Direct-driven fan structured identification method and system based on gravitational search algorithm

Abstract

The invention provides a direct-driven fan structured identification method and system based on a gravitational search algorithm, and belongs to the field of wind power generation control. The method comprises the following steps: determining control parameters needing to be identified in each LVRT control mode in an LVRT control strategy based on a control structure of a mathematical model of the direct-driven fan; training the original data set of the direct-driven fan by adopting an SVR algorithm to obtain an individual initial value and an optimization range; and identifying the LVRT control mode and control parameters of the direct-drive fan by adopting a GSA algorithm. Aiming at the defect that a traditional intelligent algorithm is easy to fall into local optimum, a preliminary result of a to-be-identified parameter is obtained through SVR before population initialization, an individual initial value is set on the basis of the preliminary result, an optimization range is narrowed, and then an optimization algorithm based on the gravitation and mass between objects in a position updating stage of a GSA algorithm is introduced. The motion process of an object in a gravitational field is simulated to search an optimal solution, and the local optimization ability of the algorithm is improved.

Description

Technical Field

[0001] The present invention relates to the field of wind power generation control, and in particular to a direct-drive wind turbine structured identification method and system based on a gravitational search algorithm. Background Art

[0002] In recent years, the continuous increase in the proportion of wind power has brought great challenges to the safe and stable operation of the power system. Low Voltage Ride Through (LVRT) capability is an important indicator for evaluating the transient performance of new energy power generation systems. However, there are many wind power equipment manufacturers in my country, and the LVRT capabilities of wind turbines of different models vary greatly. Obtaining accurate LVRT control modes and parameters and establishing accurate wind turbine models are the premise and key to analyzing the operating characteristics of wind turbines. The electromechanical transient model of wind turbines generally ignores the converter inner loop control and the dynamic process of the phase-locked loop, but it is necessary to consider the electrical characteristics of wind turbines during LVRT. Therefore, the LVRT control parameters in the electromechanical transient model are more critical than the converter proportional-integrator (PI) control parameters, and need to be obtained through parameter identification methods. At present, there have been relevant studies in the field of wind turbine LVRT control parameter identification, but there are still many difficulties in practical applications. For example, the influence of LVRT control parameters on the transient characteristics of wind turbines is not considered, the influence of active current control coefficient in LVRT control on the transient characteristics of wind turbines and the instability of identification algorithms are ignored; the traditional least squares method has poor adaptability to local operating conditions and the particle swarm optimization (PSO) algorithm is prone to fall into local optimality. Summary of the invention

[0003] In view of this, the present invention provides a structured identification method and system for a direct-drive wind turbine based on a gravitational search algorithm, which combines the structured identification methods of SVR and GSA to improve the local optimization capability of the algorithm, so as to achieve accurate identification of the control mode and parameters of the direct-drive wind turbine.

[0004] The technical solution adopted by the embodiment of the present invention to solve the technical problem is:

[0005] A structured identification method for a direct-drive wind turbine based on a gravitational search algorithm comprises:

[0006] Step S1, based on the control structure of the mathematical model of the direct-drive wind turbine, determining the control parameters that need to be identified for each LVRT control mode in the LVRT control strategy;

[0007] Step S2, using the SVR algorithm to train the original data set of the direct-drive wind turbine to obtain individual initial values ​​and optimization range;

[0008] Step S3, using the GSA algorithm to identify the LVRT control mode and control parameters of the direct-drive wind turbine.

[0009] Preferably, in step S1, the mathematical model of the direct-drive fan mathematical model includes:

[0010] Under steady-state conditions where the grid does not encounter voltage drop faults, the wind turbine maintains a constant power factor and operates in a conventional manner. The current direction flowing from the converter side to the grid side is defined as a positive value. At this time, the initial active current I flowing through the grid-connected inverter is p0 and initial reactive current I q0 They are:

[0011]

[0012] Where: P0 and Q0 are the initial active power and reactive power per unit value of the wind turbine when it is in normal operation, the reference value is the rated power, U0 is the initial voltage per unit value of the grid connection point, U0 = 1.0 pu, and the reference value is the rated voltage U of the grid connection point. N ;

[0013] Steady-state control mode of active power: In this mode, the control strategy of the system remains unchanged, that is, the current command is not switched, and the active and reactive currents under the current normal operating state are maintained. At the same time, the integral control link in the system is frozen to prevent the occurrence of integral saturation;

[0014] Steady-state control mode of reactive power: when the current command is in normal operation, the reactive current command after the integral link is frozen;

[0015] When a voltage drop occurs in the power grid, the dynamic reactive current increment of the wind turbine ΔI t Should meet:

[0016] △I t =K Uq (0.9-U t )I N (0.2≤U t ≤0.9)

[0017] Where: K Uq is the dynamic reactive current support factor, I N and U t are the per unit values ​​of the rated current of the wind turbine and the depth of the grid voltage drop, and the reference values ​​are the rated current I N and rated voltage U N ;

[0018] During the LVRT process of the wind turbine, the PMSG changes from constant power factor control to reactive current priority control, where:

[0019] The control strategy of active power during LVRT is divided into three modes: specified power control mode, specified current control mode and control mode based on the current before crossing;

[0020] Active power specified power control mode: In this mode, the system sets a specific active power target value P during low voltage ride-through LVRT , the calculation formula is:

[0021]

[0022] Where: K P,LVRT is the low wear active power coefficient; P set,LV is the low-pass active power setting value; P0 is the initial active power; Ip LVRT is the active current command; U t is the terminal voltage;

[0023] Active power specified current control mode: In this mode, the system sets a specific active power target value Ip during the LVRT period LVRT , the calculation formula is:

[0024] I pLVRT =K 1,Ip,LV U t +K 2,Ip,LV I p0 +I pset

[0025] In the formula, K 1,Ip,LV and K 2,Ip,LV All are low-power active current calculation coefficients; I pset It is the low active current setting value; I p0 is the initial active current;

[0026] Active power control mode based on pre-crossing current: In this mode, the system maintains the active current and reactive current before entering the low voltage ride-through period unchanged, and at the same time makes the active current command Ip LVRT With the initial active current I p0 Keep it the same:

[0027] I pLVRT =I p0

[0028] The control strategy of reactive power in the LVRT process is divided into two modes: specified power control mode and specified current control mode;

[0029] Specified power control mode for reactive power: Specify reactive power Q during low voltage ride-through LVRT , divided by the terminal voltage U t Get reactive current command IqLVRT , the calculation formula is:

[0030]

[0031] Where: K P,LVRT is the low-power reactive power coefficient; Q set,LV is the low-pass reactive power setting value; Q0 is the initial reactive power;

[0032] Specified current control mode for reactive power: Introducing low voltage ride-through threshold U Lin , at this time the current command Iq LVRT for:

[0033] I qLVRT =K 1,Iq,LV (U Lin -U t )+K 2,Iq,LV I q0 +I qset

[0034] Where: K 1,Iq,LV and K 2,Iq,LV All are low-power reactive current calculation coefficients; I qset It is the setting value of low reactive current; I q0 is the initial reactive current;

[0035] When the LVRT control strategy is the specified power control mode, the control parameter that needs to be identified is K P,LVRT , P set,LV , K Q,LVRT and Q set,LV ;

[0036] When the LVRT control strategy is the specified current control mode or the control mode based on the current before crossing, the control parameter that needs to be identified is K 1,Ip,LV , K 2,Ip,LV ,I pset , K 1,Iq,LV , K 2,Iq,LV and I qset .

[0037] Preferably, before step S1, the method further includes:

[0038] Step S4, establishing a PWMT electromagnetic transient simulation model based on the PLECS simulation platform, wherein the LVRT control strategy in the PWMT electromagnetic transient simulation model adopts a typical LVRT control strategy, and a low voltage generating device is used to simulate the voltage drop U at each depth. t An electromagnetic transient simulation is performed, and a machine-end response curve obtained by the electromagnetic transient simulation is used as a reference result; the electrical quantities included in the machine-end response curve are the voltage U, current I, active power P and reactive power Q at the output end of the wind turbine.

[0039] Preferably, the step S2 uses an SVR algorithm to train the original data set of the direct-drive wind turbine to obtain individual initial values ​​and an optimization range;

[0040] Step S21, data preparation: for each LVRT control strategy, simulate different voltage drop depths U t The target parameter and influencing factor data are extracted according to the fault condition. The target parameter is K 1,Ip,LV , K 2,Ip,LV ,I pset , K 1,Iq,LV , K 2,Iq,LV and I qset The influencing factors include the voltage drop depth U t , power factor PF, voltage U, current I, active power P and reactive power Q;

[0041] Step S22, data preprocessing: normalize and standardize the collected data:

[0042]

[0043] Where b i and B i are the eigenvalues ​​before and after normalization; μ is the mean, σ is the standard deviation;

[0044] Step S23, defining the objective function f(x), wherein the SVM regression algorithm is used for feature mapping:

[0045] f(x)=w·φ(x)+b

[0046] Among them, f(x) is the predicted parameter value; w is the weight vector; b is the bias term; φ(x) is the feature mapping function, which is used to map the input features to the high-dimensional feature space, and the radial basis function RBF is used as the feature mapping function φ(x) for mapping:

[0047] K(x,x′)=exp(-γ||xx′|| 2 )

[0048] Among them, γ is the kernel function parameter, K(x,x') is the kernel function;

[0049] Step S24, parameter adjustment: use the square loss function to adjust the model parameters and obtain the SVR model parameters:

[0050]

[0051] Where C is the regularization parameter; ∈ i is the slack variable; w is the weight of the SVR model, is the loss function; n is the length of the database;

[0052] Step S25, model training: using the selected kernel function and the adjusted SVR model parameters to train the SVM regression model;

[0053] Step S26, model verification: Use the test data set to verify the prediction performance of the model by calculating the prediction error and determination coefficient R of the model. 2 , the mean square error MSE indicator determines whether the model training is complete;

[0054] Step S27, model application: applying the trained SVM regression model to new data to predict target parameters.

[0055] Preferably, the step S3 uses the GSA algorithm to identify the LVRT control mode and control parameters of the direct-drive wind turbine, including:

[0056] Step S31, identifying the optimal LVRT control parameters;

[0057] Step S32: The optimal LVRT control parameter K P,LVRT , P set,LV , K Q,LVRT and Q set,LV Substitute the direct-drive fan mathematical model and execute it according to the specified power control mode to calculate the specified power error F PQ ;

[0058] Step S33: The optimal LVRT control parameter K 1,Ip,LV , K 2,Ip,LV ,I pset , K 1,Iq,LV , K 2,Iq,LV and I qset Substitute the direct-drive fan mathematical model and execute it according to the specified current control mode to calculate the specified current error F1;

[0059] Step S34, determine whether F PQ <F1, if yes, then the identification result is the specified power control mode, and the parameter to be selected is K P,LVRT , P set,LV , K Q,LVRT and Q set,LV ; If not, continue to execute step S35;

[0060] Step S35, determining whether the optimal LVRT control parameters simultaneously satisfy K 1,Ip,LV ≈0, K 2,Ip,LV ≈1, I pset ≈0, if so, the identification result is the control mode based on the current before crossing, and the parameter to be selected is K 1,Ip,LV , K2,Ip,LV ,I pset , K 1,Iq,LV , K 2,Iq,LV and I qse If yes, the identification result is the specified current control mode, and the parameter to be selected is K 1,Ip,LV , K 2,Ip,LV ,I pset , K 1,Iq,LV , K 2,Iq,LV and I qse .

[0061] Preferably, the step S31 includes:

[0062] Step S311, initialization: randomly generate N celestial bodies, each celestial body represents a set of LVRT control parameters; initialize various parameters of the algorithm;

[0063] Step S312, calculate fitness: calculate the fitness F for each celestial body i (k):

[0064]

[0065] In the formula, i represents a celestial body; n astart Indicates the initial count; n aend Indicates the end count, n aend ∈[1,N]; U i ,I i , P i and Q i The terminal voltage, current, active power and reactive power corresponding to the execution of the i-th group of LVRT control parameters;

[0066] Step S313, update the mass of the celestial body: update the mass M of each celestial body according to the fitness i , the higher the fitness, the greater the quality:

[0067]

[0068] Where M0 is the initial mass, F i is the fitness of the i-th celestial body, F min and F max are the minimum and maximum values ​​of fitness respectively;

[0069] Step S314, calculate gravity and acceleration: for each pair of celestial bodies, calculate the mutual gravity according to the mass difference and distance between them, and calculate the acceleration according to the gravity:

[0070]

[0071] Where Y ijis the gravitational force between the i-th and j-th celestial bodies, G is the gravitational constant, r ij is the distance between the two celestial bodies, where a i is the acceleration of the i-th celestial body;

[0072] Step S315, update position: update the position of each celestial body according to acceleration and velocity, i.e., LVRT control parameters:

[0073] r i (t+1)=r i (t)+v i (t)+a i (t)

[0074] In the formula, r i (t) is the position of the i-th celestial body at time t, v i (t) is the speed;

[0075] Step S316, speed update: update the speed and acceleration according to the algorithm rules, and set the maximum speed and maximum acceleration at the same time:

[0076] v i (t+1)=v i (t)+a i (t)

[0077] Step S317, checking the convergence condition: if the maximum number of iterations is reached or the error requirement is met, the algorithm ends; the error requirement means that the sum of the errors of U, I, P and Q is not higher than the error threshold;

[0078] Step S318, selecting the optimal solution: selecting the celestial body with the smallest fitness value as the optimal LVRT control parameter.

[0079] It can be seen from the above technical solution that an embodiment of the present invention provides a structured identification method and system for direct-drive wind turbines based on a gravitational search algorithm. First, based on the control structure of the mathematical model of the direct-drive wind turbine, the control parameters that need to be identified for each LVRT control mode in the LVRT control strategy are determined; the SVR algorithm is used to train the original data set of the direct-drive wind turbine to obtain individual initial values ​​and optimization ranges; based on the selected LVRT control mode, the GSA algorithm is used to identify the parameters corresponding to the LVRT control mode to obtain the optimal LVRT control parameters. The present invention takes direct-drive wind turbines as the research object. Aiming at the defect that traditional intelligent algorithms are prone to fall into local optimality, the preliminary results of the parameters to be identified are obtained through SVR before the population is initialized, and the individual initial values ​​are set based on this to narrow the optimization range. Then, the GSA algorithm is introduced in the position update stage. The optimization algorithm based on the gravity and mass between objects simulates the movement process of objects in the gravitational field to search for the optimal solution and improve the local optimization ability of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 This is the principle diagram of LVRT control parameter identification based on SVR and GSA algorithms;

[0081] Figure 2 It is a schematic diagram of steady-state operation and LVRT state switching;

[0082] Figure 3 It is the low voltage ride through active power control mode and parameter identification flow chart;

[0083] Figure 4 This is a schematic diagram of a low voltage ride through test;

[0084] Figure 5 is the error curve iteration graph;

[0085] Figure 6 It's U t Comparison chart of simulation results and test results when the value is 0.4pu;

[0086] Figure 7 It's U t Comparison chart of simulation results and test results when the value is 0.8pu;

[0087] Figure 8 The present invention is a block diagram of an electronic device for implementing the structured identification method of a direct-drive wind turbine based on a gravitational search algorithm according to an embodiment disclosed in the present invention. DETAILED DESCRIPTION

[0088] The technical scheme and technical effects of the present invention are further elaborated in detail below in conjunction with the accompanying drawings of the present invention.

[0089] refer to Figure 1 As shown, the present invention provides a direct-drive wind turbine structured identification method based on a gravitational search algorithm, the steps comprising:

[0090] Step S1, based on the control structure of the mathematical model of the direct-drive wind turbine, determining the control parameters that need to be identified for each LVRT control mode in the LVRT control strategy;

[0091] Step S2, using the SVR algorithm to train the original data set of the direct-drive wind turbine to obtain individual initial values ​​and optimization range;

[0092] Step S3, using the GSA algorithm to identify the LVRT control mode and control parameters of the direct-drive wind turbine;

[0093] Step S4: Further verify the validity of the identification results obtained based on the SVR-GSA algorithm through simulation examples.

[0094] The mathematical model of the direct-drive fan mathematical model in step S1 includes:

[0095] Under steady-state conditions where the grid does not encounter voltage drop faults, the wind turbine maintains a constant power factor and operates in a conventional manner. The current direction flowing from the converter side to the grid side is defined as a positive value. At this time, the initial active current I flowing through the grid-connected inverter is p0 and initial reactive current I q0 They are:

[0096]

[0097] Where: P0 and Q0 are the initial active power and reactive power per unit value of the wind turbine when it is in normal operation, the reference value is the rated power, U0 is the initial voltage per unit value of the grid connection point, U0 = 1.0 pu, and the reference value is the rated voltage U of the grid connection point. N ; When a voltage drop occurs in the power grid, the dynamic reactive current increment of the wind turbine is ΔI t Should meet:

[0098] △I t =K Uq (0.9-U t )I N (0.2≤U t ≤0.9) (2)

[0099] Where: K Uq is the dynamic reactive current support factor, I N and U t are the per unit values ​​of the rated current of the wind turbine and the depth of the grid voltage drop, and the reference values ​​are the rated current I N and rated voltage U N ;

[0100] The typical LVRT control strategy recommended by the PSASP simulation platform is used for analysis. Under steady-state operation, the current command is usually output by the converter-level power control model. After entering the LVRT, the converter switches to the LVRT control strategy and outputs in the "specified current" mode. Figure 2 As shown in the figure, the definitions of the parameters are shown in Table 1. During the LVRT steady-state period, the PMSG changes from constant power factor control to reactive current priority control, that is, the output requirements of reactive current are met first, and then the active current is output according to the restriction conditions.

[0101] Table 1 Control strategy parameter definition

[0102]

[0103] In the dynamic analysis of power systems, the current limiting model plays a vital role. The function of this model is that when the reactive current control command value exceeds the rated limit of the converter, the system will automatically adjust the output to the maximum allowable reactive current value. In addition, if the control command value exceeds the initially set active power threshold, the direct-drive wind turbine will no longer be able to provide the required active current. In this case, the current limiting model will intervene to ensure that the active current output by the converter is maintained within the maximum allowable range by adjusting the active current, thereby avoiding exceeding the initial active power setting value.

[0104] In electromagnetic transient analysis, the modeling process of direct-drive wind turbines needs to be based on three-phase transient values, and the dynamic response of the converter's internal control loop must be considered. In contrast, in electromechanical transient analysis, the electrical characteristics of direct-drive wind turbines are described by a fundamental frequency vector model, and in this process, the dynamic response of the converter's internal control loop is not a major consideration. Therefore, in the construction of the electromechanical transient model, the converter's PI control parameters are not a key factor, but the LVRT control parameters are particularly critical. Based on this, only the LVRT control parameters can be accurately identified and optimized.

[0105] Steady-state control mode of active power: In this mode, the control strategy of the system remains unchanged, that is, the current command is not switched, and the active and reactive currents under the current normal operating state are maintained. At the same time, in order to cope with the voltage fluctuation during the low voltage ride-through period, the integral control link in the system is frozen to prevent the occurrence of integral saturation;

[0106] Steady-state control mode of reactive power: when the current command is in normal operation, the reactive current command after the integral link is frozen;

[0107] During the LVRT process of the wind turbine, the PMSG changes from constant power factor control to reactive current priority control, where:

[0108] The active power control strategy during LVRT is divided into three modes: specified power control, specified current control, and control based on the current before crossing. The specific implementation method and control logic of each mode are as follows:

[0109] (1) Specified power control mode for active power: In this mode, the system sets a specific active power target value P during the low voltage ride-through period. LVRT , divided by the terminal voltage U t Get active current command Ip LVRT , the calculation formula is:

[0110]

[0111] Where: K P,LVRTis the low wear active power coefficient; P set,LV is the low-pass active power setting value; P0 is the initial active power; Ip LVRT is the active current command; U t is the terminal voltage; in this way, the system can maintain the preset active power output and improve the stability and reliability of the power grid.

[0112] (2) Active power specified current control mode: In this mode, the system sets a specific active power target value Ip during the LVRT period. LVRT , the calculation formula is:

[0113] I pLVRT =K 1,Ip,LV U t +K 2,Ip,LV I p0 +I pset (4)

[0114] In the formula, K 1,Ip,LV and K 2,Ip,LV All are low-power active current calculation coefficients; I pset It is the low active current setting value; I p0 is the initial active current;

[0115] (3) Active power control mode based on the current before the ride-through: In this mode, the system maintains the active current and reactive current before the ride-through during the low voltage ride-through period. This control mode can be regarded as a special case of the specified current control mode, where K 1,Ip,LV Take 0, K 2,Ip,LV Take 1, I pset Special case when taking 0:

[0116] I pLVRT =I p0 (5)

[0117] The control strategy of reactive power in the LVRT process is divided into two modes: specified power control and specified current control;

[0118] (1) Specified power control mode of reactive power: Specify the reactive power Q during low voltage ride-through LVRT , divided by the terminal voltage U t Get reactive current command Iq LVRT , the calculation formula is:

[0119]

[0120] Where: K P,LVRT is the low-power reactive power coefficient; Q set,LVis the reactive power setting value during low voltage ride-through; Q0 is the initial reactive power; this control mode allows the system to maintain the preset reactive power output during low voltage ride-through, further optimizing the operation status of the power grid.

[0121] (2) Specified current control mode of reactive power: Introducing low voltage ride-through threshold U Lin , at this time the current command Iq LVRT for:

[0122] I qLVRT =K 1,Iq,LV (U Lin -U t )+K 2,Iq,LV I q0 +I qset (7)

[0123] Where: K 1,Iq,LV and K 2,Iq,LV All are low-power reactive current calculation coefficients; I qset It is the setting value of low reactive current; I q0 is the initial reactive current;

[0124] In summary, the active and reactive power control strategy during LVRT is the key to ensure that the direct-drive wind turbine maintains stable operation when the grid voltage drops. Therefore, choosing the correct control strategy is of great significance for accurately simulating the dynamic characteristics of the direct-drive wind turbine.

[0125] When the LVRT control strategy is the specified power control mode, the control parameter that needs to be identified is K P,LVRT , P set,LV , K Q,LVRT and Q set,LV ;

[0126] When the LVRT control strategy is the specified current control mode or the control mode based on the current before crossing, the control parameter that needs to be identified is K 1,Ip,LV , K 2,Ip,LV ,I pset , K 1,Iq,LV , K 2,Iq,LV and I qset .

[0127] Establish a database for training models: Establish a PWMT electromagnetic transient simulation model based on the PLECS simulation platform. The LVRT control strategy in the PWMT electromagnetic transient simulation model adopts a typical LVRT control strategy, and a low voltage generator is used to simulate the voltage drop U at each depth. t Perform electromagnetic transient simulation and use the machine-end response curve obtained by electromagnetic transient simulation as the reference result; the electrical quantity included in the machine-end response curve is the output end of the wind turbine ( Figure 2The voltage U, current I, active power P and reactive power Q at the measuring point 1).

[0128] The SVM algorithm is a machine learning algorithm widely used in classification and regression analysis. In terms of parameter prediction, SVM is usually used for regression analysis, namely SVR. The goal of SVR is to find an optimal hyperplane that minimizes the sum of the distances from all training sample points to the hyperplane while satisfying certain constraints. Step S2 uses the SVR algorithm to train the original data set of the direct-drive wind turbine to obtain the individual initial values ​​and optimization range steps:

[0129] Step S21, data preparation: for each LVRT control strategy, simulate different voltage drop depths U t Fault conditions, extract the target parameters and influencing factor data, the target parameter is K 1,Ip,LV , K 2,Ip,LV ,I pset , K 1,Iq,LV , K 2,Iq,LV and I qset , the influencing factors include the voltage drop depth U t , power factor PF, voltage U, current I, active power P and reactive power Q;

[0130] Step S22, data preprocessing: normalize and standardize the collected data:

[0131]

[0132] Where b i and B i are the eigenvalues ​​before and after normalization; μ is the mean, σ is the standard deviation; standardization aims to eliminate the influence of data dimension on the model;

[0133] Step S23, defining the objective function f(x), wherein the SVM regression algorithm is used for feature mapping:

[0134] f(x)=w·φ(x)+b (9)

[0135] Among them, f(x) is the predicted parameter value; w is the weight vector; b is the bias term; φ(x) is the feature mapping function, which is used to map the input features to the high-dimensional feature space so as to construct a linearly separable hyperplane in the high-dimensional space. The radial basis function RBF is used as the feature mapping function φ(x) for mapping:

[0136] K(x,x′)=exp(-γ||xx′|| 2 ) (10)

[0137] Among them, γ is the kernel function parameter, K(x,x') is the kernel function;

[0138] Step S24, parameter adjustment: Use the square loss function to adjust the model parameters to obtain the SVR model parameters (the SVM regression algorithm involves multiple parameters, such as regularization parameter C, kernel function parameters, etc., which need to be adjusted by cross-validation and other methods to obtain the best model):

[0139]

[0140] Where C is the regularization parameter; ∈ i is the slack variable; w is the weight of the SVR model, is the loss function; n is the length of the database;

[0141] Step S25, model training: using the selected kernel function and the adjusted SVR model parameters to train the SVM regression model;

[0142] Step S26, model verification: Use the test data set to verify the prediction performance of the model by calculating the prediction error and determination coefficient R of the model. 2 , mean square error MSE indicator to determine whether the model is trained; (training is completed when the prediction error is lower than the prediction error threshold and the determination coefficient R 2 Higher than the determination coefficient threshold, mean square error MSE lower than the mean square error threshold)

[0143] Step S27, model application: applying the trained SVM regression model to new data to predict target parameters.

[0144] Step S28: Analyze the prediction results, evaluate the prediction accuracy and reliability of the model, and further adjust or optimize the model as needed.

[0145] Reference Figure 3 Step S3 uses the GSA algorithm to identify the LVRT control mode and control parameters of the direct-drive wind turbine, including:

[0146] Step S31, identifying the optimal LVRT control parameters;

[0147] Step S32: The optimal LVRT control parameter K P,LVRT , P set,LV , K Q,LVRT and Q set,LV Substitute the direct-drive wind turbine mathematical model and execute it according to the specified power control mode. Calculate the specified power error F according to formula (12): PQ ;

[0148]

[0149] Step S33: The optimal LVRT control parameter K1,Ip,LV , K 2,Ip,LV ,I pset , K 1,Iq,LV , K 2,Iq,LV and I qset Substitute into the mathematical model of the direct-drive wind turbine and execute according to the specified current control mode, and calculate the specified current error F1 according to formula (12);

[0150] Step S34, determine whether F PQ <F1, if yes, then the identification result is the specified power control mode, and the parameter to be selected is K P,LVRT , P set,LV , K Q,LVRT and Q set,LV ; If not, continue to execute step S35;

[0151] Step S35, determine whether the optimal LVRT control parameters simultaneously meet K 1,Ip,LV ≈0, K 2,Ip,LV ≈1, I pset ≈0, if so, the identification result is the control mode based on the current before crossing, and the parameter to be selected is K 1,Ip,LV , K 2,Ip,LV ,I pset , K 1,Iq,LV , K 2,Iq,LV and I qse If yes, the identification result is the specified current control mode, and the parameter to be selected is K 1,Ip,LV , K 2,Ip,LV ,I pset , K 1,Iq,LV , K 2,Iq,LV and I qse .

[0152] The step S31 comprises:

[0153] Step S311, initialization: randomly generate N celestial bodies (i.e., candidate solutions), each of which represents a set of LVRT control parameters; initialize various parameters of the algorithm, such as the number of search agents, the maximum number of iterations, the gravitational constant, etc.;

[0154] Step S312, calculate fitness: calculate the fitness F for each celestial body i (k):

[0155]

[0156] In the formula, i represents a celestial body; n astart Indicates the initial count; n aend Indicates the end count, n aend ∈[1,N]; U i ,I i , Pi and Q i The terminal voltage, current, active power and reactive power corresponding to the execution of the i-th group of LVRT control parameters;

[0157] Step S313, update the mass of the celestial body: update the mass M of each celestial body according to the fitness i , the higher the fitness, the greater the quality:

[0158]

[0159] Where M0 is the initial mass, F i is the fitness of the i-th celestial body, F min and F max are the minimum and maximum values ​​of fitness respectively;

[0160] Step S314, calculate gravity and acceleration: for each pair of celestial bodies, calculate the mutual gravity according to the mass difference and distance between them, and calculate the acceleration according to the gravity:

[0161]

[0162] Where Y ij is the gravitational force between the i-th and j-th celestial bodies, G is the gravitational constant, r ij is the distance between the two celestial bodies, where a i is the acceleration of the i-th celestial body;

[0163] Step S315, update position: update the position of each celestial body according to acceleration and velocity, i.e., LVRT control parameters:

[0164] r i (t+1)=r i (t)+v i (t)+a i (t) (16)

[0165] In the formula, r i (t) is the position of the i-th celestial body at time t, v i (t) is the speed;

[0166] Step S316, speed update: update the speed and acceleration according to the algorithm rules, and set the maximum speed and maximum acceleration at the same time:

[0167] v i (t+1)=v i (t)+a i (t) (17)

[0168] Step S317, check the convergence condition: if the maximum number of iterations is reached or the error requirement is met, the algorithm ends; the error requirement means that the sum of the errors of U, I, P and Q is not higher than the error threshold;

[0169] Step S318, selecting the optimal solution: selecting the celestial body with the smallest fitness value as the optimal LVRT control parameters.

[0170] In the research and practice of direct-drive wind turbine systems, the selection of control strategies and parameter optimization are key factors in improving system performance. In view of the existence of multiple control modes and corresponding parameter configurations, this paper proposes a structured identification method combining SVR and GSA to achieve accurate identification of direct-drive wind turbine control modes and parameters. The specific identification process of LVRT active power control mode is as follows: Figure 3 For example, the LVRT reactive power control mode is the same as above.

[0171] An embodiment is given below:

[0172] The PWMT electromagnetic transient simulation model is established based on the PLECS simulation platform. Figure 4 This is a schematic diagram of the LVRT test. The low voltage generator simulates voltage drops of different depths to perform electromagnetic transient simulation, and the machine-side response curve obtained by the electromagnetic transient simulation is used as the test result. The test conditions are as follows: the initial power is from 0.2pu to 1.0pu, the step size is 0.2pu, and the voltage drop depth is U t They are 0.4pu and 0.8pu respectively, and the power factors PF are 1.0 and 0.95 respectively. The electrical quantities that need to be included in the machine-side response curve are the voltage U, active power P and reactive power Q at the measuring point 1. The reactive current and active current values ​​during the LVRT steady state at different drop depths are extracted to construct the fault condition data set. The electromagnetic transient model of 1.5MWPMSG is built based on the PLECS simulation platform. The main parameters are shown in Table 2. The LVRT control strategy in the model adopts the typical LVRT control strategy.

[0173] Table 2 Main parameters of direct drive fan

[0174]

[0175] To verify the feasibility of the proposed method, the electromagnetic transient simulation results are first used as test results. Then, the LVRT control mode and parameters are identified according to the proposed identification method. Finally, the identified LVRT control parameters are substituted into the electromechanical transient simulation model established in the PSASP platform. The deviation between the electromechanical transient simulation results and the test results is calculated to verify the effectiveness of the proposed identification method.

[0176] The errors of U, I, P and Q under different control modes are shown in Table 3. According to the identification results, the selected LVRT control mode is determined to be the specified current control mode, and the selected LVRT control parameter is K. 1,Ip,LV , K 2,Ip,LV ,I pset , K 1,Iq,LV , K 2,Iq,LV and I qset .

[0177] Table 3 Errors under different control modes

[0178]

[0179] The identification results of this method are compared with those of the least squares method (LS) and PSO algorithm to verify the effectiveness of the proposed identification method. The iterative curve is shown in Figure 5 As shown. Voltage drop depth U t The output response curves when the pu is 0.4 and 0.8 are as follows: Figure 6 and Figure 7 shown.

[0180] The active current control in the electromagnetic transient model uses a specified active current to test the model. The voltage drop occurs from 1s and lasts for 0.625s. The SVR-GSA algorithm is used to identify the LVRT control mode and parameters. Considering the randomness of the optimization algorithm, the optimal value is selected as the final identification result through multiple identifications to reduce the error caused by the randomness of the algorithm.

[0181] According to the embodiments disclosed in the present invention, the present invention also provides an electronic device, a readable storage medium and a computer program product.

[0182] Figure 8 A schematic block diagram of an example electronic device 800 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0183] like Figure 8As shown, the device 800 includes a computing unit 801, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0184] A number of components in the device 800 are connected to the I / O interface 805, including: an input unit 806, such as a keyboard, a mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a disk, an optical disk, etc.; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 809 allows the device 800 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0185] The computing unit 801 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 801 performs the various methods and processes described above, such as a direct-drive wind turbine structured identification method based on a gravitational search algorithm. For example, in some embodiments, the direct-drive wind turbine structured identification method based on a gravitational search algorithm may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 808. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 800 via the ROM 802 and / or the communication unit 809. When the computer program is loaded into the RAM 803 and executed by the computing unit 801, one or more steps of the direct-drive wind turbine structured identification method based on the gravitational search algorithm described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to execute the direct-drive wind turbine structural identification method based on the gravitational search algorithm in any other appropriate manner (for example, by means of firmware).

[0186] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0187] The program code for implementing the method disclosed in the present invention can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0188] In the context disclosed by the present invention, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0189] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0190] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0191] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0192] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in the present disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution disclosed in the present invention can be achieved, and this document is not limited here.

[0193] What is disclosed above is only a preferred embodiment of the present invention, which certainly cannot be used to limit the scope of rights of the present invention. A person skilled in the art can understand that all or part of the processes of the above embodiments and equivalent changes made according to the claims of the present invention still fall within the scope of the invention.

Claims

1. A structured identification method for direct-drive wind turbines based on a gravitational search algorithm, characterized in that: include: Step S1, based on the control structure of the mathematical model of the direct-drive wind turbine, determining the control parameters that need to be identified for each LVRT control mode in the LVRT control strategy; Step S2, using the SVR algorithm to train the original data set of the direct-drive wind turbine to obtain individual initial values ​​and optimization range; Step S3, using the GSA algorithm to identify the LVRT control mode and control parameters of the direct-drive wind turbine.

2. The direct-drive wind turbine structured identification method based on the gravitational search algorithm according to claim 1, characterized in that: In step S1, the mathematical model of the direct-drive fan mathematical model includes: Under steady-state conditions where the grid does not encounter voltage drop faults, the wind turbine maintains a constant power factor and operates in a conventional manner. The current direction flowing from the converter side to the grid side is defined as a positive value. At this time, the initial active current I flowing through the grid-connected inverter is p0 and initial reactive current I q0 They are: Where: P0 and Q0 are the initial active power and reactive power per unit value of the wind turbine when it is in normal operation, the reference value is the rated power, U0 is the initial voltage per unit value of the grid connection point, U0 = 1.0 pu, and the reference value is the rated voltage U of the grid connection point. N ; Steady-state control mode of active power: In this mode, the control strategy of the system remains unchanged, that is, the current command is not switched, and the active and reactive currents under the current normal operating state are maintained. At the same time, the integral control link in the system is frozen to prevent the occurrence of integral saturation; Steady-state control mode of reactive power: when the current command is in normal operation, the reactive current command after the integral link is frozen; When a voltage drop occurs in the power grid, the dynamic reactive current increment of the wind turbine ΔI t Should meet: △I t =K Uq (0.9-U t )AND N (0.2≤U t ≤0.9) Where: K Uq is the dynamic reactive current support factor, I N and U t are the per unit values ​​of the rated current of the wind turbine and the depth of the grid voltage drop, and the reference values ​​are the rated current I N and rated voltage U N ; During the LVRT process of the wind turbine, the PMSG changes from constant power factor control to reactive current priority control, where: The control strategy of active power in the LVRT process is divided into three modes: specified power control mode, specified current control mode and control mode based on the current before crossing; Active power specified power control mode: In this mode, the system sets a specific active power target value P during low voltage ride-through LVRT , the calculation formula is: Where: K P,LVRT is the low wear active power coefficient; P set,LV is the low-pass active power setting value; P0 is the initial active power; Ip LVRT is the active current command; U t is the terminal voltage; Active power specified current control mode: In this mode, the system sets a specific active power target value Ip during the LVRT period LVRT , the calculation formula is: I pLVRT =K 1,Ip,LV U t +K 2,Ip,LV I p0 +I pset In the formula, K 1,Ip,LV and K 2,Ip,LV All are low-power active current calculation coefficients; I pset It is the low active current setting value; I p0 is the initial active current; Active power control mode based on pre-crossing current: In this mode, the system maintains the active current and reactive current before entering the low voltage ride-through period unchanged, and at the same time makes the active current command Ip LVRT With the initial active current I p0 Keep it the same: I pLVRT =I p0 The control strategy of reactive power in the LVRT process is divided into two modes: specified power control mode and specified current control mode; Specified power control mode for reactive power: Specify reactive power Q during low voltage ride-through LVRT , divided by the terminal voltage U t Get reactive current command Iq LVRT , the calculation formula is: Where: K P,LVRT is the low-power reactive power coefficient; Q set,LV is the low-pass reactive power setting value; Q0 is the initial reactive power; Specified current control mode for reactive power: Introducing low voltage ride-through threshold U Lin , at this time the current command Iq LVRT for: AND qLVRT =K 1,Iq,LV (IN Lin -IN t )+K 2,Iq,LV AND q0 +I qset Where: K 1,Iq,LV and K 2,Iq,LV All are low-power reactive current calculation coefficients; I qset It is the setting value of low reactive current; I q0 is the initial reactive current; When the LVRT control strategy is the specified power control mode, the control parameter that needs to be identified is K P,LVRT , P set,LV , K Q,LVRT and Q set,LV ; When the LVRT control strategy is the specified current control mode or the control mode based on the pre-crossing current, the control parameter that needs to be identified is K 1,Ip,LV , K 2,Ip,LV ,I pset , K 1,Iq,LV , K 2,Iq,LV and I qset .

3. The direct-drive wind turbine structured identification method based on the gravitational search algorithm according to claim 2, characterized in that: Before step S1, the method further includes: Step S4, establishing a PWMT electromagnetic transient simulation model based on the PLECS simulation platform, wherein the LVRT control strategy in the PWMT electromagnetic transient simulation model adopts a typical LVRT control strategy, and a low voltage generating device is used to simulate the voltage drop U at each depth. t An electromagnetic transient simulation is performed, and a machine-end response curve obtained by the electromagnetic transient simulation is used as a reference result; the electrical quantities included in the machine-end response curve are the voltage U, current I, active power P and reactive power Q at the output end of the wind turbine.

4. The direct-drive wind turbine structured identification method based on the gravitational search algorithm according to claim 3, characterized in that: The step S2 comprises: Step S21, data preparation: for each LVRT control strategy, simulate different voltage drop depths U t The target parameter and influencing factor data are extracted according to the fault condition. The target parameter is K 1,Ip,LV , K 2,Ip,LV ,I pset , K 1,Iq,LV , K 2,Iq,LV and I qset The influencing factors include the voltage drop depth U t , power factor PF, voltage U, current I, active power P and reactive power Q; Step S22, data preprocessing: normalize and standardize the collected data: Where b i and B i are the eigenvalues ​​before and after normalization; μ is the mean, σ is the standard deviation; Step S23, defining the objective function f(x), wherein the SVM regression algorithm is used for feature mapping: f(x)=w·φ(x)+b Among them, f(x) is the predicted parameter value; w is the weight vector; b is the bias term; φ(x) is the feature mapping function, which is used to map the input features to the high-dimensional feature space, and the radial basis function RBF is used as the feature mapping function φ(x) for mapping: K(x,x′)=exp(-γ||x-x′|| 2 ) Among them, γ is the kernel function parameter, K(x,x') is the kernel function; Step S24, parameter adjustment: use the square loss function to adjust the model parameters and obtain the SVR model parameters: Where C is the regularization parameter; ∈ i is the slack variable; w is the weight of the SVR model, is the loss function; n is the length of the database; Step S25, model training: using the selected kernel function and the adjusted SVR model parameters to train the SVM regression model; Step S26, model verification: Use the test data set to verify the prediction performance of the model by calculating the prediction error and determination coefficient R 2 , the mean square error MSE indicator determines whether the model training is complete; Step S27, model application: applying the trained SVM regression model to new data to predict target parameters.

5. The direct-drive wind turbine structured identification method based on the gravitational search algorithm according to claim 4, characterized in that: The step S3 uses the GSA algorithm to identify the LVRT control mode and control parameters of the direct-drive wind turbine, including: Step S31, identifying the optimal LVRT control parameters; Step S32: The optimal LVRT control parameter K P,LVRT , P set,LV , K Q,LVRT and Q set,LV Substitute the direct-drive fan mathematical model and execute it according to the specified power control mode to calculate the specified power error F PQ ; Step S33: The optimal LVRT control parameter K 1,Ip,LV , K 2,Ip,LV ,I pset , K 1,Iq,LV , K 2,Iq,LV and I qset Substitute the direct-drive fan mathematical model and execute it according to the specified current control mode to calculate the specified current error F1; Step S34, determine whether F PQ <F1, if yes, then the identification result is the specified power control mode, and the parameter to be selected is K P,LVRT , P set,LV , K Q,LVRT and Q set,LV ; If not, continue to execute step S35; Step S35, determining whether the optimal LVRT control parameters simultaneously satisfy K 1,Ip,LV ≈0, K 2,Ip,LV ≈1, I pset ≈0, if so, the identification result is the control mode based on the current before crossing, and the parameter to be selected is K 1,Ip,LV , K 2,Ip,LV ,I pset , K 1,Iq,LV , K 2,Iq,LV and I qse If yes, the identification result is the specified current control mode, and the parameter to be selected is K 1,Ip,LV , K 2,Ip,LV ,I pset , K 1,Iq,LV , K 2,Iq,LV and I qse .

6. The direct-drive wind turbine structured identification method based on the gravitational search algorithm according to claim 5, characterized in that: The step S31 comprises: Step S311, initialization: randomly generate N celestial bodies, each celestial body represents a set of LVRT control parameters; initialize various parameters of the algorithm; Step S312, calculate fitness: calculate the fitness F for each celestial body i (k): In the formula, i represents a celestial body; n astart Indicates the initial count; n aend Indicates the end count, n aend ∈[1,N]; U i ,I i , P i and Q i The terminal voltage, current, active power and reactive power corresponding to the execution of the i-th group of LVRT control parameters; Step S313, update the mass of the celestial body: update the mass M of each celestial body according to the fitness i , the higher the fitness, the greater the quality: Where M0 is the initial mass, F i is the fitness of the i-th celestial body, F min and F max are the minimum and maximum values ​​of fitness respectively; Step S314, calculate gravity and acceleration: for each pair of celestial bodies, calculate the mutual gravity according to the mass difference and distance between them, and calculate the acceleration according to the gravity: Where Y ij is the gravitational force between the i-th and j-th celestial bodies, G is the gravitational constant, r ij is the distance between the two celestial bodies, where a i is the acceleration of the i-th celestial body; Step S315, update position: update the position of each celestial body according to acceleration and velocity, i.e., LVRT control parameters: r i (t+1)=r i (t)+v i (t)+a i (t) In the formula, r i (t) is the position of the i-th celestial body at time t, v i (t) is the speed; Step S316, speed update: update the speed and acceleration according to the algorithm rules, and set the maximum speed and maximum acceleration at the same time: v i (t+1)=v i (t)+a i (t) Step S317, checking the convergence condition: if the maximum number of iterations is reached or the error requirement is met, the algorithm ends; the error requirement means that the sum of the errors of U, I, P and Q is not higher than the error threshold; Step S318, selecting the optimal solution: selecting the celestial body with the smallest fitness value as the optimal LVRT control parameter.

7. A direct-drive wind turbine structured identification system based on gravitational search algorithm, characterized in that: Used to execute the method according to any one of claims 1 to 6.

8. An electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method according to any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-6.

10. A computer program product, comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 6.

Images