Multi-target parameter setting method for networking fan controller

Abstract

The invention discloses a multi-target parameter setting method for a networking fan controller. The method comprises the following steps: firstly, establishing a small-signal mathematical model of the VSG type direct-driven fan, and analyzing an influence rule of a synchronous link inertia coefficient and a damping coefficient on a steady-state power control singular value peak value and a subsynchronous modal damping ratio; then constructing an optimization function containing steady-state and oscillation double targets, and cooperatively incorporating steady-state power control characteristics and oscillation suppression characteristics into parameter setting constraint conditions; and finally, searching a control parameter combination meeting a dual-target optimal solution by adopting a multi-target grey wolf optimization algorithm to realize global optimization of the inertia coefficient and the damping coefficient. The method is high in optimization efficiency and good in parameter convergence stability, does not depend on a specific operation state, is suitable for a new energy power system with multi-station access and complex operation conditions, and can remarkably improve the power grid supporting capability and operation stability of the grid-forming type direct drive fan under the weak power grid condition.

Description

Technical Field

[0001] This invention relates to the field of new energy power generation and power system control technology, and in particular to a multi-objective parameter tuning method for grid-connected wind turbine controllers. Background Technology

[0002] With the increasing proportion of new energy units such as wind power and photovoltaic power installed in the power system, traditional grid-connected converters typically achieve synchronization with the grid voltage through phase-locked loops (PLLs). However, these PLLs inherently exhibit current source characteristics and have limited capacity to support grid voltage. When large-scale new energy units are connected to weak grids, the system is more prone to broadband oscillations, impacting the safe and stable operation of the grid. To address this issue, domestic and international research has proposed various grid-connected control strategies, including Virtual Synchronous Generator (VSG) control, droop control, and matching control. These strategies enable the converter to simulate the external characteristics of a synchronous generator, achieving autonomous synchronization and actively supporting the grid.

[0003] In terms of steady-state power control, VSG control enables grid-connected inverters to possess dynamic external characteristics similar to synchronous generators. Adjustments to virtual inertia, damping coefficient, and damping compensation significantly impact the system's dynamic stability. Relevant technical specifications specify clear requirements for the active power frequency regulation response time and the inertia coefficient of the synchronization link in VSG units to ensure the safe and stable operation of the grid-connected system. Regarding subsynchronous oscillation (SSO), VSG control simulates the transient characteristics of synchronous generators by introducing a swing equation into the converter, thereby endowing the system with inertia and damping. However, this approach also introduces the oscillation characteristics of the synchronous motor rotor, making the system prone to power oscillations under disturbances. Existing research indicates that rationally optimizing the control parameters of VSG wind turbines can effectively improve the SSO mode damping ratio and enhance the stability of the grid-connected system. Therefore, it is necessary to adopt parameter optimization strategies that balance steady-state power control and oscillation suppression capabilities in practical engineering applications.

[0004] In existing technologies, parameter optimization methods for VSG direct-drive wind turbines mainly fall into two categories: changing the control structure and optimizing controller parameters. Changing the control structure can be achieved by adding control elements or improving existing control elements. Adding control elements is simple to design and has strong engineering applicability, but its general applicability is poor and depends on specific operating conditions. Optimizing controller parameters, on the other hand, does not require additional control strategies, is easy to adjust, and is highly practical. Current engineering practice often uses experimental debugging methods to tune the controller, but this process is time-consuming due to the large number of parameters in grid-connected systems. In recent years, swarm intelligence algorithms such as particle swarm optimization, MOGWO optimization, whale optimization, and Harris Eagle optimization have achieved good results in many engineering fields, but they have not yet been widely applied in the field of steady-state power control and multi-objective optimization of oscillation characteristics for VSG direct-drive wind turbines. Summary of the Invention

[0005] To address the problems of existing technologies, embodiments of the present invention provide a multi-objective parameter tuning method for a grid-connected wind turbine controller. The technical solution is as follows:

[0006] On the one hand, a multi-objective parameter tuning method for a grid-connected wind turbine controller is provided, including the following steps:

[0007] Step 1: Establish a mathematical model to characterize the dynamic response behavior of grid-connected direct-drive wind turbines. The model is based on a virtual synchronous machine control structure and is linearized to obtain a small-signal model in state-space form. Determine the state matrix and the mapping relationship between input and output variables.

[0008] Step 2: Based on the small-signal model, analyze the influence of the synchronization link control parameters on the steady-state power control characteristics and subsynchronous oscillation characteristics, and determine the inertia coefficient J and damping coefficient D of the virtual synchronizing machine synchronization link as the parameters to be tuned;

[0009] Step 3: Construct a multi-objective optimization function, including at least two objective functions: an objective function for quantifying steady-state response performance, which uses the maximum singular value of the small-signal model in the target frequency band as the quantification index; and an objective function for quantifying subsynchronous oscillation stability, which uses the minimum damping ratio of the oscillation mode as the quantification index.

[0010] Step 4: Using a multi-objective optimization algorithm, with the inertia coefficient J and damping coefficient D as optimization variables, solve the multi-objective optimization function to obtain a set of candidate parameters for forming the Pareto front;

[0011] Step 5: Based on the trade-off between the two types of objectives in the candidate parameter set, select the parameter combination as the tuning result of the grid wind turbine controller.

[0012] Further, step one includes: constructing a structural model of the grid-connected system based on the virtual synchronous machine control structure, wherein the structural model includes at least a permanent magnet synchronous generator, a generator-side converter, a grid-side converter, a filter device, a DC bus and capacitors, a step-up transformer, a grid connection line, and a grid connection point; defining the state variables of voltage, current, power, and control links in a synchronous rotating coordinate system, and determining the input and output variables; performing small-disturbance linearization on the nonlinear model at rated operating conditions to obtain a small-signal model in state-space form and extracting the state matrix and input-output mapping.

[0013] Further, step two includes: extracting the state matrix A from the small-signal model obtained in step one at the rated operating point, solving its eigenvalues ​​to identify the conjugate complex eigenvalues ​​with non-zero imaginary parts as oscillation modes, and calculating the damping ratio of each mode; while keeping other control parameters at nominal values, performing parameter scanning or sensitivity analysis on the inertia coefficient J and damping coefficient D of the virtual synchronizing link respectively, establishing the mapping relationship between the inertia coefficient J, damping coefficient D and the peak value of the maximum singular value in the low-frequency band related to active power and frequency control, as well as the minimum damping ratio of the sub-synchronization mode, and determining the inertia coefficient J and damping coefficient D as parameters to be tuned accordingly.

[0014] Furthermore, the construction of the objective function for quantifying steady-state response performance in step three includes: obtaining the frequency response of the multi-input multi-output transfer function matrix of the grid-connected system based on the small-signal model; selecting a low-frequency target frequency band related to active power and frequency control; performing singular value decomposition on the frequency response; extracting the curve of the maximum singular value changing with frequency; and using the peak value of the curve in the target frequency band as the first objective function for characterizing power overshoot and regulation capability.

[0015] Furthermore, the construction of the objective function for quantifying the stability of subsynchronous oscillations in step three includes: extracting the state matrix A based on the small-signal model obtained in step one; performing eigenvalue analysis on the state matrix to identify conjugate complex eigenvalues ​​with negative real parts and non-zero imaginary parts as oscillation modes; calculating the damping ratio of each oscillation mode; and selecting the minimum value of the damping ratio within the preset subsynchronous target frequency band as the second objective function.

[0016] Further, step four includes: within a preset feasible parameter domain, generating a candidate solution group using the inertia coefficient J and damping coefficient D as decision variables, and calculating the first objective function value and the second objective function value corresponding to each candidate solution; under the premise of satisfying stability constraints and parameter boundary constraints, forming a set of non-dominated solutions based on Pareto dominance and maintaining an external archive, iteratively updating until a preset termination condition is met, and using the non-dominated solutions in the external archive as the candidate parameter set for forming the Pareto front; wherein, the stability constraint is that the real part of all eigenvalues ​​of the small signal model is less than zero, and the parameter boundary constraint is that J is between a preset lower limit and a preset upper limit and D is between a preset lower limit and a preset upper limit.

[0017] Furthermore, step five includes: setting acceptable thresholds for the first objective function and the second objective function respectively; prioritizing the selection of non-dominated solutions that simultaneously satisfy both thresholds from the candidate parameter set of the Pareto front; when there is more than one solution that satisfies the conditions, comprehensively sorting them according to the order of the first objective function value from smallest to largest and the second objective function value from largest to smallest, and selecting the parameter combination with the highest sorting as the tuning result of the grid wind turbine controller.

[0018] Furthermore, the multi-objective optimization algorithm in step four adopts the multi-objective gray wolf optimization algorithm, which includes: maintaining an external archive to save non-dominated solutions; evaluating the density of archived solutions based on the grid partitioning of the objective function space and selecting a leader individual to guide the group search accordingly; when the external archive capacity exceeds a preset upper limit, preferentially removing solutions from the grid with higher density to maintain the diversity of the solution set.

[0019] On the other hand, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, is used to perform the method described thereon.

[0020] On the other hand, an electronic device is provided, including at least one processor and at least one memory, wherein the memory stores a computer program that, when executed, causes the processor to perform the method.

[0021] The beneficial effects of the technical solution provided by the embodiments of the present invention are as follows:

[0022] This invention provides a multi-objective parameter tuning method for a grid-connected wind turbine controller. By establishing a small-signal mathematical model of a VSG direct-drive wind turbine, the method systematically analyzes the influence of synchronization control parameters on steady-state power control characteristics and oscillation mode damping ratio. Combining a bi-objective optimization function design and a multi-objective grey wolf optimization (MOGWO) algorithm, the method performs parameter tuning, achieving synergistic optimization of steady-state performance and dynamic stability. This method effectively solves the technical problems of existing parameter tuning methods, which rely on experience-based adjustments, have long cycles, and suffer from a single optimization objective.

[0023] Compared with existing solutions that rely solely on additional control links or single parameter optimization, this invention considers two types of indicators during the optimization process: the peak value of the singularity of steady-state power control and the minimum damping ratio of oscillation characteristics. This can significantly improve the unit's damping capability under wideband disturbances, reduce the risk of power oscillation, and enhance the system's adaptability under weak grid conditions while maintaining excellent steady-state output characteristics of the system.

[0024] Furthermore, the optimization process proposed in this invention is independent of specific grid-connected operating conditions, exhibiting good versatility and portability. It can be deployed and applied in various operating environments, such as industrial control servers, power grid dispatch center simulation platforms, and new energy equipment operation and maintenance terminals. This method boasts high computational efficiency and good parameter convergence stability, making it suitable for new energy power systems with multiple substations connected and complex operating conditions. It possesses significant engineering promotion value and application prospects. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a flowchart of a multi-objective parameter tuning method for a grid-connected wind turbine controller according to an embodiment of the present invention.

[0027] Figure 2 This is a schematic diagram of a typical scenario of wind power collection and grid connection after GFM access according to an embodiment of the present invention; wherein, Figure (a) is the structure of VSG type direct drive wind turbine, and Figure (b) is the input-output relationship of small signal model;

[0028] Figure 3 This is a flowchart of the MOGWO algorithm according to an embodiment of the present invention;

[0029] Figure 4 This is a schematic diagram illustrating the verification of the small-signal model step response according to an embodiment of the present invention.

[0030] Figure 5 This is a schematic diagram of singular value curves when the parameters are different according to an embodiment of the present invention;

[0031] Figure 6 This is a schematic diagram of the damping ratio as a function of parameters in an embodiment of the present invention;

[0032] Figure 7 This is a schematic diagram of the Pareto optimal frontier in an embodiment of the present invention;

[0033] Figure 8This is a schematic diagram of the active power step response characteristics before and after optimization in an embodiment of the present invention;

[0034] Figure 9 This is a schematic diagram of the damping ratio before and after optimization in an embodiment of the present invention. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0036] This embodiment provides a method for tuning multi-objective parameters of a grid-connected wind turbine controller. (See also...) Figure 1 This includes the following steps:

[0037] Step 1: Establish a mathematical model to characterize the dynamic response behavior of grid-connected direct-drive wind turbines. The model is based on a virtual synchronous machine control structure and is linearized to obtain a small-signal model in state-space form. Determine the state matrix and the mapping relationship between input and output variables.

[0038] Specifically, step one includes: constructing a structural model of the grid-connected system based on a virtual synchronous machine control structure. The structural model includes at least a permanent magnet synchronous generator, a generator-side converter, a grid-side converter, a filter device, a DC bus and capacitors, a step-up transformer, a grid connection line, and a grid connection point; defining the state variables of voltage, current, power, and control loops in a synchronous rotating coordinate system, and determining the input and output variables; linearizing the nonlinear model under rated operating conditions with small disturbances to obtain a small-signal model in state-space form, and extracting the state matrix and input-output mapping.

[0039] Step 2: Based on the small-signal model, analyze the influence of the synchronization link control parameters on the steady-state power control characteristics and subsynchronous oscillation characteristics, and determine the inertia coefficient J and damping coefficient D of the virtual synchronizing machine synchronization link as the parameters to be tuned;

[0040] Specifically, step two includes: extracting the state matrix A from the small-signal model obtained in step one at the rated operating point, solving its eigenvalues ​​to identify the conjugate complex eigenvalues ​​with non-zero imaginary parts as oscillation modes, and calculating the damping ratio of each mode; while keeping other control parameters at nominal values, performing parameter scanning or sensitivity analysis on the inertia coefficient J and damping coefficient D of the virtual synchronizing machine synchronization link respectively, establishing the mapping relationship between the inertia coefficient J, damping coefficient D and the peak value of the maximum singular value in the low-frequency band related to active power and frequency control, as well as the minimum damping ratio of the subsynchronous mode, and determining the inertia coefficient J and damping coefficient D as parameters to be tuned accordingly.

[0041] Step 3: Construct a multi-objective optimization function, including at least two objective functions: an objective function for quantifying steady-state response performance, which uses the maximum singular value of the small-signal model in the target frequency band as the quantification index; and an objective function for quantifying subsynchronous oscillation stability, which uses the minimum damping ratio of the oscillation mode as the quantification index.

[0042] Specifically, the construction of the objective function for quantifying steady-state response performance in step three includes: obtaining the frequency response of the multi-input multi-output transfer function matrix of the grid-connected system based on the small-signal model; selecting a low-frequency target frequency band related to active power and frequency control; performing singular value decomposition on the frequency response; extracting the curve of the maximum singular value changing with frequency; and using the peak value of the curve in the target frequency band as the first objective function for characterizing power overshoot and regulation capability.

[0043] Furthermore, the construction of the objective function for quantifying the stability of subsynchronous oscillations in step three includes: extracting the state matrix A based on the small-signal model obtained in step one; performing eigenvalue analysis on the state matrix to identify conjugate complex eigenvalues ​​with negative real parts and non-zero imaginary parts as oscillation modes; calculating the damping ratio of each oscillation mode; and selecting the minimum value of the damping ratio within the preset subsynchronous target frequency band as the second objective function.

[0044] Step 4: Using a multi-objective optimization algorithm, with the inertia coefficient J and damping coefficient D as optimization variables, solve the multi-objective optimization function to obtain a set of candidate parameters for forming the Pareto front;

[0045] Specifically, step four includes: within a preset feasible parameter domain, generating a candidate solution group using the inertia coefficient J and damping coefficient D as decision variables, and calculating the first objective function value and the second objective function value corresponding to each candidate solution; under the premise of satisfying stability constraints and parameter boundary constraints, forming a set of non-dominated solutions based on Pareto dominance and maintaining an external archive, iteratively updating until a preset termination condition is met, and using the non-dominated solutions in the external archive as the candidate parameter set for forming the Pareto front; wherein, the stability constraint is that the real part of all eigenvalues ​​of the small signal model is less than zero, and the parameter boundary constraint is that J is between a preset lower limit and a preset upper limit and D is between a preset lower limit and a preset upper limit.

[0046] Furthermore, the multi-objective optimization algorithm in step four adopts the multi-objective gray wolf optimization algorithm, which includes: maintaining an external archive to save non-dominated solutions; evaluating the density of archived solutions based on the grid partitioning of the objective function space and selecting a leader individual to guide the group search accordingly; when the external archive capacity exceeds a preset upper limit, preferentially removing solutions from the grid with higher density to maintain the diversity of the solution set.

[0047] Step 5: Based on the trade-off between the two types of objectives in the candidate parameter set, select the parameter combination as the tuning result of the grid wind turbine controller.

[0048] Specifically, step five includes: setting acceptable thresholds for the first objective function and the second objective function respectively; prioritizing the selection of non-dominated solutions that simultaneously satisfy both thresholds from the candidate parameter set of the Pareto front; when there is more than one solution that satisfies the conditions, sorting them comprehensively according to the order of the first objective function value from smallest to largest and the second objective function value from largest to smallest, and selecting the parameter combination with the highest ranking as the tuning result of the grid wind turbine controller.

[0049] To facilitate a better understanding of the technical solution of this invention by those skilled in the art, the implementation process of this invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The technical steps and parameter values ​​described in the embodiments are for better illustrating the principle and preferred method of this invention. In actual application, the specific parameters, calculation process and algorithm iteration conditions can be adjusted accordingly based on the system's rated capacity, grid connection conditions, control strategy configuration, and operating conditions, without affecting the overall idea and effect of the method of this invention.

[0050] The technical solution of this invention is a parameter tuning method for a grid-connected wind turbine controller that takes into account both steady-state power control and subsynchronous oscillation. The main steps include: establishing a small-signal model of a VSG-type direct-drive wind turbine; analyzing the influence of the unit's synchronization control parameters on its subsynchronous mode damping ratio; based on the small-signal model, analyzing the influence of the unit's synchronization control parameters on its steady-state power control characteristics; and finally, optimizing the steady-state power control characteristics and oscillation characteristics of the VSG-type direct-drive wind turbine using the MOGWO algorithm.

[0051] The VSG type direct-drive wind turbine grid-connected system mainly consists of modules such as a permanent magnet synchronous generator, a turbine-side converter, a grid-side converter, a filter device, and a measurement and control system. The MSC uses constant DC voltage control, and the GSC uses a virtual synchronous machine control. The wind turbine undergoes two stages of voltage boosting: boosting to 35kV for collection, and boosting to 345kV for grid connection at the PCC point. When studying the SSO problem of a wind power system, the wind farm can be considered as a whole, dynamically interacting with other systems within the system. Figure 2 The variables of the VSG direct-drive wind turbine grid-connected system subsystem are defined as follows: Lg and Cg are the filter inductor and capacitor, respectively; u td u tq i gd i gq The outlet voltage and current of the fan; u gd u gq The LC filter capacitor voltages L1, R1, L2, and R2 are 35kV and 345kV respectively, and the line impedances are 345kV; C and U dc These are the DC capacitor and its voltage value.

[0052] Based on the correspondence between the sub-modules of the VSG direct-drive wind turbine grid-connected system, a corresponding small-signal model is established to obtain... Figure 2 (b) is a diagram showing the small-signal input-output relationship of the entire system.

[0053] I. Singular Value-Quantization Steady-State Power Control Characteristics

[0054] According to GB / T19963.1-2021, the voltage fluctuation range at the grid connection point of a wind farm shall not exceed ±10% (under normal operation) or -15% to +10% of the rated voltage. Excessive overshoot can lead to transient voltage exceeding the limit, potentially triggering protection devices or even causing grid disconnection. Grid-connected wind turbines need to actively support the system frequency; power oscillations caused by overshoot will exacerbate frequency fluctuations, violating the limit of ±0.2Hz (normal) / ±0.5Hz (fault).

[0055] GB / T38983 stipulates that after a dispatch command is issued, the actual power of a grid-connected wind farm should reach 90% of the commanded value within 10 seconds. Grid-connected wind turbines need to respond quickly to power shortages; excessive adjustment time will exacerbate frequency drops.

[0056] According to the requirements of the national standard GB / T38983, the overshoot and adjustment time of the steady-state power control of the VSG direct-drive fan grid-connected system are quantified using singular value curves.

[0057] Singular value decomposition (SVD) is an effective method for studying the influence of parameters on the resonant stability of multi-input multi-output (MIMO) systems, and it can accurately determine the resonant frequencies within the system. For an m×n matrix M, its singular value decomposition can be expressed as:

[0058] M=UΣV H (1)

[0059] Where U and V are m×n and n×n identity orthogonal matrices, respectively, and V H Let Σ be the conjugate transpose of V, and let Σ be the matrix containing singular values ​​σ1…σ2. q A diagonal matrix Σ, where q = min{m,n}, has diagonal elements arranged in a decreasing order. Matrix M can also be represented as a column vector of vectors U and V:

[0060] Mv i =σ i u i (2)

[0061] Where u i It outputs singular vectors, v i It is the input singular vector.

[0062] As can be seen from the above formula, if the direction of the input vector is v iThen the direction of the output vector is u. i The gain of the corresponding vector length is σ. i The maximum gain in any input direction is equal to the maximum singular value. Therefore, the resonant characteristics of a multi-input multi-output system can be studied by analyzing the frequency response of the maximum singular value.

[0063] Correspondence between peak singular values ​​and step response: If at a certain frequency ω_r, the maximum singular value σ - The presence of a significant peak (far exceeding the gains on either side) in (G(jω)) indicates resonance in the system near this frequency. A large peak foreshadows significant overshoot in the step response. The greater the overshoot, the longer the settling time.

[0064] II. Damping Ratio-Quantization Oscillation Characteristics

[0065] Based on the small-signal disturbance model established in the previous chapter, its state matrix A is extracted. Solving the characteristic equation (λE-A)X = 0 yields the eigenvalues ​​of the state matrix A of the small-signal disturbance state-space model. The eigenvalues ​​can be real or complex. In classical control theory, if the real parts of all eigenvalues ​​are negative, the system is considered asymptotically stable; if even one eigenvalue is non-negative, the system is considered unstable, i.e., the oscillations are divergent. For an N-order matrix, there are N eigenvalues. If the state matrix is ​​a real matrix, then the complex eigenvalues ​​always exist in pairs, conjugates. This means that each pair of complex eigenvalues ​​corresponds to an oscillation mode. For any pair of eigenvalues ​​of the system, we have:

[0066] λ i,i+1 =α i ±jβ i (3)

[0067] The real part determines the damping degree of the corresponding oscillation mode, while the imaginary part represents the oscillation frequency of the corresponding oscillation mode. When β i When >, λ i,i+1 This is a pair of oscillation modes. When β i When λ = 0, the characteristic roots only have real parts. i,i+1 This is a non-oscillating mode. If the real part is negative, it indicates that the oscillation is a damped oscillation with positive damping; if the real part is positive, it indicates that the oscillation is an amplified oscillation with negative damping; and if the real part is zero, it indicates that the oscillation is a constant amplitude oscillation with no damping.

[0068] via β i The corresponding oscillation frequency can be calculated using the following formula:

[0069]

[0070] The ratio of the real part to the characteristic root modulus is the damping ratio ξ. iIt can reflect the damping characteristics of oscillations, and its calculation formula is:

[0071]

[0072] The characteristics of a system's oscillation modes primarily depend on the damping ratio. When the damping ratio is greater than zero, the corresponding oscillation mode is stable, and the larger the damping ratio, the faster the oscillation amplitude decays. When the damping ratio is less than zero, the corresponding oscillation mode is unstable. When the damping ratio is equal to zero, the oscillation mode is said to be boundary stable.

[0073] II. Multi-objective optimization based on the Grey Wolf algorithm

[0074] The Multi-Objective Grey Wolf Optimizer (MOGWO) is a metaheuristic algorithm that simulates the social hierarchy and hunting behavior of grey wolves, specifically designed for solving multi-objective optimization problems. Its core idea is to utilize an external archive to store and maintain non-dominated solutions discovered during the optimization process, and to guide the population towards the Pareto Optimal Front through a special leader selection mechanism and archive maintenance strategy.

[0075] The overall optimization process for MOGWO is as follows: Figure 3 As shown, the algorithm first randomly initializes the gray wolf population and calculates the objective function value for each individual. Then, the non-dominated solutions in the population are stored in an external archive. During the iteration process, the algorithm selects three leaders (α, β, δ) from the archive to guide the other individuals in the population to update their positions. The updated population is merged with the archive, and a new set of non-dominated solutions is selected from it to update the archive again. When the archive exceeds a preset capacity, a pruning mechanism is initiated to remove solutions from the most crowded regions to ensure solution diversity. This process is repeated until the maximum number of iterations is reached.

[0076] The distance surrounding the prey is shown in equation (6):

[0077] D = |C·X p (t)-X(t)| (6)

[0078] Where D is the distance between an individual wolf and its prey; X p Let X be the location of the prey, and let C be the location of the wolf pack. C is the coefficient vector, as shown in equation (7).

[0079] C = 2·r² (7) where r is a random number in the range [0,1]. The wolf pack position is updated as shown in equation (8):

[0080] X(t+1)=X p (t)-A·D(8)

[0081] Where A is the coefficient vector, as shown in equation (9):

[0082] A = 2a·r1-a (9) r is a random number in [0,1], and a is the convergence factor as the number of iterations decreases from 2 to 0.

[0083] When the gray wolves spot their prey, the β and δ wolves, led by the α wolf, will surround the prey. The location of the individual gray wolves tracking the prey is shown in (10):

[0084]

[0085] Among them, D α D β D δ Let X represent the distances of α wolf, β wolf, and δ wolf to other individuals, respectively; α X β X δ Let α, β, and δ represent the current positions of wolves α, β, and δ, respectively. Wolf ω updates its position under the guidance of wolves α, β, and δ, as shown in equation (11).

[0086]

[0087] Leader selection mechanism: To ensure the breadth and uniformity of the Pareto front, leader selection is not simply based on individual fitness ranking, but rather employs a grid-based roulette wheel selection strategy. This strategy first divides the objective function space into a hypercubic grid and counts the number N of non-dominated solutions falling into each grid. i Then, the fitness F of each grid is calculated according to the following formula. i :

[0088] F i =e-βN i (12)

[0089] Here, the parameter β > 1 serves as the selection pressure. The design of this fitness function gives a higher selection probability to grids with lower candidate solution density (i.e., sparser), thereby motivating the algorithm to search for regions in the target space that have not yet been fully explored.

[0090] Archive Pruning Mechanism: When the external archive reaches its preset capacity, an archive pruning process will be initiated to maintain its size. This process employs a roulette wheel mechanism that contradicts the leader selection logic. Specifically, the probability P of each grid being selected and having an individual removed from it is... i It is positively correlated with its internal solution density:

[0091] P i =Σj=1MeγN j eγN i (13)

[0092] In the formula, γ is the preset deletion pressure parameter, and M is the total number of grids containing at least one solution. This mechanism ensures that individuals are removed from the regions with the densest concentration of solutions, thereby effectively maintaining the diversity of non-dominated solutions in the archive.

[0093] To verify the effectiveness of the fault ride-through control strategy proposed in this invention, a system was built in PSCAD / EMTDC. Figure 2 The VSG grid-connected simulation model shown is built in MATLAB / SIMULINK as follows. Figure 2 The main simulation parameters of the small-signal model shown are shown in Tables 1 and 2.

[0094] Table 1 Electrical parameters of VSG type direct drive fan

[0095]

[0096] Table 2 Control Parameters of VSG Type Direct Drive Fan

[0097]

[0098] To verify the correctness of the small-signal modeling, the active power was set to jump from 2MW to 2.3MW. The step responses of the electromagnetic transient simulation model and the small-signal model were compared. The red dashed line represents the waveform of the small-signal model, and the blue curve represents the waveform of the electromagnetic transient simulation. Figure 4 As shown.

[0099] according to Figure 4 It can be seen that the step response of the small-signal model is basically consistent with the trend of the electromagnetic transient simulation waveform, which verifies the accuracy of the small-signal model.

[0100] I. Steady-state power control characteristics and oscillation characteristics

[0101] Based on the singular value decomposition method, taking the active frequency element of a VSG direct-drive fan as an example, the changes in the singular value curves when the virtual inertia coefficient J and damping coefficient D change are shown in the figure.

[0102] Based on eigenvalue analysis, the subsynchronous modal damping ratios are screened. The minimum damping ratios of the system when the virtual inertia coefficient J and damping coefficient D change are respectively as follows: Figure 5 and Figure 6 As shown.

[0103] As shown in the graph, with the increase of the virtual inertia coefficient J, the peak of the singular value curve in the low-frequency band decreases, indicating that the overshoot of the active-frequency link decreases; with the increase of the virtual inertia coefficient J, the minimum damping ratio of the system increases. As shown in the graph, with the increase of the damping coefficient D, the peak of the singular value curve in the low-frequency band decreases, indicating that the overshoot of the active-frequency link decreases; with the increase of the damping coefficient D, the minimum damping ratio of the system first increases and then decreases.

[0104] The results above show that, within certain parameter ranges, there is a mutual constraint between steady-state performance and oscillation suppression performance regarding system oscillation and steady-state power control characteristics.

[0105] II. Simulation Verification of Multi-Objective Optimization Results Based on Grey Wolf Algorithm

[0106] The optimization problem in this study is biobjective, with the decision variables being the controller parameters x = [J, D]. (J is transformed into T) J The value ranges from 3 to 12 seconds, and the damping coefficient D takes values ​​from [0.0003 to 0.0008].

[0107] Objective 1: Minimize peak singular values. By performing singular value decomposition on the system's state-space model, find its peak value within a specific frequency band and minimize it.

[0108] f1(x)=max(20log10(σ(sys(x),ω))) (14)

[0109] Where σ represents singular values.

[0110] Objective 2: Maximize the minimum damping ratio. This objective aims to improve the dynamic stability of the system, particularly suppressing subsynchronous oscillations. This is achieved by calculating the system's eigenvalues, identifying oscillation modes within a specific frequency range, and maximizing the minimum damping ratio ξ among them.

[0111] f2(x)=min(-abs(λ i Re(λ) i (15)

[0112] Where λi is the characteristic value of the system in the subsynchronous frequency band.

[0113] According to the settings in the main.m script, the algorithm ran 10 iterations, searching for 30 agents. The runtime log shows that 87 non-dominated solutions were ultimately found in the archive, and the result was... Figure 7 As shown.

[0114] Figure 7This diagram illustrates the Pareto optimal frontier ultimately found by the MOGWO algorithm. The x-axis represents "Objective 1: Peak Singularity (dB)" and the y-axis represents "Objective 2: Minimum Damping Ratio." Each asterisk in the diagram represents a non-dominated solution. This frontier clearly reveals the constraint between the two objectives: reducing the peak singularity (enhancing robustness) leads to a decrease in the minimum damping ratio (weaker stability), and vice versa. For example, according to the log, the optimal robust solution (Objective 1: 6.64 dB) has a damping ratio of 0.2509; while the optimal damping ratio solution (damping ratio: 0.0437) has a peak singularity of 21.55 dB. This curve provides decision-makers with a range of optimal parameter combinations to choose from, each with different performance preferences.

[0115] The final parameter combination is [T] J =4.28s, D =0.000465).

[0116] Simulation verification: Based on the algorithm optimization results, the optimized equilibrium solution is compared with the steady-state power control characteristics and oscillation characteristics of the initial parameters. The unit's steady-state power control characteristics and oscillation characteristics are as follows: Figure 8 , Figure 9 As shown.

[0117] Depend on Figure 8 It can be seen that, compared with the initial parameters, the VSG direct-drive fan under the equilibrium point parameter optimization scheme has a smaller overshoot and a shorter active power adjustment time when the active power increases by a step, thus optimizing the steady-state power control characteristics of the VSG direct-drive fan.

[0118] Depend on Figure 9 It can be seen that, compared with the initial parameters, the minimum oscillation mode damping ratio of the equilibrium point parameters is greater than that of the initial parameters, which effectively suppresses the subsynchronous oscillation of the VSG direct-drive fan and optimizes the unit's oscillation characteristics.

[0119] In summary, the iterative optimization scheme for equilibrium point parameters satisfies the optimization objective of suppressing unit oscillations while effectively optimizing its steady-state power control characteristics, thus comprehensively optimizing both the unit's steady-state power control and oscillation characteristics. The results show that the proposed MOGWO algorithm can achieve parameter optimization for VSG direct-drive wind turbines that balances steady-state power control and oscillation characteristics.

[0120] In some embodiments, a computer-readable storage medium is provided, on which a computer program is stored. When executed by a processor, the program is used to implement the steps of the multi-objective parameter tuning method for the grid-connected wind turbine controller described in this embodiment of the invention. The program may include: establishing a mathematical model to characterize the dynamic response behavior of the grid-connected direct-drive wind turbine; analyzing the influence of the synchronous link control parameters on the steady-state power control characteristics and subsynchronous oscillation characteristics based on the model; constructing a multi-objective optimization function that simultaneously quantifies steady-state response performance and oscillation stability; solving for the Pareto front candidate parameter set using a multi-objective optimization algorithm; and selecting the final tuning parameters based on the objective trade-off relationship. These functional modules support automated tuning and comprehensive evaluation of the steady-state and dynamic performance under different unit control parameter configurations.

[0121] In other embodiments, an electronic device is proposed, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of the multi-objective parameter tuning method for the grid-connected wind turbine controller described in this embodiment of the invention. The electronic device can be a wind farm control strategy optimization platform, a wind turbine controller parameter tuning terminal, or an edge operation and maintenance module with wind turbine simulation and analysis capabilities. It can perform integrated operations such as constructing a small-signal model of the turbine, identifying the inertia and damping coefficient of the synchronization link, calculating the steady-state response and oscillation mode characteristics, performing multi-objective optimization solutions, and outputting tuning results, thereby improving the control performance and system stability of the grid-connected wind turbine under different operating scenarios.

[0122] The embodiments of the present invention may also include, but are not limited to, the forms described above. The computer program can be deployed in the centralized control system of a wind farm, a regional power grid operation optimization platform, or the parameter debugging terminal of a wind turbine manufacturer, to realize multi-objective parameter tuning and performance evaluation of the grid-connected direct-drive wind turbine controller under multiple operating conditions and multiple unit access scenarios, and has good versatility, portability, and engineering scalability.

[0123] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for tuning multi-objective parameters of a grid-connected wind turbine controller, characterized in that, Includes the following steps: Step 1: Establish a mathematical model to characterize the dynamic response behavior of grid-connected direct-drive wind turbines. The model is based on a virtual synchronous machine control structure and is linearized to obtain a small-signal model in state-space form. Determine the state matrix and the mapping relationship between input and output variables. Step 2: Based on the small-signal model, analyze the influence of the synchronization link control parameters on the steady-state power control characteristics and subsynchronous oscillation characteristics, and determine the inertia coefficient J and damping coefficient D of the virtual synchronizing machine synchronization link as the parameters to be tuned; Step 3: Construct a multi-objective optimization function, including at least two objective functions: an objective function for quantifying steady-state response performance, which uses the maximum singular value of the small-signal model in the target frequency band as the quantification index; And the objective function used to quantify the stability of subsynchronous oscillations, which uses the minimum damping ratio of the oscillation modes as the quantification index; Step 4: Using a multi-objective optimization algorithm, with the inertia coefficient J and damping coefficient D as optimization variables, solve the multi-objective optimization function to obtain a set of candidate parameters for forming the Pareto front; Step 5: Based on the trade-off between the two types of objectives in the candidate parameter set, select the parameter combination as the tuning result of the grid wind turbine controller.

2. The method according to claim 1, characterized in that, Step one includes: constructing a structural model of the grid-connected system based on a virtual synchronous machine control structure. The structural model includes at least a permanent magnet synchronous generator, a generator-side converter, a grid-side converter, a filter device, a DC bus and capacitors, a step-up transformer, a grid connection line, and a grid connection point; defining the state variables of voltage, current, power, and control loops in a synchronous rotating coordinate system, and determining the input and output variables; linearizing the nonlinear model under rated operating conditions with small disturbances to obtain a small-signal model in state-space form, and extracting the state matrix and input-output mapping.

3. The method according to claim 1, characterized in that, Step two includes: extracting the state matrix A from the small-signal model obtained in step one at the rated operating point, solving its eigenvalues ​​to identify the conjugate complex eigenvalues ​​with non-zero imaginary parts as oscillation modes, and calculating the damping ratio of each mode; while keeping other control parameters at nominal values, performing parameter scanning or sensitivity analysis on the inertia coefficient J and damping coefficient D of the virtual synchronizing machine synchronization link respectively, establishing the mapping relationship between the inertia coefficient J, damping coefficient D and the peak value of the maximum singular value in the low-frequency band related to active power and frequency control, as well as the minimum damping ratio of the subsynchronous mode, and determining the inertia coefficient J and damping coefficient D as parameters to be tuned accordingly.

4. The method according to claim 1, characterized in that, The construction of the objective function for quantifying steady-state response performance in step three includes: obtaining the frequency response of the multi-input multi-output transfer function matrix of the grid-connected system based on the small-signal model; selecting a low-frequency target frequency band related to active power and frequency control; performing singular value decomposition on the frequency response; extracting the curve of the maximum singular value changing with frequency; and using the peak value of the curve in the target frequency band as the first objective function for characterizing power overshoot and regulation capability.

5. The method according to claim 1, characterized in that, The construction of the objective function for quantifying the stability of subsynchronous oscillations in step three includes: extracting the state matrix A based on the small-signal model obtained in step one; performing eigenvalue analysis on the state matrix to identify conjugate complex eigenvalues ​​with negative real parts and non-zero imaginary parts as oscillation modes; calculating the damping ratio of each oscillation mode; and selecting the minimum value of the damping ratio within the preset subsynchronous target frequency band as the second objective function.

6. The method according to claim 1, characterized in that, Step four includes: within a preset feasible parameter domain, generating a candidate solution group using the inertia coefficient J and damping coefficient D as decision variables, and calculating the first objective function value and the second objective function value corresponding to each candidate solution; under the premise of satisfying stability constraints and parameter boundary constraints, forming a set of non-dominated solutions based on Pareto dominance and maintaining an external archive, iteratively updating until a preset termination condition is met, and using the non-dominated solutions in the external archive as the candidate parameter set for forming the Pareto front; wherein, the stability constraint is that the real part of all eigenvalues ​​of the small signal model is less than zero, and the parameter boundary constraint is that J is between a preset lower limit and a preset upper limit and D is between a preset lower limit and a preset upper limit.

7. The method according to claim 1, characterized in that, Step five includes: setting acceptable thresholds for the first objective function and the second objective function respectively; prioritizing the selection of non-dominated solutions that simultaneously satisfy both thresholds from the candidate parameter set of the Pareto front; when there is more than one solution that satisfies the conditions, sorting them in order of increasing value of the first objective function and decreasing value of the second objective function, and selecting the parameter combination with the highest ranking as the tuning result of the grid wind turbine controller.

8. The method according to claim 1, characterized in that, The multi-objective optimization algorithm in step four is the multi-objective gray wolf optimization algorithm, which includes: maintaining an external archive to save non-dominated solutions; evaluating the density of archived solutions based on the grid partitioning of the objective function space and selecting a leader individual to guide the group search accordingly; when the external archive capacity exceeds a preset upper limit, preferentially removing solutions from the grid with higher density to maintain the diversity of the solution set.

9. A computer-readable storage medium having a computer program stored thereon, the program being executed by a processor to perform the method according to any one of claims 1 to 8.

10. An electronic device comprising at least one processor and at least one memory, the memory storing a computer program that, when executed, causes the processor to perform the method according to any one of claims 1 to 8.

Images