Non-cascade permanent magnet synchronous motor model prediction torque control method based on improved wild horse algorithm

Abstract

The invention discloses a non-cascade permanent magnet synchronous motor model prediction torque control method based on an improved wild horse algorithm, and the method comprises the steps: building a stator voltage and current equation and an electromagnetic torque equation under a synchronous rotating coordinate system, constructing an MPC controller and an MTPA controller, and obtaining a current and torque prediction equation through discretization; a dynamic weight coefficient is introduced to optimize a non-cascade MPC-MTPA controller; and constructing a fitness function, carrying out preferential selection on the weight coefficient by using an improved wild horse algorithm, avoiding a local optimal condition according to a population hybridization behavior, outputting an optimal weight coefficient after a set termination condition is met, and determining a cost function of the non-cascaded MPC-MTPA controller subjected to final preferential selection so as to improve a motor control effect. Compared with the prior art, the method has the advantages that better target torque tracking, MTPA control and current constraint satisfaction effects are achieved, the motor operation is more stable and efficient, and the dynamic response characteristic is better.

Description

Technical Field

[0001] The invention relates to the field of motor control, and in particular to a method for predicting torque control of a non-cascaded permanent magnet synchronous motor model based on an improved Mustang algorithm. Background Art

[0002] Permanent magnet synchronous motors have the advantages of high efficiency, high power density, high torque density, and good dynamic response. They are widely used in industrial automation and renewable energy fields. Especially in electric vehicles, it has become a common type of drive motor due to its low energy consumption, high acceleration performance, and ability to provide longer distances. The control system of permanent magnet synchronous motors needs to have fast and accurate speed control capabilities and good dynamic response characteristics. Electric vehicles mainly use vector control combined with PI controllers to control motor speed and current. This method controls the three-phase AC motor equivalent to a DC motor through vector transformation, and the PI controller is used to adjust the deviation of speed and current to achieve motor control. The parameter setting process of vector control combined with PI controller is cumbersome and needs to be adjusted according to the specific parameters and operating conditions of the motor, which increases the complexity and difficulty of control. When faced with changes in operating conditions such as load mutations, the dynamic response speed of this control method is slow, resulting in poor anti-interference ability and low working efficiency of the motor.

[0003] Model predictive torque control is an advanced motor control strategy. It predicts the future system behavior based on the mathematical model of the motor and selects the optimal control input by optimizing the cost function, thereby achieving precise control of the motor torque. The cost function is at the core of MTPA. It comprehensively considers multiple control objectives, such as torque tracking error, current limit, switching frequency limit, etc. By adjusting the weights in the cost function, the priority between different control objectives can be balanced, thereby optimizing the control performance of the motor. The traditional method of determining the weight of the cost function is mainly based on experience or trial and error. This method has many disadvantages, lacks systematicity, relies on manual experience, lacks a systematic theoretical basis and optimization process, and is difficult to fully consider the impact of various factors on control performance. Local optimal problems may occur, and it is easy to fall into the local optimal solution. It is impossible to guarantee that the global optimal weight combination is found, which may lead to poor motor control performance, such as large torque fluctuations, current overload, and slow system response.

[0004] The wild horse algorithm is a meta-heuristic optimization algorithm inspired by the social life behavior of wild horses. It simulates the social life behavior of wild horses, and realizes information exchange between individuals and population evolution by simulating the grazing, mating, and movement behaviors of wild horse groups. It has a strong bionics background and intuitiveness. The multi-population parallel search and the random fluctuation mechanism of individuals enable it to explore extensively in the search space, increase the possibility of finding the global optimal solution, and avoid falling into the local optimum. Compared with some other complex algorithms, it requires fewer parameters to be set, which reduces the difficulty of algorithm parameter adjustment and is convenient for practical application. For permanent magnet synchronous motors running at high speed, its control effect is better. Summary of the invention

[0005] Purpose of the invention: In response to the problems in the background technology, the present invention discloses a method for predicting torque control of a non-cascaded permanent magnet synchronous motor model based on the Mustang algorithm, which achieves better target torque tracking, MTPA control and current constraint satisfaction effects, making the motor run more stable and efficient, and having better dynamic response characteristics.

[0006] Technical solution: The present invention discloses a method for predicting torque control of a non-cascaded permanent magnet synchronous motor model based on an improved Mustang algorithm, comprising the following steps:

[0007] Step 1: Establish the stator voltage and current equation and the electromagnetic torque equation in the synchronous rotating coordinate system, construct the model predictive control MPC controller and the maximum torque current ratio control MTPA controller in the permanent magnet synchronous motor model predictive torque control, and obtain the current and torque prediction equations through discretization;

[0008] Step 2: Introduce the dynamically allocated weight coefficient to optimize the non-cascade MPC-MTPA controller and determine the cost function of the final non-cascade MPC-MTPA controller;

[0009] Step 3: Construct the fitness function, use the improved Mustang algorithm to optimize the weight coefficients, and determine the cost function of the non-cascaded MPC-MTPA controller after the final optimization.

[0010] Furthermore, the stator voltage and current equation and the electromagnetic torque equation established in step 1 are:

[0011] 1) The stator voltage and current equation is:

[0012]

[0013] 2) The electromagnetic torque equation is: T e =1.5p n [ψ f +(L d -L q )i d ]i q ;

[0014] Among them, i d 、i q is the d-axis and q-axis components of the actual current, R s is the motor phase resistance, L d , L q is the d and q axis inductance, ω e is the electrical angular velocity, ψ f is the permanent magnet flux, u d 、u q is the d and q axis voltage, p n is the pole pair number;

[0015] Furthermore, the model predictive control MPC is specifically as follows:

[0016] Model predictive control MPC uses the forward Euler method to discretize the voltage and current state equations in the dq coordinate system to obtain the prediction equations for the d-axis and q-axis currents:

[0017]

[0018] Among them, i d (k+1) is the predicted value of the d-axis current at time k+1, T s is the system control period, is the feedback d-axis current value at time k, i q (k+1) is the predicted value of the q-axis current at time k+1, is the feedback q-axis current value at time k, is the d-axis voltage value fed back at time k, is the feedback q-axis voltage value at time k.

[0019] Furthermore, the maximum torque current ratio controls the MTPA controller to solve the minimum stator current i when the electromagnetic torque is the same. d 、i q The combination is as follows:

[0020] The stator current is expressed in polar coordinate form:

[0021]

[0022] in, represents the amplitude of the stator current, β is the angle between the current vector and the α axis; substituting the current into the electromagnetic torque equation, we get:

[0023]

[0024] when i s When the torque T e is a function of the current angle β, for T eFind the partial derivative about the current angle β:

[0025]

[0026] Let the partial derivative equal to 0 and simplify to get:

[0027]

[0028] Using the double angle formula cos2β=2cos 2 Beta-1:

[0029] 2(L d -L q )i s cos 2 β+ψ f cosβ-(L d -L q )i s =0

[0030] Solve cosβ according to the quadratic equation root formula:

[0031]

[0032] Since the value range of β is usually [0, π],

[0033]

[0034] Will Substituting into the function of MTPA is:

[0035]

[0036] Among them, sign is the sign function,

[0037] Furthermore, the step 2 optimizes the non-cascaded MPC-MTPA controller, and uses the cost function in the MPC to simultaneously complete the torque control and the MTPA control, specifically:

[0038] The cost function g of the current control scheme of model predictive control MPC imax Considering the tracking of torque characterizing current, optimizing the ratio of torque to current and limiting the current amplitude, the cost function g imax for:

[0039]

[0040] in, i max is the maximum input current;

[0041] Determine the torque tracking cost function:

[0042]

[0043] Among them, g Te is the torque tracking cost function, is the torque target value, is the predicted value of electromagnetic torque at time k+1;

[0044] According to the MTPA control constraints on dq axis current:

[0045]

[0046] Cost function g of the maximum torque current ratio control MTPA controller MTPA :

[0047]

[0048] Considering the safety factor to limit the maximum current value, the constraint item of the maximum current value is set to g lim_MTPA :

[0049]

[0050] in, is the predicted value of the d-axis current at time k+1;

[0051] The formula for determining the final cost function g of the non-cascaded MPC-MTPA controller is:

[0052] g=λ1g Te +λ2g MTPA +λ3(g imax +g lim_MTPA )

[0053] Among them, λ1, λ2, λ3 represent weight coefficients, g imax represents the cost function of model predictive control MPC, g MTPA represents the cost function of the maximum torque current ratio control MTPA controller, g lim_MTPA A constraint term representing the maximum current value.

[0054] Furthermore, the fitness function constructed in step 3 is:

[0055]

[0056] Among them, the integral of the torque error is selected As an indicator of torque tracking accuracy, T e is the actual torque, T e，ref For the reference torque, select the integral of the deviation between the current and the MTPA curve As an indicator for evaluating the control effect of MTPA, i d 、i q is the d-axis and q-axis components of the actual current, i d，MTPA 、i q,MTPA are the d-axis and q-axis components of the ideal current obtained according to the MTPA control principle; the time t when the current exceeds the constraint range is selected over As a measure of whether the current constraint is satisfied.

[0057] Furthermore, the improved Mustang algorithm optimizes the weight coefficients as follows:

[0058] Step 3.1: Set the population size to N = 30 and the stallion individual proportion P s =0.2, population hybridization rate P c =0.1;

[0059] Step 3.2: According to the population size N and stallion individual proportion P s , calculate the number of stallions G = N × P s , and the number of remaining individuals in the population, and then randomly divide the population into G groups, each group is led by a stallion, the initial position of the stallion is randomly selected within the value range of the decision variable, and the initial positions of the remaining individuals are also randomly initialized within the corresponding range. The decision variables are weight coefficients λ1, λ2, λ3;

[0060] Step 3.3: For each individual in the population, a set of weight coefficient combinations Substitute it into the cost function of the MPC-MTPA controller and calculate the corresponding fitness value in combination with the actual operation model of the motor. The specific calculation process is as follows: according to the current state and control requirements of the motor, the predicted torque and current parameters are calculated using the cost function, and then the individual fitness value Fitness is calculated according to the definition of the fitness function.

[0061] Step 3.4: The herding behavior of the group is led by the stallion to move the entire group, and the remaining individuals in the group perform grazing search with the stallion as the center. The position of the remaining individuals is updated as follows:

[0062]

[0063] in, is the current position of the remaining individuals, are the updated positions of the remaining individuals, is the current position of the stallion, R is a random number in [-2,2], and Z is an adaptive parameter, which is calculated as follows:

[0064]

[0065] IDX=(P=0)

[0066] P=R1<T tdr

[0067] Among them, R1 and R3 are random variables uniformly distributed in [0,1], T tdr is an adaptive parameter that starts at 1 and gradually decreases as the algorithm runs, and becomes 0 when the algorithm ends: T is the current iteration number, T max is the maximum number of iterations;

[0068] Step 3.5: Population hybridization behavior means that when the foals in the population mature, they will leave the current group and mate with new foals in other groups to produce new individuals. The position information of the hybrid individuals is updated as follows:

[0069]

[0070] Crossover=Mean

[0071] in, is the position of the new individual p in the population k after mating, is the position information of the two individuals in the parent generation, Crossover = Mean means the mean crossover method is adopted;

[0072] Step 3.6: Population movement behavior means that the stallion individual, as the leader of the population, will lead the population to find a better habitat, and the other individuals will move with the stallion individual; the position update method of the stallion individual is shown in the formula:

[0073]

[0074] Among them, W H is the location of the “water hole” in the habitat, that is, the location corresponding to the currently found optimal weight coefficient combination, N SGi is the current position of the stallions in the herd, is the updated position of the stallion in the group, R3 is a random number in [0,1], which determines the clockwise or counterclockwise direction of the group's advance, Q1 and Q2 are random numbers in [-1,1], and ω is the dynamic inertia weight coefficient, which is calculated as follows:

[0075]

[0076] Among them, f i (t) is the fitness value of the stallion individual at the current iteration number, f min (t) is the minimum fitness value of all individuals in the population in the current iteration number, f avg (t) is the average fitness of all stallion individuals in the current iteration, ωmax is the upper limit of the dynamic weight, ω min is the lower limit of the dynamic weight; the population moves towards the optimal solution and continuously optimizes the combination of weight coefficients;

[0077] Step 3.7: The competition behavior within the population refers to the random selection of stallion particles in the initialization stage of the algorithm. As the algorithm runs, stallion individuals are selected according to their fitness information, and the individuals with the smallest fitness are selected; the competition behavior within the population is shown in the formula:

[0078]

[0079] Among them, cost is the fitness function, x G,i For the position of group members, through internal competition, it is ensured that the stallion individuals are always the individuals with better fitness in the population, leading the population to search in a better direction;

[0080] Step 3.8: Set the termination condition. If the termination condition is met, the iteration will be terminated. After the iteration is completed, the combination with the smallest fitness function value is selected from all the recorded optimal weight coefficient combinations. As the final optimal solution, the optimal weight coefficient combination is applied to the cost function of the MPC-MTPA controller.

[0081] Beneficial effects:

[0082] 1. The present invention optimizes the weight of the cost function in the prediction torque control of the permanent magnet synchronous motor model based on the Mustang algorithm, and has many effects. The motor control performance can be improved by improving the torque control accuracy, stabilizing and accurately outputting the torque, and optimizing the current control. The stator current harmonic content is reduced, the power conversion efficiency is improved, and the energy loss is reduced, which can improve the motor operation efficiency. Because the torque, current and other parameters of the motor are better controlled, unnecessary energy loss is reduced, and the electric energy can be more efficiently converted into mechanical energy, thereby improving the overall operation efficiency of the motor. In application scenarios such as electric vehicles, the cruising range can be extended.

[0083] 2. Advantages of the cost function designed in the present invention: Non-cascade control uses the cost function in MPC to simultaneously complete torque control and MTPA control, simplifying the control algorithm structure. It reduces the occupation of computing resources and the complexity of the algorithm, improves the stability and dynamic response characteristics of the control algorithm, and makes the system easier to implement and maintain.

[0084] 3. Advantages of the fitness function of the present invention: The fitness function constructed comprehensively considers the torque error integral, the current and MTPA curve deviation integral, and the time when the current exceeds the constraint range, and comprehensively measures the influence of the weight coefficient combination on the motor control performance. It can accurately guide the algorithm to search for the optimal weight coefficient, improve the optimization effect, and ensure that the motor achieves balance and optimization in many aspects of performance.

[0085] 4. Advantages of the algorithm of the present invention: The improved Mustang optimization algorithm has a strong bionics background and intuitiveness. The multi-population parallel search and individual random fluctuation mechanism increase the possibility of finding the global optimal solution and avoid falling into the local optimum. The parameter setting is relatively simple, which reduces the difficulty of algorithm parameter adjustment, facilitates application in actual engineering, and improves the practicality and effectiveness of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 Design FFT graph analysis for traditional weight coefficients (weight coefficient 32);

[0087] Figure 2 Design FFT graph analysis for traditional weight coefficients (weight coefficient 80);

[0088] Figure 3 Design FFT graph analysis for new weight coefficients;

[0089] Figure 4 This is the change of the overall simulation waveform of the control under the traditional weight design;

[0090] Figure 5 This is the change of the overall simulation waveform of the control under the new weight design. DETAILED DESCRIPTION

[0091] The following examples may enable those skilled in the art to more fully understand the present invention, but are not intended to limit the present invention in any way.

[0092] The present invention discloses a method for forming a cost function weight optimization in a non-cascaded permanent magnet synchronous motor model predictive torque control based on a Mustang algorithm, comprising the following steps:

[0093] Step 1: Establish the stator voltage and current equation and the electromagnetic torque equation in the synchronous rotating coordinate system, construct the model predictive control MPC controller and the maximum torque current ratio control MTPA controller in the permanent magnet synchronous motor model predictive torque control, and obtain the current and torque prediction equations through discretization.

[0094] Establish a mathematical model and establish the stator voltage and current equation and electromagnetic torque equation in the synchronous rotating coordinate system (dq coordinate system). The stator voltage and current equation is:

[0095]

[0096] The electromagnetic torque equation is: T e =1.5p n [ψ f +(L d -L q )i d ]i q , where i d 、i q is the d-axis and q-axis components of the actual current, R s is the motor phase resistance, L d , L q is the d and q axis inductance, ω e is the electrical angular velocity, ψ f is the permanent magnet flux, u d 、u q is the d and q axis voltage, p n is the pole pair number.

[0097] Model predictive control MPC uses the forward Euler method to discretize the voltage and current state equations in the dq coordinate system to obtain the prediction equations for the d-axis and q-axis currents:

[0098]

[0099] Among them, i d (k+1) is the predicted value of the d-axis current at time k+1, T s is the system control period, is the feedback d-axis current value at time k. q (k+1) is the predicted value of the q-axis current at time k+1, is the feedback q-axis current value at time k, is the d-axis voltage value fed back at time k, is the feedback q-axis voltage value at time k.

[0100] The cost function of the model predictive current control scheme needs to consider tracking the torque-characterizing current, optimizing the ratio of torque to current, and limiting the current amplitude. The cost function g imax for:

[0101]

[0102] in,

[0103] MTPA control is to find the minimum stator current i when the electromagnetic torque is the same d 、i q In the synchronous rotating coordinate system (dq coordinate system), the electromagnetic torque equation of the permanent magnet synchronous motor is:

[0104] Te =1.5p n [ψ f +(L d -L q )i d ]i q

[0105] The stator current is expressed in polar coordinate form:

[0106]

[0107] in, represents the amplitude of the stator current, and β is the angle between the current vector and the α axis.

[0108] Substituting the above current expression into the electromagnetic torque equation, we can obtain:

[0109]

[0110] when i s When the torque T e is a function of the current angle β. e Find the partial derivative about the current angle β:

[0111]

[0112] Let the above partial derivative equal to 0 and simplify to get

[0113]

[0114] Using the double angle formula cos2β=2cos 2 β-1 available

[0115] ψ f cosβ+(L d -L q )i s (2cos 2 β-1)=0

[0116] Arrangement gives the quadratic equation for cosβ:

[0117] 2(L d -L q )i s cos 2 β+ψ f cosβ-(L d -L q )i s =0

[0118] Solve cosβ according to the quadratic equation root formula:

[0119]

[0120] Since the value range of β is usually [0, π],

[0121]

[0122] Will Substituting in the function of MTPA, we can get:

[0123]

[0124] Where sign is the sign function,

[0125] Step 2: Introduce the dynamically assigned weight coefficient to optimize the non-cascade MPC-MTPA controller and determine the cost function of the final non-cascade MPC-MTPA controller.

[0126] A non-cascade MPC-MTPA controller is constructed. Non-cascade control uses the cost function in MPC to simultaneously complete torque control and MTPA control, which simplifies the control algorithm structure and improves the stability and dynamic response characteristics of the control algorithm.

[0127] g Te is the torque tracking cost function, is the torque target value, is the predicted value of electromagnetic torque at time k+1. According to the MTPA control constraint relationship of dq axis current: The cost function g can be obtained MTPA :

[0128]

[0129] Considering the safety factor to limit the maximum current value, the constraint item of the maximum current value is set to g lim_MTPA

[0130]

[0131] In summary, the cost function g of the non-cascaded MPC-MTPA controller is:

[0132] g=λ1g Te +λ2g MTPA +λ3(g imax +g lim_MTPA )

[0133] Among them, λ1, λ2, λ3 represent weight coefficients, g imax represents the cost function of model predictive control MPC, g MTPA represents the cost function of the maximum torque current ratio control MTPA controller, g lim_MTPAA constraint term representing the maximum current value.

[0134] Step 3: Construct the fitness function, use the improved Mustang algorithm to optimize the weight coefficients, and determine the cost function of the non-cascaded MPC-MTPA controller after the final optimization.

[0135] Construct a fitness function:

[0136]

[0137] Where the integral of the torque error is selected As an indicator of torque tracking accuracy, T e is the actual torque, T e，ref is the reference torque. Select the deviation integral of the current and the MTPA curve As an indicator for evaluating the control effect of MTPA, i d 、i q is the dq axis component of the actual current, i d，MTPA 、i q,MTPA is the dq axis component of the ideal current obtained according to the MTPA control principle. The time t when the current exceeds the constraint range is selected over As a measure of whether the current constraint is satisfied.

[0138] Set the population size to N = 30. The proportion of stallions P s =0.2. Population hybridization rate P c =0.1.

[0139] First, according to the population size N and the stallion individual proportion P s , calculate the number of stallions G = N × P s , and the number of remaining individuals in the population. The population is then randomly divided into G groups, each led by a stallion. The initial position of the stallion is randomly selected within the range of decision variables (weight coefficients λ1, λ2, λ3), and the initial positions of the remaining individuals are also randomly initialized within the corresponding range.

[0140] For each individual in the population (i.e. a set of weight coefficient combinations ) is substituted into the objective function of the MPC-MTPA control method, and the corresponding fitness value is calculated in combination with the actual operation model of the motor. The specific calculation process is: according to the current state and control requirements of the motor, the predicted torque, current and other parameters are calculated using the objective function, and then the fitness value Fitness of the individual is calculated according to the definition of the fitness function.

[0141] The herding behavior is one of the most important behaviors in the algorithm. The stallion leads the entire herd to move, and the rest of the individuals in the herd perform herding searches with the stallion as the center. The position update method of the rest of the individuals is as follows:

[0142]

[0143] in, is the current position of the remaining individuals, are the updated positions of the remaining individuals, is the current position of the stallion, R is a random number in [-2,2], and Z is an adaptive parameter, which is calculated as follows:

[0144]

[0145] IDX=(P=0)

[0146] P=R1<T tdr

[0147] Among them, R1 and R3 are random variables uniformly distributed in [0,1], R1 is a random number in the range of [0,1], and T tdr It is an adaptive parameter of the algorithm. It starts from 1, decreases gradually as the algorithm runs, and becomes 0 when the algorithm ends.

[0148]

[0149] T is the current iteration number. In this way, the remaining individuals can search around the stallion within a certain range and explore different weight coefficient combinations.

[0150] Population hybridization behavior refers to the fact that when foals in a population mature, they will leave the current group and mate with new foals in other groups to produce new individuals. The position information of hybrid individuals is updated as follows:

[0151]

[0152] Crossover=Mean

[0153] in is the position of the new individual p in the population k after mating, is the position information of the two individuals in the parent generation, Crossover = Mean means the mean crossover method is adopted. Through hybridization, the diversity of the population can be increased, which helps the algorithm to jump out of the local optimal solution.

[0154] The population movement behavior means that the stallion individual, as the leader of the population, will lead the population to find a better habitat ("water hole", that is, the optimal solution), and the rest of the individuals will move with the stallion individual. The position update method of the stallion individual is shown in the formula:

[0155]

[0156] Where W H is the location of the “water hole” in the habitat (i.e., the location corresponding to the currently found optimal weight coefficient combination), N SGi is the current position of the stallions in the herd, is the updated position of the stallion in the group, R3 is a random number in [0,1], which determines the clockwise or counterclockwise direction of the group's advance, Z is an adaptive parameter, Q1 and Q2 are random numbers in [-1,1], and ω is the dynamic inertia weight coefficient, which is calculated as follows:

[0157]

[0158] f i (t) is the fitness value of the stallion individual at the current iteration number, f min (t) is the minimum fitness value of all individuals in the population in the current iteration number, f avg (t) is the average fitness of all stallion individuals in the current iteration. In this way, the population can move towards the optimal solution and continuously optimize the combination of weight coefficients.

[0159] The competition behavior within the population refers to the random selection of stallion particles in the initialization stage of the algorithm. As the algorithm runs, stallion individuals are selected according to their fitness information, and the individual with the smallest fitness is selected. The competition behavior within the population is shown in the formula:

[0160]

[0161] Where cost is the fitness function, x G,i Through internal competition, it ensures that stallions are always individuals with better fitness in the population, leading the population to search in a better direction.

[0162] Termination condition setting, judging whether the termination condition is met, setting the maximum number of iterations to K = 50. At the same time, when the fitness value change of the optimal solution in L = 5 consecutive iterations is less than a certain threshold, it is also considered to meet the convergence condition.

[0163] After the iteration is completed, the combination with the smallest fitness function value is selected from all recorded optimal weight coefficient combinations As the final optimal solution. The optimal weight coefficient combination is applied to the cost function. This improves the control performance of the motor under various working conditions, achieves better target torque tracking, MTPA control and current constraint satisfaction, and makes the motor run more stable, efficient and with better dynamic response characteristics. If the termination condition is not met, continue to iterate until the condition is met.

[0164] The present invention is further described below in conjunction with the accompanying drawings, wherein the model is combined with vector control to verify the feasibility of the improved model, but it is only used to explain the present invention rather than to limit the scope of the present invention.

[0165] Figure 1 The FFT diagram analysis for the traditional weight coefficient design (weight coefficient 32) is shown. The traditional weight coefficient is obtained by trial and error, which is inaccurate. It can be seen from the figure that when the weight coefficient is designed to be 32, the harmonic content of the stator current is large. Figure 2 FFT diagram analysis for traditional weight coefficient design (weight coefficient 80). It can be seen from the figure that as the weight coefficient increases, the harmonic content of the stator current decreases. Figure 3 Design FFT diagram analysis for new weight coefficients. With the new weight method, the harmonics of the stator current are very small and the waveform of the electronic current tends to be sinusoidal. Figure 4 Changes in the overall simulation waveform of control under traditional weight design. Figure 5 Changes in the overall simulation waveform of control under the new weight design.

[0166] Those skilled in the art should understand that the above embodiments are merely exemplary embodiments and that various changes, substitutions, and alterations may be made without departing from the spirit and scope of the present application.

Claims

1. A method for predictive torque control of a non-cascaded permanent magnet synchronous motor model based on an improved Mustang algorithm, characterized in that: The steps include: Step 1: Establish the stator voltage and current equation and the electromagnetic torque equation in the synchronous rotating coordinate system, construct the model predictive control MPC controller and the maximum torque current ratio control MTPA controller in the permanent magnet synchronous motor model predictive torque control, and obtain the current and torque prediction equations through discretization; Step 2: Introduce the dynamically allocated weight coefficient to optimize the non-cascade MPC-MTPA controller and determine the cost function of the final non-cascade MPC-MTPA controller; Step 3: Construct the fitness function, use the improved Mustang algorithm to optimize the weight coefficients, and determine the cost function of the non-cascaded MPC-MTPA controller after the final optimization.

2. The method for predicting torque control of a non-cascaded permanent magnet synchronous motor model based on an improved Mustang algorithm according to claim 1 is characterized in that: The stator voltage and current equation and electromagnetic torque equation established in step 1 are: 1) The stator voltage and current equation is: 2) The electromagnetic torque equation is: T e =1.5p n [ψ f +(L d -L q )i d ]i q ; Among them, i d 、i q is the d-axis and q-axis components of the actual current, R s is the motor phase resistance, L d , L q is the d and q axis inductance, ω e is the electrical angular velocity, ψ f is the permanent magnet flux, u d 、u q is the d and q axis voltage, p n is the pole pair number.

3. The method for predicting torque control of a non-cascaded permanent magnet synchronous motor model based on an improved Mustang algorithm according to claim 2 is characterized in that: The model predictive control MPC is specifically as follows: Model predictive control MPC uses the forward Euler method to discretize the voltage and current state equations in the dq coordinate system to obtain the prediction equations for the d-axis and q-axis currents: Among them, i d (k+1) is the predicted value of the d-axis current at time k+1, T s is the system control period, is the feedback d-axis current value at time k, i q (k+1) is the predicted value of the q-axis current at time k+1, is the feedback q-axis current value at time k, is the d-axis voltage value fed back at time k, is the feedback q-axis voltage value at time k.

4. The method for predicting torque control of a non-cascaded permanent magnet synchronous motor model based on an improved Mustang algorithm according to claim 2 is characterized in that: The maximum torque current ratio controls the MTPA controller to solve the minimum stator current i when the electromagnetic torque is the same d 、i q The combination is as follows: The stator current is expressed in polar coordinate form: in, represents the amplitude of the stator current, β is the angle between the current vector and the α axis; substituting the current into the electromagnetic torque equation, we get: when i s When the torque T e is a function of the current angle β, for T e Find the partial derivative about the current angle β: Let the partial derivative equal to 0 and simplify to get: Using the double angle formula cos2β=2cos 2 Beta-1: 2(L d -L q )i s cos 2 b+ψ f cosβ-(L d -L q )i s =0 Solve cosβ according to the quadratic equation root formula: Since the value range of β is usually [0, π], Will Substituting into the function of MTPA is: Among them, sign is the sign function, 5. The method for predicting torque control of a non-cascaded permanent magnet synchronous motor model based on an improved Mustang algorithm according to claim 1, characterized in that: The step 2 optimizes the non-cascaded MPC-MTPA controller and uses the cost function in the MPC to simultaneously complete the torque control and the MTPA control, specifically: The cost function g of the current control scheme of model predictive control MPC imax Considering the tracking of torque characterizing current, optimizing the ratio of torque to current and limiting the current amplitude, the cost function g imax for: in, i max is the maximum input current; Determine the torque tracking cost function: Among them, g T e is the torque tracking cost function, is the torque target value, T e p (k+1) is the predicted value of electromagnetic torque at time k+1; According to the MTPA control constraints on dq axis current: Cost function g of the maximum torque current ratio control MTPA controller MTPA : Considering the safety factor to limit the maximum current value, the constraint item of the maximum current value is set to g lim_MTPA : in, is the predicted value of the d-axis current at time k+1; The formula for determining the final cost function g of the non-cascaded MPC-MTPA controller is: g=λ1g Te +λ2g MTPA +λ3(g imax +g lim_MTPA ) Among them, λ1, λ2, λ3 represent weight coefficients, g imax represents the cost function of model predictive control MPC, g MTPA represents the cost function of the maximum torque current ratio control MTPA controller, g lim_MTPA A constraint term representing the maximum current value.

6. The method for predicting torque control of a non-cascaded permanent magnet synchronous motor model based on an improved Mustang algorithm according to claim 1, characterized in that: The fitness function constructed in step 3 is: Among them, the integral of the torque error is selected As an indicator of torque tracking accuracy, T e is the actual torque, T e，ref For the reference torque, select the integral of the deviation between the current and the MTPA curve As an indicator for evaluating the control effect of MTPA, i d 、i q is the d-axis and q-axis components of the actual current, i d，MTPA 、i q,MTPA are the d-axis and q-axis components of the ideal current obtained according to the MTPA control principle; the time t when the current exceeds the constraint range is selected over As a measure of whether the current constraint is satisfied.

7. The method for predicting torque control of a non-cascaded permanent magnet synchronous motor model based on an improved Mustang algorithm according to claim 6 is characterized in that: The improved wild horse algorithm selects the optimal weight coefficient as follows: Step 3.1: Set the population size to N = 30 and the stallion individual proportion P s =0.2, population hybridization rate P c =0.1; Step 3.2: According to the population size N and stallion individual proportion P s , calculate the number of stallions G = N × P s , and the number of remaining individuals in the population, and then randomly divide the population into G groups, each group is led by a stallion, the initial position of the stallion is randomly selected within the value range of the decision variable, and the initial positions of the remaining individuals are also randomly initialized within the corresponding range. The decision variables are weight coefficients λ1, λ2, λ3; Step 3.3: For each individual in the population, a set of weight coefficient combinations Substitute it into the cost function of the MPC-MTPA controller and calculate the corresponding fitness value in combination with the actual operation model of the motor. The specific calculation process is as follows: according to the current state and control requirements of the motor, the predicted torque and current parameters are calculated using the cost function, and then the individual fitness value Fitness is calculated according to the definition of the fitness function. Step 3.4: The herding behavior of the group is led by the stallion to move the entire group, and the rest of the individuals in the group perform grazing search with the stallion as the center. The position of the rest of the individuals is updated as follows: in, is the current position of the remaining individuals, are the updated positions of the remaining individuals, is the current position of the stallion, R is a random number in [-2,2], and Z is an adaptive parameter, which is calculated as follows: Z=R2⊙IDX+R3⊙(~IDX) IDX=(P=0) P=R1<T tdr Among them, R1 and R3 are random variables uniformly distributed in [0,1], T tdr is an adaptive parameter that starts at 1 and gradually decreases as the algorithm runs, and becomes 0 when the algorithm ends: T is the current number of iterations, T max is the maximum number of iterations; Step 3.5: Population hybridization behavior means that when the foals in the population mature, they will leave the current group and mate with new foals in other groups to produce new individuals. The position information of the hybrid individuals is updated as follows: Crossover=Mean in, is the position of the new individual p in the population k after mating, is the position information of the two individuals in the parent generation, Crossover = Mean means the mean crossover method is adopted; Step 3.6: Population movement behavior means that the stallion individual, as the leader of the population, will lead the population to find a better habitat, and the other individuals will move with the stallion individual; the position update method of the stallion individual is shown in the formula: Among them, W H is the location of the "water hole" in the habitat, that is, the location corresponding to the optimal weight coefficient combination currently found, is the current position of the stallions in the herd, is the updated position of the stallion in the group, R3 is a random number in [0,1], which determines the clockwise or counterclockwise direction of the group's advance, Q1 and Q2 are random numbers in [-1,1], and ω is the dynamic inertia weight coefficient, which is calculated as follows: Among them, f i (t) is the fitness value of the stallion individual at the current iteration number, f min (t) is the minimum fitness value of all individuals in the population in the current iteration number, f avg (t) is the average fitness of all stallion individuals in the current iteration, ω max is the upper limit of the dynamic weight, ω min is the lower limit of the dynamic weight; the population moves towards the optimal solution and continuously optimizes the combination of weight coefficients; Step 3.7: The competition behavior within the population refers to the random selection of stallion particles in the initialization stage of the algorithm. As the algorithm runs, stallion individuals are selected according to their fitness information, and the individuals with the smallest fitness are selected; the competition behavior within the population is shown in the formula: Among them, cost is the fitness function, x G,i For the position of group members, through internal competition, the stallion individuals are always guaranteed to be the individuals with better fitness in the population, leading the population to search in a better direction; Step 3.8: Set the termination condition. If the termination condition is met, the iteration will be terminated. After the iteration is completed, the combination with the smallest fitness function value is selected from all the recorded optimal weight coefficient combinations. As the final optimal solution, the optimal weight coefficient combination is applied to the cost function of the MPC-MTPA controller.

Images