Weight coefficient adaptive optimization synchronous reluctance motor model prediction control method

Abstract

The invention discloses a synchronous reluctance motor model predictive control method with a self-adaptive weight coefficient optimization function, and belongs to the field of synchronous reluctance motor model predictive control. The method is characterized by comprising the following steps: step a, establishing a synchronous reluctance motor prediction equation; b, establishing a d-q axis inductance model of the synchronous reluctance motor; c, performing three-vector model predictive current control; d, designing a value function; e, executing weight coefficient optimization by adopting a cuckoo algorithm; and step f, executing multi-objective optimization. According to the weight coefficient adaptive optimization synchronous reluctance motor model prediction control method, a minimum value function is designed, constraint conditions are met, and dynamic characteristics under different working conditions are considered when the value function of model prediction control is designed. And the weight coefficient is dynamically updated to adapt to the change of the system under different working conditions, so that the control performance is improved.

Description

Technical Field

[0001] The invention discloses a synchronous reluctance motor model predictive control method with adaptive weight coefficient optimization, belonging to the field of synchronous reluctance motor model predictive control. Background Art

[0002] Model predictive control of synchronous reluctance motors is an advanced control strategy that establishes an accurate mathematical model of the motor, uses the model to predict future states, and uses optimization algorithms to find the optimal control strategy to achieve precise control of state variables such as motor current, speed or position. This control method takes into account the nonlinear characteristics and parameter changes of the motor, and aims to improve the accuracy and robustness of control while balancing the real-time requirements and computational complexity of the controller.

[0003] The weight coefficient is an important factor in the predictive control of synchronous reluctance motor model. The setting of the weight coefficient is essentially a parameter optimization problem. To select a suitable weight coefficient:

[0004] Reference 1: Yan Liming, Guo Xin, Zhao Dongdong. Predictive torque control of permanent magnet motor based on a new weight analytical method [J]. Acta Aeronautica Sinica, 2022, 43(12): 486-497. Based on the intrinsic relationship between the electromagnetic torque, stator flux and stator current of the motor system, the analytical expression of the weight coefficient in the cost function is derived.

[0005] Reference 2: Yuan Qingqing, Bu Fan, Xie Xiaotong, et al. Improved model predictive voltage control of NPC type three-level dual three-phase PMSM based on vector classification [J]. Proceedings of the CSEE, 2024. Through the mathematical model of the motor and the characteristic analysis of the inverter, the weight coefficient limit that satisfies the midpoint potential balance is calculated to meet the harmonic suppression and midpoint potential balance requirements of the motor current.

[0006] However, although the above analytical method can directly derive the relationship between motor parameters and weight coefficients, it has limitations such as high model accuracy requirements and the need to set ideal conditions. To solve the above problems:

[0007] Reference 3: Sakamoto Y, Xu Yihao, Wang Bingnan, et al. Multi-Objective Motor Design Optimization with Physics-Assisted Neural Network Model [C]. IEEE International Electric Machines & Drives Conference, 2023: 1-7. The multi-objective average ranking method is used to convert the value function calculation into ranking value ranking, thereby avoiding the selection of weight coefficients.

[0008] Reference 4: Li Yaohua, Qin Hui, Su Jinshi. Fuzzy adaptive variable switching number weight coefficient model predictive torque control of permanent magnet synchronous motor [J]. Journal of Electric Machines and Control, 2021, 25(02): 102-112. Fuzzy theory is introduced into model predictive control, and the weight coefficient value in the value function is adjusted in real time through the motor system state information. However, the calculation steps of the above method are too cumbersome, which greatly increases the online computing burden of the model predictive control algorithm.

[0009] At the same time, a technical solution is disclosed in a Chinese invention patent with application number 201611005682.0, application date November 15, 2016, and patent name "Model Predictive Control Method for Synchronous Reluctance Motor Driven by Two-Stage Matrix Converter". In this technical solution, a model prediction value function is established based on torque, flux, and current tracking error. Although it performs well in motor performance, this will cause the switching frequency of the motor inverter to be too high. Excessive switching frequency causes the switching elements in the inverter to switch frequently, increasing switching losses, which causes the equipment to heat up and affects the system's heat dissipation capacity and component life. At high frequencies, switching losses increase significantly, which may lead to a decrease in overall energy efficiency, especially under low or light load conditions, where efficiency losses are more obvious.

[0010] A Chinese invention patent with application number 202410646961.3 and application date May 23, 2024, and patent name “New Flux Observer Model Prediction Speed ​​Control Method for Synchronous Reluctance Motor” discloses a technical solution. In this technical solution, in the weight coefficient setting of the value function for model prediction, there is a lack of a systematic method for the appropriate weight coefficient setting, and the weight coefficient is usually set statically, while there may be dynamic changes during the operation of the system. Fixed weights cannot adapt to changes in the system state, affecting the control effect. In multivariable systems, the nonlinear and coupling characteristics of the system make it difficult to achieve the expected control performance through simple weight adjustments alone. There are contradictions between different control objectives, and it is difficult to balance the various objectives through a single weight setting, resulting in performance degradation. Once the weight coefficient is set, subsequent verification and adjustment may require a long time and a large amount of data support, which increases the difficulty of implementation.

[0011] In summary, the following defects are generally present in the synchronous reluctance motor model predictive control in the prior art:

[0012] (1) Nonlinear problem: Under different working conditions, the dynamic characteristics of the system will change. The fixed weight coefficient cannot adapt to all working conditions, resulting in a decrease in control performance.

[0013] (2) Difficulty in multi-objective optimization: When multiple performance indicators (such as torque pulsation and inverter switching frequency) need to be optimized simultaneously, the setting of weight coefficients may cause certain objectives to be ignored or conflict, making it difficult to achieve global optimization.

[0014] (3) Lack of systematic methods: There is usually a lack of systematic methods for adjusting weight coefficients, resulting in a lack of scientific basis and standardization in the adjustment process. The selection of weight coefficients usually relies on experience and experiments, and the adjustment process requires multiple experiments to find the best combination, which is time-consuming and not necessarily stable. Summary of the invention

[0015] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a synchronous reluctance motor model predictive control method that designs a minimized value function while satisfying constraints, taking into account the dynamic characteristics under different working conditions when designing the value function of the model predictive control, and dynamically updating the weight coefficient to adapt to the changes of the system under different working conditions, thereby improving the control performance.

[0016] The technical solution adopted by the present invention to solve the technical problem is: the weight coefficient adaptively optimized synchronous reluctance motor model predictive control method is characterized by comprising the following steps:

[0017] Step a, establishing a synchronous reluctance motor prediction equation;

[0018] Step b, establishing a dq-axis inductance model of a synchronous reluctance motor;

[0019] Step c, performing three-vector model predictive current control;

[0020] Step d, value function design;

[0021] Step e, using cuckoo algorithm to perform weight coefficient optimization;

[0022] Step f, perform multi-objective optimization.

[0023] Preferably, in step d, the expression of the value function is:

[0024] e(i)=e1(i)+λe2(i)+e3(i)

[0025]

[0026] e2=(S a (k+1)-S a (k))+

[0027] (S b (k+1)-S b (k))+(S c (k+1)-Sc (k))

[0028]

[0029] Among them, λ is the weight coefficient, F is the sum of the changes between the switch states at time k and time k+1, S a , S b , S c is the output state of the switch quantity of the two-level voltage type three-phase inverter. At present, the setting of λ is generally obtained by engineering experience data, but with the continuous development of motor control and the complex working conditions in the actual operation of the motor, it is necessary to design a more accurate weight coefficient. T p is the torque ripple, T plim Sets an upper limit for torque ripple.

[0030] Preferably, the torque ripple T p The expression is:

[0031]

[0032] Among them, max(T p ) is the torque ripple T p The maximum value, min(T p ) is the torque ripple T p The minimum value, T avg is the average torque.

[0033] Preferably, step f comprises the following steps:

[0034] Step f-1, initialize the number of bird nests, step size, number of iterations, and probability of bird nests being found of the cuckoo algorithm;

[0035] Step f-2, calculating the fitness value of the bird's nest;

[0036] Step f-3, determine whether the termination number has been reached, if so, execute step f-7, if not, execute step f-4;

[0037] Step f-4, record and update the Pareto optimal solution set;

[0038] Step f-5, updating the step size factor and discovery probability of the cuckoo algorithm;

[0039] Step f-6, record the weight coefficient value under the best solution, and output the motor torque and inverter switching times at this time. Return to step 1006-3.

[0040] Step f-7, output the final solution set.

[0041] Preferably, in step f-5, the updating formula of the step size factor is:

[0042]

[0043] Among them, α(t) is the value of α at the tth iteration, T max is the total number of iterations, α max and α min are the upper and lower limits of the step scale factor.

[0044] Preferably, in step f-5, the update formula of the discovery probability is:

[0045]

[0046] in, is the value of the probability of discovery at the tth iteration, g p is the probability learning factor for discovery, and R is the population improvement rate.

[0047] Preferably, in step f-2, the updating formula of the bird's nest fitness value is:

[0048] fitness=βT p +F

[0049] Among them, T p is the torque pulsation, F is the sum of the changes between the switch states at time k and time k+1, and β is 100000.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] In this weight coefficient adaptive optimization synchronous reluctance motor model predictive control method, a minimized value function is designed while satisfying constraints. When designing the value function of the model predictive control, the dynamic characteristics under different working conditions are considered. The weight coefficient is dynamically updated to adapt to the changes of the system under different working conditions, thereby improving the control performance.

[0052] In this synchronous reluctance motor model predictive control method with adaptive weight coefficient optimization, current pulsation and inverter switching frequency are taken as two main objectives for multi-objective optimization. Through dynamic adjustment of weight coefficients, it is ensured that each objective is balanced during the optimization process, avoiding any objective being ignored or conflicting, and striving to achieve global optimality.

[0053] In this method of predictive control of synchronous reluctance motor with adaptive weight coefficient optimization, a standardized process for weight coefficient setting is clearly formulated, including initial range setting, iteration criteria and convergence conditions. Combined with the adaptive cuckoo algorithm, a systematic optimization framework is formed to ensure that the whole process has a scientific basis and operational specifications. The adaptive cuckoo algorithm is used to reduce the number of manual experiments through intelligent algorithm search rather than relying solely on experience. This method automatically adjusts the weight coefficient and can find a better solution without a large number of experiments.

[0054] In this model predictive control method for synchronous reluctance motor with adaptive weight coefficient optimization, an improved cuckoo algorithm is proposed. Through adaptive step size and discovery probability, the algorithm's search capability is increased to avoid falling into local optimum. In each control cycle, the optimization problem is solved in real time according to the current state, and the optimal control input is obtained through the numerical optimization algorithm. It can effectively improve the control performance of the synchronous reluctance motor, adapt to complex working conditions, and achieve efficient and accurate control effects.

[0055] In this weight coefficient adaptively optimized synchronous reluctance motor model predictive control method, the state space model of the system is established according to the electrical and mechanical characteristics of the synchronous reluctance motor, and the nonlinear characteristics of the synchronous reluctance motor, especially the influence of magnetic saturation on the performance, are considered. The inductance-current query table is established through ANSYS software, and the inductance parameters are identified; in each control cycle, the optimization problem is solved in real time according to the current state, and the optimal control input is obtained through the numerical optimization algorithm, which can effectively improve the control performance of the synchronous reluctance motor, adapt to complex working conditions, and achieve efficient and accurate control effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 Flowchart of the synchronous reluctance motor model predictive control method with adaptive optimization of weight coefficients.

[0057] Figure 2 Principle equation diagram of the improved model predictive control system for synchronous reluctance motor model predictive control with adaptive optimization of weight coefficients.

[0058] Figure 3 Weight coefficient optimization flow chart of synchronous reluctance motor model predictive control method with adaptive optimization of weight coefficients.

[0059] Figure 4 The waveform diagram of the fitness value change of the synchronous reluctance motor model predictive control method with adaptive optimization of weight coefficients.

[0060] Figure 5 The waveform diagram of weight coefficient change of synchronous reluctance motor model predictive control method with adaptive optimization of weight coefficient.

[0061] Figure 6Pareto solution waveform of synchronous reluctance motor model predictive control method with adaptive optimization of weight coefficients.

[0062] Figure 7 This is the electromagnetic torque waveform before optimization.

[0063] Figure 8 This is the electromagnetic torque diagram after cuckoo algorithm optimization.

[0064] Fig. 9 Electromagnetic torque diagram after optimization for the improved cuckoo algorithm. DETAILED DESCRIPTION

[0065] Figures 1 to 9 The best embodiment of the present invention is shown below in conjunction with the attached Figures 1 to 9 The present invention is further described.

[0066] like Figure 1 As shown, the weight coefficient adaptively optimized synchronous reluctance motor model predictive control method (hereinafter referred to as the predictive control method) includes the following steps:

[0067] Step 1001, establishing a synchronous reluctance motor prediction equation;

[0068] Assuming that the three-phase stator windings of the synchronous reluctance motor are symmetrically distributed, the axes differ by 120 degrees, the stator slot effect is ignored, the rotor surface is smooth and the reluctance is continuous, then the voltage equation in the two-phase rotating coordinate system is:

[0069]

[0070] Among them, u d ,u q are the motor d-axis and q-axis voltages respectively; i d ,i q are the motor d-axis and q-axis currents, R s is the winding resistance, ω e is the electrical angular velocity, Ψ d ,Ψ q They are the d-axis and q-axis magnetic flux respectively.

[0071] The magnetic flux equation of the synchronous reluctance motor can be expressed as:

[0072]

[0073] Among them, L d , L q are the motor d-axis and q-axis inductances respectively. The torque equation of the synchronous reluctance motor is:

[0074]

[0075] Among them, T eis the electromagnetic torque, and p is the number of motor pole pairs. According to equations (1) to (3), the differential equations of the motor's d and q axis currents are shown in equation (4):

[0076]

[0077] The continuous system is discretized by the forward Euler algorithm, and equation (5) is obtained:

[0078]

[0079] Among them, T s is the sampling time. According to equations (1)-(5), the flux, current and torque prediction equations of the synchronous reluctance motor are established. The flux prediction equation is:

[0080]

[0081] Among them, d (k+1) is the predicted value of magnetic flux on the d-axis at time k+1, Ψ q (k+1) is the predicted value of magnetic flux on the q axis at time k+1, Ψ d (k) is the flux value of d-axis at time k, Ψ q (k) is the magnetic flux value of the q axis at time k, u d (k) is the voltage value of the d-axis at time k, u q (k) is the voltage value of the q axis at time k, ω e (k) is the electrical angular velocity at time k, i d (k) is the current value of the d-axis at time k, i q (k) is the current value of the q axis at time k. The current prediction equation is:

[0082]

[0083] Among them, i d (k+1) is the predicted current value of the d-axis at time k+1, i q (k+1) is the predicted current value of the q-axis at time k+1. The torque prediction equation is:

[0084]

[0085] Step 1002, establishing a dq-axis inductance model of a synchronous reluctance motor;

[0086] As the stator current changes, the synchronous reluctance motor will have a serious magnetic field saturation effect, causing the inductance parameters to change. The inductance has an important influence on the torque characteristics of the motor. Therefore, accurately identifying the inductance parameters of the synchronous reluctance motor can optimize the control performance of the motor. In order to more accurately simulate the operating effect of the synchronous reluctance motor under actual working conditions, a two-dimensional inductance-current query table is constructed in this control method. By using the linear interpolation method to update the inductance data in the synchronous reluctance motor in real time, the accuracy of the motor parameters is greatly improved, and more accurate model parameters can be provided for subsequent control to achieve better control effects. In order to take into account both the torque calculation accuracy and real-time performance, a two-dimensional table data model of the dq-axis inductance is established. The bilinear interpolation method is used in each cycle, and the d and q-axis currents are used to calculate the inductance value in real time, as follows:

[0087]

[0088] Where: Q 11 =(x1,y1),Q 12 =(x1,y2),Q 21 =(x2,y1),Q 22 =(x2,y2) represents a known data point in two-dimensional space.

[0089] Step 1003, performing three-vector model predictive current control;

[0090] Since the predicted value at the next sampling moment is equal to the given value in deadbeat control, the d and q axis current prediction formulas can be rewritten as:

[0091]

[0092] in, is the given value of the d-axis current, is the given value of the q-axis current. Considering the deadbeat tracking of the d-axis and q-axis currents, according to equations (1) and (2), the d-axis and q-axis current slopes S when zero voltage is applied are oda ,S oda The calculation formulas are:

[0093]

[0094] From equation (11), we can get two adjacent effective voltage vectors u a ,u b When in action, the slopes of the d and q axis currents are:

[0095]

[0096] where u da ,u qa ,u db ,u dbfor u a ,u b For the voltage components of the d and q axes, the voltage action time formula can be obtained by combining the above equations (9) to (12):

[0097]

[0098] T s =t a +t b +t z

[0099] S=s oq0 s odb +s oqa s od0 +s oqb s oda -s oqa s odb -s oqb s od0 -s oq0 s oda

[0100] Combination Figure 2 , use the sensor to obtain the current value at the current moment and calculate the current slope. Then select the basic voltage vector, calculate the action time of the three basic voltage vectors respectively to obtain the d and q axis voltage components of the 6 expected voltage vectors. Substitute the calculated d and q axis voltage components of the 6 expected voltage vectors into equations (9)-(13) to obtain the d and q axis current prediction values ​​corresponding to the 6 expected voltage vectors. The core of the traditional MPCC method is to use the discrete mathematical model of the motor to predict the current response of the inverter under different switch combinations. The voltage vector that minimizes the error between the predicted current and the target current is selected through the value function as the optimal control strategy.

[0101] Step 1004, designing a value function;

[0102] In order to simultaneously meet the requirements of high motor control accuracy and low switching loss and ensure motor performance, a new model is designed in this control method to predict the value function of current control as shown in formula (14):

[0103]

[0104] Among them, λ is the weight coefficient, F is the sum of the changes between the switch states at time k and time k+1, S a , S b , S c is the output state of the switch quantity of the two-level voltage type three-phase inverter. At present, the setting of λ is generally obtained by engineering experience data, but with the continuous development of motor control and the complex working conditions in the actual operation of the motor, it is necessary to design a more accurate weight coefficient. Tp is the torque ripple, T plim The value function is divided into three items: the first item is the current error control item, the second item is the switching frequency control item, and the third item is the maximum torque pulsation upper limit. In this control method, the upper limit is set to 15%, where the torque pulsation is defined as shown in formula (15):

[0105]

[0106] Among them, max(T p ) is the torque ripple T p The maximum value, min(T p ) is the torque ripple T p The minimum value, T avg for

[0107] Step 1005, performing weight coefficient optimization;

[0108] In this control method, the cuckoo search algorithm is used. The cuckoo search algorithm is a heuristic optimization algorithm based on foraging behavior, as shown in (16):

[0109]

[0110] in, is the position of the i-th nest in the t-th iteration, α is the step size scaling factor, s is the step size, and p a is the discovery probability, H is the unit step function, and ε is a random number extracted from a uniform distribution. Although the cuckoo algorithm has been widely used, it still has the problem of slow convergence and easy to fall into local optimality. α is a key parameter in the search process, which affects the movement range of the algorithm in the search space. A larger α helps to explore new solutions extensively, while a smaller α focuses on detailed searches near known excellent solutions. To optimize the effect, the improved algorithm will dynamically adjust α. On the other hand, p a Determines whether the current solution is replaced by a new solution, a higher p a Promote diversity. You can set a higher p in the early stage a To enhance exploration, lower p later a Focus on high-quality solutions to avoid falling into local optimality too early. Dynamic parameters are divided into deterministic dynamics and adaptive dynamics. The former is determined before running, and the latter evolves with the population.

[0111] In this control method, α is processed dynamically and p is aAs an adaptive dynamic process, to optimize the characteristics of the offspring. Therefore, the step size factor and the discovery probability in the cuckoo algorithm are optimized. The step size used in the early stage of the search is used to expand the search range and cover a wider solution space. As the iteration proceeds, the algorithm increases the accuracy of the local search by reducing the step size. The optimization of the step size factor is shown in formula (17):

[0112]

[0113] Among them, α(t) is the value of α at the tth iteration, T max is the total number of iterations, α max and α min is the upper and lower limits of the step size scale factor. Define R as the population improvement rate. When the improvement rate exceeds 0.2, the search space is smoother and the algorithm is easier to find a better solution.

[0114] To improve efficiency, the step size can be appropriately increased to expand the search range. When the improvement rate is lower than 0.2, the search space is complex and the probability of finding a better solution is reduced. The step size should be reduced to improve the search accuracy. When the improvement rate is 0.2, the step size setting is ideal and no adjustment is required. Therefore, the discovery probability is adjusted as shown in formula (18):

[0115]

[0116] in, is the value of the probability of discovery at the tth iteration, g p is the discovery probability learning factor. Through the above optimization of step size and discovery probability, the search ability of the cuckoo algorithm can be significantly increased to avoid falling into the local optimum. The specific algorithm parameters are shown in Table 1:

[0117] Table 1 Algorithm parameters

[0118]

[0119] Step 1006, performing multi-objective optimization;

[0120] like Figure 3 As shown, multi-objective optimization includes the following steps:

[0121] Step 1006-1, initialization;

[0122] Initialize the number of bird nests, step size, number of iterations, and the probability of a bird nest being found.

[0123] Step 1006-2, calculating the fitness value of the bird's nest;

[0124] Call the simulation model parameters, normalize the number of switching times and torque ripple, and calculate the fitness value of each individual, which is the combination of the number of switching times and torque ripple per unit time in this control method. Record the position x of the current best individual best That is, the weight coefficient λ and the fitness value of the position are the values ​​corresponding to formula (19), where β is taken as 100000 in this control method to eliminate the influence of the difference in the order of magnitude of the two indicators.

[0125] fitness=βT p +F (19)

[0126] Among them, T p is the torque pulsation, and F is the sum of the changes between the switch states at time k and time k+1.

[0127] Step 1006-3, determine whether the termination number has been reached;

[0128] Determine whether the termination number has been reached. If so, execute step 1006-7. If not, execute step 1006-4. The termination number in this control method is 100 times.

[0129] Step 1006-4, record and update the Pareto optimal solution set;

[0130] Reducing the switching frequency of the inverter will lead to increased torque ripple, because lower switching frequency means rougher waveforms of current and voltage, which increases torque fluctuations. Conversely, increasing the switching frequency can reduce torque ripple, but it will also increase the loss and cost of the inverter. However, in practical applications, there are still some dominated solutions, such as: there is λ a ,λ b , when T pλa <T pλb And F λa <F λb When the weight coefficient is a, that is, the torque pulsation and switching frequency are both smaller than the weight coefficient b. In this control method, the weight coefficient b is discarded through Pareto optimality.

[0131] For each individual, a new solution is generated according to formula (19). The motor Simulink model is called once in each iteration. If the new solution is better than the previous solution, the previous solution is replaced and the number of iterations is increased.

[0132] Step 1006-5, updating the step size factor and discovery probability of the cuckoo algorithm;

[0133] Calculate the individual improvement rate of this iteration, and calculate the new step size and discovery probability according to equations (17)-(18).

[0134] Step 1006-6, update the bird's nest position and fitness function

[0135] The weight coefficient value under the best solution is recorded, and the motor torque and inverter switching times at this time are output. Return to step 1006-3.

[0136] Step 1006-7, output the final solution set.

[0137] The following example verifies the above-mentioned synchronous reluctance motor model predictive control method with adaptive optimization of weight coefficients:

[0138] The simulation model of the synchronous reluctance motor is built using the Simulink platform of MATLAB, and its model predictive control system is established. The motor parameters are shown in Table 2, and the speed is 1000r / min, the load torque is 10Nm, and the three-vector model predictive control of the synchronous reluctance motor based on the combination of Pareto optimality and improved cuckoo algorithm is modeled to verify that the control algorithm proposed in this patent significantly reduces the switching frequency while maintaining relatively superior control performance compared to the algorithm before the improvement.

[0139] Table 2 Prototype parameters

[0140]

[0141] After introducing the switching frequency control item, the contradiction between the switching frequency and the torque pulsation can be adaptively adjusted. The switching frequency of the system can also be reduced without significantly increasing the torque pulsation. This is an advantage that the traditional three-vector model predictive control does not have. Figures 4-5 This is the optimization process diagram. It can be seen that the cuckoo algorithm after optimizing the step size and probability is easier to avoid falling into the local optimum. Under the same number of iterations, a lower objective function value and a faster convergence speed can be obtained. This is because in the early stage of the algorithm, smaller α and p a It helps to explore the search space extensively and avoid falling into the local optimum. As the iteration proceeds, the fitness value tends to be stable and the population improvement rate is relatively low. Relatively small α and p a Making the search more focused helps to find a better solution. The specific results are shown in Table 3.

[0142] Table 3 Simulation results

[0143]

[0144] Figure 6 is the Pareto optimal solution set. Through this graph, the lowest motor inverter switching frequency can be selected under different torque ripple requirements. Figure 7 This is the motor torque waveform before optimization. Figure 8 This is the motor torque waveform after cuckoo algorithm optimization. Fig. 9Motor torque waveform after optimization to improve the cuckoo algorithm.

[0145] In this control method, an evaluation index is designed based on formula (19). The lower the index value, the better the overall performance under a specific weight coefficient. Specifically, when λ is increased, the weight of the e2 item is increased, and the inverter switching frequency is reduced, but this will affect the motor control performance and increase the current and torque pulsation; while reducing λ will increase the weight of the e1 item, reduce the torque pulsation, but increase the switching frequency.

[0146] according to Figures 4 to 9 The results in Table 3 show that after optimization by the cuckoo algorithm, the torque ripple of the motor increased from 7.294% to 8.715%, while the number of switches within 1 second decreased from 11598 to 9511; after optimization by the improved cuckoo algorithm, the torque ripple increased from 7.294% to 13.045%, but the number of switches within 1 second was further reduced to 4917. Both methods significantly reduce the number of inverter switches per unit time while maintaining low torque ripple. In addition, the evaluation index of the improved algorithm is lower, proving its superior comprehensive performance. Compared with the original cuckoo algorithm, which requires 87 iterations to converge, the improved algorithm only requires 43 iterations and the fitness function value at convergence is also lower than that of the original algorithm, indicating that the improved algorithm has faster convergence speed and higher convergence accuracy.

[0147] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

Claims

1. A synchronous reluctance motor model predictive control method with adaptive weight coefficient optimization, characterized in that: The steps include: Step a, establishing a synchronous reluctance motor prediction equation; Step b, establishing a dq-axis inductance model of a synchronous reluctance motor; Step c, performing three-vector model predictive current control; Step d, value function design; Step e, using cuckoo algorithm to perform weight coefficient optimization; Step f, perform multi-objective optimization.

2. The synchronous reluctance motor model predictive control method with adaptive optimization of weight coefficients according to claim 1 is characterized in that: In step d, the value function is expressed as: e(i)=e1(i)+λe2(i)+e3(i) e2=(S a (k+1)-S a (k))+ (S b (k+1)-S b (k))+(S c (k+1)-S c (k)) Among them, λ is the weight coefficient, F is the sum of the changes between the switch states at time k and time k+1, S a , S b , S c is the output state of the switch quantity of the two-level voltage type three-phase inverter. At present, the setting of λ is generally obtained by engineering experience data, but with the continuous development of motor control and the complex working conditions in the actual operation of the motor, it is necessary to design a more accurate weight coefficient. T p is the torque ripple, T plim Sets an upper limit for torque ripple.

3. The synchronous reluctance motor model predictive control method with adaptive weight coefficient optimization according to claim 2 is characterized in that: Torque ripple T p The expression is: Among them, max(T p ) is the torque ripple T p The maximum value, min(T p ) is the torque ripple T p The minimum value, T avg is the average torque.

4. The synchronous reluctance motor model predictive control method with adaptive weight coefficient optimization according to claim 1 is characterized in that: Step f comprises the following steps: Step f-1, initialize the number of bird nests, step size, number of iterations, and probability of bird nests being found of the cuckoo algorithm; Step f-2, calculating the fitness value of the bird's nest; Step f-3, determine whether the termination number has been reached, if so, execute step f-7, if not, execute step f-4; Step f-4, record and update the Pareto optimal solution set; Step f-5, updating the step size factor and discovery probability of the cuckoo algorithm; Step f-6, record the weight coefficient value under the best solution, and output the motor torque and inverter switching times at this time. Return to step 1006-3. Step f-7, output the final solution set.

5. The synchronous reluctance motor model predictive control method with adaptive optimization of weight coefficients according to claim 4 is characterized in that: In step f-5, the update formula of the step factor is: Among them, α(t) is the value of α at the tth iteration, T max is the total number of iterations, α max and α min are the upper and lower limits of the step scale factor.

6. The synchronous reluctance motor model predictive control method with adaptive weight coefficient optimization according to claim 4 is characterized in that: In step f-5, the update formula of the discovery probability is: in, is the value of the probability of discovery at the tth iteration, g p is the probability learning factor for discovery, and R is the population improvement rate.

7. The synchronous reluctance motor model predictive control method with adaptive optimization of weight coefficients according to claim 4 is characterized in that: In step f-2, the update formula of the bird's nest fitness value is: fitness=βT p +F Among them, T p is the torque pulsation, F is the sum of the changes between the switch states at time k and time k+1, and β is 100000.

Images