Permanent magnet synchronous wind driven generator parameter identification method based on improved grey wolf algorithm

Abstract

The invention discloses a permanent magnet synchronous wind driven generator parameter identification method based on an improved grey wolf algorithm, and the method comprises the steps: carrying out the optimization of an initial population through employing Cat chaotic mapping and a reverse learning strategy, enabling the traversal to be uniform, achieving a higher iteration speed, introducing a nonlinear convergence factor of cosine change, and strengthening the global search in an earlier stage, thereby achieving the recognition of the parameters of a permanent magnet synchronous wind driven generator. The local search is enhanced in the later stage, the diversity of populations is ensured, the search range is enlarged, the global search capability and the local search capability are balanced, the probability of falling into local optimum is reduced, the identification speed is accelerated, then the tube voltage drop nonlinearity of the voltage source type inverter is compensated, the voltage error between the output voltage and the reference voltage of the inverter is reduced, and the identification accuracy is improved. And finally, learning is carried out by combining memory self-learning and the individual experience of the grey wolf, so that the problems of low parameter identification precision, easiness in falling into local optimum and low speed of the parameter identification method of the permanent magnet synchronous motor are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of motor parameter identification, and more specifically, to a permanent magnet synchronous wind turbine generator parameter identification method based on an improved grey wolf algorithm. Background Art

[0002] Permanent magnet synchronous wind turbine generator (PMSWG) has the advantages of good control performance, simple structure and high power density. Its parameter identification has always been a key issue. Accurate parameters are crucial to achieving optimal control performance and control prediction.

[0003] The Gray Wolf Algorithm (GWO) is a meta-heuristic optimization algorithm that simulates the collaborative predation of gray wolf groups. It has the characteristics of simple structure, few parameters to be adjusted, and easy implementation. There are adaptively adjustable convergence factors and information feedback mechanisms, which can achieve a balance between local optimization and global search. Therefore, it has good performance in terms of problem solving accuracy and convergence speed compared with other intelligent algorithms. The A value in the traditional gray wolf algorithm is used to represent the global search ability and local search ability of the wolf group in searching for prey. When |A|>1, the gray wolf expands the search range and conducts a global search in order to find prey. When |A|<1, the wolf group will shrink the encirclement and attack the prey, and the gray wolf conducts a local search. The size of the A value is determined by the convergence factor a. As the number of iterations increases, the convergence factor a decreases linearly from 2 to 0. However, in fact, the gray wolf algorithm changes nonlinearly during the search process. Therefore, the use of a linear convergence factor cannot reflect the actual search process.

[0004] In 2021100548815, a method for parameter identification of permanent magnet synchronous motor based on improved gray wolf optimization algorithm was disclosed. The improved gray wolf optimization algorithm is as follows: first, the initial population is generated by using Fuch mapping and reverse learning strategy; second, a nonlinear decreasing convergence factor update formula is proposed, and the individual memory function in the particle swarm algorithm is introduced to enhance the local search ability of the algorithm and accelerate its convergence speed; finally, the adaptive normal cloud model is used to update and deeply develop the wolf pack position, further improving the ability of the gray wolf optimization algorithm to easily fall into the local optimum. The parameter identification of the permanent magnet synchronous motor is effectively realized by improving the gray wolf optimization algorithm. The improved gray wolf optimization algorithm is superior to the traditional gray wolf optimization algorithm in identification accuracy and stability. Although the patent adopts a nonlinear decreasing convergence factor to enhance the local search ability of the algorithm and accelerate its convergence speed, it is easy to fall into the probability of local optimum, and the voltage error between the inverter output voltage and the reference voltage leads to a decrease in parameter identification accuracy. Summary of the invention

[0005] The main technical problem to be solved by the present invention is to provide a permanent magnet synchronous wind turbine parameter identification method based on an improved grey wolf algorithm in view of the shortcomings of the permanent magnet synchronous motor parameter identification method in the prior art, such as low parameter identification accuracy, easy to fall into local optimum and slow speed.

[0006] The purpose of the present invention is achieved through the following technical solutions:

[0007] A method for identifying parameters of a permanent magnet synchronous wind turbine generator using an improved grey wolf algorithm, comprising the following steps:

[0008] S1. Collect electrical signals;

[0009] S2. Generate the initial population using Cat chaotic mapping and reverse learning strategy;

[0010] S3. Compensate for the voltage error caused by the nonlinearity of the tube voltage drop. The compensation is expressed as:

[0011]

[0012] Among them, V dc is the DC voltage, V CE is the on-state saturation voltage drop of the switch tube, V D is the freewheeling diode voltage drop;

[0013] S4. Calculate the nonlinear cosine convergence factor a, and coefficient matrices A and C;

[0014] S5. Calculate the distance and position of each wolf pack;

[0015] S6. Combined with memory self-learning, according to Update the position of the gray wolf, where r4 is a random number between [0, 1], is the best position for the i-th gray wolf to pass through, k1 and k2 are constants between [0, 1], which are the group communication coefficient and individual memory coefficient respectively;

[0016] S7. Calculate the fitness value of each wolf pack;

[0017] S8. After the iteration is completed, the optimal solution is output, which is the identification result of the resistance, inductance and permanent magnet flux.

[0018] Furthermore, the electrical signal includes i d =-2 and i d =0 when current, voltage and angular velocity.

[0019] Furthermore, the Cat mapping expression is:

[0020]

[0021] Where a=b=1, k represents the number of times the chaotic sequence is generated.

[0022] Further, the cosine convergence factor a is expressed as:

[0023]

[0024] Among them, t represents the current iteration number, t max is the maximum number of iterations, n is a constant greater than 1, a i and a f They represent the initial value and final value of the control parameter a respectively.

[0025] Furthermore, the coefficient matrices A and C are expressed as:

[0026]

[0027] in, are two random number vectors with values ​​in [0,1]. The value of a decreases linearly from 2 to 0 as the number of iterations increases.

[0028] Furthermore, the distance and position of the wolf pack are expressed as:

[0029]

[0030] in, It indicates the position that the ω gray wolf individual needs to adjust in the end. Respectively represent the step length and direction of approaching the α wolf pack, β wolf pack, and δ wolf pack,

[0031] Furthermore, the It is expressed as:

[0032]

[0033] in, They represent the distances of other ω gray wolf individuals from the α wolf pack, β wolf pack, and δ wolf pack respectively.

[0034] Furthermore, the It is expressed as:

[0035]

[0036] Among them, α, β, and δ represent the optimal solution, the second-best solution, and the third-best solution, respectively. is the current position of the gray wolf, They represent the positions of the α wolf pack, β wolf pack, and δ wolf pack respectively.

[0037] Furthermore, the fitness value calculation formula is:

[0038]

[0039] Where c1, c2, c3, c4 are constants in the range (0, 1), i d 、i q is the stator current of the dq axis. The symbols represented by "0" or "1" in the subscripts indicate that the dq axis is injected into i d =-2 and i d =0, are the corresponding estimated values.

[0041] Furthermore, if the fitness is less than the last iteration value, the value of each wolf pack is updated; if the fitness is greater than the last iteration value, return to S4 to update the nonlinear cosine convergence factor a.

[0042] Compared with the prior art, the beneficial effects are:

[0043] The present invention firstly optimizes the initial population by using Cat chaos mapping and reverse learning strategy, traverses evenly, ensures the diversity of the initial population, has a faster iteration speed, and introduces the nonlinear convergence factor of cosine variation, which can strengthen the global search in the early stage and the local search in the later stage, thereby ensuring the diversity of the population, increasing the search range, balancing the global search capability and the local search capability, reducing the probability of falling into the local optimum, and accelerating the identification speed. Then, the nonlinear voltage drop of the voltage source inverter tube is compensated, the voltage error between the inverter output voltage and the reference voltage is reduced, and the parameter identification accuracy is improved. Finally, the memory self-learning is combined to learn the individual experience of the gray wolf, and the influence of the individual optimal position is added when the position is updated each time, so as to strengthen the communication between the individual and the group, take into account the experience gained by the individual in the search process, avoid searching for repeated positions in the search process, and further improve the accuracy and speed of the gray wolf search. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a flow chart of the parameter identification method of permanent magnet synchronous wind turbine generator based on the improved Grey Wolf algorithm.

[0045] Figure 2 It is a single-phase current path diagram.

[0046] Figure 3 This is the voltage compensation schematic diagram.

[0047] Figure 4 It is the resistance identification curve under general working conditions.

[0048] Figure 5 This is the inductance identification curve under normal working conditions.

[0049] Figure 6 It is the permanent magnet flux identification curve under general working conditions.

[0050] Figure 7 It is the resistance identification curve under dynamic conditions.

[0051] Figure 8 It is the inductance identification curve under dynamic conditions.

[0052] Fig. 9 It is the permanent magnet flux identification curve under dynamic conditions.

[0053] in, Figures 4 to 9 In the figure, a represents the traditional grey wolf algorithm, b represents the grey wolf algorithm considering voltage compensation, c represents the chaotic reverse learning grey wolf algorithm without combining memory self-learning, and d represents the method proposed in the present invention. DETAILED DESCRIPTION

[0054] The present invention will be further explained and illustrated below in conjunction with the embodiments, but the specific embodiments do not limit the present invention in any form.

[0055] Example 1

[0056] This embodiment provides a method for identifying parameters of a permanent magnet synchronous wind turbine generator using an improved grey wolf algorithm. Figure 1 As shown, the steps include:

[0057] S1. Collection d =-2 and i d = 0 when the current, voltage and angular velocity are measured.

[0058] S2. Use Cat chaotic mapping and reverse learning strategy to generate the initial population. The Cat mapping expression is:

[0059]

[0060] Where a=b=1, k represents the number of times the chaotic sequence is generated.

[0061] Generate N initial solutions X from Cat chaotic map i , which produces the corresponding inverse solution for the initial solution: Among them, OP k Yes X k The corresponding reverse solution, r3 is a random number between [0, 1], and are the minimum and maximum values ​​of the initial solution of the dth dimension respectively.

[0062] Combine the initial solution and the reverse solution and arrange them in ascending order of fitness value, and select the top N optimal solutions as the initial population.

[0063] S3. Compensate voltage error;

[0064] The current flows through the following four paths. Taking the a-phase bridge arm as an example, the current flows through the following four paths: Figure 2 When the gate signal of the switch tube is turned on at different times, the voltage error caused by the tube voltage drop is different. The effect of the tube voltage drop on the output voltage is shown in Figure 3 As shown. Current i k The direction of is defined as the positive direction, when i k >0, upper arm G m The gate signal is 1, the switch tube S m If the current is conducting, the current path is Figure 1 W in the middle. G in the upper arm m The gate signal is 0, the switch tube S m Not conducting, diode D2 conducting, the current path is Figure 1 Z in. When i k <0, lower arm G m The gate signal is 1, the switch tube S n The current path is Figure 1 Y in the middle. Lower arm G n The gate signal is 0, the switch tube S n Not conducting, diode D1 conducting, the current path is Figure 1 The voltage error caused by the nonlinearity of the tube voltage drop is compensated, and the compensation is expressed as:

[0065]

[0066] Among them, V dc is the DC voltage, V CE is the on-state saturation voltage drop of the switch tube, V D is the freewheeling diode voltage drop;

[0067] S4. Calculate the nonlinear cosine convergence factor a, expressed as:

[0068]

[0069] Among them, t represents the current iteration number, t max is the maximum number of iterations, n is a constant greater than 1, a i and a f They represent the initial value and final value of the control parameter a. The initial value is set to 2, and the final value is set to 0.

[0070] Generate coefficient matrices A and C based on the obtained nonlinear cosine convergence factor a:

[0071]

[0072] in, are two random number vectors with values ​​in [0,1]. The value of a decreases linearly from 2 to 0 as the number of iterations increases.

[0073] S5. Calculate the distance and position of each wolf pack as follows:

[0074]

[0075] in, It indicates the position that the ω gray wolf individual needs to adjust in the end. They represent the step length and direction of approaching the α wolf pack, β wolf pack, and δ wolf pack, respectively, and are expressed as:

[0076]

[0077] in, They represent the distances of other ω gray wolf individuals from the α wolf pack, β wolf pack, and δ wolf pack, respectively, and are expressed as:

[0078]

[0079] Among them, α, β, and δ represent the optimal solution, the second-best solution, and the third-best solution, respectively. is the current position of the gray wolf, They represent the positions of the α wolf pack, β wolf pack, and δ wolf pack respectively.

[0080] S6. Combined with memory self-learning, according to Update the position of the gray wolf, where r4 is a random number between [0, 1], is the best position for the i-th gray wolf to pass through, k1 and k2 are constants between [0, 1], which are the group communication coefficient and individual memory coefficient respectively.

[0081] S7. According to Calculate the fitness value of each wolf pack, where c1, c2, c3, c4 are constants in the range (0, 1), i d 、i q is the stator current of the dq axis. The symbols represented by "0" or "1" in the subscripts indicate that the dq axis is injected into i d =-2 and i d =0, is the corresponding estimated value. If the fitness is less than the last iteration value, the values ​​of each wolf pack are updated; if the fitness is greater than the last iteration value, return to S4 to update the nonlinear cosine convergence factor a.

[0082] S8. After the iteration is completed, the optimal solution is output, which is the identification result of the resistance, inductance and permanent magnet flux.

[0083] Example 2

[0084] This embodiment provides a simulation experiment under general working conditions, and uses the RT-LAB hardware-in-the-loop simulation (HILS) experimental platform to verify the feasibility of the proposed method.

[0085] In this embodiment, the torque is set to 12N·m and the speed is set to 1000r / min. Under this working condition, four algorithms are used to identify the motor parameters: the traditional gray wolf algorithm, the gray wolf algorithm considering voltage compensation, the chaotic reverse learning gray wolf algorithm without memory self-learning, and the method proposed by the present invention.

[0086] Depend on Figure 4 As shown in the figure, the convergence time of the four algorithms is 0.2s, 0.15s, 0.1s, and 0.08s respectively. After considering the nonlinearity of the inverter tube voltage drop, the convergence speed of the Grey Wolf algorithm is effectively improved. Figure 5 As we know, the identification errors of the algorithms are 4.50%, 3.39%, 2.65% and 1.81% respectively. The algorithm for tube voltage drop compensation is implemented, and the identification errors are reduced. The method proposed in the present invention has the highest accuracy. Figure 6 It can be seen that compared with the traditional gray wolf algorithm, the gray wolf algorithm considering voltage compensation, the chaotic reverse learning gray wolf algorithm without memory self-learning, and the method proposed in the present invention have faster identification speed and reduced overshoot. The error of the traditional gray wolf algorithm is 4.192%, and the convergence time is 0.21s. The identification error of the method proposed in the present invention is only 1.198%, and the convergence time is only 0.09s.

[0087] Table 1

[0088]

[0089] From the experimental identification average values ​​under general working conditions shown in Table 1, it can be seen that the maximum error of the traditional gray wolf algorithm identification is 4.494%, while the maximum error of the algorithm taking into account the inverter nonlinear tube voltage drop compensation does not exceed 3.393%, and the identification accuracy of the gray wolf algorithm is improved. The maximum error of the method proposed in the present invention does not exceed 1.198%, and the convergence speed is the fastest. Therefore, after compensating for the nonlinearity of the inverter tube voltage drop, the accuracy of the algorithm is improved. After considering the individual experience and reverse learning of the gray wolf, the global search ability and local search ability of the algorithm are improved, and the identification accuracy and speed are further improved.

[0090] Example 3

[0091] This embodiment provides a simulation experiment under dynamic conditions with variable parameters, using four algorithms to identify motor parameters: the traditional gray wolf algorithm, the gray wolf algorithm considering voltage compensation, the chaotic reverse learning gray wolf algorithm without memory self-learning, and the method proposed in the present invention. Figure 7It can be seen that the resistance identification curve under dynamic conditions fluctuates. The identification error of the traditional gray wolf algorithm increased by 0.46% compared with the general working condition, and the gray wolf algorithm considering voltage compensation increased by 0.97%. The identification errors of the chaotic reverse learning gray wolf algorithm without memory self-learning and the method proposed in the present invention were reduced by 0.41% and 0.18% respectively. Therefore, the identification error of the method proposed in the present invention under dynamic conditions is 0.67%, and the anti-interference performance is good. Figure 8 It can be seen that compared with the general working conditions, the recognition time of the four methods under dynamic working conditions is affected differently, which are 0.25s, 0.15s, 0.1s, and 0.08s respectively. The traditional gray wolf algorithm and the gray wolf algorithm considering voltage compensation are extended by 0.1s and 0.03s respectively, while the chaotic reverse learning gray wolf algorithm without memory self-learning and the recognition time of the method proposed in the present invention are shortened by 0.02s. Therefore, the response speed of the method proposed in the present invention is better and is suitable for fields with high requirements for speed. Fig. 9 It can be seen that the identification error of the traditional gray wolf algorithm under dynamic conditions is 8.642%, the gray wolf algorithm considering voltage compensation is 6.790%, and the identification error of the chaotic reverse learning gray wolf algorithm without memory self-learning is 5.556%; the identification error of the method proposed in the present invention is 4.321%, and the identification accuracy under dynamic conditions is better.

[0092] Table 2

[0093]

[0094] From the experimental identification average values ​​under dynamic conditions shown in Table 2, it can be seen that the maximum error of the traditional gray wolf algorithm is 7.139%, while the maximum error of the algorithm taking into account the inverter nonlinear tube voltage drop compensation is 5.226%, and the maximum error of the chaotic reverse learning gray wolf algorithm without memory self-learning is 4.066%. The maximum error of the method proposed in the present invention does not exceed 2.861%, and the accuracy is significantly improved.

[0095] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A method for identifying parameters of a permanent magnet synchronous wind turbine generator using an improved grey wolf algorithm, characterized in that the steps include: S1. Collect electrical signals; S2. Generate the initial population using Cat chaotic mapping and reverse learning strategy; S3. Compensate for the voltage error caused by the nonlinearity of the tube voltage drop. The compensation is expressed as: Among them, V dc is the DC voltage, V CE is the on-state saturation voltage drop of the switch tube, V D is the freewheeling diode voltage drop; S4. Calculate the nonlinear cosine convergence factor a, and coefficient matrices A and C; S5. Calculate the distance and position of each wolf pack; S6. Combined with memory self-learning, according to Update the position of the gray wolf, where r4 is a random number between [0, 1], is the best position for the i-th gray wolf to pass through, k1 and k2 are constants between [0, 1], which are the group communication coefficient and individual memory coefficient respectively; S7. Calculate the fitness value of each wolf pack; S8. After the iteration is completed, the optimal solution is output, which is the identification result of the resistance, inductance and permanent magnet flux.

2. The method for identifying parameters of a permanent magnet synchronous wind turbine generator using an improved grey wolf algorithm according to claim 1 is characterized in that: The electrical signal includes i d =-2 and i d =0 when current, voltage and angular velocity.

3. The method for identifying parameters of a permanent magnet synchronous wind turbine generator using an improved grey wolf algorithm according to claim 1, characterized in that: The Cat mapping expression is: Where a=b=1, k represents the number of times the chaotic sequence is generated.

4. The method for identifying parameters of a permanent magnet synchronous wind turbine generator using an improved grey wolf algorithm according to claim 1, characterized in that: The cosine convergence factor a is expressed as: Among them, t represents the current iteration number, t max is the maximum number of iterations, n is a constant greater than 1, a i and a f They represent the initial value and final value of the control parameter a respectively.

5. The method for identifying parameters of a permanent magnet synchronous wind turbine generator using an improved grey wolf algorithm according to claim 1, characterized in that: The coefficient matrices A and C are expressed as: in, are two random number vectors with values ​​in [0,1]. The value of a decreases linearly from 2 to 0 as the number of iterations increases.

6. The method for identifying parameters of a permanent magnet synchronous wind turbine generator using an improved grey wolf algorithm according to claim 1, characterized in that: The distance and position of the wolf pack are expressed as: in, It indicates the position that the ω gray wolf individual needs to adjust in the end. They represent the step size and direction of approaching the α wolf pack, β wolf pack, and δ wolf pack respectively.

7. The method for identifying parameters of a permanent magnet synchronous wind turbine generator using an improved grey wolf algorithm according to claim 6 is characterized in that: Said It is expressed as: in, They represent the distances of other ω gray wolf individuals from the α wolf pack, β wolf pack, and δ wolf pack respectively.

8. The method for identifying parameters of a permanent magnet synchronous wind turbine generator using an improved grey wolf algorithm according to claim 7, characterized in that: Said It is expressed as: Among them, α, β, and δ represent the optimal solution, the second-best solution, and the third-best solution, respectively. is the current position of the Gray Wolf.

9. The method for identifying parameters of a permanent magnet synchronous wind turbine generator using an improved grey wolf algorithm according to claim 1, characterized in that: The fitness value calculation formula is: Where c1, c2, c3, c4 are constants in the range (0, 1), i d 、i q is the stator current of the dq axis. The symbols represented by "0" or "1" in the subscripts indicate that the dq axis is injected into i d =-2 and i d =0, are the corresponding estimated values.

10. The method for identifying parameters of a permanent magnet synchronous wind turbine generator using an improved grey wolf algorithm according to claim 1, characterized in that: If the fitness is less than the last iteration value, the values ​​of each wolf pack are updated; if the fitness is greater than the last iteration value, return to S4 to update the nonlinear cosine convergence factor a.

Images