Double-fed fan parameter identification method based on improved sparrow algorithm

Abstract

The invention discloses a doubly-fed fan parameter identification method based on an improved sparrow algorithm, and the method comprises the steps: analyzing the sensitivity of doubly-fed fan parameters, selecting key control parameters, building a doubly-fed fan low-voltage ride-through transient model in a PSCAD (Power System Computer Aided Design), obtaining a plurality of groups of output data through adjusting the key control parameters, and carrying out the identification of the doubly-fed fan parameters. Constructing a multi-target error function of the original model and the identification model; the improved sparrow algorithm is adopted to optimize and solve the established multi-target error function based on the key control parameters and the output data, the identification result of the to-be-identified control parameters of the doubly-fed fan is obtained, an identification model obtained through identification and an original model are compared and verified, the key control parameters in the transient model of the doubly-fed fan can be accurately and reliably identified, and the identification accuracy of the to-be-identified control parameters of the doubly-fed fan is improved. The problems that a traditional doubly-fed fan parameter identification method is low in identification precision, poor in adaptability to extreme working conditions and low in consistency of multiple identification results are effectively solved.

Description

Technical Field

[0001] The invention belongs to the technical field of parameter identification of renewable energy power generation, and in particular to a method for identifying parameters of a doubly-fed wind turbine based on an improved sparrow algorithm. Background Art

[0002] In the study of grid stability issues, current grid analysis mainly relies on digital simulation or hybrid simulation. With the continuous increase in the proportion of renewable energy, the impact of model simulation accuracy on grid stability cannot be ignored. Correct simulation models and parameters are the basis for studying the fault characteristics and protection technology of renewable energy grid-connected inverters. Therefore, parameter identification of wind turbine controllers is of great research significance.

[0003] In the existing ultra-high voltage AC / DC hybrid power grid, large-scale renewable energy is connected at the same time, making the operation characteristics of the power grid more complicated and the control more difficult. The simulation of large power grid faces severe challenges. In addition, there are many manufacturers of renewable energy power generation equipment connected to the grid in the existing power grid, and there are more than a hundred renewable energy power generation models. The equipment structures and grid-connected characteristics of different models from different manufacturers are significantly different. The typical model parameter simulation cannot accurately simulate the key grid-connected characteristics of different models, and it is difficult to meet the needs of power grid simulation calculations. Therefore, it is necessary to design a doubly fed wind turbine parameter identification method based on the improved sparrow algorithm to solve the above problems. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a doubly fed wind turbine parameter identification method based on an improved sparrow algorithm. For a doubly fed wind turbine low voltage ride-through identification model established in PSCAD, the control parameters of the PSCAD doubly fed wind turbine transient model are identified based on the sparrow algorithm, and the key control parameters in the doubly fed wind turbine transient model including the grid-side converter PI controller control parameters can be accurately and reliably identified.

[0005] In order to achieve the above technical effects, the technical solution adopted by the present invention is: A method for identifying parameters of a doubly-fed wind turbine based on an improved sparrow algorithm, comprising: S1, divide the control parameters of the doubly fed wind turbine into key control parameters and non-key control parameters. The non-key control parameters use fixed typical parameters, and the key parameters are identified by combining test data: K p5 , k p6 , k p7 , k i5 , k i6 , k i7 is selected as the parameter to be identified, where k p5 is the proportional coefficient of the GSC voltage controller, k i5 is the integral control coefficient of the GSC voltage controller, k p6is the proportional control coefficient of the inner loop of the d-axis current of GSC, k i6 is the GSC d-axis current inner loop integral control coefficient, k p7 is the q-axis current inner loop proportional control coefficient of GSC, k i7 is the q-axis current inner loop integral control coefficient of GSC; S2, build a doubly-fed wind turbine control system model in PSCAD, analyze parameter sensitivity to identify the best observation, obtain multiple sets of output data by adjusting key control parameters, and construct the error function between the original model and the identification model; S3, by improving the sparrow algorithm, the error function established is optimized based on the key control parameters and output data, and the optimization result is used as the identification result of the key control parameters of the identification model to obtain the identification result of the control parameters to be identified of the doubly fed wind turbine; S4, compare and verify the identified model and the original model to determine the reliability of the identified model.

[0006] Preferably, the method of step S2 is: S201, a three-phase short circuit fault is set at the beginning of the system grid connection, the parameter sensitivity is analyzed to identify the best observation value, and the corresponding disturbance and observation value are selected according to the sensitivity value of each parameter; the trajectory sensitivity value calculation formula and parameter curve are: ; ; In the formula, S j is the sensitivity value of the jth parameter, y(﹒) is the output of the model, N is the number of sampling times, y0 is the steady-state value of the output, K r To divide K j The exact values ​​of other parameters except K j0 K j The given value of ΔK j K j The change in ΔK is preferably 10%. j =0.1 K j0 ; For each parameter K j , calculate the change in system output y when increasing and decreasing the disturbance amplitude i (K r K j +ΔK j,n ) and y i (K r K j0 -ΔK j,n ), substitute the above result into S j The specific sensitivity value can be obtained; According to the sensitivity calculation results, select active power P, reactive power Q and reactive current I qAs the observed variable under voltage sag disturbance, the identifiability of control parameters is determined; S202, construct the error function of the original model and the identification model: The fitness values ​​of all optimization values ​​are calculated using the sum of the variances of the three sets of output signals according to the calculated multi-objective error function formula: ; In the formula, represents the value at the xth time point, k represents how many data there are in a set of data, Indicates the actual running results in rtlab, It represents the value filled into the simulation model according to the predicted parameters, and the square of the difference between the two is used as the fitness value of the parameter identification method.

[0007] Preferably, the method of step S3 is: S301, establish a function in which the optimization process is affected by the individual optimal value and the population optimal value; the population individual position update formula is: ; In the formula, is the position of the i-th sparrow in the d-dimensional search space; θ is a random constant between (0,1]; i termax is the maximum number of iterations; R2 represents the warning value; S safe is the safety value; Q is a normally distributed random number; L is a d-dimensional unit column vector; when the foraging environment is safe, that is, R2 <R safe When the foraging environment is dangerous, the discoverer will sound an alarm, signaling the population to go to other locations for hunting; As the number of searches increases, the search step should be more refined to explore potential optimization directions. The dynamic adjustment formula for the search step is as follows: ; ; In the formula, step α1 is R2; α is the adaptive adjustment factor, 0<α<1, step α0 is the initial value of the search step; H is the current search number; H max is the maximum number of searches; In the predation behavior, when it is considered to be closer to the optimal solution, the predator's siege step length should also be gradually reduced, as shown in the following formula: ; ; In the formula, step c1 For S safe; β is the adjustment factor, 0<β<1, rand is a randomly generated number between (-1,1), step c0 is the initial value of the search step, n is the current number of iterations, n max is the maximum number of iterations; S302, initialize basic parameters of individual iteration, set population individual number, total iteration number, particle dimension, individual X maximum value MaxX , minimum value MinX and critical fitness value ε; S303, introduce the Logistic-Tent composite chaotic map to initialize the position of the sparrow group in the solution space: ; In the formula, Z i+1 is the generated composite chaotic sequence, Z i The range is [0,1]; r is the value range of the control parameter, which is [0,4]; modl is the modulo operation; S303, adjusting the individual search step length and the predator siege step length according to the number of iterations, and updating the individual position in the population; The composite chaotic sequence generated by the above formula is used to generate the initial solution position sequence of the sparrow individuals in the search area: ; Where M i is the individual position sequence of the sparrow after initialization; X i,max , X i,min They correspond to the upper and lower bounds of the solution space respectively; S304, in order to prevent the algorithm from falling into a local optimal solution, after each iteration, a part of the individuals are randomly migrated: ; In the formula, represents the global optimal position of generation t; represents the global worst position of the tth generation; S305, update X according to the update formula of the individual position of the sparrow i,j position, and X i,j Bring in Calculate the fitness of sparrow population; S306, if satisfied <ε or the number of iterations is reached, the iteration is stopped and the optimal solution is output as the identification result of the control parameters. Otherwise, steps S301-S305 are repeated until <ε or the number of iterations is reached, the iteration is stopped.

[0008] Preferably, step S4 includes comparing and verifying the identification result obtained by optimizing the ISSA algorithm with hardware-in-the-loop experimental data at voltage drop levels of 20%, 40%, 60% and 80%.

[0009] Furthermore, in step S4, in order to further verify the effectiveness and practicability of the proposed model identification, the ISSA algorithm is compared with the PSO algorithm and the SSA algorithm, and the same measured data are used for calculation using the above algorithms.

[0010] Furthermore, four types of deviations of the identification results at each fault period under 20%, 40%, 60% and 80% voltage drop levels are calculated respectively to measure the accuracy of the parameter identification results; the specific steps are as follows: S401, select the number of iterations as 60, perform 60 simulations, take the average value as the result of parameter identification, and obtain the objective function convergence curves of three different optimization algorithms, namely, PSO algorithm, SSA algorithm, and ISSA algorithm, when solving control parameters; S402, respectively compare the output signal of the original model with the output signal of the final identification model. The deviation types include: average deviation, average absolute deviation, maximum deviation and weighted average deviation. According to the model verification standard, the output characteristics of the DFIG wind power system during the LVRT period can be divided into three periods, namely, before the fault a, during the fault b, and after the fault c. Among them, for the steady-state intervals A, B2, and C2 of the output characteristics of each period, it is necessary to calculate the steady-state average deviation , steady-state mean absolute deviation , and the maximum steady-state deviation ; For transient intervals B1 and C1, it is necessary to calculate the transient average deviation and transient mean absolute deviation ; The deviations of the three periods a, b, and c are weighted averaged according to the weights of 10%, 60%, and 30% respectively to obtain the weighted average absolute deviation of the entire period ; The specific formula is as follows: ; Where: y obs and sim are the test value and simulation value respectively; ε avg and ε abs are the mean deviation and mean absolute deviation of each interval respectively; ε G is the weighted mean absolute deviation; max is the maximum allowable value of the maximum deviation; if it meets the maximum allowable value of the deviation specified in the electromechanical transient model verification standard, the identification result is valid and meets the engineering requirements.

[0011] Compared with the prior art, the beneficial effects of the present invention include: 1. Based on the preliminary identification values ​​of GSC control parameters, this method uses the improved sparrow algorithm to accurately identify the GSC control parameters of the doubly fed wind turbine under different operating conditions, effectively solving the problems of low identification accuracy, poor adaptability to extreme working conditions and low consistency of multiple identification results of the traditional doubly fed wind turbine parameter identification method.

[0012] 2. In order to improve the ability of SSA algorithm in dealing with nonlinear parameter identification research, this method makes nonlinear improvements to the search step size and search range, and introduces a step of randomly migrating the control parameter solution position after each iteration, which effectively avoids the problem that traditional SSA is prone to falling into local optimality, and improves the global search ability and convergence speed of the SSA algorithm.

[0013] 3. This method uses the trajectory sensitivity method to analyze the correlation between each GSC control parameter and different observations, and selects observations with higher sensitivity to each control parameter as the construction index of the fitness objective function of the improved sparrow algorithm, which effectively improves the parameter identification accuracy.

[0014] 4. This method uses the weighted mean absolute deviation to measure the response curve error, calculates the response error between the identification curve and the actual curve under the three-phase short-circuit fault condition, and further verifies the accuracy of the identification result. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The present invention will be further described below in conjunction with the accompanying drawings and embodiments: Figure 1 It is a flow chart of control parameter identification of the doubly-fed wind turbine converter of the present invention; Figure 2 It is a flowchart of the steps of the ISSA parameter identification algorithm of the present invention; Figure 3 It is a trajectory sensitivity calculation result diagram between different observation quantities and GSC control parameters in the embodiment; Figure 4 It is a graph of iterative convergence of PSO\SSA\ISSA algorithms in the embodiment; Figure 5 This is a comparison diagram of the actual curves and identification curve responses of active power, reactive power, reactive current, and terminal voltage when LVRT is 0.8 and P is 0.9, i.e., the wind turbine is in high-power operation state; Figure 6 This is a comparison diagram of the actual curves and identification curve responses of active power, reactive power, reactive current, and terminal voltage when LVRT is 0.6 and P is 0.9, i.e., the wind turbine is in high-power operation state; Figure 7 It is a comparison diagram of the actual curves and the identification curve responses of the active power, reactive power, reactive current, and terminal voltage under the operating state when LVRT is 0.4 in the embodiment; Figure 8 It is a comparison diagram of the actual curves and the identification curve responses of the active power, reactive power, reactive current, and terminal voltage under the operating state when LVRT is 0.2 in the embodiment; Fig. 9 The actual curves of reactive current, reactive power, active power, and terminal voltage and the identification curve response comparison diagram of LVRT 0.8, P=0.2, i.e., low-power operation state in the embodiment; Fig.10 The actual curves of reactive current, reactive power, active power, and terminal voltage and the identification curve response comparison diagram of LVRT 0.6, P=0.2, i.e., low-power operation state in the embodiment; Fig.11 The actual curves of reactive current, reactive power, active power, and terminal voltage and the identification curve response comparison diagram of LVRT 0.4, P=0.2, i.e., low-power operation state in the embodiment; Fig.12 LVRT 0.2, P = 0.2 in the embodiment, that is, the actual curves of reactive current, reactive power, active power, and terminal voltage and the identification curve response comparison diagram under the low power operation state. DETAILED DESCRIPTION

[0016] Embodiment 1: like Figure 1 As shown, a method for identifying parameters of a doubly-fed wind turbine based on an improved sparrow algorithm comprises: S1, divide the control parameters of the doubly fed wind turbine into key control parameters and non-key control parameters. The non-key control parameters use fixed typical parameters, and the key parameters are identified by combining test data: K p5 , k p6 , k p7 , k i5 , k i6 , k i7 is selected as the parameter to be identified, where k p5 is the proportional coefficient of the GSC voltage controller, k i5 is the integral control coefficient of the GSC voltage controller, k p6 is the proportional control coefficient of the inner loop of the d-axis current of GSC, k i6 is the GSC d-axis current inner loop integral control coefficient, k p7 is the q-axis current inner loop proportional control coefficient of GSC, k i7 is the q-axis current inner loop integral control coefficient of GSC; S2, build a doubly-fed wind turbine control system model in PSCAD, analyze parameter sensitivity to identify the best observation, obtain multiple sets of output data by adjusting key control parameters, and construct the error function between the original model and the identification model; S3, by improving the sparrow algorithm, the error function established is optimized based on the key control parameters and output data, and the optimization result is used as the identification result of the key control parameters of the identification model to obtain the identification result of the control parameters to be identified of the doubly fed wind turbine; S4, compare and verify the identified model and the original model to determine the reliability of the identified model.

[0017] Preferably, the method of step S2 is: S201, a three-phase short circuit fault is set at the beginning of the system grid connection, the parameter sensitivity is analyzed to identify the best observation value, and the corresponding disturbance and observation value are selected according to the sensitivity value of each parameter; the trajectory sensitivity value calculation formula and parameter curve are: ; ; In the formula, S j is the sensitivity value of the jth parameter, y(﹒) is the output of the model, N is the number of sampling times, y0 is the steady-state value of the output, K r To divide K j The exact values ​​of other parameters except K j0 K j The given value of ΔK j K j The change in ΔK is preferably 10%. j =0.1 K j0 ; For each parameter K j , calculate the change in system output y when increasing and decreasing the disturbance amplitude i (K r K j +ΔK j,n ) and y i (K r K j0 -ΔK j,n ), substitute the above result into S j The specific sensitivity value can be obtained; According to the sensitivity calculation results, select active power P, reactive power Q and reactive current I q As the observed variable under voltage sag disturbance, the identifiability of control parameters is determined; S202, construct the error function of the original model and the identification model: The fitness values ​​of all optimization values ​​are calculated using the sum of the variances of the three sets of output signals according to the calculated multi-objective error function formula: ; In the formula, represents the value at the xth time point, k represents how many data there are in a set of data, Indicates the actual running results in rtlab, It represents the value filled into the simulation model according to the predicted parameters, and the square of the difference between the two is used as the fitness value of the parameter identification method.

[0018] like Figure 2 As shown, preferably, the method of step S3 is: S301, establish a function in which the optimization process is affected by the individual optimal value and the population optimal value; the population individual position update formula is: ; In the formula, is the position of the i-th sparrow in the d-dimensional search space; θ is a random constant between (0,1]; i termax is the maximum number of iterations; R2 represents the warning value; S safe is the safety value; Q is a normally distributed random number; L is a d-dimensional unit column vector; when the foraging environment is safe, that is, R2 <R safe When the foraging environment is dangerous, the discoverer will sound an alarm, signaling the population to go to other locations for hunting; As the number of searches increases, the search step should be more refined to explore potential optimization directions. The dynamic adjustment formula for the search step is as follows: ; ; In the formula, step α1 is R2; α is the adaptive adjustment factor, 0<α<1, step α0 is the initial value of the search step; H is the current search number; H max is the maximum number of searches; In the predation behavior, when it is considered to be closer to the optimal solution, the predator's siege step length should also be gradually reduced, as shown in the following formula: ; ; In the formula, step c1 For S safe ; β is the adjustment factor, 0<β<1, rand is a randomly generated number between (-1,1), step c0 is the initial value of the search step, n is the current number of iterations, n max is the maximum number of iterations; S302, initialize basic parameters of individual iteration, set population individual number, total iteration number, particle dimension, individual X maximum value MaxX , minimum value MinX and critical fitness value ε; S303, introduce the Logistic-Tent composite chaotic map to initialize the position of the sparrow group in the solution space: ; In the formula, Z i+1 is the generated composite chaotic sequence, Z i The range is [0,1]; r is the value range of the control parameter, which is [0,4]; modl is the modulo operation; S303, adjusting the individual search step length and the predator siege step length according to the number of iterations, and updating the individual position in the population; The composite chaotic sequence generated by the above formula is used to generate the initial solution position sequence of the sparrow individuals in the search area: ; Where M i is the individual position sequence of the sparrow after initialization; X i,max , X i,min They correspond to the upper and lower bounds of the solution space respectively; S304, in order to prevent the algorithm from falling into a local optimal solution, after each iteration, a part of the individuals are randomly migrated: ; In the formula, represents the global optimal position of generation t; represents the global worst position of the tth generation; S305, update X according to the update formula of the individual position of the sparrow i,j position, and X i,j Bring in Calculate the fitness of sparrow population; S306, if satisfied <ε or the number of iterations is reached, the iteration is stopped and the optimal solution is output as the identification result of the control parameters. Otherwise, steps S301-S305 are repeated until <ε or the number of iterations is reached, the iteration is stopped.

[0019] Preferably, step S4 includes comparing and verifying the identification result obtained by optimizing the ISSA algorithm with hardware-in-the-loop experimental data at voltage drop levels of 20%, 40%, 60% and 80%.

[0020] Furthermore, in step S4, in order to further verify the effectiveness and practicability of the proposed model identification, the ISSA algorithm is compared with the PSO algorithm and the SSA algorithm, and the same measured data are used for calculation using the above algorithms.

[0021] Furthermore, four types of deviations of the identification results at each fault period under 20%, 40%, 60% and 80% voltage drop levels are calculated respectively to measure the accuracy of the parameter identification results; the specific steps are as follows: S401, select the number of iterations as 60, perform 60 simulations, take the average value as the result of parameter identification, and obtain the objective function convergence curves of three different optimization algorithms, namely, PSO algorithm, SSA algorithm, and ISSA algorithm, when solving control parameters; S402, respectively compare the output signal of the original model with the output signal of the final identification model. The deviation types include: average deviation, average absolute deviation, maximum deviation and weighted average deviation. According to the model verification standard, the output characteristics of the DFIG wind power system during the LVRT period can be divided into three periods, namely, before the fault a, during the fault b, and after the fault c. Among them, for the steady-state intervals A, B2, and C2 of the output characteristics of each period, it is necessary to calculate the steady-state average deviation , steady-state mean absolute deviation , and the maximum steady-state deviation ; For transient intervals B1 and C1, it is necessary to calculate the transient average deviation and transient mean absolute deviation ; The deviations of the three periods a, b, and c are weighted averaged according to the weights of 10%, 60%, and 30% respectively to obtain the weighted average absolute deviation of the entire period ; The specific formula is as follows: ; Where: y obs and sim are the test value and simulation value respectively; ε avg and ε abs are the mean deviation and mean absolute deviation of each interval respectively; ε G is the weighted mean absolute deviation; max is the maximum allowable value of the maximum deviation; if it meets the maximum allowable value of the deviation specified in the electromechanical transient model verification standard, the identification result is valid and meets the engineering requirements.

[0022] Embodiment 2: This embodiment implements this method based on the electromechanical transient modeling of the doubly fed wind turbine model. Its GSC control parameters should be as adaptable to all its working conditions as possible, so that the model can meet the model verification deviation index under all working conditions. Therefore, the adaptability of the identification results to extreme working conditions should be given priority. The identification results are brought into different working conditions, the errors between the simulation and the test values ​​are calculated, and compared with the fitting results of the ordinary SSA algorithm. U , I , I q and Q The deviation of the model is verified to verify the accuracy of the modeling. Taking U as an example, the disturbance process is divided into five intervals: A, B1, B2, C1, C2, where A is the steady-state period from the start of the simulation to the occurrence of the fault, B1 is the fault occurrence period, B2 is the period from the occurrence of the fault to the beginning of the fault recovery period, C1 is the fault recovery period, and C2 is the steady-state period after the fault recovery.

[0023] Select 4 groups of working conditions including the power in the high wind state, that is, high power (P0>0.9pu); and 4 groups of working conditions under the low wind state, that is, low power (P0<0.3pu), and calculate the deviations of the model B1 and C1 intervals; like Figure 3 As shown in the figure, Step 1: divide the control parameters of the doubly fed wind turbine into key control parameters and non-key control parameters. The non-key control parameters use fixed typical parameters, and the key parameters are identified by combining test data. p5 , k p6 , k p7 , k i5 , k i6 , k i7 is selected as the parameter to be identified, where k p5 is the proportional coefficient of the GSC voltage controller, k i5 is the integral control coefficient of the GSC voltage controller, k p6 is the proportional control coefficient of the inner loop of the d-axis current of GSC, k i6 is the GSC d-axis current inner loop integral control coefficient, k p7 is the q-axis current inner loop proportional control coefficient of GSC, k i7 is the q-axis current inner loop integral control coefficient of GSC; Step 2: Build a low voltage ride-through transient model of a doubly-fed wind turbine in PSCAD, obtain multiple sets of output data by adjusting key control parameters, and construct a multi-objective error function between the original model and the identification model: S201, analyze parameter sensitivity to identify the best observation value, select appropriate disturbance and observation value according to the sensitivity value of each parameter, and the trajectory sensitivity value calculation formula and parameter curve are as follows: ; ; In the formula, S j is the sensitivity value of the jth parameter, y(﹒) is the output of the model, N is the number of sampling times, y0 is the steady-state value of the output, Kr is the exact value of other parameters except Kj, Kj0 is the given value of Kj, ΔKj is the change of Kj, which is taken as 10% in this paper, that is, ΔKj=0.1 Kj0.

[0024] For each parameter Kj, calculate the change in system output when increasing and decreasing the disturbance amplitude: yi(KrKj0+ΔKj,n), yi(KrKj0-ΔKj,n); substitute the above results into Sj to obtain the specific sensitivity value.

[0025] According to the sensitivity calculation results, active power P, reactive power Q, and reactive current Iq are selected as observation variables under voltage sag disturbance to determine the identifiability of control parameters; S202, constructing a multi-objective error function of the original model and the identification model: as shown in the following formula, the fitness values ​​of all optimization values ​​are calculated according to the calculated multi-objective error function formula with the sum of the variances of the three groups of output signals; ; Substituting each observation into the following formula: ; In the formula, , , They are the active power, reactive power, and reactive current obtained by substituting the parameters into the PSCAD model; , , The measured active power, reactive power, and reactive current; Step 3: The improved sparrow algorithm is used to optimize and solve the error function established based on the key control parameters and output data. The optimization result is the identification result of the key control parameters of the identification model, and the identification result of the control parameters to be identified of the doubly fed wind turbine is obtained; the specific steps are as follows: Step 301, introduce the Logistic-Tent composite chaotic mapping to initialize the positions of six parameters to be identified, i.e., kp5, kp6, kp7, ki5, ki6, and ki7, in the solution space. The chaotic mapping formula is as follows: ; In the formula, Z i+1 is the generated composite chaotic sequence, Z i The range is [0,1];r The range is [0,4]; modl is the modulo operation.

[0026] The composite chaotic sequence generated by the above formula is used to generate the initial solution position sequence of each control parameter in the solution space. The formula is shown as follows: ; Where Mi is the position sequence of each control parameter after initialization; X i,max 、X i,min They correspond to the upper and lower bounds of the solution space respectively.

[0027] S302, initialize the upper bound of the solution space X i,max and the lower bound X i,min , and the critical fitness value ε ; S303, the optimization process is affected by the optimal value of the control parameter and the optimal value of the control parameter group. The update formula of each parameter position in the parameter group is as follows: ; In the formula, For the i The position of the control parameters in the d-dimensional search space; θ is a random constant between (0,1]; i termax is the maximum number of iterations; R 2 represents the search step length; S safe is the search step range; Q is a normally distributed random number; L is a d-dimensional unit column vector. When the parameter is within the set range, that is, R 2< Rsafe When , the control parameter solution position can be searched here; when , then the search for the control parameter solution position will not continue in this area.

[0028] As the number of searches increases, the search step should be gradually reduced to explore potential optimization directions. The dynamic adjustment formula for the search step is as follows: ; ; In the formula, stepα1 is R2 in the above formula; α is the adaptive adjustment factor, 0<α<1, stepα0 is the initial value of the search step; H is the current search number; H max is the maximum number of searches; When it is considered to be closer to the optimal solution, the control parameter range should also be gradually reduced, as shown in the following formula: ; ; In the formula, step c1 is S in the above formula safe ; β is the adjustment factor, 0<β<1, rand is a randomly generated number between (-1,1), step c0 is the initial value of the search step, n is the current number of iterations, n max is the maximum number of iterations; S304, in order to prevent the algorithm from falling into a local optimal solution, after each iteration, a part of the solution is randomly migrated. The migration formula is as follows: ; In the formula, represents the global optimal position of generation t; represents the global worst position of the tth generation; S305, updating the position of each control parameter according to the updating formula of the control parameter position, and substituting the position into the above formula to calculate the fitness of the control parameter group; S306, if satisfied < ε Or the number of iterations is reached, the iteration is stopped and the optimal solution is output as the identification result of the control parameters. Otherwise, steps S301-S305 are repeated until the number of iterations is satisfied. < ε Or when the number of iterations is reached, the iteration is stopped; like Figure 5~Figure 12 As shown, S4: the identification results obtained by optimizing the ISSA algorithm are compared and verified with the hardware-in-the-loop experimental data at voltage drop levels of 20%, 40%, 60% and 80%.

[0029] Step 4, in order to further verify the effectiveness and practicality of the proposed model identification, the ISSA algorithm of the present invention is compared with the PSO algorithm and the SSA algorithm, and the same measured data is used for calculation using the above algorithms; further, four types of deviations of the identification results at each fault period under 20%, 40%, 60% and 80% voltage drop levels are calculated respectively to measure the accuracy of the parameter identification results; the specific steps are as follows: S401, select the number of iterations as 60, perform 60 simulations, and take the average value as the result of parameter identification. Figure 4 The objective function convergence curves of three different optimization algorithms, such as PSO algorithm, SSA algorithm, and ISSA algorithm, when solving the control parameters are shown in Figure 4 As shown; the error between the true parameter identification result and the true value is as shown in Table 1: Table 1: ISSA parameter identification results and errors;

[0030] S402, respectively compare the output signal of the original model with the output signal of the final identification model. The deviation types include: average deviation, average absolute deviation, maximum deviation and weighted average deviation. According to the model verification standard, the output characteristics of the DFIG wind power system during LVRT can be divided into three periods, namely before the fault a, during the fault b, and after the fault c; among them, for the steady-state intervals A, B2, and C2 of the output characteristics of each period, it is necessary to calculate the steady-state average deviation , steady-state mean absolute deviation , and the maximum steady-state deviation ; For transient intervals B1 and C1, it is necessary to calculate the transient average deviation and transient mean absolute deviation ; The deviations of the three periods a, b, and c are weighted averaged according to the weights of 10%, 60%, and 30% respectively to obtain the weighted average absolute deviation of the entire period ; The specific formula is as follows: ; In the formula, y obs and y sim are the test value and the simulation value respectively; ε avg and ε abs are the mean deviation and mean absolute deviation of each interval respectively; ε G is the weighted mean absolute deviation; ε max is the maximum allowable value of the maximum deviation; if it meets the maximum allowable value of the deviation specified in the electromechanical transient model verification standard, the identification result is valid and meets the engineering requirements; the deviation calculation results of the obtained simulation model under various working conditions are shown in Tables 2 to 9 below: Table 2: Simulation model deviation calculation table (working condition 1):

[0031] Table 3: Simulation model deviation calculation table (working condition 2):

[0032] Table 4: Simulation model deviation calculation table (working condition 3):

[0033] Table 5: Simulation model deviation calculation table (working condition 4):

[0034] Table 6: Simulation model deviation calculation table (working condition 5):

[0035] Table 7: Simulation model deviation calculation table (working condition 6):

[0036] Table 8: Simulation model deviation calculation table (working condition 7):

[0037] Table 9: Simulation model deviation calculation table (working condition 8):

[0038] As shown in Tables 2 to 9 above, the eight selected operating conditions are representative because they cover a range of voltage levels from nominal to extreme low voltage scenarios, simulating the LVRT capabilities of wind farms under voltage sags. These conditions not only verify the accuracy of the identified control parameters, but also evaluate the adaptability of the system to extreme conditions. By testing the deviations at different voltage levels, the effectiveness and robustness of the parameter identification method in complex power systems can be fully evaluated.

Claims

1. A method for identifying parameters of a doubly-fed wind turbine based on an improved sparrow algorithm, characterized in that: include: S1, divide the control parameters of the doubly fed wind turbine into key control parameters and non-key control parameters. The non-key control parameters use fixed typical parameters, and the key parameters are identified by combining test data: K p5 , k p6 , k p7 , k i5 , k i6 , k i7 is selected as the parameter to be identified, where k p5 is the proportional coefficient of the GSC voltage controller, k i5 is the integral control coefficient of the GSC voltage controller, k p6 is the proportional control coefficient of the inner loop of the d-axis current of GSC, k i6 is the GSC d-axis current inner loop integral control coefficient, k p7 is the q-axis current inner loop proportional control coefficient of GSC, k i7 is the q-axis current inner loop integral control coefficient of GSC; S2, build a doubly-fed wind turbine control system model in PSCAD, analyze parameter sensitivity to identify the best observation, obtain multiple sets of output data by adjusting key control parameters, and construct the error function between the original model and the identification model; S3, by improving the sparrow algorithm, the error function established is optimized based on the key control parameters and output data, and the optimization result is used as the identification result of the key control parameters of the identification model to obtain the identification result of the control parameters to be identified of the doubly fed wind turbine; S4, comparing and verifying the identified model and the original model to determine the reliability of the identified model.

2. The method for identifying parameters of a doubly-fed wind turbine based on an improved sparrow algorithm according to claim 1 is characterized in that: The method of step S2 is: S201, a three-phase short circuit fault is set at the beginning of the system grid connection, the parameter sensitivity is analyzed to identify the best observation value, and the corresponding disturbance and observation value are selected according to the sensitivity value of each parameter; the trajectory sensitivity value calculation formula and parameter curve are: ； ； In the formula, S j is the sensitivity value of the jth parameter, y(﹒) is the output of the model, N is the number of sampling times, y0 is the steady-state value of the output, K r To divide K j The exact values ​​of other parameters except K j0 K j The given value of ΔK j K j The change in value; for each parameter K j , calculate the change in system output y when increasing and decreasing the disturbance amplitude i (K r K j +ΔK j,n ) and y i (K r K j0 -ΔK j,n ), substitute the above result into S j The specific sensitivity value can be obtained; According to the sensitivity calculation results, select active power P, reactive power Q and reactive current I q As the observed variable under voltage sag disturbance, the identifiability of control parameters is determined; S202, construct the error function of the original model and the identification model: The fitness values ​​of all optimization values ​​are calculated using the sum of the variances of the three sets of output signals according to the calculated multi-objective error function formula: ； In the formula, represents the value at the xth time point, k represents how many data there are in a set of data, Indicates the actual running results in rtlab, It represents the value filled into the simulation model according to the predicted parameters, and the square of the difference between the two is used as the fitness value of the parameter identification method.

3. The method for identifying parameters of a doubly-fed wind turbine based on an improved sparrow algorithm according to claim 2 is characterized in that: The method of step S3 is: S301, establishing a function in which the optimization process is affected by the individual optimal value and the population optimal value; S302, initialize basic parameters of individual iteration, set population individual number, total iteration number, particle dimension, individual X maximum value MaxX , minimum value MinX and critical fitness value ε; S303, introduce the Logistic-Tent composite chaotic map to initialize the position of the sparrow group in the solution space: ； In the formula, Z i+1 is the generated composite chaotic sequence, Z i The range is [0,1]; r is the value range of the control parameter, which is [0,4]; modl is the modulo operation; S303, adjusting the individual search step length and the predator siege step length according to the number of iterations, and updating the individual position in the population; The composite chaotic sequence generated by the above formula is used to generate the initial solution position sequence of the sparrow individuals in the search area: ； Where M i is the individual position sequence of the sparrow after initialization; X i,max , X i,min They correspond to the upper and lower bounds of the solution space respectively; S304, after each iteration, a portion of individuals are randomly migrated: ； In the formula, represents the global optimal position of generation t; represents the global worst position of the tth generation; S305, update X according to the update formula of the individual position of the sparrow i,j position, and X i,j Bring in Calculate the fitness of sparrow population; S306, if satisfied <ε or the number of iterations is reached, the iteration is stopped and the optimal solution is output as the identification result of the control parameters. Otherwise, steps S301-S305 are repeated until <ε or the number of iterations is reached, the iteration is stopped.

4. The method for identifying parameters of a doubly-fed wind turbine based on an improved sparrow algorithm according to claim 3 is characterized in that: The step S4 includes comparing and verifying the identification result obtained by optimizing the ISSA algorithm with the hardware-in-the-loop experimental data at voltage drop levels of 20%, 40%, 60% and 80%.

Images