Improved artificial rabbit optimization algorithm based on model predictive control of static variable frequency driving system

Abstract

The invention provides an improved artificial rabbit optimization algorithm based on model predictive control of a static variable frequency driving system. A model prediction control method is adopted for the LCI static variable frequency driving system, the main steps of an objective function, a cost function, rolling optimization and the like are mainly involved, solving of the cost function relates to a quadratic programming solving problem, and an optimization algorithm is especially key in the aspects of convergence speed and calculation speed, so that the problem that the algorithm is difficult to carry out is solved. The invention provides an improved artificial rabbit optimization algorithm suitable for a model prediction control method of a static variable frequency driving system, which has very important significance.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of synchronous machines, and relates to synchronous machine speed regulation control, in particular to an improved artificial rabbit optimization algorithm, and particularly to a method for model predictive control of a static frequency conversion driving system by using the improved artificial rabbit optimization algorithm. BACKGROUND

[0002] A load commutated inverter (LCI) static frequency conversion driving synchronous machine control system has the advantages of high control efficiency, strong reliability and low cost, and has become the preferred solution for large-power driving control of a hot continuous rolling mill unit, a reversible unit of a pumped storage power station, a blast system of an extra-large blast furnace and the like. In order to cope with special and complex application scenarios and higher control performance requirements, scholars at home and abroad have proposed a model predictive control (MPC) method. The model predictive control method adopts multi-objective optimization, and solves a quadratic programming (QP) problem of an optimal value. However, by using a traditional interior point method, a Newton method and a gradient descent method and the like, such general solution methods often depend on numerical iteration or need to solve complex equations, and have a large amount of calculation, which may cause insufficient calculation time and fail to meet the real-time requirement in engineering. Therefore, compared with the traditional numerical calculation method, the swarm intelligence algorithm has stronger adaptability and solving efficiency, and is more and more welcomed in challenging and high-demand engineering fields.

[0003] In the aspect of model predictive control of a synchronous machine, most scholars adopt a limited control set solving method, and solve an optimal voltage vector by generating a large number of virtual vectors according to current tracking errors and the same sector. The setting of the virtual voltage is based on the turn-on and turn-off of a fully controlled switching device, while the load commutated static frequency conversion system adopts a thyristor semi-controlled device, and the thyristor is turned on according to a certain time sequence. Therefore, the optimization algorithm using the limited set is not applicable to the load commutated static frequency conversion driving system.

[0004] In the aspect of model predictive control of a static frequency conversion system driving a synchronous machine, a cost function is constructed according to real-time collected state quantities, an optimal vector of the cost function is solved by using an optimization algorithm, the obtained first control vector is applied to the system, and the above steps are repeated at the arrival of the next moment for rolling optimization. The algorithm needs to be able to solve such a linear MPC problem by using embedded hardware within milliseconds or even microseconds. However, the traditional optimization algorithm has a long calculation period, which may cause a delay error of control. The optimization algorithm for solving the quadratic programming is particularly critical in terms of convergence speed and calculation speed. Therefore, it is of great significance to propose an improved artificial rabbit optimization algorithm suitable for the model predictive control method of the static frequency conversion driving system. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application adopts an improved artificial rabbit optimization algorithm (ARO) solving algorithm without additionally increasing the computing device, and comprises the following steps:

[0006] S1: constructing an LCI model prediction cost function: constructing an LCI system single-vector or double-vector model prediction control incremental constraint quadratic programming function to determine a cost function based on the LCI driving system;

[0007] S2: determining the control input increment dimension: determining the dimension of the control input increment Δu of the cost function as d dimensions;

[0008] S3: initializing the rabbit position and control parameters: initializing the rabbit position and control parameters, randomly initializing the population, setting the population size n, the maximum number of iterations T, the space dimension d, the search upper bound Up and the search lower bound Low according to the initialized rabbit population;

[0009] S4: calculating the energy factor A: updating the value of the current iteration number t, determining the value of the energy factor formula A and judging:

[0010] When A(t)>1, a detour foraging strategy is selected, the influence of the current optimal solution position is enhanced to accelerate the convergence speed of the algorithm, and a disturbance R is added in the foraging process to expand the search range, and the position update is used to represent the position at the next moment;

[0011] When A(t)≤1, a random hiding strategy is selected, and a random cave is generated when hiding randomly;

[0012] S5: calculating the precision to determine the position: calculating the fitness of the i-th rabbit at the t+1 iteration position, and comparing it with the fitness at the t iteration position; if the t+1 iteration fitness is better, the rabbit position is updated to replace the last fitness position; if the t+1 iteration fitness is worse, the position of the rabbit remains unchanged;

[0013] S6: judging whether the stopping condition is reached: if the maximum number of iterations is reached or the stopping requirement is met, the optimal control sequence of the artificial rabbit optimization algorithm is output, the first sequence Δu(k) is added to u(k-1) at the last moment to obtain u(k) as the control input at the current moment, and the system is acted on; if the stopping requirement is not met, the process starts from S4 again;

[0014] S7: updating the system state: when the next moment comes, the system state changes, the cost function is reconstructed, and the cost function is solved from S1 again to realize the rolling optimization of motor control.

[0015] Further, the cost function of the S1: LCI driving system is as follows:

[0016]

[0017] s.t.h i (x)=a i x-b i ≥0,i=1,2,…,m,

[0018] Wherein, f(x) is the cost function, Q is a symmetric matrix, c is a p×1 matrix, Δu is the control input increment, the feasible region D={x∈R n |h i (x)≥0,i=1,2,…,m} is the parameter of the feasible region inequality. The smaller the cost function value is, the better the control effect of the control vector is.

[0019] According to the prediction step p and the control step m set by the load commutation driving system model predictive control, the dimension of the cost function control input Δu is determined to be d, and the prediction model matrix is constructed according to the load commutation system.

[0020] According to the system state data collected, the model predictive control cost function expression is constructed as an incremental constraint quadratic programming problem, and the cost function of the LCI driving system is obtained.

[0021] The artificial rabbit optimization algorithm of swarm intelligence is used to solve the optimal value of the cost function. The rabbit position and control parameters are initialized, each rabbit has its own foraging area and d caves (d is the problem dimension), and the population is randomly initialized.

[0022] Further, the S4: an energy factor A is introduced, and the energy factor A expression is calculated as:

[0023]

[0024] Wherein, T is the maximum number of iterations; t is the current number of iterations; r is a random number between 0 and 1.

[0025] Further, the detour foraging strategy, the position information update expression of the detour foraging exploration process is:

[0026]

[0027] i,j=1,...n and j≠i,n1~N(0,1)R=L·c

[0028]

[0029] Wherein, is the original position vector of the i-th rabbit. is the original current best position vector; is the candidate position of the i-th rabbit at the t+1 iteration; is the position vector of any j-th rabbit other than the i-th rabbit; R is the disturbance added in the detour foraging process; T is the maximum number of iterations, t is the current iteration number; L is the step length of the rabbit movement at the current iteration number when the rabbit is detouring foraging; c is a mapping vector representing an integer between 0 and 1; n is the number of the rabbit group, and d is the problem dimension; r1 and r2 are random numbers between 0 and 1; and n1 is a standard normal distribution.

[0030] Further, the random hiding strategy, the random hiding development process position information updating expression is:

[0031]

[0032] wherein, indicates that the i-th rabbit randomly selects a cave; r4 and r5 are also random numbers in the range of (0, 1), and H is a hidden parameter, the value of which is related to the iteration number; the size of the position range of the cave is linearly related to the hidden parameter and related to the iteration number; when the iteration number increases, the hidden parameter gradually decreases, that is, the range of the cave generation is reduced; in order to avoid being preyed upon by natural enemies, the rabbit will generate multiple caves in its own field and select one of them to hide.

[0033] Further, the S5: calculation precision fitness comparison expression is:

[0034]

[0035] wherein, x is the position at the iteration number, t indicates the iteration number, and p is the candidate position of the i-th rabbit at the iteration.

[0036] The characteristics of the present application are:

[0037] 1. The present application is based on a load commutation static frequency conversion system model predictive control method, which does not additionally increase the calculation device, adopts an improved artificial rabbit optimization algorithm to solve the motor drive control algorithm, and the optimal control vector of the motor can be obtained through the algorithm, so as to reduce the speed fluctuation caused by load disturbance, make it quickly return to the predetermined speed, reduce the regulation time, improve the dynamic response performance of the system, and reduce the calculation cost.

[0038] 2. The improved artificial rabbit optimization algorithm of the present application adopts an enhanced current optimal solution position, accelerates the convergence speed of the algorithm, has a fast solving speed, greatly improves the speed of solving the optimal vector, and improves the dynamic response ability, control precision and calculation efficiency of the system under load disturbance.

[0039] 3. The improved artificial rabbit optimization algorithm is used for exploration and development in an energy convergence balance mode, the exploration step is shortened with time, long step is beneficial for global exploration, and short step is beneficial for development, strategy conversion can be well performed, the problem of falling into local optimum is avoided, and the accuracy of solving optimal solution is improved.

[0040] 4. The improved artificial rabbit optimization algorithm can be used for model predictive control of load commutation static frequency conversion driving system, and compared with other methods, the dynamic performance of single vector model predictive control and double vector model predictive control is improved. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiment or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0042] Figure 1 It is a single vector model predictive control block diagram based on LCI system;

[0043] Figure 2 It is a double vector model predictive control block diagram based on LCI system;

[0044] Figure 3 It is a flowchart of solving cost function by improved artificial rabbit optimization algorithm;

[0045] Figure 4 It is a contraction function graph of energy factor A;

[0046] Figure 5 It is a comparison diagram of single vector artificial rabbit optimization algorithm based on LCI system and interior point method;

[0047] Figure 6 It is a speed waveform diagram under 33% step load disturbance of single vector MPC;

[0048] Figure 7 It is a speed waveform diagram under 66% step load disturbance of single vector MPC;

[0049] Figure 8 It is a speed waveform diagram under 100% step load disturbance of single vector MPC;

[0050] Figure 9 It is a comparison diagram of double vector artificial rabbit optimization algorithm based on LCI system and interior point method;

[0051] Figure 10The speed waveform diagram under the double-vector MPC 33% step load disturbance;

[0052] Figure 11 The speed waveform diagram under the double-vector MPC 66% step load disturbance;

[0053] Figure 12 The speed waveform diagram under the double-vector MPC 100% step load disturbance;

[0054] Figure 13 The comparison diagram of the convergence speed of the interior point method and the ARO optimization algorithm;

[0055] Figure 14 The comparison diagram of the calculation time of the interior point method and the ARO optimization algorithm. DETAILED DESCRIPTION

[0056] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the present application.

[0057] Embodiment:

[0058] For the single-vector and double-vector models based on the LCI system, wherein, Figure 1 The prediction control block diagram of the single-vector model based on the LCI system; Figure 2 The prediction control block diagram of the double-vector model based on the LCI system, which adopts an improved artificial rabbit optimization algorithm to solve the algorithm, and the flowchart of solving the cost function of the improved artificial rabbit optimization algorithm is shown in Figure 3 , and includes the following steps:

[0059] S1: Constructing the LCI model prediction cost function: constructing the LCI system single-vector or double-vector model prediction control incremental constraint quadratic programming function, and determining the cost function based on the LCI driving system;

[0060] The cost function of the LCI driving system is as follows:

[0061]

[0062] s.t.h i (x)=a i x-b i ≥0,i=1,2,…,m,

[0063] Wherein, f(x) is the cost function, Q is a symmetric matrix, c is a p×1 matrix, Δu is the control input increment, the feasible region D={x∈R n |hi (x)≥0, i=1,2,…,m}, where a and b are the parameters of the feasible region inequality. The smaller the cost function value, the better the control effect of the control vector.

[0064] S2: Determine the dimension of the control input increment: Determine that the dimension of the control input increment Δu of the cost function is d-dimensional; Based on the prediction step size p and control step size m of the load-switching drive system model predictive control settings, determine that the dimension of the control input quantity Δu of the cost function is d-dimensional, and construct the prediction model matrix based on the load-switching system.

[0065] S3: Initialize rabbit positions and control parameters: Initialize rabbit positions and control parameters, randomly initialize the population, and based on the initialized rabbit population, set the population size n, maximum number of iterations T, spatial dimension d, upper search bound Up, and lower search bound Low;

[0066] S4: Calculate the energy factor A: Update the value of the current iteration number t, determine the energy factor formula, calculate the value of A, and then judge:

[0067] The expression for calculating the energy factor A is:

[0068]

[0069] Where T is the maximum number of iterations; t is the current number of iterations; and r is a random number between 0 and 1. Figure 4 The graph shows the contraction function of energy factor A.

[0070] When A(t)>1, a detour foraging strategy is chosen. By enhancing the influence of the current optimal solution position, the convergence speed of the algorithm is accelerated. A perturbation R is added during the foraging process to expand the search range. Its position update is used to represent the position at the next moment.

[0071] The detour foraging strategy, the location information update expression during the detour foraging exploration process is:

[0072]

[0073] i,j=1,...n and j≠i,n1~N(0,1)R=L·c

[0074]

[0075] in, Let be the original position vector of the i-th rabbit; This is the original current optimal position vector; It is the candidate position of the i-th rabbit at iteration t+1; is the position vector of any jth rabbit except the ith rabbit; R is the disturbance added in the process of detour foraging; T is the maximum number of iterations, t is the current iteration number; L is the step length of the rabbit at the current iteration number when the rabbit detours for food; c is a mapping vector, which represents an integer between 0 and 1; n is the number of the rabbit group, d is the problem dimension; r1 and r2 are random numbers between 0 and 1; n1 is a standard normal distribution.

[0076] When A(t)≤1, a random hiding strategy is selected, and the development stage is in a random hiding state, and a random cave is generated when hiding randomly;

[0077] The random hiding strategy, the position information update expression of the random hiding development process is:

[0078]

[0079] wherein, indicates the cave randomly selected by the ith rabbit; r4 and r5 are also random numbers in the range of (0, 1), and H is a hidden parameter, the value of which is related to the iteration number; the size of the position range of the cave is linearly related to the hidden parameter, and is related to the iteration number; when the iteration number increases, the hidden parameter gradually decreases, that is, the range of the cave generated is reduced; in order to avoid being preyed upon by natural enemies, the rabbit will generate multiple caves in its own territory and select one of them to hide.

[0080] S5: Calculate the accuracy of the position: calculate the fitness of the position of the ith rabbit at the t+1th iteration, and compare it with the fitness of the position at the tth iteration;

[0081] The calculation accuracy fitness comparison expression is:

[0082]

[0083] wherein, x is the position at the iteration number, t indicates the iteration number, and p is the candidate position of the ith rabbit at the iteration.

[0084] If the t+1th fitness is better, the position of the rabbit is updated to replace the last fitness position;

[0085] If the t+1th fitness is worse, the position of the rabbit remains unchanged;

[0086] S6: Determine whether the stopping condition is reached: if the maximum iteration number is reached or the stopping requirement is met, output the optimal control sequence of the artificial rabbit optimization algorithm, add the first sequence Δu(k) to u(k-1) at the last moment to obtain u(k) as the control input at the current moment, and act on the system; if the stopping requirement is not met, start again from S4;

[0087] S7: updating system state: when the next time comes, the system state changes, the cost function is reconstructed, and the cost function is solved again from S1 to realize the rolling optimization of motor control.

[0088] For the single vector model based on the LCI system, the improved artificial rabbit optimization algorithm of the application is compared with the interior point method:

[0089] Figure 5 The figure is a comparison of the single vector artificial rabbit optimization algorithm and the interior point method based on the LCI system. The upper speed graph is a comparison of the speed using the artificial rabbit algorithm and the interior point method, and the lower part is a load test using 33%, 66%, and 100% rated load waveforms.

[0090] Figure 6 The figure is a speed waveform graph under the step load disturbance of the single vector MPC 33%. The upper part of the graph is a comparison of the speed waveform using the artificial rabbit algorithm (ARO) and the interior point method, and the lower part indicates that the rated load of 33% is used in the load test.

[0091] Figure 7 The figure is a speed waveform graph under the step load disturbance of the single vector MPC 66%. The upper part of the graph is a comparison of the speed waveform using the artificial rabbit algorithm (ARO) and the interior point method, and the lower part indicates that the rated load of 66% is used in the load test.

[0092] Figure 8 The figure is a speed waveform graph under the step load disturbance of the single vector MPC 100%. The upper part of the graph is a comparison of the speed waveform using the artificial rabbit algorithm (ARO) and the interior point method, and the lower part indicates that the rated load of 100% is used in the load test.

[0093] Comparing the ARO algorithm and the interior point method algorithm, from the single vector MPC simulation waveform comparison, it can be seen that when the system is tested with 33% rated load, the maximum drop speed under load is reduced by 28%, and the maximum rise speed under unloading is reduced by 26%; when the system is tested with 66% rated load, the maximum drop speed under load is reduced by 30%, and the maximum rise speed under unloading is reduced by 22%; when the system is tested with 100% rated load, the maximum drop speed under load is reduced by 12%, and the maximum rise speed under unloading is reduced by 22%. The adjustment time and overshoot amplitude of the ARO algorithm are slightly better than those of the interior point method, and the overall dynamic response effect is better.

[0094] For the double vector model based on the LCI system, the improved artificial rabbit optimization algorithm of the application is compared with the interior point method:

[0095] Figure 9The figure is the comparison of double vector artificial rabbit optimization algorithm based on LCI system and interior point method. The upper speed diagram is the comparison of speed of artificial rabbit optimization algorithm (ARO) and interior point method, and the lower is the load test with 33%, 66%, 100% rated load waveform.

[0096] Figure 10 The figure is the speed waveform diagram of double vector MPC under 33% step load disturbance. The upper waveform is the comparison of speed of artificial rabbit optimization algorithm (ARO) and interior point method, and the lower is the load test with 33% rated load.

[0097] Figure 11 The figure is the speed waveform diagram of double vector MPC under 66% step load disturbance. The upper waveform is the comparison of speed of artificial rabbit optimization algorithm (ARO) and interior point method, and the lower is the load test with 66% rated load.

[0098] Figure 12 The figure is the speed waveform diagram of double vector MPC under 100% step load disturbance. The upper waveform is the comparison of speed of artificial rabbit optimization algorithm (ARO) and interior point method, and the lower is the load test with 100% rated load.

[0099] The figure is the comparison of double vector MPC waveform. ARO algorithm is better than interior point method. When the system is tested with 33% rated load, the maximum drop speed under load is reduced by 18%, the load adjustment time is reduced by 16%; the maximum rise speed under unloading is reduced by 22%, and the unloading time is reduced by 9%. When the system is tested with 66% rated load, the maximum drop speed under load is reduced by 25%, the load adjustment time is reduced by 10%; the maximum rise speed under unloading is reduced by 11%, and the unloading time is reduced by 20%. When the system is tested with 100% rated load, the maximum drop speed under load is reduced by 5%, the load adjustment time is reduced by 13%; the maximum rise speed under unloading is reduced by 10%, and the unloading time is reduced by 7%.

[0100] Figure 13 The figure is the comparison of convergence speed of interior point method and ARO optimization algorithm. The upper waveform represents the convergence speed of interior point method, and the lower waveform represents the convergence speed of artificial rabbit optimization algorithm. Figure 14 The figure is the comparison of calculation time of interior point method and ARO optimization algorithm. The upper waveform represents the calculation time of interior point method, and the lower waveform represents the calculation time of artificial rabbit optimization algorithm. It can be seen from the figure that ARO algorithm has better performance in convergence speed and solution speed compared with traditional algorithm. In LCI static frequency conversion system model predictive control method, ARO optimization algorithm has better dynamic performance in motor control, with smaller speed fluctuation and shorter adjustment time.

[0101] It will be appreciated by those skilled in the art that features of the various embodiments and / or claims of the present application can be combined or / and integrated, even if such combinations or integrations are not expressly disclosed in the present application. In particular, features of the various embodiments and / or claims of the present application can be combined or / and integrated in any number of ways, without departing from the spirit and scope of the present application. All such combinations and / or integrations are within the scope of the present application.

Claims

1. An improved artificial rabbit optimization algorithm based on model predictive control of a static variable frequency drive system, characterized in that, Includes the following steps: S1: Construct the LCI model prediction cost function: Construct the incremental constrained quadratic programming function for predictive control of the LCI system single-vector or double-vector model, and determine the cost function based on the LCI-driven system; S2: Determine the dimension of the control input increment: Determine the dimension of the control input increment Δu of the cost function to be d-dimensional; S3: Initialize rabbit positions and control parameters: Initialize rabbit positions and control parameters, randomly initialize the population, and based on the initialized rabbit population, set the population size n, maximum number of iterations T, spatial dimension d, upper search bound Up, and lower search bound Low; S4: Calculate the energy factor A: Update the value of the current iteration number t, determine the energy factor formula, calculate the value of A, and then judge: When A(t)>1, a detour foraging strategy is chosen. By enhancing the influence of the current optimal solution position, the convergence speed of the algorithm is accelerated. A perturbation R is added during the foraging process to expand the search range. Its position update is used to represent the position at the next moment. When A(t)≤1, the random hiding strategy is selected, which is in the development stage. Random caves will be generated when hiding randomly. S5: Calculate the fitness of the position of the i-th rabbit at iteration t+1 and compare it with the fitness of the position at iteration t: If the fitness is better in time t+1, then update the rabbit's position and replace the previous fitness position. If the fitness is worse in time t+1, the rabbit's position remains unchanged; S6: Determine if the stopping condition is met: If the maximum number of iterations is reached or the stopping requirement is met, output the optimal control sequence of the artificial rabbit optimization algorithm, add the first sequence Δu(k) to u(k-1) of the previous time step to obtain u(k) as the control input of the current time step, and apply it to the system; if the stopping requirement is not met, start again from S4. S7: Update system state: When the next moment arrives, the system state changes, the cost function is reconstructed, and the cost function is solved again starting from S1 to achieve rolling optimization of motor control.

2. The improved artificial rabbit optimization algorithm based on model predictive control of a static variable frequency drive system according to claim 1, characterized in that: The cost function of the S1: LCI driving system is as follows: s.t.h i (x)=a i x-b i ≥0,i=1,2,…,m, Where f(x) is the cost function, Q is a symmetric matrix, c is a p×1 dimensional matrix, Δu is the control input increment, and the feasible region D = {x∈R} n |h i (x)≥0,i=1,2,…,m}, where a and b are the parameters of the feasible region inequality.

3. The improved artificial rabbit optimization algorithm based on model predictive control of a static variable frequency drive system according to claim 1, characterized in that: The expression for calculating energy factor A in S4 is: Where T is the maximum number of iterations; t is the current number of iterations; and r is a random number between 0 and 1.

4. The improved artificial rabbit optimization algorithm based on model predictive control of a static variable frequency drive system according to claim 1, characterized in that: The detour foraging strategy in S4, the location information update expression during the detour foraging exploration process is: i,j=1,...n and j≠i,n1~N(0,1) R = L·c in, Let be the original position vector of the i-th rabbit; This is the original current optimal position vector; It is the candidate position of the i-th rabbit at iteration t+1; is the position vector of any j-th rabbit other than the i-th rabbit; R is the perturbation added during the detour foraging process; T is the maximum number of iterations, t is the current number of iterations; L is the step size of the rabbit at the current number of iterations when the rabbit detours foraging; c is a mapping vector representing integers between {0, 1}; n is the number of rabbits in the group, d is the problem dimension; r1 and r2 are random numbers between 0 and 1 respectively; n1 is a standard normal distribution.

5. The improved artificial rabbit optimization algorithm based on model predictive control of a static variable frequency drive system according to claim 1, characterized in that: The random hiding strategy in S4, the location information update expression during the random hiding development process is: in, R represents the burrow randomly selected by the i-th rabbit; r4 and r5 are also random numbers in the range (0,1); H is the hidden parameter, the value of which is related to the number of iterations; the size of the location range where the burrow is generated is linearly related to the hidden parameter and the number of iterations; as the number of iterations increases, the hidden parameter gradually decreases, that is, the range where the burrow is generated shrinks; in order to avoid being preyed upon by predators, rabbits will generate multiple burrows in their own territory and choose one of them to hide in.

6. The improved artificial rabbit optimization algorithm based on model predictive control of a static variable frequency drive system according to claim 1, characterized in that: The expression for comparing the accuracy fitness of calculation in S5 is as follows: Where x represents the position at the iteration number, t represents the iteration number, and p represents the candidate position of the i-th rabbit at iteration number.

Images