms of the grid-connected and network-connected controllers differ fundamentally. Existing hard handover or fixed-time transition methods cannot adaptively match the degree of state mismatch under different operating conditions, easily generating huge voltage, current, and power jumps at the moment of handover, thus failing to achieve truly bumpless handover. Summary of the Invention

[0009] The purpose of this invention is to provide a wind power grid connection switching control method based on stability margin and safety constraints, in order to solve the technical problems existing in the current wind power grid connection and grid connection switching control, such as delayed switching, easy switching chattering, lack of equipment safety constraints, and large switching transient impact.

[0010] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0011] A wind power grid switching control method based on stability margin and safety constraints includes the following steps:

[0012] S10. Real-time acquisition of grid connection point voltage, grid connection point current and unit operating status of wind turbine units, and online assessment of grid strength indicators, available rapid support energy and current margin;

[0013] S20. Within the set prediction time domain, the future power grid state is predicted in multiple steps based on the rate of change of the power grid strength index, and active power, reactive power, available fast support energy and current margin are predicted simultaneously to form a prediction state vector.

[0014] S30. Based on the offline constructed stability domain model and the predicted state vector, calculate the predicted stability margins of the grid-following mode and the grid-building mode in the prediction time domain, and perform uncertainty safety correction on the predicted stability margins in combination with the grid strength estimation standard deviation to obtain the robust worst-case stability margins of the two control modes.

[0015] S40. Construct a safe and feasible set based on the equipment safety constraints, eliminate control modes that do not meet the robust worst-case stability margin or equipment safety constraints in the prediction time domain, and calculate the prediction performance cost among the remaining safe and feasible control modes.

[0016] S50. Construct a switching function based on the predictive performance cost of the two control modes, and determine the target control mode by combining the dynamic hysteresis width and minimum dwell time based on the online adjustment of the variance and rate of change of the power grid strength estimation.

[0017] S60. Calculate the adaptive transition time based on the state mismatch before and after the switch and the current current load rate, and complete the disturbance-free control transfer between the grid-following mode and the grid-building mode through state mapping and a cubic smoothing weighting function.

[0018] Specifically, in step S30, the process of obtaining the robust worst-case stability margin includes:

[0019] For the two control modes of grid-following mode and grid-building mode, a unified predictive stability margin is defined, and small-signal damping, impedance phase margin, current margin and energy margin are mapped to the same dimensionless scale. The standard deviation of grid strength estimation is introduced as an uncertainty penalty term.

[0020] The minimum value of the prediction stability margin among all prediction steps in the prediction time domain is selected as the worst-case stability margin of the model.

[0021] Based on the worst-case stability margin of the aforementioned model, the product of the standard deviation of the margin prediction error and the confidence coefficient is subtracted to obtain the robust worst-case stability margin.

[0022] Specifically, in step S40, the specific process of constructing the safe and feasible set and calculating the predicted performance cost includes:

[0023] Determine whether the robust worst-case stability margin of both the grid-following mode and the grid-building mode in the prediction time domain is not less than zero, and whether the DC bus voltage, wind turbine speed and converter current meet the safety boundary constraints.

[0024] Control patterns that do not meet any of the above conditions are removed from the decision set to form a safe and feasible set.

[0025] If the safe and feasible set is empty, then directly enter the network flow limiting safety state and reduce active power support and virtual inertia;

[0026] If the safe and feasible set is not empty, then in the safe and feasible control mode, the weighted sum of the stability margin violation penalty, voltage deviation, frequency deviation, current over-limit penalty, energy deviation penalty, and mode switching penalty in the prediction time domain is used as the prediction performance cost.

[0027] Specifically, in step S50, the process of determining the target control mode includes:

[0028] The difference in predictive performance cost between the network configuration mode and the network following mode is used as the switching function;

[0029] The dynamic hysteresis width at the current moment is calculated based on the basic hysteresis, uncertainty, grid change rate, and power change rate.

[0030] When transitioning from follow-the-network mode to build-the-network mode, the switching function must be greater than the dynamic hysteresis width, the worst-case robustness stability margin of follow-the-network mode must be close to the warning value, and this condition must be continuously satisfied N times. on One sampling period;

[0031] When returning from network construction mode to network following mode, the switching function is required to be less than the negative dynamic hysteresis width, meet the minimum dwell time, and simultaneously meet the pre-synchronization conditions of phase angle error and frequency deviation between the phase-locked loop and the network construction controller.

[0032] Specifically, in step S60, the specific process of calculating the adaptive transition time includes:

[0033] Extract the internal phase angle error, angular frequency error, voltage amplitude error, and current load rate of the two controllers at the switching moment;

[0034] After weighted summing of the absolute values ​​of the aforementioned state mismatches, a nonlinear mapping is performed using a limiting function. The larger the state deviation or the closer the current is to the limit, the longer the adaptive transition time T of the mapped output. tr The longer;

[0035] The adaptive transition time T tr As a time scale for a cubic smoothing function, it ensures the continuity of the first derivative at the transition start and end points, thereby achieving a smooth transition of voltage, current, and power during the control transfer process.

[0036] Specifically, in step S60, the process of transferring control without disturbance includes:

[0037] During network-following mode operation, the network controller is kept in shadow mode operation; during network-building mode operation, the phase-locked loop is kept in low-bandwidth shadow synchronization.

[0038] At the moment of switching, the internal phase angle of the grid controller is initialized to the phase angle of the phase-locked loop output, the internal angular frequency is initialized to the angular frequency of the phase-locked loop output, the grid voltage amplitude is initialized to the voltage amplitude of the grid connection point at the moment of switching, and the status of the active power filter and reactive power filter is initialized to the current actual output value.

[0039] During the adaptive transition time, the voltage commands output by the grid-following mode and the grid-connected mode are gradually weighted based on a cubic smoothing function to complete the smooth transfer of DC bus control from the grid-side converter to the machine-side converter.

[0040] Specifically, after determining the target control mode as the network construction mode, the method further includes a network construction parameter coordinated scheduling step:

[0041] Real-time monitoring of the robust worst-case stability margin and available rapid support energy of the network configuration;

[0042] When the robust worst-case stability margin decreases, the virtual damping is increased online, and the virtual resistance and reactive-voltage droop coefficient are adjusted according to the same principle.

[0043] The virtual inertia is adaptively tuned based on the available fast support energy constraints to avoid excessive inertia support causing the wind turbine speed or DC bus voltage to exceed the limit.

[0044] Specifically, during the execution of the grid configuration mode, a DC-side energy coordination control step is also included:

[0045] The DC bus voltage is controlled by the generator-side converter, and the reference value of the input power required to maintain the DC side energy balance is calculated based on the dynamic equation of the DC bus voltage.

[0046] The input power reference value is converted into a generator electromagnetic torque reference value to adjust the wind turbine capture power;

[0047] The available rapid support energy, consisting of wind turbine rotational kinetic energy, DC side energy, and additional energy storage, is assessed in real time. The complete grid configuration is maintained only if the available rapid support energy meets the expected support requirements.

[0048] Specifically, the return from grid-building mode to grid-following mode also includes a current reference value inverse calculation step:

[0049] Obtain the actual active power output and actual reactive power output at the moment of switching;

[0050] Under voltage orientation conditions, the reference value of the d-axis current when returning to grid-following mode is calculated based on the d-axis component of the grid connection point voltage, and the reference value of the q-axis current is calculated based on the d-axis component of the grid connection point voltage.

[0051] The control is transferred to the following network mode by using a gradual weighting method that is the opposite of the network entry mode.

[0052] Specifically, after entering the network-based rate limiting security state, fault rate limiting control steps are also included:

[0053] Real-time monitoring of converter current; when converter current exceeds a set threshold, dynamically increase virtual resistance and virtual inductance.

[0054] During low voltage ride-through, the set current reference value must meet the converter's maximum allowable current constraint. The q-axis current takes priority in supporting reactive power, and the remaining current capacity is used to allocate the d-axis active current under the premise of prioritizing reactive power.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] (1) This invention achieves a leap from static instantaneous judgment to predictive look-ahead switching by adopting multi-step prediction in the finite time domain and combining it with robust worst-case stability margin for decision-making, thereby effectively avoiding the problem of delayed switching when the power grid strength deteriorates rapidly and reducing the risk of transient voltage instability.

[0057] (2) This invention introduces the standard deviation of power grid strength estimation as an uncertainty penalty term in the stability margin, and combines it with the hysteresis width dynamically adjusted based on variance and rate of change, thereby achieving strong robust suppression of measurement noise and identification error, thus completely eliminating the switching chattering phenomenon near the boundary and avoiding invalid switching.

[0058] (3) By constructing a safe and feasible set including DC voltage, speed, current and energy, the present invention achieves deep decoupling and forced binding of mode selection and equipment physical boundary, thereby avoiding secondary failures caused by insufficient support capacity and ensuring the lifeline safety of wind turbine under all operating conditions.

[0059] (4) By introducing an adaptive transition time and a cubic smooth function state mapping based on state mismatch and current load rate, this invention achieves a perfect match between the control transfer rate and the current operating condition, thereby effectively controlling the active power fluctuation and current peak during the switching process, and achieving the engineering standard of disturbance-free switching. Attached Figure Description

[0060] Figure 1 This is a schematic diagram of the method process of an embodiment of the present invention.

[0061] Figure 2 This is a simplified flowchart of the switching control strategy of an embodiment of the present invention. Detailed Implementation

[0062] The present invention will be further described below with reference to the accompanying drawings and embodiments. The embodiments of the present invention include, but are not limited to, the following embodiments.

[0063] Example

[0064] like Figures 1 to 2 As shown, this embodiment provides a wind power grid-connected switching control method based on stability margin and safety constraints, deployed in the main controller of the wind turbine, and adopts the idea of ​​rolling optimization to execute the following steps in each control cycle:

[0065] Step S10: Real-time acquisition of the grid connection point voltage V of the wind turbine generator. PCC Grid connection point current I PCC And the unit operating status variables, which include active power P, reactive power Q, and DC bus voltage V. dc Wind turbine speed ω r and wind speed v w It also assesses grid strength indicators, available rapid support energy, and current margin online; this step provides high-precision real-time basic data support for subsequent prediction and safety verification.

[0066] Specifically, the real-time data collected by the input acquisition module is transmitted to the online grid strength assessment module and the energy / current margin assessment module. In the online grid strength assessment module, the recursive least squares (RLS) method or Kalman filter (KF) is used to measure the voltage increment ΔV at the grid connection point. PCC Current disturbance increment ΔI PCC Estimating the Thevenin equivalent impedance of the power grid , can be represented as

[0067]

[0068] Then, the short-circuit ratio and the comprehensive grid strength index (GSI) are calculated, and the estimated standard deviation σ of the GSI is output. GSI The rate of change can be expressed as:

[0069]

[0070]

[0071]

[0072] In the above formula, S sc For the short-circuit capacity of the grid connection point, V n The rated voltage at the grid connection point, S is the estimated short-circuit ratio. n The rated capacity of the wind turbine unit, This represents the normalized short-circuit ratio. Indicates phase margin, Indicates modal damping, This represents the voltage-reactive power sensitivity, where w1, w2, w3, and w4 are all weighting coefficients.

[0073] In the energy / current margin assessment module, the available rapid support energy E, consisting of the wind turbine rotational kinetic energy, DC-side capacitor energy, and additional energy storage, is calculated. av And based on the current output current With the maximum allowable current Ilim Calculate the current margin I by the difference. margin Current margin is used to characterize the remaining capacity that the converter can continue to support.

[0074] Step S20: Within the set prediction time domain, perform multi-step prediction of the future power grid state based on the rate of change of the power grid strength index, and extrapolate the power grid state. A first-order prediction model can be used, expressed as follows:

[0075]

[0076] In the above formula, N p To predict the number of steps, T s For the sampling period, GSI k The rate of change of the power grid strength index can be obtained by differential filtering or a state observer. It is the first derivative of the rate of change; this first-order prediction model has low computational cost and is suitable for real-time execution by embedded controllers. It can also be replaced with ARX or Kalman prediction models according to actual application requirements.

[0077] Simultaneous prediction of active power P, reactive power Q, and available rapid support energy E av and current margin I margin This forms a predicted state vector. The role of the prediction layer is not to accurately reproduce the future trajectory, but to identify in advance the operating trend that is "currently stable but may cross the stability boundary in a short time".

[0078] Step S30: Based on the offline constructed stability domain model and the predicted state vector, calculate the predicted stability margins of the grid-following mode and the grid-building mode in the prediction time domain, and perform uncertainty safety correction on the predicted stability margins in combination with the standard deviation of the grid strength estimation to obtain the robust worst-case stability margins of the two control modes; the purpose of this step is to explicitly consider the impact of identification error on stability and improve the robustness of decision-making.

[0079] Specifically, in the offline phase, parameter scanning is performed in advance within the feasible range of grid strength, active power, reactive power, and control parameters to obtain the minimum damping ratio, impedance phase margin, and peak current for both the grid-connected mode (GFL) and the grid-connected mode (GFM). A stability margin model that can be queried online is then formed using polynomial response surface methodology, radial basis function, or Gaussian process regression. In the online phase, the predicted state vector is input into this model to obtain the stability margin sequences of the two control modes (GFL and GFM) in the prediction time domain. For mode m∈{GFL, GFM}, the predicted stability margin μ at step j is... m In the calculation, small-signal damping, impedance phase margin, current margin, and energy margin are mapped to the same dimensionless scale, and the standard deviation σ of the grid strength estimation is introduced. GSIAs an uncertainty penalty term to prevent incorrect switching caused by grid impedance identification errors, it can be expressed as:

[0080]

[0081] In the above formula, This indicates the small-signal damping of mode m. Indicates the small-signal damping reference. This represents the small-signal damping coefficient. This represents the impedance phase margin of mode m. Indicates the impedance phase margin reference. This represents the impedance phase margin coefficient. Indicates the current margin factor. Indicates the energy margin coefficient. Indicates the energy margin benchmark. Indicates the rated energy margin. This represents the standard deviation coefficient of the power grid strength estimate.

[0082] The minimum prediction stability margin among all prediction steps in the prediction time domain is selected as the worst-case stability margin of the model, so that the switching decision is based on the most unfavorable prediction state, rather than solely on the current instantaneous state. This can be expressed as:

[0083]

[0084] In the above formula, This represents the worst-case stability margin of the model.

[0085] Based on the worst-case stability margin of the aforementioned model, subtracting the product of the standard deviation of the margin prediction error and the confidence coefficient yields the robust worst-case stability margin, which can be expressed as:

[0086]

[0087] In the above formula, This represents the worst-case stability margin of pattern m. This represents the standard deviation of the margin prediction error. This represents the confidence coefficient. The above correction is equivalent to reserving a safety margin for model error and measurement noise.

[0088] Step S40: Construct a safe and feasible set based on the equipment safety constraints, eliminate control modes that do not meet the robust worst-case stability margin or equipment safety constraints in the prediction time domain, and calculate the predicted performance cost among the remaining safe and feasible control modes; the purpose of this step is to form a two-layer decision-making mechanism of "hard safety constraints and soft performance optimization" to eliminate the risk of exceeding the limits.

[0089] Specifically, a decision set is first constructed based on the robust worst-case stability margin and equipment safety constraints, which may include the robust worst-case stability margin. DC bus voltage V dc Wind turbine speed ω r Converter current I g wait.

[0090] Then, it is determined whether the robust worst-case stability margins of both the grid-following mode and the grid-connected mode in the prediction time domain are not less than zero, and whether the DC bus voltage, wind turbine speed, and converter current meet the safety boundary constraints; among which, the DC bus voltage V dc Wind turbine speed ω r The safety boundary is limited by its corresponding safety allowable set, and the converter current I g The safety boundary is the maximum allowable current I lim Limited to.

[0091] Control modes that do not meet any of the above conditions are removed from the decision set, forming a safe and feasible set M. feas ; can be represented as:

[0092]

[0093] In the above formula, This represents the safe allowable set of DC bus voltages. This represents the safe allowable set of wind turbine rotation speeds.

[0094] If the safe and feasible set is empty, it means that no mode can operate safely. In this case, the system directly enters the network-limited current safety state GFM-CL, triggers the fault current limiting module, and reduces active power support and virtual inertia.

[0095] If the safe and feasible set is not empty, then in the safe and feasible control mode, the weighted sum of the stability margin violation penalty, voltage deviation, frequency deviation, current over-limit penalty, energy deviation penalty, and mode switching penalty in the prediction time domain is used as the prediction performance cost. , can be represented as

[0096]

[0097] In the above formula, φ(x) = max(0,x) is the stability margin violation penalty function. Indicates voltage deviation. Indicates frequency deviation. This indicates a penalty for exceeding the current limit. w represents the energy deviation penalty. s w v w f w i and w e These are the corresponding weighting coefficients. Indicates the penalty for mode switching. The mode switching penalty weight; where the mode switching penalty term can suppress ineffective switching that only brings minor performance improvements.

[0098] The online controller selects only the mode with the lowest predictive performance cost among safe and feasible control modes, which can be expressed as:

[0099]

[0100] In the above formula, Indicates the selection mode.

[0101] Step S50: Construct a switching function based on the predicted performance cost of the two control modes, and determine the target control mode by combining the dynamic hysteresis width and minimum dwell time based on the online adjustment of the variance and rate of change of the power grid strength estimation. The purpose of this step is to fundamentally suppress switching chattering caused by noise and ensure the reliability of the system's long-term operation.

[0102] Specifically, the difference in predictive performance cost between the network construction mode and the network following mode is first calculated as the switching function S. k , can be represented as

[0103]

[0104] In the above formula, To account for the performance cost of prediction in the following network mode, The performance cost of predicting network configuration patterns; when the switching function S k When the value is greater than 0, the overall cost of the network configuration mode GFM in the predictive sense is lower; when the switching function S k When the value is less than 0, the overall cost of the network-following mode GFL in the predictive sense is lower.

[0105] The dynamic hysteresis width h at the current moment is calculated based on the fundamental hysteresis, uncertainty, grid change rate, and power change rate. k To overcome noise, it can be expressed as

[0106]

[0107] In the above formula, h0 is the basic hysteresis loop. , and Uncertainty Grid change rate and power change rate The correction factor.

[0108] When transitioning from follow-the-network mode to build-the-network mode, the switching function must be greater than the dynamic hysteresis width, the worst-case robustness stability margin of follow-the-network mode must be close to the warning value, and this condition must be continuously satisfied N times.on Each sampling period can be expressed as

[0109]

[0110] In the above formula, Indicates the dynamic hysteresis width in the open state. This indicates the worst-case stability margin of the robustness of the network model. N represents the worst-case stability margin warning value for robustness. c This indicates the number of consecutive sampling periods.

[0111] When returning from network construction mode to network following mode, the switching function must be less than the negative dynamic hysteresis width, satisfy the minimum dwell time, and simultaneously meet the pre-synchronization conditions of phase angle error and frequency deviation between the phase-locked loop and the network construction controller. This can be expressed as follows:

[0112]

[0113]

[0114]

[0115] In the above formula, The dynamic hysteresis width represents the closed state. T represents the robust worst-case stability margin safety value. res Indicates the mode return time, T hold Indicates the minimum stay time; Indicates the phase angle of the phase-locked loop. Indicates the phase angle of the network controller. This represents the phase angle pre-synchronization parameter. Indicates the phase-locked loop frequency. Indicates the network control frequency. This represents the frequency pre-synchronization parameters.

[0116] Step S60: Calculate the adaptive transition time based on the state mismatch before and after the switch and the current current load rate. Complete the disturbance-free control transfer between the grid-following mode and the grid-building mode through state mapping and a cubic smoothing weighting function. The purpose of this step is to dynamically adjust the smoothness of the transition according to the actual operating conditions and completely eliminate the electrical shock at the moment of switching.

[0117] Specifically, firstly, data such as the internal phase angle error, angular frequency error, voltage amplitude error, and current load rate of the two controllers at the switching moment are extracted; after weighted summation of the absolute values ​​of the above state mismatches, a nonlinear mapping is performed through a limiting function. The larger the state deviation or the closer the current is to the limit, the larger the adaptive transition time T of the mapped output. trThe longer the duration, the lower the instantaneous power and current surge; when the state deviation is small, the transition time can be shortened, improving the support response speed, which can be expressed as...

[0118]

[0119] In the above formula, sat(·) is the amplitude limiting function, and T0 represents the base transition time. Indicates internal phase angle error. Indicates angular frequency error. Indicates voltage amplitude error. Indicates current load factor. , , , These are the corresponding weighting coefficients. Indicates the lower limit of the transition time. This indicates the upper limit of the transition time.

[0120] The adaptive transition time T tr As a time scale for the cubic smoothing function α(τ)=3τ²-2τ³, it ensures the continuity of the first derivative at the transition start and end points, thereby achieving a smooth transition of voltage, current and power during the control transfer process.

[0121] Specifically, the process of uninterrupted transfer of control includes:

[0122] During network follow mode operation, the network controller is kept in shadow mode; during network construction mode operation, the phase-locked loop is kept in low-bandwidth shadow synchronization.

[0123] At the moment t during the switchover execution s The internal phase angle of the network controller Initialize to PLL output phase angle , is represented as: ; internal angular frequency Initialize to the output angular frequency of the phase-locked loop. , is represented as: ; Grid voltage amplitude Initialized to the grid connection point voltage amplitude at the moment of switching. , is represented as: Active power filter and reactive power filter The state is initialized to the current actual output values ​​P and Q, represented as: , ;

[0124] During the adaptive transition time, the voltage commands output in grid-following mode and grid-connected mode are gradually weighted based on a cubic smoothing function to complete the smooth transfer of DC bus control from the grid-side converter to the generator-side converter, which can be expressed as:

[0125]

[0126]

[0127]

[0128] In the above formula, This refers to the final output grid-side converter voltage command vector during the switching transition process. This is the voltage command vector output by the grid controller. The voltage command vector output by the network controller. It is a cubic smoothing function. Let t be the normalized transition time, and t be the current time. s The start time of mode switching This is the reference value for the active power of the generator-side converter. The active power reference value of the machine-side converter generated for maximum power point tracking control. The active power reference value of the machine-side converter generated by the DC bus voltage controller.

[0129] When returning from grid-building mode to grid-following mode, the process also includes a current reference value inverse calculation step:

[0130] Get the moment of switching t s Actual active power output and actual reactive power output ;

[0131] Under voltage-oriented conditions, the d-axis current reference value for returning to grid-following mode is calculated based on the d-axis component of the grid-connected point voltage, and the q-axis current reference value is calculated based on the d-axis component of the grid-connected point voltage, which can be expressed as follows:

[0132]

[0133]

[0134] In the above formula, This indicates the reference value for the d-axis current. This indicates the reference value for the q-axis current. Indicates directional voltage;

[0135] The transfer of control to the following network mode is accomplished by using a gradual weighting method, which is the opposite of the entry network mode. This can be expressed as: .

[0136] Specifically, after determining the target control mode as the network construction mode, the method further includes a network construction parameter coordinated scheduling step:

[0137] Real-time monitoring of the robust worst-case stability margin and available rapid support energy of the network configuration;

[0138] When the robust worst-case stability margin decreases, increasing the virtual damping online can be expressed as

[0139]

[0140] In the above formula, For virtual damping after online tuning, Based on virtual damping, This is the adaptive adjustment coefficient between stability margin and virtual damping. This serves as a reference value for the stability margin of the network configuration. For the robust worst-case stability margin of the network construction mode, and These are the lower and upper limits of the virtual damping, respectively; and the virtual resistance and reactive power-voltage droop coefficient can be adjusted according to the same principle.

[0141] The virtual inertia is adaptively tuned based on the available fast support energy constraints to avoid excessive inertia support causing the wind turbine speed or DC bus voltage to exceed the limit.

[0142] Specifically, the grid-following mode is implemented through the GFL control module. Under strong grid conditions, the grid-side converter adopts grid-following control based on a phase-locked loop (PLL). Its core is to obtain the synchronization phase angle through the PLL, and then establish a power outer loop and a current inner loop, which can be expressed as the following formula:

[0143]

[0144] In the above formula, ω pll ω is the output angular frequency of the phase-locked loop, ω0 is the rated angular frequency, and v q Let q be the q-axis component of the grid connection point voltage in a synchronous rotating coordinate system. and These are the proportional gain and integral gain of the phase-locked loop (PLL) PI regulator, used to adjust the PLL output angular frequency based on the q-axis voltage deviation at the grid connection point.

[0145]

[0146] In the above formula, θ pll This is the phase angle output of the phase-locked loop.

[0147]

[0148] In voltage orientation v dUnder the condition of neglecting filter losses, the active power P and the d-axis current i d It approximately satisfies the above equation.

[0149]

[0150] In the above formula, Q represents reactive power, and i q This is the q-axis current.

[0151]

[0152] In the above formula, P * This is a reference value for active power.

[0153]

[0154] In the above formula, Q * This is a reference value for reactive power.

[0155]

[0156] In the above formula, and These represent the proportional gain and integral gain of the outer-loop PI controller for the DC bus voltage, respectively. This is the reference value for the DC bus voltage. This is the actual DC bus voltage; in engineering implementation, the grid-side converter typically generates the d-axis current reference value from the DC voltage outer loop.

[0157]

[0158] In the above formula, For d-axis converter voltage command, R f With L f These are the filter resistor and the filter inductor, respectively.

[0159]

[0160] In the above formula, This is the voltage command for the q-axis converter.

[0161]

[0162] In the above formula, G i (s) is the current inner loop PI controller. and These are the proportional gain and integral gain of the current inner-loop PI controller, respectively.

[0163] Specifically, the grid-connected mode is implemented through the GFM control module. When the grid strength decreases or the stability margin of the grid-connected mode is insufficient, the grid-side converter switches to grid-connected control. The grid-connected controller no longer relies on the phase-locked loop to establish the control coordinate system, but instead forms voltage source control through the internal power synchronization loop, which can be expressed as the following formula:

[0164]

[0165] The above equation is the active-frequency control equation for a virtual synchronous machine, where H v For virtual inertia, D v For virtual damping, Δω is the frequency deviation.

[0166]

[0167] In the above formula, θ gfm This refers to the phase angle inside the network controller.

[0168]

[0169] The above equation is the reactive power-voltage droop control equation, where E * For reference of grid voltage amplitude, E0 is the rated voltage amplitude, K q It is the reactive power-voltage droop coefficient, used to correct reactive power deviations into a reference value for grid voltage amplitude.

[0170]

[0171] In the above formula, H min and H max These are the lower and upper limits of the virtual inertia, respectively; to balance support capability and energy security, the virtual inertia is determined based on the energy margin index Γ. E Adaptive tuning.

[0172]

[0173] In the above formula, K q0 K represents the baseline value of the reactive power-voltage droop factor under the benchmark grid strength. qg Γ is the adaptive adjustment coefficient of the reactive power-voltage droop factor as a function of grid strength. g This is a normalized grid strength index. The weaker the grid, the higher the voltage support capacity can be.

[0174]

[0175] In the above formula, a virtual impedance Z is introduced into the voltage command. v This can improve the damping and power decoupling capabilities of parallel systems.

[0176]

[0177] In the above formula, R v For virtual resistance, L v Let s be the virtual inductance and s be the Laplace operator.

[0178]

[0179] In the above formula, R v0 k is the base value of the virtual resistance under the reference grid strength. rv This is the adaptive adjustment coefficient for the virtual resistance as it changes with the grid strength, adjusting the virtual resistance according to the grid strength.

[0180]

[0181] In the above formula, L v0 k is the base value of the virtual inductance under the reference grid strength. lv This is the adaptive adjustment coefficient of the virtual inductance as the grid strength changes, and the virtual inductance is adjusted according to the grid strength.

[0182] Specifically, during the execution of the grid-connected mode, a DC-side energy coordination control step is also included. The generator-side converter handles the DC bus voltage control to address situations where the AC-side output power may deviate from the wind turbine input power for a short period. This can be represented as...

[0183]

[0184] The above formula describes the impact of the imbalance between generator-side power and grid-side power on the dynamics of DC bus voltage. The specific process is as follows:

[0185] The DC bus voltage is controlled by the generator-side converter, and the reference input power required to maintain DC-side energy balance is calculated based on the DC bus voltage dynamic equation. , can be represented as

[0186]

[0187] In the above formula, P gsc This refers to the active power output by the grid-side converter to the AC grid. When converter losses are neglected, it can also be considered as the active power absorbed by the grid-side converter from the DC bus. and These are the proportional gain and integral gain of the DC bus voltage regulator of the machine-side converter, respectively.

[0188] The input power reference value is converted into a generator electromagnetic torque reference value. Adjusting the wind turbine's capture power can be expressed as

[0189]

[0190] In the above formula, This refers to the mechanical angular velocity of the generator rotor;

[0191] Real-time assessment of available rapid support energy consisting of wind turbine rotational kinetic energy, DC-side energy, and additional energy storage. , can be represented as

[0192]

[0193] In the above formula, J is the equivalent moment of inertia of the wind turbine-generator transmission system referred to the selected speed reference side. The current wind speed, The minimum permissible wind turbine speed, This is the equivalent capacitance of the DC bus. This is the minimum allowable voltage for the DC bus. The additional energy storage device can currently release effective energy for rapid support;

[0194] Maintaining the complete network configuration mode can be expressed as follows: (This can be represented as...)

[0195]

[0196] In the above formula, This is the additional support power required relative to the normal operating power reference. To ensure the expected rapid support duration, An energy safety margin reserved to address prediction errors, control losses, and subsequent dynamic processes.

[0197] Specifically, after entering the network current-limiting safety state GFM-CL, a fault current-limiting control step is also included to avoid serious overcurrent that may be caused by excessive voltage difference. The process is as follows:

[0198] Real-time monitoring of converter current I; when the converter current exceeds the set threshold I... th At that time, dynamically increasing the virtual resistance and virtual inductance can be expressed as follows:

[0199]

[0200]

[0201] In the above formula, R v and These are the virtual resistance values ​​before and after adjustment, L. v and These are the virtual inductance values ​​before and after adjustment, respectively. and These are the adjustment coefficients for virtual resistance and virtual inductance, respectively;

[0202] During low voltage ride-through, the set current reference value must meet the converter's maximum allowable current I. lim Under the constraint that reactive power support is primarily undertaken by the q-axis current, and with reactive power priority given to it, the remaining current capacity is used to allocate active current to the d-axis, which can be expressed as:

[0203]

[0204]

[0205]

[0206] In the above formula, It is a low-voltage ride-through reactive current support factor, used to generate a reactive current reference value based on the degree of voltage drop at the grid connection point. This is a reference value for active power. This is a sign function used to determine the positive and negative directions of the d-axis current reference value based on the direction of the active power reference value.

[0207] To facilitate implementation in Simulink / Stateflow, the control system can be divided into five states: grid-following operation state S0, grid-following to grid-building transition state S1, grid-building operation state S2, grid-building current-limiting state S3, and grid-building to grid-following transition state S4. The mode switching logic is as follows: when the grid is weak and the energy margin is sufficient, S0→S1→S2; when a fault occurs, S2→S3; after the fault is cleared and recovered, S3→S2; when the grid recovers and the pre-synchronization conditions are met, S2→S4→S0.

[0208]

[0209] The above equation defines the state variables at discrete time k.

[0210]

[0211] The above equation defines the control variable at discrete time k.

[0212]

[0213] The offline optimization objective function comprehensively considers stability, voltage and frequency deviation, current overrun, energy deviation, load reduction cost, and mode switching cost.

[0214]

[0215] Offline optimization must meet the DC bus voltage safety constraints.

[0216]

[0217] Offline optimization must meet speed constraints.

[0218]

[0219] Offline optimization must meet the converter current constraint.

[0220]

[0221] The required ratio of grid-connected turbines for a wind farm can be adaptively set according to the system status.

[0222]

[0223] The network priority index of the i-th generating unit can be used for site-level coordination and ranking.

[0224] To enhance the theoretical support of the method, the stability of the switching system can be analyzed using multiple Lyapunov functions and the average residence time method. Assume that each continuous control mode has a Lyapunov function V. m (x) satisfies the decay condition within the safe operating domain of this mode, and the transitions between different mode Lyapunov functions at the moment of switching are bounded.

[0225]

[0226] This condition indicates that GFL, GFM, and GFM-CL have exponential convergence characteristics within their respective feasible operating domains.

[0227]

[0228] This condition limits the Lyapunov function jump caused by mode switching; state mapping and perturbationless transition can reduce κ. mn .

[0229]

[0230] When the average residence time τ a When the above equation is satisfied, the overall stability of the hybrid switching system can be guaranteed. This condition is the minimum residence time T. hold This provides a theoretical basis for tuning.

[0231] The above embodiments are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any changes made based on the design principles of the present invention, or any non-creative modifications made thereon, shall fall within the scope of protection of the present invention.

Claims

1. A wind power grid switching control method based on stability margin and safety constraints, characterized in that, Includes the following steps: S10. Real-time acquisition of grid connection point voltage, grid connection point current and unit operating status of wind turbine units, and online assessment of grid strength indicators, available rapid support energy and current margin; S20. Within the set prediction time domain, the future power grid state is predicted in multiple steps based on the rate of change of the power grid strength index, and active power, reactive power, available fast support energy and current margin are predicted simultaneously to form a prediction state vector. S30. Based on the offline constructed stability domain model and the predicted state vector, calculate the predicted stability margins of the grid-following mode and the grid-building mode in the prediction time domain, and perform uncertainty safety correction on the predicted stability margins in combination with the grid strength estimation standard deviation to obtain the robust worst-case stability margins of the two control modes. S40. Construct a safe and feasible set based on the equipment safety constraints, eliminate control modes that do not meet the robust worst-case stability margin or equipment safety constraints in the prediction time domain, and calculate the prediction performance cost among the remaining safe and feasible control modes. S50. Construct a switching function based on the predictive performance cost of the two control modes, and determine the target control mode by combining the dynamic hysteresis width and minimum dwell time based on the online adjustment of the variance and rate of change of the power grid strength estimation. S60. Calculate the adaptive transition time based on the state mismatch before and after the switch and the current current load rate, and complete the disturbance-free control transfer between the grid-following mode and the grid-building mode through state mapping and a cubic smoothing weighting function.

2. The wind power grid connection switching control method based on stability margin and safety constraints according to claim 1, characterized in that, In step S30, the specific process of obtaining the robust worst-case stability margin includes: For the two control modes of grid-following mode and grid-building mode, a unified predictive stability margin is defined, and small-signal damping, impedance phase margin, current margin and energy margin are mapped to the same dimensionless scale. The standard deviation of grid strength estimation is introduced as an uncertainty penalty term. The minimum value of the prediction stability margin among all prediction steps in the prediction time domain is selected as the worst-case stability margin of the model. Based on the worst-case stability margin of the aforementioned model, the product of the standard deviation of the margin prediction error and the confidence coefficient is subtracted to obtain the robust worst-case stability margin.

3. The wind power grid connection switching control method based on stability margin and safety constraints according to claim 2, characterized in that, In step S40, the specific process of constructing the safe and feasible set and calculating the predicted performance cost includes: Determine whether the robust worst-case stability margin of both the grid-following mode and the grid-building mode in the prediction time domain is not less than zero, and whether the DC bus voltage, wind turbine speed and converter current meet the safety boundary constraints. Control patterns that do not meet any of the above conditions are removed from the decision set to form a safe and feasible set. If the safe and feasible set is empty, then directly enter the network flow limiting safety state and reduce active power support and virtual inertia; If the safe and feasible set is not empty, then in the safe and feasible control mode, the weighted sum of the stability margin violation penalty, voltage deviation, frequency deviation, current over-limit penalty, energy deviation penalty, and mode switching penalty in the prediction time domain is used as the prediction performance cost.

4. The wind power grid switching control method based on stability margin and safety constraints according to claim 3, characterized in that, In step S50, the specific process of determining the target control mode includes: The difference in predictive performance cost between the network configuration mode and the network following mode is used as the switching function; The dynamic hysteresis width at the current moment is calculated based on the basic hysteresis, uncertainty, grid change rate, and power change rate. When transitioning from follow-the-network mode to build-the-network mode, the switching function must be greater than the dynamic hysteresis width, the worst-case robustness stability margin of follow-the-network mode must be close to the warning value, and this condition must be continuously satisfied N times. on One sampling period; When returning from network construction mode to network following mode, the switching function is required to be less than the negative dynamic hysteresis width, meet the minimum dwell time, and simultaneously meet the pre-synchronization conditions of phase angle error and frequency deviation between the phase-locked loop and the network construction controller.

5. The wind power grid switching control method based on stability margin and safety constraints according to claim 4, characterized in that, In step S60, the specific process of calculating the adaptive transition time includes: Extract the internal phase angle error, angular frequency error, voltage amplitude error, and current load rate of the two controllers at the switching moment; After weighted summing of the absolute values ​​of the aforementioned state mismatches, a nonlinear mapping is performed using a limiting function. The larger the state deviation or the closer the current is to the limit, the longer the adaptive transition time T of the mapped output. tr The longer; The adaptive transition time T tr As a time scale for a cubic smoothing function, it ensures the continuity of the first derivative at the transition start and end points, thereby achieving a smooth transition of voltage, current, and power during the control transfer process.

6. The wind power grid switching control method based on stability margin and safety constraints according to claim 5, characterized in that, In step S60, the specific process of the disturbance-free control transfer includes: During network-following mode operation, the network controller is kept in shadow mode operation; during network-building mode operation, the phase-locked loop is kept in low-bandwidth shadow synchronization. At the moment of switching, the internal phase angle of the grid controller is initialized to the phase angle of the phase-locked loop output, the internal angular frequency is initialized to the angular frequency of the phase-locked loop output, the grid voltage amplitude is initialized to the voltage amplitude of the grid connection point at the moment of switching, and the status of the active power filter and reactive power filter is initialized to the current actual output value. During the adaptive transition time, the voltage commands output by the grid-following mode and the grid-connected mode are gradually weighted based on a cubic smoothing function to complete the smooth transfer of DC bus control from the grid-side converter to the machine-side converter.

7. The wind power grid connection switching control method based on stability margin and safety constraints according to claim 6, characterized in that, After determining the target control mode as the network construction mode, the method further includes a network construction parameter coordinated scheduling step: Real-time monitoring of the robust worst-case stability margin and available rapid support energy of the network configuration; When the robust worst-case stability margin decreases, the virtual damping is increased online, and the virtual resistance and reactive-voltage droop coefficient are adjusted according to the same principle. The virtual inertia is adaptively tuned based on the available fast support energy constraints to avoid excessive inertia support causing the wind turbine speed or DC bus voltage to exceed the limit.

8. The wind power grid switching control method based on stability margin and safety constraints according to claim 7, characterized in that, During the execution of the grid configuration mode, a DC-side energy coordination control step is also included: The DC bus voltage is controlled by the generator-side converter, and the reference value of the input power required to maintain the DC side energy balance is calculated based on the dynamic equation of the DC bus voltage. The input power reference value is converted into a generator electromagnetic torque reference value to adjust the wind turbine capture power; The available rapid support energy, consisting of wind turbine rotational kinetic energy, DC side energy, and additional energy storage, is assessed in real time. The complete grid configuration is maintained only if the available rapid support energy meets the expected support requirements.

9. The wind power grid connection switching control method based on stability margin and safety constraints according to claim 8, characterized in that, When returning from grid-building mode to grid-following mode, the process also includes a current reference value inverse calculation step: Obtain the actual active power output and actual reactive power output at the moment of switching; Under voltage orientation conditions, the reference value of the d-axis current when returning to grid-following mode is calculated based on the d-axis component of the grid connection point voltage, and the reference value of the q-axis current is calculated based on the d-axis component of the grid connection point voltage. The control is transferred to the following network mode by using a gradual weighting method that is the opposite of the network entry mode.

10. The wind power grid connection switching control method based on stability margin and safety constraints according to claim 9, characterized in that, After entering the network-based rate limiting safety state, fault rate limiting control steps are also included: Real-time monitoring of converter current; when converter current exceeds a set threshold, dynamically increase virtual resistance and virtual inductance. During low voltage ride-through, the set current reference value must meet the converter's maximum allowable current constraint. The q-axis current takes priority in supporting reactive power, and the remaining current capacity is used to allocate the d-axis active current under the premise of prioritizing reactive power.

Images