PMSM rotor position estimation method based on model prediction PLL

By combining the model prediction method and PLL, MP-PLL structure is designed and discrete search algorithm is used to optimize the solution, which solves the problems of poor robustness and complex calculations in PMSM rotor position estimation, and achieves a more efficient and accurate rotor position Estimation and motor control.

CN118944515APending Publication Date: 2024-11-12SHENYANG LIGONG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410980230.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-22
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

In the PMSM rotor position estimation, traditional PLL has poor robustness due to fixed PI parameters, complex calculations and large estimation errors, making it difficult to adapt to motor parameters and operating conditions changes.

Method used

Combining the model prediction method (MP) and PLL, the model predicts the phase-locked loop MP-PLL structure is designed, and the rotor position search algorithm is used based on discrete search, and the cost function is constructed to optimize the solution, and the fixed gain PI controller is cancelled to improve the calculation efficiency and Estimate accuracy.

Benefits of technology

It improves the accuracy of rotor position estimation and the robustness of the system, reduces the computational complexity, strong adaptability, strong anti-interference ability, reduces system costs, and optimizes motor control performance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118944515A_ABST
    Figure CN118944515A_ABST
Patent Text Reader

Abstract

The invention provides a PMSM rotor position estimation method based on a model prediction PLL, relates to the technical field of rotor position estimation, and aims to solve the problem that the rotor position estimation robustness is poor due to the fact that the PMSM parameters can change along with actual working conditions in the conventional PLL for estimating the rotor position of the PMSM. The method is provided for solving the problems that an existing improved method is complex in calculation and large in estimation error. Firstly, a PMSM model prediction control principle is analyzed, then a model prediction method is combined with a PLL, an MP-PLL structure is provided for rotor position estimation, and targeted design is carried out for improving MP-PLL calculation efficiency. Finally, the effectiveness of the proposed method is verified by comparing the traditional PLL and MP-PLL rotor position estimation methods.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of rotor position estimation, and in particular to a PMSM rotor position estimation method based on model prediction PLL. Background Art

[0002] The sensorless control technology of permanent magnetic synchronous machine (PMSM) is an effective means to reduce the cost and volume of PMSM drive system and improve system reliability. With the increasing level of electrical automation, the permanent magnetic synchronous motor drive system based on traditional position sensorless control technology has been difficult to meet the needs of driving conditions with high dynamic performance requirements. Since PMSM parameters vary with actual working conditions, the use of fixed controller parameters will lead to reduced accuracy of rotor position estimation.

[0003] As an effective prediction method in the field of predictive control, model prediction (MP) has been widely used in power electronics technology, motor transmission and other fields. The main components of MP include prediction model, feedback correction and rolling optimization. It has the advantages of simple structure, good dynamic performance and no modulation. In each discrete control cycle, MP uses the output measurement value of the controlled object of the system at the current moment, combined with the prediction model of the system's future multiple control cycles, and uses the preset cost function to solve the open-loop optimization problem in the finite time domain online. The first control input in the optimal control input sequence is applied to the system to achieve rolling optimization. In the next discrete control cycle, MP continues to repeat the above process, that is, the optimization problem is refreshed and solved again with new measurement values. The biggest feature that distinguishes MP from traditional control methods is that it can achieve closed-loop control by solving the open-loop optimization problem online and repeatedly.

[0004] In the field of motor drive, the application of MP can be divided into two categories: continuous-control-set model prediction (CCS-MP) and finite-control-set model prediction (FCS-MP). The system structure and implementation process of CCS-MP are as follows: Figure 1 and Figure 2As shown in the figure. In CCS-MP, based on the internal model of the system, first, the behavior of the system within the unit cycle is predicted; then, within the spatial plane range of the continuously rotatable voltage vector, the optimal voltage vector is selected based on the evaluation of the future response by the cost function; finally, the output voltage vector is synthesized through PWM modulation technology to realize motor control. Since the input in the continuous set is an infinite number of continuously adjustable voltage vectors, it is impossible to calculate all voltage vectors one by one using the exhaustive optimization method, so different optimization problems need to be constructed to solve the optimal output voltage vector, which increases the complexity of CCS-MP calculation.

[0005] Due to the discrete switching characteristics of the inverter and the nonlinear characteristics of the motor system, FCS-MP has become a unique method in the field of motor drives. The prediction structure and implementation process of the finite control set model are as follows: Figure 3 and Figure 4 As shown. The basic principle of FCS-MP is to use the switching state of power devices such as inverters as control actions, predict the system output measurement value, and use this to construct a cost function. The system is optimized by selecting the input of the cost function that minimizes the constraint conditions. FCS-MP is a simple and efficient control technology. FCS-MP eliminates the PI regulator and modulation stage, and can solve the optimization problem without the assistance of PWM modulation technology.

[0006] Considering the strong coupling and nonlinearity of the PMSM system model, PMSM is usually combined with FCS-MP. According to the linear or nonlinear constraints of the multi-objective control variables, the dynamic and steady-state performance of the PMSM drive system can be effectively improved. FCS-MP is a method for solving nonlinear problems. It simplifies the optimization problem solving process through a finite number of switching states. When dealing with multivariable dynamic system problems, FCS-MP has high solution efficiency and does not require PWM modulation process, and has strong robustness to system parameters. However, in PMSM applications, FCS-MP has the problem of large computational complexity.

[0007] In traditional PLL, fixed gain PI controller is usually used to estimate the rotor position. Considering the constant changes in motor parameters and operating conditions, traditional PLL is no longer suitable for PMSM rotor position estimation. Therefore, combined with the characteristics of FCS-MP, FCS-MP can be combined with PLL to improve the accuracy of rotor position estimation. Summary of the invention

[0008] In view of the shortcomings of the prior art, the present invention provides a PMSM rotor position estimation method based on model prediction PLL. Traditional PLL usually uses fixed PI parameters to estimate the PMSM rotor position, but since the PMSM parameters will change with the actual working conditions, the robustness of the rotor position estimation is poor. The existing improved methods have the problems of complex calculation and large estimation error. First, the PMSM model predictive control principle is analyzed, and then the model prediction method is combined with PLL, and an MP-PLL structure is proposed for rotor position estimation, and a targeted design is performed to improve the calculation efficiency of MP-PLL. Finally, by comparing the traditional PLL and MP-PLL rotor position estimation methods, the effectiveness of the proposed method is verified.

[0009] A PMSM rotor position estimation method based on model prediction PLL comprises the following steps:

[0010] Step 1: Design a model prediction phase-locked loop MP-PLL structure that combines a model prediction method MP with a phase-locked loop PLL to estimate the position of the PMSM rotor;

[0011] The model prediction phase-locked loop MP-PLL structure is specifically as follows: the fixed gain PI controller in the phase-locked loop PLL is cancelled, the best estimated rotor position is found, and the optimized solution is obtained by minimizing the constructed cost function: a rotor position search algorithm based on discrete search is specifically adopted to discretize the rotor position of the PMSM, and the cost function is used for solution;

[0012] Step 1.1: Calculate the discrete rotor position; assume the sampled current and voltage Define the initial rotor position θ in,i [k], construct the cost function g in [k]; use i and j to perform nested loop iterations, i and j represent the number of iterations of the inner and outer loops respectively, then the discrete rotor position is expressed as:

[0013] θ ri,j [k] = θ in,i [k]+(j-4)Δθ i [k] (1)

[0014] in:

[0015]

[0016] Step 1.2: Using the discrete rotor position, calculate the back EMF The d-axis component of the rotor is then constructed to perform the best estimation of the rotor position. The rotor position that minimizes the cost function in the back-EMF calculation is selected, and the inner and outer nested iteration loops are performed according to the rotor position search algorithm. When the maximum number of iterations N = 7, after 64 iterations, the estimated value of the best rotor position is finally obtained. The cost function is expressed as:

[0017]

[0018] The rotor position search algorithm based on discrete search, when and g in When [k] = ∞, is the k-1th rotor position angle estimate. According to formula (2), the first iteration of the outer loop is performed, that is, when i = 0, Δθ0[k] = π / 4rad. Then, according to formula (1), 8 discrete rotor positions are generated, namely: 0rad, π / 4rad, π / 2rad, 3π / 4rad, -πrad, -3π / 4rad, -π / 2rad, -π / 4rad. Using the discrete rotor position θ ri,j [k] Calculate the back EMF Then, the rotor position angle that minimizes the cost function is selected as the optimal rotor position angle θ r,opt [k], and also serves as the initial angle θ for the second iteration in,1 [k]; Assuming that the best rotor position angle calculated after the first iteration is π / 2rad, when the second iteration of the outer loop is performed, that is, i=1, Δθ1[k]=π / 8rad; then, according to formula (1), 8 new discrete rotor positions are generated; with continuous iterations, it will converge to the best rotor position; finally, in the last iteration, that is, when i=N=7, the estimated value of the best rotor position is obtained;

[0019] Since the output of MP-PLL is an estimate of the optimal rotor position, when calculating the rotor speed, the change in rotor position during the previous sampling period is calculated and divided by the sampling time T s , that is, the estimated rotor speed is:

[0020]

[0021] Among them, k and k-1 are the actual sampling time and the previous sampling time respectively;

[0022] If the speed change of the motor system using MP-PLL is less than the set threshold, the number of loop iterations N is reduced, and the rotor position at the last sampling moment is used as the initial rotor position, that is, Reduce the number of iterations of the rotor position search algorithm;

[0023] When the motor speed change is greater than the set threshold, the rotor position search algorithm is used. The rotor position search algorithm analyzes the factors affecting the calculation amount: it consists of two inner and outer nested loops to discretize the rotor position;

[0024] Step 2: Improve the rotor position search algorithm based on discrete search to achieve PMSM rotor position estimation based on model prediction PLL;

[0025] The improved rotor position search algorithm consists of inner and outer nested loops, which reduces the number of iterations and rewrites formula (2) as follows:

[0026]

[0027] The beneficial effects of adopting the above technical solution are:

[0028] The present invention provides a PMSM rotor position estimation method based on model prediction PLL, which has significant beneficial effects in the field of motor control. The following is a detailed description of the beneficial effects produced:

[0029] 1. Improve the accuracy of rotor position estimation: Traditional PMSM control methods usually rely on position sensors to obtain rotor position information, which may be affected by factors such as sensor accuracy, installation location, and working environment, resulting in errors in position information. In contrast, the rotor position estimation method based on model prediction PLL can more accurately estimate the rotor position by sampling and processing the current and voltage signals of the motor in real time, using advanced signal processing algorithms and control strategies. This not only improves the accuracy of motor control, but also helps to improve the performance and stability of the entire system.

[0030] 2. Reduce computational complexity: By optimizing algorithms and mathematical models, the amount of computation is reduced and the demand for hardware resources is reduced, making the method implementable in embedded systems and suitable for real-time control applications.

[0031] 3. Strong adaptability: This method is not only suitable for rotor position estimation under various speed and load conditions, but also can adapt to different types of PMSM, and has broad application prospects and applicability.

[0032] 4. Strong anti-interference ability: Through the combination of model prediction and PLL technology, it can effectively filter out system noise and interference, improve the robustness and anti-interference ability of the estimation results, and ensure the stable operation of the system under complex working conditions

[0033] 5. Reduce system cost: Since there is no need to rely on position sensors, the rotor position estimation method based on model prediction PLL can reduce the hardware cost of the system. At the same time, since the possibility of sensor failure is reduced, the maintenance and repair costs of the system are also reduced. This has very important economic significance for industries such as industry and automobiles that mass produce and apply PMSM.

[0034] 6. Improve the adaptability and robustness of the system: The rotor position estimation method based on the model prediction PLL can adaptively adjust the control strategy and parameter settings according to different motor operating conditions and working environments, thereby achieving more stable and reliable control. Especially in the case of low-speed operation and large disturbances, this method can effectively reduce the estimation error and improve the robustness of the system through means such as phase compensation.

[0035] 7. Optimize motor control performance: Accurate rotor position information is the key to efficient and stable operation of the motor. The rotor position estimation method based on model prediction PLL can provide accurate rotor position information, so that the motor control system can more accurately control the motor's current, voltage and other parameters, thereby achieving more efficient energy conversion and more stable operation. This helps to improve the efficiency of the motor, reduce energy consumption, and extend the service life of the motor.

[0036] This invention promotes the development of motor control technology. The rotor position estimation method based on model prediction PLL is an innovative motor control method, which not only improves the accuracy and stability of motor control, but also provides new ideas and methods for the development of motor control technology. The proposal and application of this method not only provides an efficient and reliable solution for the control and application of PMSM, but also helps to promote the continuous progress and innovation of motor control technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 Predict structural diagrams for continuous control set models;

[0038] Figure 2 Implementing process diagram for continuous control set model prediction;

[0039] Figure 3 Predict structural diagrams for finite control set models;

[0040] Figure 4 Implementation process diagram for finite control set model prediction;

[0041] Figure 5 It is a structural diagram of the MP-PLL of the present invention;

[0042] Figure 6 It is a block diagram of PMSM rotor position estimation based on MP-PLL in the present invention;

[0043] Figure 7 It is a waveform diagram of the measured speed and the estimated speed under the speed step jump in the present invention;

[0044] Figure 8 This is a waveform diagram of rotor speed error under speed step jump in the present invention;

[0045] Fig. 9 This is a waveform diagram of the rotor position estimation error under a speed step jump in the present invention;

[0046] Fig.10 It is the electromagnetic torque waveform diagram under the speed step jump in the present invention;

[0047] Fig.11 This is a d-axis current waveform diagram under a speed step jump in the present invention;

[0048] Fig.12 This is a q-axis current waveform diagram under a speed step jump in the present invention;

[0049] Fig.13 This is a d-axis / q-axis voltage waveform diagram under a speed step jump in the present invention;

[0050] Fig.14 It is a waveform diagram of measured speed and estimated speed under reference torque step jump in the present invention;

[0051] Fig.15 The rotor speed error waveform diagram under the reference torque step jump in the present invention;

[0052] Fig.16 It is a waveform diagram of rotor position estimation error under reference torque step jump in the present invention;

[0053] Fig.17 It is the electromagnetic torque waveform diagram under the reference torque step jump in the present invention;

[0054] Fig.18 This is a d-axis current waveform diagram under a reference torque step jump in the present invention;

[0055] Fig.19 It is the q-axis current waveform diagram under the reference torque step jump in the present invention;

[0056] Fig. 20 It is a d-axis / q-axis voltage waveform diagram under reference torque step jump in the present invention;

[0057] Fig.21 It is a waveform diagram of the rotation speed measured in a steady state in the present invention;

[0058] Fig. 22 is a waveform diagram of rotor speed estimation error in steady state in the present invention;

[0059] Fig.23is a waveform diagram of rotor position estimation error in a steady state in the present invention;

[0060] Fig.24 It is the electromagnetic torque waveform diagram in steady state in the present invention;

[0061] Fig.25 This is a d-axis current waveform diagram in a steady state in the present invention;

[0062] Fig.26 It is the q-axis current waveform diagram in steady state in the present invention;

[0063] Fig. 27 It is a d-axis / q-axis voltage waveform diagram in steady state in the present invention;

[0064] Fig.28 The waveform diagram of the measured speed and the estimated speed under the step change of the stator resistance ±50% in the present invention;

[0065] Fig.29 This is a speed error waveform diagram under a ±50% step change of the stator resistance in the present invention;

[0066] Fig.30 This is a waveform diagram of the rotor position estimation error under a ±50% step change of the stator resistance in the present invention;

[0067] Fig.31 This is the electromagnetic torque waveform diagram under the stator resistance ±50% step jump in the present invention;

[0068] Fig.32 This is a d-axis current waveform diagram under a ±50% step change of the stator resistance in the present invention;

[0069] Fig.33 This is a q-axis current waveform diagram under a ±50% step change of the stator resistance in the present invention;

[0070] Fig.34 This is a d-axis / q-axis voltage waveform diagram under a ±50% step change of the stator resistance in the present invention;

[0071] Fig.35 The waveform diagram of the measured speed and the estimated speed under the step change of the stator inductance ±50% in the present invention;

[0072] Fig.36 This is a waveform diagram of rotor error under a ±50% step change of the stator inductance in the present invention;

[0073] Fig.37 This is a waveform diagram of the rotor position estimation error under a ±50% step change of the stator inductance in the present invention;

[0074] Fig.38 The waveform diagram of the electromagnetic torque under the step change of the stator inductance of ±50% in the present invention;

[0075] Fig.39 This is a d-axis current waveform diagram under a ±50% step change of the stator inductance in the present invention;

[0076] Fig.40 It is the q-axis current waveform diagram under the stator inductance ±50% step jump in the present invention;

[0077] Fig.41 The d-axis / q-axis voltage waveform diagram of the stator inductance under ±50% step jump in the present invention;

[0078] in Figure 7-Figure 41 In the figure, (a) is the model predicted phase-locked loop MP-PLL structure, and (b) is the traditional phase-locked loop PLL. DETAILED DESCRIPTION

[0079] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0080] A PMSM rotor position estimation method based on model prediction PLL comprises the following steps:

[0081] Step 1: Design a model prediction phase-locked loop MP-PLL structure that combines a model prediction method MP with a phase-locked loop PLL, which is used to estimate the position of the PMSM rotor; the MP-PLL structure proposed in this embodiment is shown in FIG. Figure 5 shown.

[0082] The model predicts that the phase-locked loop MP-PLL structure is specifically:

[0083] The fixed gain PI controller in the phase-locked loop PLL is cancelled, and the best estimated rotor position is found from the limited rotor positions, and the optimized solution is obtained by minimizing the constructed cost function: However, since the rotor position changes continuously between 0 and 2π, it does not have a discrete state like the switching action of the inverter. Therefore, in this embodiment, a rotor position search algorithm based on discrete search is specifically adopted to discretize the rotor position of the PMSM, similar to the 8 switching combination states of the two-level inverter, and the possible rotor position is obtained by continuously iterating 8 positions, and it is used in solving the cost function; the algorithm flow in this embodiment is shown in Table 1.

[0084] Table 1 Rotor position search algorithm

[0085]

[0086]

[0087] Step 1.1: Calculate the discrete rotor position; assume the sampled current and voltage Define the initial rotor position θ in,i [k], construct the cost function g in [k]; In order to obtain a discrete finite number of rotor positions, i and j are used to perform nested loop iterations, where i and j represent the number of iterations of the inner and outer loops, respectively. The discrete rotor position is expressed as:

[0088] θ ri,j [k] = θ in,i [k]+(j-4)Δθ i [k] (1)

[0089] in:

[0090]

[0091] Step 1.2: Using the discrete rotor position, calculate the back EMF The d-axis component of Figure 5 As shown. Then, a cost function is constructed to optimally estimate the rotor position. The rotor position that minimizes the cost function in the back EMF calculation is selected, and the inner and outer nested iteration loops are performed according to the rotor position search algorithm. When the maximum number of iterations N = 7, after 64 iterations, the estimated value of the optimal rotor position is finally obtained; wherein the cost function is expressed as:

[0092]

[0093] The rotor position search algorithm based on discrete search, when and g in When [k] = ∞, is the k-1th rotor position angle estimate. According to formula (2), the first iteration of the outer loop is performed, that is, when i = 0, Δθ0[k] = π / 4rad. Then, according to formula (1), 8 discrete rotor positions are generated, namely: 0rad, π / 4rad, π / 2rad, 3π / 4rad, -πrad, -3π / 4rad, -π / 2rad, -π / 4rad. Using the discrete rotor position θ ri,j [k] Calculate the back EMF Then, the rotor position angle that minimizes the cost function is selected as the optimal rotor position angle θ r,opt [k], and also serves as the initial angle θ for the second iteration in,1[k]; Assuming that the best rotor position angle calculated after the first iteration is π / 2rad, when the second iteration of the outer loop is performed, that is, i=1, Δθ1[k]=π / 8rad; compared with the first outer loop iteration, the accuracy of the proposed discrete search algorithm is improved by 2 times. Then, according to formula (1), 8 new discrete rotor positions are generated; with continuous iterations, it will converge to the best rotor position; finally, in the last iteration, that is, when i=N=7, the estimated value of the best rotor position is obtained;

[0094] According to formula (2), the accuracy of rotor position estimation is obtained as

[0095] Since the output of MP-PLL is an estimate of the optimal rotor position, when calculating the rotor speed, the change in rotor position during the previous sampling period is calculated and divided by the sampling time T s , that is, the estimated rotor speed is:

[0096]

[0097] Among them, k and k-1 are the actual sampling time and the previous sampling time respectively; it can be seen that since the iterative discrete search method is used to determine the rotor position, the robustness of this method to the motor parameters is improved, and the estimation accuracy can be improved by increasing the number of iterations N.

[0098] The main feature of the proposed MP-PLL is that it does not require adjustment of the PI controller. It has the advantages of simple design and easy implementation, and can be applied to any type of motor. However, if higher accuracy is required, multiple loop iterations are required, which consumes a long running time. If the speed change of the motor system using MP-PLL is less than the set threshold and does not change much between two sampling moments, then reduce the number of loop iterations N and use the rotor position at the previous sampling moment as the initial rotor position, that is, Reduce the number of iterations of the rotor position search algorithm;

[0099] When the motor speed change is greater than the set threshold, the rotor position search algorithm is used. A smaller number of loop iterations N will result in an insufficient position search and a larger estimation error. When N is increased, although the position search range will be closer to the actual rotor position, the amount of calculation will increase significantly, which may seriously affect the execution speed of the algorithm. Therefore, a compromise must be made between accuracy and calculation.

[0100] Analysis of the factors affecting the computational amount of the rotor position search algorithm: The rotor position search algorithm consists of two nested loops inside and outside, which discretize the rotor position; in the first round of iteration, Δθ0=π / 4rad, generating 8 discrete rotor positions. Assuming that the best rotor position after the first iteration is π / 2rad, according to the algorithm, in the second round of iteration, the algorithm will perform discrete searches at 8 discrete positions of 0rad, π / 8rad, π / 4rad, 3π / 8rad, π / 2rad, 5π / 8rad, 3π / 4rad, and 7π / 8rad. However, considering that the best rotor position π / 2rad after the first iteration is between 3π / 8rad and 5π / 8rad, during the second round of iteration, the discrete search range should be limited to between 3π / 8rad and 5π / 8rad rather than within the range of 0rad to 7π / 8rad. This can effectively reduce the computational amount of the algorithm and quickly find the optimal position of the rotor. Therefore, it is necessary to find a method to limit the search range of the next round of iterations to improve the efficiency of the search.

[0101] Step 2: Improve the rotor position search algorithm based on discrete search to achieve PMSM rotor position estimation based on model prediction PLL;

[0102] The improved rotor position discrete search algorithm is similar to the algorithm mentioned in the previous section. The improved rotor position search algorithm consists of inner and outer nested loops. In order to improve the algorithm calculation efficiency and reduce the number of iterations, formula (2) is rewritten as:

[0103]

[0104] As the number of iterations N increases, the accuracy of discrete search will be improved. For example, in this embodiment, the value range of i is set to 0 to N = 5, and the value range of j is set to 0 to 5. Assume that the best position angle θ obtained by the first iteration is in,1 When the improved algorithm performs the first iteration of the outer loop, the discrete search range of the rotor position is between π / 12 and -π / 12. When the outer loop is iterated for the second time, that is, when i=1, Δθ1 is π / 16rad, which is more than 2.6 times more accurate than the first iteration. In contrast, the accuracy of the discrete search algorithm in the second iteration is only 2 times higher than that in the first iteration, indicating that the improved algorithm has a faster discrete search speed. In the second iteration, there is θ r1,0 =-π / 4rad is not within the discrete search range, but a single iteration does not significantly increase the amount of calculation. According to the algorithm flow, the improved algorithm only needs 36 iterations to find the optimal rotor position.

[0105] According to formula (5), the accuracy of rotor position estimation is obtained as Compared with previous algorithms, it has similar rotor position estimation accuracy.

[0106] Experimental results and analysis

[0107] The performance of the proposed improved MP-PLL based on discrete search algorithm is compared with that of the traditional PLL. In order to reflect the performance of the proposed MP-PLL in a wide speed range, the speed range is set to 4-100 rad / s. The switching frequency and sampling frequency of the power converter are both set to 4kHz. The PI controller parameter k of the SRF-PLL is set to p 303,k i is 1100. PMSM uses Figure 6 The field-oriented control (FOC) technology shown in the figure, the parameters of the PMSM are shown in Table 2.

[0108] Table 2 PMSM parameter table

[0109]

[0110]

[0111] (1) Rotor speed step jump experiment

[0112] In order to verify the effectiveness of the proposed MP-PLL rotor position estimation method when the rotor speed jumps step, the measured speed ω r , measure the rotation speed ω r With estimated speed The speed estimation error Δω r , actual rotor position θ r With the estimated rotor position The rotor position estimation error Δθ r The three aspects of the test analysis were compared with the performance of the traditional PLL. The results are as follows Figure 7-Figure 9 shown.

[0113] like Figure 7 As shown in the figure, at t = 4.5s, the rotor speed changes from 50r / min to 2600r / min. Figure 7 In (a), since the proposed MP-PLL method eliminates the fixed gain PI controller in the traditional PLL, the measured speed ω r With estimated speed The error between them is small. Figure 7 In (b), the measured speed ω of the traditional PLL method r With estimated speed The error between them is large. When measuring the speed ω r When it jumps to 2600r / min, the estimated speed The MP-PLL method has better transient performance than the traditional PLL method. The amplitude is smaller and the oscillation is eliminated quickly. The rotor speed error and rotor position estimation error under speed step jump are as follows: Figure 8 and 9 shown.

[0114] like Figure 8 As shown, when the speed jumps, the rotor speed error Δω r also changes, and the speed error tends to zero after about 1s. Figure 8 In (a), the rotor speed error Δω in the MP-PLL method r First it increases and then decreases, and the maximum speed error is 60r / min; Figure 8 In (b), the rotor speed error Δω in the PLL method r It is in the form of oscillation, and the maximum speed error is 200r / min. By comparison, the maximum speed error Δω of MP-PLL is r Smaller, speed error Δω r Converges to 0 faster.

[0115] like Fig. 9 As shown in the figure, when the speed jumps, the rotor position estimation error Δθ r changes. Fig. 9 In (a), the rotor position estimation error Δθ in the MP-PLL method r Change from 0rad to -0.05rad and stabilize; Fig. 9 In (b), the rotor position estimation error Δθ in the PLL method r It is in the form of oscillation, and the maximum rotor position estimation error is 0.19rad. Rotor position estimation error Δθ r The theoretical value should be 0.003rad, but Fig. 9 It can be seen that the rotor position estimation error Δθ r The value of is 0.05rad, which is larger than the theoretical value. This is due to the PMSM model error, nonlinear inverter, and estimated back EMF harmonics. Nevertheless, the performance of the proposed MP-PLL is better than that of the traditional PLL.

[0116] In order to compare the performance of the FOC strategy using MP-PLL and traditional PLL when the rotor speed jumps, experiments are carried out in four aspects: torque, d-axis current, q-axis current and d-axis / q-axis voltage. The results are shown in Figure 2. Figure 10-13 shown.

[0117] like Fig.10 As shown, the rotor measured speed ω r When the step jump occurs, the electromagnetic torque Te changes. Fig.10 (a) The electromagnetic torque T in the MP-PLL method e After 0.5s of change, it tends to be stable, with the maximum change amplitude being 4N·m. Fig.10 In (b), the electromagnetic torque T in the PLL method e After 1s of oscillation, it becomes stable with a maximum amplitude of 8N·m.

[0118] like Fig.11 and 12 As shown, the rotor measured speed ω r When the step jump occurs, the d-axis current and q-axis current changes. Fig.11 (a) and 12(a), the d-axis current in the MP-PLL method and q-axis current After 1s of change, it tends to be stable; Fig.11 (b) and 12(b), the d-axis current in the PLL method and q-axis current After 1s of oscillation, it becomes stable with a maximum amplitude of 2A.

[0119] like Fig.13 As shown, the rotor measured speed ω r When the step jump occurs, the q-axis voltage of the two methods is A step jump also occurs, from 0V to 80V. No significant change, d-axis voltage in PLL method Oscillations are generated with a maximum amplitude of 10 V. Compared with the traditional PLL, the proposed MP-PLL does not generate oscillations in torque, current, and voltage when the rotor speed steps, and the performance of the FOC strategy using MP-PLL is significantly better than that using the traditional PLL.

[0120] (2) Reference torque step jump experiment

[0121] To verify the reference torque The dynamic performance of the proposed MP-PLL rotor position estimation method under step jump. This section starts from the measured speed ω r , measure the rotation speed ω r With estimated speed The speed estimation error Δω r , actual rotor position θ r With the estimated rotor position The rotor position estimation error Δθ r The three aspects are analyzed and compared with the dynamic performance of traditional PLL. The results are as follows Figure 14-16 shown.

[0122] like Fig.14 As shown, at t = 6s, the reference torque of the simulated PMSM control system From -10N·m to -38N·m. Fig.14 In (a), ω r MP-PLL estimated speed by reducing the speed from 900r / min to 650r / min The speed is reduced from 900r / min to 700r / min, and then restored to 900r / min after 1.6s. The speed is measured. r With estimated speed The error between them is small. Fig.14 In (b), the conventional PLLω r From 900r / min to 620r / min, the estimated speed The oscillation changes with a maximum amplitude of 400r / min, which recovers to 900r / min after 1.6s, and the estimated error is relatively large.

[0123] The estimated error under the reference torque step jump is as follows: Fig.15 and 16 shown. Fig.15 When the reference torque When the step jump occurs, the rotor speed error Δω r changes. Fig.15 In (a), the rotor speed error Δω in the MP-PLL method r It increases first and then decreases, and the maximum speed error is -20r / min; Fig.15 In (b), the rotor speed error Δω in the traditional PLL method r It is in the form of oscillation, and the maximum speed error is -200r / min. By comparison, the maximum speed error Δω of MP-PLL is r Smaller, the speed error Δω of the two methods r The time to converge to 0r / min is close.

[0124] like Fig.16 As shown, when the reference torque When the step jump occurs, the rotor position estimation error Δθ r changes. Fig.16 In (a), the rotor position estimation error Δθ in the MP-PLL method r Change from 0rad to 0.02rad and stabilize; Fig.16 In (b), the rotor position estimation error Δθ in the traditional PLL method r It is in the form of oscillation, and the maximum rotor position estimation error is 0.09rad. When a step jump occurs, the proposed MP-PLL does not oscillate when estimating the speed and position, and has better dynamic response performance than the traditional PLL. The performance of the FOC strategy using MP-PLL and traditional PLL during step jumps was tested in terms of torque, d-axis current, q-axis current and d-axis / q-axis voltage. The results are shown in Figure 2. Figure 17-Figure 20 shown.

[0125] like Fig.17 As shown, the reference torque When the step jump occurs, the electromagnetic torque T e changes. Fig.17 (a) The electromagnetic torque T in the MP-PLL method e With this reference torque The error is small. Fig.17 In (b), when the reference torque When the step jump occurs, the electromagnetic torque T in the traditional PLL method e Overall change vs. reference torque Similar, but the electromagnetic torque T e Oscillation occurs and becomes stable after 0.4s.

[0126] like Fig.18 and 19 As shown, the reference torque When the step jump occurs, the d-axis current and q-axis current Also changes. Fig.18 (a) and 19(a), the d-axis current in the MP-PLL method and q-axis current With reference current and It changes with the change, and the error is small. Fig.18 (b) and 19(b), the conventional PLL d-axis current and q-axis current Both produce oscillations, and the d-axis current The maximum amplitude is 10A, the q-axis current The maximum amplitude is 2A.

[0127] like Fig. 20 As shown, the reference torque When the step jump occurs, the d-axis voltage of the two methods and q-axis voltage changes. Fig. 20 (a) The d-axis voltage in the MP-PLL method No significant change, q-axis voltage After falling from 30V to 0V, it recovers to 30V. Fig. 20 In (b), the d-axis voltage in the traditional PLL method Oscillation occurs, the q-axis voltage After dropping from 30V to -30V, it recovers to 30V. Compared with the traditional PLL, the proposed MP-PLL has a better reference torque. There is no oscillation of torque, current and voltage during step jumps, and the performance of the FOC strategy of MP-PLL is better than that of traditional PLL.

[0128] (3) Steady-state performance analysis

[0129] In order to verify the steady-state performance of the proposed MP-PLL, the speed ω r , measure the rotation speed ω r With estimated speed The speed estimation error Δω r , actual rotor position θ r With the estimated rotor position The rotor position estimation error Δθ r The three aspects of the experiment were analyzed and compared with the traditional PLL. The results are as follows Figure 21-23 In the control system of the experimental platform, the rotor speed is 1200r / min and the reference torque is It is -60N·m.

[0130] exist Fig.21 In (a), the MP-PLL method measures the rotation speed ω r The maximum is 1215r / min, the minimum is 1185r / min, and the measured speed is ω r The range of change is 30r / min. Fig.21 In (b), the traditional PLL method measures the speed ω r The maximum is 1210r / min, the minimum is 1190r / min, and the measured speed is ω r The range of change is 20r / min. Compared with the traditional method, the MP-PLL method measures the speed ω r The variation range is relatively large, but it generally meets the performance requirements. The rotor speed error and rotor position estimation error are as follows: Fig. 22 and Fig.23 shown.

[0131] like Fig. 22 As shown in the figure, under the steady-state performance test, the rotor speed error Δω of the two methods is r Approximately. Fig. 22 In (a), the rotor speed error Δω in the MP-PLL method r -20r / min to 20r / min; Fig. 22In (a), the rotor speed error Δω in the traditional PLL method r The rotor speed error Δω in the MP-PLL method is -18r / min to 18r / min. r Relatively large.

[0132] like Fig.23 As shown in the figure, under the steady-state performance test, the rotor position estimation error Δθ r changes. Fig.23 In (a), the rotor position estimation error Δθ in the MP-PLL method r The fluctuation range is -0.02rad to 0.02rad; Fig.23 In (b), the rotor position estimation error Δθ in the traditional PLL method r The fluctuation range is -0.01rad to 0.01rad, and the maximum rotor position estimation error is 0.02rad. By comparison, under the steady-state performance test, the ripple of the MP-PLL method is slightly higher than that of the traditional PLL, and the fluctuation range is larger, but the overall performance meets the requirements.

[0133] In order to compare the performance of the FOC strategy using MP-PLL and traditional PLL in steady state, experiments were conducted in four aspects: torque, d-axis current, q-axis current, and d-axis / q-axis voltage. The results are shown in Figure 2. Figure 24-Figure 27 shown.

[0134] like Fig.24 As shown, under the steady-state performance test, the electromagnetic torque T e Changes within a certain range. Reference torque When it is -60N·m, the electromagnetic torque T e The fluctuation range is -61N·m to -59N·m. The electromagnetic torque T in the traditional PLL method and the MP-PLL method e The fluctuation range is approximate.

[0135] like Fig.25 and 26 As shown, in the steady-state performance test, the d-axis current and q-axis current Fluctuates within a range. Fig.25 (a) The d-axis current in the MP-PLL method The fluctuation range is -0.3A to 0.3A. Fig.25 In (b), the d-axis current in the conventional PLL method The fluctuation range is -0.2A to 0.2A. Fig.26 (a) The q-axis current in the MP-PLL method The fluctuation range is -35.5A to -34.5A, and the q-axis current in the traditional PLL method The fluctuation range is similar to that of the MP-PLL method.

[0136] like Fig. 27 As shown in the figure, the d-axis voltage of the two methods and q-axis voltage Compare the q-axis voltage in 28(a) and 28(b) It can be seen that under the steady-state performance test, the q-axis voltage fluctuation range of the two methods is similar, both of which are 34V to 37V. Stable in the range of 2V to 3V, the d-axis voltage in the traditional PLL method It is stable in the range of 12 V to 13 V. When estimating the rotor position, the steady-state performance of MP-PLL is similar to that of the traditional PLL, which meets the requirements of rotor position estimation.

[0137] (4) Motor parameter sensitivity test

[0138] In order to verify the robustness of the proposed MP-PLL, the measured speed ω r ,ω r With estimated speed The estimated error Δω r , actual rotor position θ r With the estimated rotor position The rotor position estimation error Δθ r The three aspects are analyzed and compared with the traditional PLL. The results are as follows Figure 28-Figure 30 As shown. At 3-4s and 5-6s, the stator resistance is R s ; At 4-5s and 6-7s, the stator resistance is 0.5R respectively s and 1.5R s .

[0139] like Fig.28 As shown in the figure, when the stator resistance changes step by step, the speed ω is measured. r In the experiment, the motor speed is 1500r / min and the reference torque is -40N·m. Fig.28 In (a), the MP-PLL method measures the speed ω when the stator resistance jumps at 4s, 5s, and 6s. r There was no significant change, and the fluctuation range remained between 1490r / min and 1510r / min within 3-7s. Fig.28 In (b), when the stator resistance changes stepwise, the speed ω is measured. rThe fluctuation is large, with the maximum fluctuation range being 1485r / min to 1515r / min.

[0140] like Fig.29 As shown in the figure, when the stator resistance changes step by step, the rotor speed error Δω r changes. Fig.29 In (a), the rotor speed error Δω in the MP-PLL method r The minimum is -10r / min and the maximum is 10r / min. Fig.29 In (b), the rotor speed error Δω in the traditional PLL method r It is greatly affected by the step change of stator resistance. When the stator resistance step changes at 4s, 5s and 6s, the speed error Δω r The maximum error is 15r / min.

[0141] like Fig.30 As shown in Figure 2, when the stator resistance changes step by step, the rotor position estimation error Δθ r changes. Fig.30 In (a), the rotor position estimation error Δθ in the MP-PLL method r The fluctuation range is -0.02rad to 0.02rad and is not affected by the step change of stator resistance. Fig.30 In (b), the rotor position estimation error Δθ in the traditional PLL method r The stator resistance is greatly affected by the step change. At t = 4s, the stator resistance changes from R s Jump to 0.5R s , rotor position estimation error Δθ r The fluctuation range of the stator resistance decreases accordingly; at t = 5s and t = 6s, the stator resistance changes from 0.5R s Jump to R s , and then jumps to 1.5R s , rotor position estimation error Δθ r The fluctuation range of stator resistance R s A ±50% step change has little effect on the response of the MP-PLL. However, the traditional PLL is more sensitive to parameter uncertainty and has poor response performance.

[0142] In order to compare the performance of FOC strategy using MP-PLL and traditional PLL under stator resistance step jump, experiments were conducted in four aspects: torque, d-axis current, q-axis current and d-axis / q-axis voltage. The results are shown in Figure 2. Figure 31-Figure 34 shown.

[0143] like Fig.31 As shown, the stator resistance R s ±50% step jump, electromagnetic torque Te Changes within a certain range. Reference torque is -40N·m, and the electromagnetic torque T in the proposed MP-PLL method e The fluctuation range is -41N·m to -39N·m. The electromagnetic torque T in the traditional PLL method and the MP-PLL method e The fluctuation range is approximate.

[0144] like Fig.32 and 33 As shown, the stator resistance R s ±50% step jump, d-axis current and q-axis current Fluctuates within a range. Fig.32 (a) The d-axis current in the MP-PLL method The fluctuation range is -0.3A to 0.3A. Fig.32 In (b), the d-axis current in the traditional PLL method The fluctuation range is -0.2A to 0.2A. Fig.33 (a) The q-axis current in the MP-PLL method The fluctuation range is -24A to -23A. The q-axis current in the traditional PLL method The fluctuation range is similar to that of the MP-PLL method.

[0145] like Fig.34 As shown in the figure, the d-axis voltage of the two methods and q-axis voltage Fluctuations occur. Fig.34 (a) and Fig.34 (b) The q-axis voltage It can be seen that the stator resistance R s When a step jump occurs at ±50%, the q-axis voltage fluctuation range in the two methods is similar, both of which are 56V to 63V. Stable in the range of -3V to 0V, the d-axis voltage in the traditional PLL method It is stable in the range of 12 V to 15 V. When estimating the rotor position, the proposed MP-PLL has a similar response performance to the traditional PLL, which meets the requirements of PMSM rotor position estimation.

[0146] Also considered is the stator inductance L s Under the jump change, from the measured speed ω r With estimated speed The speed estimation error Δω r , actual rotor position θ r With the estimated rotor position The rotor position estimation error Δθ rThe response performance of MP-PLL and traditional PLL is compared and analyzed. The results are as follows: Figure 35-Figure 37 As shown. At 3-4s and 5-6s, the stator inductance is L s , at 4-5s and 6-7s, the stator inductance is 0.5L respectively s and 1.5L s .

[0147] like Fig.35 As shown, when the stator inductance changes step by step, the speed ω is measured. r In the experiment, the speed was 1500r / min and the reference torque was It is -40N·m. Fig.35 In (a), the MP-PLL method measures the speed ω when the stator resistance jumps at 4s, 5s, and 6s. r There was no significant change, and the fluctuation range remained between 154r / min and 156r / min within 3-7s. Fig.35 In (b), when the stator resistance changes stepwise, the speed ω is measured. r Oscillation occurs, minimum measured speed ω r The maximum measuring speed is 145r / min. r It is 165r / min.

[0148] like Fig.36 As shown in Figure 2, when the stator inductance undergoes a step jump, the rotor speed error Δω r changes. Fig.36 In (a), the rotor speed error Δω in the MP-PLL method r The minimum is -139r / min and the maximum is 141r / min. Fig.36 In (b), the rotor speed error Δω in the traditional PLL method r Oscillation occurs due to the stator inductance step jump. Maximum rotor speed error Δω r is 140r / min, the minimum rotor speed error Δω r It is -139r / min.

[0149] like Fig.37 As shown in Figure 2, when the stator inductance changes step by step, the rotor position estimation error Δθ r changes. Fig.37 In (a), the rotor position estimation error Δθ in the MP-PLL method r The fluctuation range is -0.02rad to 0.02rad and is not affected by the step jump of the stator inductance. Fig.37 In (b), the rotor position estimation error Δθ in the traditional PLL method rThe stator inductance is greatly affected by the step change. At t = 4s, the stator inductance changes from L s Jump to 0.5L s , rotor position estimation error Δθ r becomes larger; at t=5s and t=6s, the stator resistance increases from 0.5L s Jump to L s , and then jumped to 1.5L s , rotor position estimation error Δθ r It also decreases and oscillates when the stator inductance jumps. In contrast, the proposed MP-PLL has good response performance when the stator inductance changes step by step, and the measured speed ω r , speed estimation error Δω r and the rotor position estimation error Δθ r No oscillation was observed.

[0150] In order to compare the performance of the FOC strategy using MP-PLL and traditional PLL under stator inductance step jump, experiments were conducted in four aspects: torque, dq axis current and dq axis voltage. The results are shown in Figure 2. Figure 38-Figure 41 shown.

[0151] like Fig.38 As shown, the stator inductance L s ±50% step jump, electromagnetic torque T e Varies within a certain range. Fig.38 (a) The electromagnetic torque T in the MP-PLL method e Not affected by stator inductance step jump, the fluctuation range is maintained at -41N·m to -39N·m. Fig.38 In (b), the stator inductance L s When the step jump occurs, the electromagnetic torque T e Will produce large fluctuations, the maximum electromagnetic torque T e is -37N·m, the minimum electromagnetic torque T e It is -42N·m.

[0152] like Fig.39 and Fig.40 As shown, the stator inductance L s ±50% step jump, d-axis current and q-axis current Fluctuates within a range. Fig.39 (a) The d-axis current in the MP-PLL method The fluctuation range is -0.1A to 0.1A, Fig.39 In (b), when the stator inductance L s When the step jump occurs, the d-axis current in the traditional PLL method Oscillation occurs, the maximum d-axis current The minimum d-axis current is 1.7A It is -1.7A. Fig.40 In the case where the stator inductance does not undergo a step change, the q-axis current of the two methods The fluctuation range is -24A to -23A. s When the step jump occurs, the traditional PLL method The fluctuation is large, the maximum is -22A and the minimum is -25A.

[0153] like Fig.41 As shown, the stator inductance L s ±50% step jump, the d-axis voltage of the two methods and q-axis voltage There are fluctuations. Fig.41 (a) and Fig.41 (b) The q-axis voltage It can be seen that when the stator inductance jumps, the q-axis voltage fluctuation range in the two methods is similar, both from 56V to 63V. Stable in the range of -3V to 0V, the d-axis voltage in the traditional PLL method Oscillation occurs when the stator inductance jumps. By comparison, it can be seen that the system performance of the proposed MP-PLL is better than that of the traditional PLL. When the stator inductance jumps, there is no oscillation in the torque, d-axis current, q-axis current and d-axis / q-axis voltage.

[0154] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, the above-mentioned features are replaced with the technical features with similar functions disclosed in the embodiments of the present disclosure (but not limited to) to form a technical solution.

Claims

1. A PMSM rotor position estimation method based on model prediction PLL, characterized in that: The following steps are involved: Step 1: Design a model prediction phase-locked loop (MP-PLL) structure that combines a model prediction method (MP) with a phase-locked loop (PLL) to estimate the position of the PMSM rotor. The model prediction phase-locked loop (MP-PLL) structure specifically includes: removing the fixed-gain PI controller in the phase-locked loop (PLL), finding the best estimated rotor position, and optimizing the solution by minimizing the constructed cost function: specifically, a rotor position search algorithm based on discrete search is adopted to discretize the rotor position of the PMSM and solve it using the cost function; Step 2: Improve the rotor position search algorithm based on discrete search to realize PMSM rotor position estimation based on model prediction PLL.

2. The PMSM rotor position estimation method based on model prediction PLL according to claim 1, characterized in that: The step 1 specifically includes the following steps: Step 1.1: Calculate the discrete rotor position; assume the sampled current and voltage Define the initial rotor position θ in,i [k], construct cost function g in [k]; Use i and j to perform nested loop iterations, where i and j represent the number of iterations of the inner and outer loops, respectively. The discrete rotor position is expressed as: i ri,j [k]=θ in,i [k]+(j-4)Δθ i [k] (1) in: Step 1.2: Using the discrete rotor position, calculate the back EMF The d-axis component of the rotor is obtained, and then a cost function is constructed to perform the best estimation of the rotor position. The rotor position that minimizes the cost function in the back EMF calculation is selected, and an inner and outer nested iterative loop is performed according to the rotor position search algorithm. When the maximum number of iterations N = 7, after 64 iterations, the estimated value of the best rotor position is finally obtained. The cost function is expressed as:

3. The PMSM rotor position estimation method based on model prediction PLL according to claim 2, characterized in that: The rotor position search algorithm based on discrete search, when and g in When [k] = ∞, is the k-1th rotor position angle estimate. According to formula (2), the first iteration of the outer loop is performed, that is, when i = 0, Δθ0[k] = π / 4rad. Then, according to formula (1), 8 discrete rotor positions are generated, namely: 0rad, π / 4rad, π / 2rad, 3π / 4rad, -πrad, -3π / 4rad, -π / 2rad, -π / 4rad. Using the discrete rotor position θ ri,j [k] Calculate the back EMF Then, the rotor position angle that minimizes the cost function is selected as the optimal rotor position angle θ r,opt [k], and serves as the initial angle θ for the second iteration in,1 [k]; Assuming that the optimal rotor position angle calculated after the first iteration is π / 2rad, when the second iteration of the outer loop is performed, that is, i=1, Δθ1[k]=π / 8rad; then, according to formula (1), 8 new discrete rotor positions are generated; with continuous iterations, the optimal rotor position will be converged; finally, in the last iteration, that is, when i=N=7, the estimated value of the optimal rotor position is obtained; Since the output of the MP-PLL is an estimate of the optimal rotor position, when calculating the rotor speed, the change in the rotor position during the previous sampling period is calculated and divided by the sampling time T s , that is, the estimated rotor speed is: Among them, k and k-1 are the actual sampling time and the previous sampling time respectively; If the speed change of the motor system using MP-PLL is less than the set threshold, the number of loop iterations N is reduced, and the rotor position at the last sampling moment is used as the initial rotor position, that is, Reduce the number of iterations of the rotor position search algorithm; When the motor speed change is greater than a set threshold, a rotor position search algorithm is used. The rotor position search algorithm analyzes the factors affecting the calculation amount: it consists of two inner and outer nested loops and discretizes the rotor position.

4. The PMSM rotor position estimation method based on model prediction PLL according to claim 1, characterized in that: The improved rotor position search algorithm described in step 2 consists of inner and outer nested loops, which reduces the number of iterations and rewrites formula (2) as: