Induction motor weight factor adaptive model prediction direct speed control method

Abstract

The invention discloses an induction motor weight factor adaptive model prediction direct speed control method, which comprises the following steps of: firstly, establishing a mathematical model under a two-phase static coordinate system of an induction motor by taking stator current and rotor flux linkage as state variables; then, in a mathematical model under a two-phase static coordinate system, assuming that the current moment is k moment, predicting the stator flux linkage and the rotating speed at the (k + 1) moment, and obtaining a stator flux linkage predicted value and a rotating speed predicted value; designing a cost function according to the stator flux linkage and the rotating speed predicted value at the k + 1 moment in the model prediction control algorithm; introducing a genetic algorithm into model prediction direct speed control, and adaptively adjusting a weight factor in a cost function; and finally, an adaptive crossover operator and an adaptive mutation operator are introduced to improve a genetic algorithm, adaptive adjustment of a weight factor is finally realized, and the induction motor is controlled. According to the method, the self-adaptive adjustment of the weight factor is realized, the calculation amount is greatly reduced, and the prediction precision and the motor control capability are effectively improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of high-performance induction motor control, and in particular relates to a direct speed control method for an induction motor using a weight factor adaptive model prediction method. Background Art

[0002] The induction motor is a simple motor type. Due to its low cost, easy maintenance, and high reliability, it is widely used in industrial and commercial applications, providing a stable power source. With technological advancements, the demand for motor performance continues to increase, and high-performance AC speed control technology is gradually developing, providing more possibilities for the application of induction motors.

[0003] In recent years, model predictive control (MPC) has gained widespread attention in the field of motor control as an advanced control strategy. By applying MPC to induction motors, higher levels of dynamic performance, accuracy, and stability can be achieved to meet the ever-increasing industrial demands.

[0004] In model predictive control (MPC), the cost function can include multiple objectives and constraints. For example, the cost function for MPC (Model Predictive Direct Speed Control) includes both speed error and stator flux error. However, because speed and stator flux have different dimensions, it is necessary to design appropriate weighting factors to achieve simultaneous optimal control of both. Currently, the design of weighting factors is still immature, and most methods rely on extensive simulations, experiments, or trial-and-error approaches to optimize the weighting factors. Therefore, adaptive adjustment of weighting factors is of great significance for MPC (Model Predictive Direct Speed Control) of induction motors. Summary of the Invention

[0005] The purpose of the present invention is to provide an induction motor weight factor adaptive model prediction direct speed control method, which realizes adaptive adjustment of the weight factor, greatly reduces the amount of calculation, and effectively improves the prediction accuracy and the control ability of the motor.

[0006] The technical solution adopted by the present invention is a direct speed control method of an induction motor weight factor adaptive model prediction, which is specifically implemented according to the following steps:

[0007] Step 1: Using the stator current and rotor flux as state variables, establish a mathematical model of the induction motor in a two-phase stationary coordinate system;

[0008] Step 2: Based on the mathematical model of the induction motor in the two-phase stationary coordinate system obtained in step 1, assuming that the current moment is moment k, the stator flux and speed at moment k+1 are predicted to obtain a stator flux prediction value and a speed prediction value;

[0009] Step 3: Design a cost function based on the stator flux and speed prediction values at time k+1 in the model predictive control algorithm obtained in step 2;

[0010] Step 4: Introduce the genetic algorithm into the model predictive direct speed control and adaptively adjust the weight factor in the cost function of step 3;

[0011] Step 5: Based on the traditional genetic algorithm described in step 4, the genetic algorithm is improved by introducing an adaptive crossover operator and an adaptive mutation operator, thereby finally achieving adaptive adjustment of the weight factor and controlling the induction motor.

[0012] The present invention is also characterized in that:

[0013] In step 1, the induction motor is modeled to obtain a mathematical model of the induction motor in a two-phase stationary coordinate system, as follows:

[0014] The stator current and rotor flux are selected as state variables, and the mathematical model of the induction motor is expressed as a complex vector:

[0015]

[0016] in, x=[i s ψ r ] T , u=u s =[u sα u sβ ] T ,

[0017] The stator flux is calculated from the stator current and the rotor flux:

[0018]

[0019] Induction motor electromagnetic torque T e for:

[0020]

[0021] Where u s is the voltage vector, i s is the stator current, i r is the rotor current, ψ s is the stator flux, ψ r is the rotor flux, T e is the electromagnetic torque, T l is the load torque, ω e is the synchronous speed of the motor, ω r is the rotor speed, R s is the stator resistance, Rr is the rotor resistance, L s is the stator inductance, L r is the rotor inductance, L m is the mutual inductance, J is the moment of inertia, n p is the pole pair number.

[0022] In step 2, assuming that the current moment is k, the stator flux and current estimated by the full-order observer predict the stator flux at time k+1, and calculate the electromagnetic torque, specifically:

[0023] Step 201: Estimate the stator flux and current of the induction motor using a full-order observer based on the mathematical model of the induction motor, and calculate the electromagnetic torque. The equations are shown in formulas (4) to (6):

[0024]

[0025] Where, is the estimated value of ●, The operation represents the multiplication of corresponding elements of two vectors. G1 and G2 are calculated as follows:

[0026]

[0027] Where, k is the proportionality coefficient;

[0028] Step 202: Based on the forward Euler discretization formula, discretize formulas (4) to (5) to obtain the stator flux prediction value at time k+1 in the model predictive control method as shown in formula (7), and the stator current prediction value as shown in formula (8):

[0029]

[0030] Among them, T s is the sampling period;

[0031] Step 203: Obtain the predicted speed value at time k+1 based on the electromagnetic torque and load torque at time k:

[0032]

[0033] Where, is the predicted speed value at time k+1, is the speed at time k, is the electromagnetic torque at time k.

[0034] In step 3, the cost function is designed based on the predicted flux and speed at time k+1, specifically:

[0035]

[0036] Where λ is the weight factor; the controller compensates for the delay by determining the voltage vector at time k+1 one step ahead.

[0037] Step 4 is as follows:

[0038] Step 401: Initialize the candidate set of weight factors:

[0039] Within the set weight factor variation range, N numbers are randomly generated as the initial population of the genetic algorithm;

[0040] Step 402: Calculate the fitness value F(λ i );

[0041] Step 403: Select an operation.

[0042] Step 404: Cross-exchange operation.

[0043] Step 402 is specifically as follows:

[0044] First, for each λ, the optimal voltage vector corresponding to λ is calculated by equation (10). Then, for each λ, the adaptation value F(λ i ):

[0045]

[0046] Where S A is the error surface,

[0047]

[0048] in, y is an assumed variable, * is the given value of y.

[0049] Step 403 is specifically as follows:

[0050] The next generation population is selected according to the roulette wheel algorithm. The specific operations are:

[0051] (1) First, calculate the selection probability of each λ:

[0052]

[0053] (2) Generate a random probability P between 0 and 1 s , select The first established i As an individual of the next generation group, N lambdas are selected as the next generation group.

[0054] Step 404 is specifically as follows:

[0055] The crossover operation generates offspring from two parents. The specific operation is:

[0056] (1) Encoding λ: Using real-valued encoding, λ ranges from 1 to 99, that is, the ones and tens digits of each λ are regarded as "chromosomes";

[0057] (2) First, randomly select two λ as parents, and then generate a random crossover probability P c , if it is greater than the given crossover probability The units digit of one parent is exchanged with the tens digit of the other parent, and the two new individuals obtained replace the parents as individuals in the group.

[0058] Step 405: Mutation operation:

[0059] The specific operations are as follows:

[0060] First, randomly generate a mutation probability P of 0 to 1 m , if it is greater than the given mutation probability, that is Then randomly select an individual and generate a new λ to replace the individual;

[0061] Step 406: Selection of optimal weight factor and optimal switching vector:

[0062] The group is iterated according to steps 401 to 405. When the number of iterations reaches a given value or the number of identical individuals in the group reaches 90% of the total, the mode of the group is selected as the optimal weight factor and the corresponding optimal switching vector is selected.

[0063] In step 5, an adaptive crossover operator is introduced to replace the traditional fixed crossover operator, specifically:

[0064] (1) If the number of modes in the current group n satisfies n>0.5N, let the crossover probability

[0065] (2) When n>0.5N, perform the following cross-interchange operation:

[0066] Choose two parents at random, if the random probability Then only the units digits of the two parents are exchanged, and the tens digit remains unchanged.

[0067] In step 5, the adaptive mutation operator is introduced as follows:

[0068]

[0069] Where, is a fixed mutation probability, is the adaptive mutation probability, F max is the maximum value of the fitness function in the t generation population, ASD t is the mean square error of the fitness value of the t generation group

[0070]

[0071] Among them, F v is the average fitness value of the t generation group

[0072] Therefore, introducing the adaptive crossover operator and mutation operator into the genetic algorithm can avoid premature phenomenon, accelerate the convergence of the algorithm, and ultimately achieve adaptive adjustment of the weight factor.

[0073] For a two-level voltage source inverter, there are eight basic switching states. By improving the genetic algorithm, the optimal weight factor is obtained, and the optimal switching vector at the next moment is also obtained.

[0074] The present invention has the beneficial effect of introducing a genetic algorithm to adaptively adjust the weight factors of the cost function of model predictive direct speed control for induction motors. The invention also employs adaptive crossover and mutation operators to accelerate algorithm convergence, reduce computational complexity, and avoid premature convergence. This overcomes the difficulty of weight factor design, enhances the control performance of model predictive control, and thus improves the operating performance of the motor. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 This is a control block diagram of the induction motor model prediction direct speed control method of the present invention;

[0076] Figure 2 It is the basic voltage vector block diagram of the two-level inverter in the present invention;

[0077] Figure 3 It is an algorithm flow chart of the genetic algorithm in the present invention;

[0078] Figure 4 This is a block diagram of a two-level voltage source inverter in the present invention. DETAILED DESCRIPTION

[0079] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0080] The invention provides an induction motor weight factor adaptive model prediction direct speed control method, such as Figure 1 The figure shows the control block diagram of the induction motor model prediction direct speed control method: First, the reconstructed stator voltage and measured stator current are used as the input of the full-order observer to estimate the rotor flux and electromagnetic torque; then, the discrete mathematical model is used to predict the following: Figure 2The eight basic voltage vectors shown in the figure correspond to the stator flux and motor speed at the next moment; the design cost function includes stator flux error and speed error, and an improved genetic algorithm is used to adaptively adjust the weight factor. The specific flow chart of the genetic algorithm is as follows: Figure 3 As shown, the optimal weight factor and its corresponding optimal voltage vector are obtained through the algorithm, and the corresponding switching signal is output in the next control cycle to act on the Figure 4 The two-level inverter shown is used to drive the motor. Since the weight factors are adaptively adjusted, the weight factor design process is avoided and the control accuracy of the motor is improved.

[0081] The induction motor weight factor adaptive model prediction direct speed control method of the present invention is combined with Figure 1 , specifically follow the steps below:

[0082] Step 1: Using the stator current and rotor flux as state variables, establish a mathematical model of the induction motor in a two-phase stationary coordinate system;

[0083] In step 1, the induction motor is modeled to obtain a mathematical model of the induction motor in a two-phase stationary coordinate system, as follows:

[0084] The stator current and rotor flux are selected as state variables, and the mathematical model of the induction motor is expressed as a complex vector:

[0085]

[0086] in, x=[i s ψ r ] T , u=u s =[u sα u sβ ] T ,

[0087] The stator flux is calculated from the stator current and the rotor flux:

[0088]

[0089] Induction motor electromagnetic torque T e for:

[0090]

[0091] Where u s is the voltage vector, i s is the stator current, i r is the rotor current, ψ s is the stator flux, ψ ris the rotor flux, T e is the electromagnetic torque, T l is the load torque, ω e is the synchronous speed of the motor, ω r is the rotor speed, R s is the stator resistance, R r is the rotor resistance, L s is the stator inductance, L r is the rotor inductance, L m is the mutual inductance, J is the moment of inertia, n p is the pole pair number.

[0092] Step 2: Based on the mathematical model of the induction motor in the two-phase stationary coordinate system obtained in step 1, assuming that the current moment is moment k, the stator flux and speed at moment k+1 are predicted to obtain a stator flux prediction value and a speed prediction value;

[0093] In step 2, assuming that the current moment is k, the stator flux and current estimated by the full-order observer predict the stator flux at time k+1, and calculate the electromagnetic torque, specifically:

[0094] Step 201: Estimate the stator flux and current of the induction motor using a full-order observer based on the mathematical model of the induction motor, and calculate the electromagnetic torque. The equations are shown in formulas (4) to (6):

[0095]

[0096]

[0097] Where, is the estimated value of , The operation represents the multiplication of corresponding elements of two vectors. G1 and G2 are calculated as follows:

[0098]

[0099] Where, k is the proportionality coefficient;

[0100] Step 202: Based on the forward Euler discretization formula, discretize formulas (4) to (5) to obtain the stator flux prediction value at time k+1 in the model predictive control method as shown in formula (7), and the stator current prediction value as shown in formula (8):

[0101]

[0102] Among them, T s is the sampling period;

[0103] Step 203: Obtain the predicted speed value at time k+1 based on the electromagnetic torque and load torque at time k:

[0104]

[0105] Where, is the predicted speed value at time k+1, is the speed at time k, is the electromagnetic torque at time k.

[0106] Step 3: Design a cost function based on the stator flux and speed prediction values at time k+1 in the model predictive control algorithm obtained in step 2;

[0107] In step 3, the cost function is designed based on the predicted flux and speed at time k+1, specifically:

[0108]

[0109] Among them, λ is the weight factor;

[0110] Because digital control systems have a one-beat delay in practical applications, the voltage vector that should be applied at the current time k is not updated until the next time k+1. To eliminate the impact of this one-beat delay on control effectiveness, the controller compensates for this delay by determining the voltage vector at time k+1 one step ahead.

[0111] Step 4: Introduce the genetic algorithm into the model predictive direct speed control and adaptively adjust the weight factor in the cost function of step 3;

[0112] Step 4 is as follows:

[0113] Step 401: Initialize the candidate set of weight factors:

[0114] Within the set weight factor variation range, N numbers are randomly generated as the initial population of the genetic algorithm;

[0115] Step 402: Calculate the fitness value F(λ i );

[0116] Step 402 is specifically as follows:

[0117] First, for each λ, the optimal voltage vector corresponding to λ is calculated by equation (10). Then, for each λ, the adaptation value F(λ i ):

[0118]

[0119] Where S A is the error surface,

[0120]

[0121] in, y is an assumed variable, * is the given value of y.

[0122] Step 403: Select an operation.

[0123] Step 403 is specifically as follows:

[0124] Selection is the process of selecting superior individuals from a population and eliminating inferior ones. It is based on fitness assessment. Individuals with greater fitness are more likely to be selected, and their "offspring" will have more children in the next generation. The selected individuals are then placed in a pairing pool. Common selection methods include the roulette wheel method, the best individual retention method, the expected value method, the ranked selection method, the competitive method, and the linear normalization method. Here, the next generation population is selected based on the roulette wheel algorithm. The specific operations are:

[0125] (1) First, calculate the selection probability of each λ:

[0126]

[0127] (2) Generate a random probability P between 0 and 1 s , select The first established i As an individual of the next generation group, N lambdas are selected as the next generation group.

[0128] Step 404: Cross-exchange operation.

[0129] Step 404 is specifically as follows:

[0130] The crossover operation produces offspring from two parents, reflecting the process of biological reproduction. The specific operations are:

[0131] (1) Encoding λ: Common encoding methods include binary encoding and real-valued encoding. Here, real-valued encoding is used, and the value of λ ranges from 1 to 99. That is, the ones and tens digits of each λ are regarded as "chromosomes";

[0132] (2) First, randomly select two λ as parents, and then generate a random crossover probability P c , if it is greater than the given crossover probability The units digit of one parent is exchanged with the tens digit of the other parent, and the two new individuals obtained replace the parents as individuals in the group.

[0133] Step 405: Mutation operation:

[0134] Mutation operation is a manifestation of biodiversity. It generally occurs with a small probability and plays an important role in the genetic process. The specific operation is as follows:

[0135] First, randomly generate a mutation probability P of 0 to 1 m , if it is greater than the given mutation probability, that is Then randomly select an individual and generate a new λ to replace the individual;

[0136] Step 406: Selection of optimal weight factor and optimal switching vector:

[0137] The group is iterated according to steps 401 to 405. When the number of iterations reaches a given value or the number of identical individuals in the group reaches 90% of the total, the mode of the group is selected as the optimal weight factor and the corresponding optimal switching vector is selected.

[0138] Step 5: Based on the traditional genetic algorithm described in step 4, the genetic algorithm is improved by introducing an adaptive crossover operator and an adaptive mutation operator, thereby finally achieving adaptive adjustment of the weight factor and controlling the induction motor.

[0139] In step 5, in order to speed up the convergence of the genetic algorithm, an adaptive crossover operator is introduced to replace the traditional fixed crossover operator, specifically:

[0140] (1) If the number of modes in the current group n satisfies n>0.5N, let the crossover probability

[0141] (2) When n>0.5N, perform the following cross-interchange operation:

[0142] Choose two parents at random, if the random probability Then only the units digits of the two parents are exchanged, and the tens digit remains unchanged.

[0143] In step 5, in order to avoid premature phenomenon in the iterative process of genetic algorithm, an adaptive mutation operator is introduced, as follows:

[0144]

[0145] Where, is a fixed mutation probability, is the adaptive mutation probability, F max is the maximum value of the fitness function in the t generation population, ASD t is the mean square error of the fitness value of the t generation group

[0146]

[0147] Among them, F v is the average fitness value of the t generation group

[0148] Therefore, introducing the adaptive crossover operator and mutation operator into the genetic algorithm can avoid premature phenomenon, accelerate the convergence of the algorithm, and ultimately achieve adaptive adjustment of the weight factor.

[0149] For a two-level voltage source inverter, there are eight basic switching states. By improving the genetic algorithm, the optimal weight factor is obtained, and the optimal switching vector at the next moment is also obtained.

[0150] Example 1

[0151] The induction motor weight factor adaptive model prediction direct speed control method of the present invention is combined with Figure 1 , specifically follow the steps below:

[0152] Step 1: Using the stator current and rotor flux as state variables, establish a mathematical model of the induction motor in a two-phase stationary coordinate system;

[0153] Step 2: Based on the mathematical model of the induction motor in the two-phase stationary coordinate system obtained in step 1, assuming that the current moment is moment k, the stator flux and speed at moment k+1 are predicted to obtain a stator flux prediction value and a speed prediction value;

[0154] Step 3: Design a cost function based on the stator flux and speed prediction values at time k+1 in the model predictive control algorithm obtained in step 2;

[0155] Step 4: Introduce the genetic algorithm into the model predictive direct speed control and adaptively adjust the weight factor in the cost function of step 3;

[0156] Step 5: Based on the traditional genetic algorithm described in step 4, the genetic algorithm is improved by introducing an adaptive crossover operator and an adaptive mutation operator, thereby finally achieving adaptive adjustment of the weight factor and controlling the induction motor.

[0157] Example 2

[0158] The induction motor weight factor adaptive model prediction direct speed control method of the present invention is combined with Figure 1 , specifically follow the steps below:

[0159] Step 1: Using the stator current and rotor flux as state variables, establish a mathematical model of the induction motor in a two-phase stationary coordinate system;

[0160] In step 1, the induction motor is modeled to obtain a mathematical model of the induction motor in a two-phase stationary coordinate system, as follows:

[0161] The stator current and rotor flux are selected as state variables, and the mathematical model of the induction motor is expressed as a complex vector:

[0162]

[0163] in, x=[i s ψ r ] T , u=u s =[u sα u sβ ] T ,

[0164] The stator flux is calculated from the stator current and the rotor flux:

[0165]

[0166] Induction motor electromagnetic torque T e for:

[0167]

[0168] Where u s is the voltage vector, i s is the stator current, i r is the rotor current, ψ s is the stator flux, ψ r is the rotor flux, T e is the electromagnetic torque, T l is the load torque, ω e is the synchronous speed of the motor, ω r is the rotor speed, R s is the stator resistance, R r is the rotor resistance, L s is the stator inductance, L r is the rotor inductance, L m is the mutual inductance, J is the moment of inertia, n p is the pole pair number.

[0169] Step 2: Based on the mathematical model of the induction motor in the two-phase stationary coordinate system obtained in step 1, assuming that the current moment is moment k, the stator flux and speed at moment k+1 are predicted to obtain a stator flux prediction value and a speed prediction value;

[0170] Step 3: Design a cost function based on the stator flux and speed prediction values at time k+1 in the model predictive control algorithm obtained in step 2;

[0171] Step 4: Introduce the genetic algorithm into the model predictive direct speed control and adaptively adjust the weight factor in the cost function of step 3;

[0172] Step 5: Based on the traditional genetic algorithm described in step 4, the genetic algorithm is improved by introducing an adaptive crossover operator and an adaptive mutation operator, thereby finally achieving adaptive adjustment of the weight factor and controlling the induction motor.

[0173] Example 3

[0174] The induction motor weight factor adaptive model prediction direct speed control method of the present invention is combined with Figure 1 , specifically follow the steps below:

[0175] Step 1: Using the stator current and rotor flux as state variables, establish a mathematical model of the induction motor in a two-phase stationary coordinate system;

[0176] In step 1, the induction motor is modeled to obtain a mathematical model of the induction motor in a two-phase stationary coordinate system, as follows:

[0177] The stator current and rotor flux are selected as state variables, and the mathematical model of the induction motor is expressed as a complex vector:

[0178]

[0179] in, x=[i s ψ r ] T , u=u s =[u sα u sβ ] T ,

[0180] The stator flux is calculated from the stator current and the rotor flux:

[0181]

[0182] Induction motor electromagnetic torque T e for:

[0183]

[0184] Where u s is the voltage vector, i s is the stator current, i r is the rotor current, ψ s is the stator flux, ψ r is the rotor flux, T e is the electromagnetic torque, T l is the load torque, ω e is the synchronous speed of the motor, ω r is the rotor speed, R s is the stator resistance, Rr is the rotor resistance, L s is the stator inductance, L r is the rotor inductance, L m is the mutual inductance, J is the moment of inertia, n p is the pole pair number.

[0185] Step 2: Based on the mathematical model of the induction motor in the two-phase stationary coordinate system obtained in step 1, assuming that the current moment is moment k, the stator flux and speed at moment k+1 are predicted to obtain a stator flux prediction value and a speed prediction value;

[0186] In step 2, assuming that the current moment is k, the stator flux and current estimated by the full-order observer predict the stator flux at time k+1, and calculate the electromagnetic torque, specifically:

[0187] Step 201: Estimate the stator flux and current of the induction motor using a full-order observer based on the mathematical model of the induction motor, and calculate the electromagnetic torque. The equations are shown in formulas (4) to (6):

[0188]

[0189] Where, is the estimated value of , The operation represents the multiplication of corresponding elements of two vectors. G1 and G2 are calculated as follows:

[0190]

[0191] Where, k is the proportionality coefficient;

[0192] Step 202: Based on the forward Euler discretization formula, discretize formulas (4) to (5) to obtain the stator flux prediction value at time k+1 in the model predictive control method as shown in formula (7), and the stator current prediction value as shown in formula (8):

[0193]

[0194] Among them, T s is the sampling period;

[0195] Step 203: Obtain the predicted speed value at time k+1 based on the electromagnetic torque and load torque at time k:

[0196]

[0197] Where, is the predicted speed value at time k+1, is the speed at time k, is the electromagnetic torque at time k.

[0198] Step 3: Design a cost function based on the stator flux and speed prediction values at time k+1 in the model predictive control algorithm obtained in step 2;

[0199] Step 4: Introduce the genetic algorithm into the model predictive direct speed control and adaptively adjust the weight factor in the cost function of step 3;

[0200] Step 5: Based on the traditional genetic algorithm described in step 4, the genetic algorithm is improved by introducing an adaptive crossover operator and an adaptive mutation operator, thereby finally achieving adaptive adjustment of the weight factor and controlling the induction motor.

[0201] Example 4

[0202] The induction motor weight factor adaptive model prediction direct speed control method of the present invention is combined with Figure 1 , specifically follow the steps below:

[0203] Step 1: Using the stator current and rotor flux as state variables, establish a mathematical model of the induction motor in a two-phase stationary coordinate system;

[0204] Step 2: Based on the mathematical model of the induction motor in the two-phase stationary coordinate system obtained in step 1, assuming that the current moment is moment k, the stator flux and speed at moment k+1 are predicted to obtain a stator flux prediction value and a speed prediction value;

[0205] In step 2, assuming that the current moment is k, the stator flux and current estimated by the full-order observer predict the stator flux at time k+1, and calculate the electromagnetic torque, specifically:

[0206] Step 201: Estimate the stator flux and current of the induction motor using a full-order observer based on the mathematical model of the induction motor, and calculate the electromagnetic torque. The equations are shown in formulas (4) to (6):

[0207]

[0208] Where, is the estimated value of , The operation represents the multiplication of corresponding elements of two vectors. G1 and G2 are calculated as follows:

[0209]

[0210] Where, k is the proportionality coefficient;

[0211] Step 202: Based on the forward Euler discretization formula, discretize formulas (4) to (5) to obtain the stator flux prediction value at time k+1 in the model predictive control method as shown in formula (7), and the stator current prediction value as shown in formula (8):

[0212]

[0213] Among them, T s is the sampling period;

[0214] Step 203: Obtain the predicted speed value at time k+1 based on the electromagnetic torque and load torque at time k:

[0215]

[0216] Where, is the predicted speed value at time k+1, is the speed at time k, is the electromagnetic torque at time k.

[0217] Step 3: Design a cost function based on the stator flux and speed prediction values at time k+1 in the model predictive control algorithm obtained in step 2;

[0218] Step 4: Introduce the genetic algorithm into the model predictive direct speed control and adaptively adjust the weight factor in the cost function of step 3;

[0219] Step 5: Based on the traditional genetic algorithm described in step 4, the genetic algorithm is improved by introducing an adaptive crossover operator and an adaptive mutation operator, thereby finally achieving adaptive adjustment of the weight factor and controlling the induction motor.

[0220] Example 5

[0221] The induction motor weight factor adaptive model prediction direct speed control method of the present invention is combined with Figure 1 , specifically follow the steps below:

[0222] Step 1: Using the stator current and rotor flux as state variables, establish a mathematical model of the induction motor in a two-phase stationary coordinate system;

[0223] In step 1, the induction motor is modeled to obtain a mathematical model of the induction motor in a two-phase stationary coordinate system, as follows:

[0224] The stator current and rotor flux are selected as state variables, and the mathematical model of the induction motor is expressed as a complex vector:

[0225]

[0226] in, x=[i s ψ r ] T , u=u s =[u sα u sβ ]T ,

[0227]

[0228] The stator flux is calculated from the stator current and the rotor flux:

[0229]

[0230] Induction motor electromagnetic torque T e for:

[0231]

[0232] Where u s is the voltage vector, i s is the stator current, i r is the rotor current, ψ s is the stator flux, ψ r is the rotor flux, T e is the electromagnetic torque, T l is the load torque, ω e is the synchronous speed of the motor, ω r is the rotor speed, R s is the stator resistance, R r is the rotor resistance, L s is the stator inductance, L r is the rotor inductance, L m is the mutual inductance, J is the moment of inertia, n p is the pole pair number.

[0233] Step 2: Based on the mathematical model of the induction motor in the two-phase stationary coordinate system obtained in step 1, assuming that the current moment is moment k, the stator flux and speed at moment k+1 are predicted to obtain a stator flux prediction value and a speed prediction value;

[0234] In step 2, assuming that the current moment is k, the stator flux and current estimated by the full-order observer predict the stator flux at time k+1, and calculate the electromagnetic torque, specifically:

[0235] Step 201: Estimate the stator flux and current of the induction motor using a full-order observer based on the mathematical model of the induction motor, and calculate the electromagnetic torque. The equations are shown in formulas (4) to (6):

[0236]

[0237] Where, is the estimated value of ●, The operation represents the multiplication of corresponding elements of two vectors. G1 and G2 are calculated as follows:

[0238]

[0239] Where, k is the proportionality coefficient;

[0240] Step 202: Based on the forward Euler discretization formula, discretize formulas (4) to (5) to obtain the stator flux prediction value at time k+1 in the model predictive control method as shown in formula (7), and the stator current prediction value as shown in formula (8):

[0241]

[0242] Among them, T s is the sampling period;

[0243] Step 203: Obtain the predicted speed value at time k+1 based on the electromagnetic torque and load torque at time k:

[0244]

[0245] Where, is the predicted speed value at time k+1, is the speed at time k, is the electromagnetic torque at time k.

[0246] Step 3: Design a cost function based on the stator flux and speed prediction values at time k+1 in the model predictive control algorithm obtained in step 2;

[0247] In step 3, the cost function is designed based on the predicted flux and speed at time k+1, specifically:

[0248]

[0249] Among them, λ is the weight factor;

[0250] Because digital control systems have a one-beat delay in practical applications, the voltage vector that should be applied at the current time k is not updated until the next time k+1. To eliminate the impact of this one-beat delay on control effectiveness, the controller compensates for this delay by determining the voltage vector at time k+1 one step ahead.

[0251] Step 4: Introduce the genetic algorithm into the model predictive direct speed control and adaptively adjust the weight factor in the cost function of step 3;

[0252] Step 5: Based on the traditional genetic algorithm described in step 4, the genetic algorithm is improved by introducing an adaptive crossover operator and an adaptive mutation operator, thereby finally achieving adaptive adjustment of the weight factor and controlling the induction motor.

[0253] Example 6

[0254] The induction motor weight factor adaptive model prediction direct speed control method of the present invention is combined with Figure 1 , specifically follow the steps below:

[0255] Step 1: Using the stator current and rotor flux as state variables, establish a mathematical model of the induction motor in a two-phase stationary coordinate system;

[0256] In step 1, the induction motor is modeled to obtain a mathematical model of the induction motor in a two-phase stationary coordinate system, as follows:

[0257] The stator current and rotor flux are selected as state variables, and the mathematical model of the induction motor is expressed as a complex vector:

[0258]

[0259] in, x=[i s ψ r ] T , u=u s =[u sα u sβ ] T ,

[0260] The stator flux is calculated from the stator current and the rotor flux:

[0261]

[0262] Induction motor electromagnetic torque T e for:

[0263]

[0264] Where u s is the voltage vector, i s is the stator current, i r is the rotor current, ψ s is the stator flux, ψ r is the rotor flux, T e is the electromagnetic torque, T l is the load torque, ω e is the synchronous speed of the motor, ω r is the rotor speed, R s is the stator resistance, R r is the rotor resistance, L s is the stator inductance, L r is the rotor inductance, L m is the mutual inductance, J is the moment of inertia, n p is the pole pair number.

[0265] Step 2: Based on the mathematical model of the induction motor in the two-phase stationary coordinate system obtained in step 1, assuming that the current moment is moment k, the stator flux and speed at moment k+1 are predicted to obtain a stator flux prediction value and a speed prediction value;

[0266] Step 3: Design a cost function based on the stator flux and speed prediction values at time k+1 in the model predictive control algorithm obtained in step 2;

[0267] Step 4: Introduce the genetic algorithm into the model predictive direct speed control and adaptively adjust the weight factor in the cost function of step 3;

[0268] Step 4 is as follows:

[0269] Step 401: Initialize the candidate set of weight factors:

[0270] Within the set weight factor variation range, N numbers are randomly generated as the initial population of the genetic algorithm;

[0271] Step 402: Calculate the fitness value F(λ i );

[0272] Step 402 is specifically as follows:

[0273] First, for each λ, the optimal voltage vector corresponding to λ is calculated by equation (10). Then, for each λ, the adaptation value F(λ i ):

[0274]

[0275] Where S A is the error surface,

[0276]

[0277] in, y is an assumed variable, * is the given value of y.

[0278] Step 403: Select an operation.

[0279] Step 403 is specifically as follows:

[0280] Selection is the process of selecting superior individuals from a population and eliminating inferior ones. It is based on fitness assessment. Individuals with greater fitness are more likely to be selected, and their "offspring" will have more children in the next generation. The selected individuals are then placed in a pairing pool. Common selection methods include the roulette wheel method, the best individual retention method, the expected value method, the ranked selection method, the competitive method, and the linear normalization method. Here, the next generation population is selected based on the roulette wheel algorithm. The specific operations are:

[0281] (1) First, calculate the selection probability of each λ:

[0282]

[0283] (2) Generate a random probability P between 0 and 1 s , select The first established i As an individual of the next generation group, N lambdas are selected as the next generation group.

[0284] Step 404: Cross-exchange operation.

[0285] Step 404 is specifically as follows:

[0286] The crossover operation produces offspring from two parents, reflecting the process of biological reproduction. The specific operations are:

[0287] (1) Encoding λ: Common encoding methods include binary encoding and real-valued encoding. Here, real-valued encoding is used, and the value of λ ranges from 1 to 99. That is, the ones and tens digits of each λ are regarded as "chromosomes";

[0288] (2) First, randomly select two λ as parents, and then generate a random crossover probability P c , if it is greater than the given crossover probability The units digit of one parent is exchanged with the tens digit of the other parent, and the two new individuals obtained replace the parents as individuals in the group.

[0289] Step 405: Mutation operation:

[0290] Mutation operation is a manifestation of biodiversity. It generally occurs with a small probability and plays an important role in the genetic process. The specific operation is as follows:

[0291] First, randomly generate a mutation probability P of 0 to 1 m , if it is greater than the given mutation probability, that is Then randomly select an individual and generate a new λ to replace the individual;

[0292] Step 406: Selection of optimal weight factor and optimal switching vector:

[0293] The group is iterated according to steps 401 to 405. When the number of iterations reaches a given value or the number of identical individuals in the group reaches 90% of the total, the mode of the group is selected as the optimal weight factor and the corresponding optimal switching vector is selected.

[0294] Step 5: Based on the traditional genetic algorithm described in step 4, the genetic algorithm is improved by introducing an adaptive crossover operator and an adaptive mutation operator, thereby finally achieving adaptive adjustment of the weight factor and controlling the induction motor.

[0295] In step 5, in order to speed up the convergence of the genetic algorithm, an adaptive crossover operator is introduced to replace the traditional fixed crossover operator, specifically:

[0296] (1) If the number of modes in the current group n satisfies n>0.5N, let the crossover probability

[0297] (2) When n>0.5N, perform the following cross-interchange operation:

[0298] Choose two parents at random, if the random probability Then only the units digits of the two parents are exchanged, and the tens digit remains unchanged.

[0299] The present invention discloses a method for direct speed control of an induction motor using a weight factor adaptive model prediction. Aiming at the problem that the weight factor of a traditional model prediction direct speed control algorithm is difficult to adjust, an optimization algorithm of an improved genetic algorithm is adopted to adaptively adjust the weight factor, and an adaptive crossover operator and a mutation operator are introduced to accelerate the convergence speed, reduce the amount of calculation, avoid premature phenomenon, improve the practicability of the model prediction direct speed control, and improve the control performance of the motor.

Claims

1. A direct speed control method based on an adaptive model predictive method for an induction motor with weight factors, characterized in that: Please follow the steps below to implement it: Step 1: Using the stator current and rotor flux as state variables, establish a mathematical model of the induction motor in a two-phase stationary coordinate system; Step 2: Based on the mathematical model of the induction motor in the two-phase stationary coordinate system obtained in step 1, assuming that the current moment is moment k, the stator flux and speed at moment k+1 are predicted to obtain a stator flux prediction value and a speed prediction value; Step 3: Design a cost function based on the stator flux and speed prediction values at time k+1 in the model predictive control algorithm obtained in step 2; Step 4: Introduce the genetic algorithm into the model predictive direct speed control and adaptively adjust the weight factor in the cost function of step 3; Step 5: Based on the traditional genetic algorithm described in step 4, the genetic algorithm is improved by introducing an adaptive crossover operator and an adaptive mutation operator, thereby finally achieving adaptive adjustment of the weight factor and controlling the induction motor.

2. The induction motor weight factor adaptive model predictive direct speed control method according to claim 1, characterized in that: In step 1, the induction motor is modeled to obtain a mathematical model of the induction motor in a two-phase stationary coordinate system, which is as follows: The stator current and rotor flux are selected as state variables, and the mathematical model of the induction motor is expressed as a complex vector: Among them, x=[i s ψ r ] T ,u=u s =[in sα in sβ ] T , The stator flux is calculated from the stator current and the rotor flux: Induction motor electromagnetic torque T e for: Where u s is the voltage vector, i s is the stator current, i r is the rotor current, ψ s is the stator flux, ψ r is the rotor flux, T e is the electromagnetic torque, T l is the load torque, ω e is the synchronous speed of the motor, ω r is the rotor speed, R s is the stator resistance, R r is the rotor resistance, L s is the stator inductance, L r is the rotor inductance, L m is the mutual inductance, J is the moment of inertia, n p is the pole pair number.

3. The induction motor weight factor adaptive model predictive direct speed control method according to claim 2, characterized in that: In step 2, assuming that the current moment is k, the stator flux and current estimated by the full-order observer predict the stator flux at k+1, and calculate the electromagnetic torque, specifically: Step 201: Estimate the stator flux and current of the induction motor using a full-order observer based on the mathematical model of the induction motor, and calculate the electromagnetic torque. The equations are shown in formulas (4) to (6): Where, is the estimated value of ●, The operation represents the multiplication of corresponding elements of two vectors. G1 and G2 are calculated as follows: Where, k is the proportionality coefficient; Step 202: Based on the forward Euler discretization formula, discretize formulas (4) to (5) to obtain the stator flux prediction value at time k+1 in the model predictive control method as shown in formula (7), and the stator current prediction value as shown in formula (8): Among them, T s is the sampling period; Step 203: Obtain the predicted speed value at time k+1 based on the electromagnetic torque and load torque at time k: Where, is the predicted speed value at time k+1, is the speed at time k, is the electromagnetic torque at time k.

4. The induction motor weight factor adaptive model predictive direct speed control method according to claim 3, characterized in that: In step 3, a cost function is designed based on the predicted flux and speed at time k+1, specifically: Where λ is the weight factor; the controller compensates for the delay by determining the voltage vector at time k+1 one step ahead.

5. The induction motor weight factor adaptive model predictive direct speed control method according to claim 4, characterized in that: The step 4 is specifically as follows: Step 401: Initialize the candidate set of weight factors: Within the set weight factor variation range, N numbers are randomly generated as the initial population of the genetic algorithm; Step 402: Calculate the fitness value F(λ i ); Step 403: Select an operation. Step 404: Cross-exchange operation.

6. The induction motor weight factor adaptive model predictive direct speed control method according to claim 5, characterized in that: The step 402 is specifically as follows: First, for each λ, the optimal voltage vector corresponding to λ is calculated by equation (10). Then, for each λ, the adaptation value F(λ i ): Where S A is the error surface, in, y is an assumed variable, * is the given value of y.

7. The induction motor weight factor adaptive model predictive direct speed control method according to claim 6, characterized in that: The step 403 is specifically as follows: The next generation population is selected according to the roulette wheel algorithm. The specific operations are: (1) First, calculate the selection probability of each λ: (2) Generate a random probability P between 0 and 1 s , select The first established i As an individual of the next generation group, N lambdas are selected as the next generation group.

8. The induction motor weight factor adaptive model predictive direct speed control method according to claim 7, characterized in that: The step 404 is specifically as follows: The crossover operation generates offspring from two parents. The specific operation is: (1) Encoding λ: Using real-valued encoding, λ ranges from 1 to 99, that is, the ones and tens digits of each λ are regarded as "chromosomes"; (2) First, randomly select two λ as parents, and then generate a random crossover probability P c , if it is greater than the given crossover probability Then the units digit of one parent is exchanged with the tens digit of the other parent, and the two new individuals obtained replace the parents as individuals in the group; Step 405: Mutation operation: The specific operations are as follows: First, randomly generate a mutation probability P of 0 to 1 m , if it is greater than the given mutation probability, that is Then randomly select an individual and generate a new λ to replace the individual; Step 406: Selection of optimal weight factor and optimal switching vector: The group is iterated according to steps 401 to 405. When the number of iterations reaches a given value or the number of identical individuals in the group reaches 90% of the total, the mode of the group is selected as the optimal weight factor and the corresponding optimal switching vector is selected.

9. The induction motor weight factor adaptive model predictive direct speed control method according to claim 8, characterized in that: In step 5, an adaptive crossover operator is introduced to replace the traditional fixed crossover operator, specifically: (1) If the number of modes in the current group n satisfies n>0.5N, let the crossover probability (2) When n>0.5N, perform the following cross-interchange operation: Choose two parents at random, if the random probability Then only the units digit of the two parents is exchanged, and the tens digit remains unchanged.

10. The induction motor weight factor adaptive model predictive direct speed control method according to claim 9, characterized in that: In step 5, an adaptive mutation operator is introduced, as follows: Where, is a fixed mutation probability, is the adaptive mutation probability, F max is the maximum value of the fitness function in the t generation population, ASD t is the mean square error of the fitness value of the t generation group Among them, F v is the average fitness value of the t generation group Therefore, introducing the adaptive crossover operator and mutation operator into the genetic algorithm can avoid premature phenomenon, speed up the convergence of the algorithm, and ultimately achieve adaptive adjustment of the weight factor; For a two-level voltage source inverter, there are eight basic switching states. By improving the genetic algorithm, the optimal weight factor is obtained, and the optimal switching vector at the next moment is also obtained.

Images