A method for designing a vector decoupling angle of an open-winding permanent magnet synchronous motor

Abstract

The application belongs to the technical field of motor control, and aims at the problem that current researches are mostly focused on static analysis of fixed decoupling angle, and provides a vector decoupling angle design method for open-winding permanent magnet synchronous motor, which comprises the following steps: constructing a mathematical model and a cost function of the open-winding permanent magnet synchronous motor; taking torque error, total motor loss and total harmonic current distortion rate as a first data set, and dividing a training set and a verification set; optimizing the weight and bias coefficient of the feedforward neural network; training the optimized feedforward neural network through the training set, and verifying through the verification set to obtain a trained neural network model; taking different working conditions and vector decoupling angles as a second data set, inputting into the trained neural network model, and outputting predicted performance index data; constructing a database; and realizing dynamic optimization of the vector decoupling angle of the open-winding permanent magnet synchronous motor under different working conditions through a look-up table method.

Description

Technical Field

[0001] This invention belongs to the field of motor control technology, specifically relating to a vector decoupling angle design method for an open-winding permanent magnet synchronous motor. Background Technology

[0002] Open-winding permanent magnet synchronous motors (PMSMs) are a special type of PMSM structure where the stator windings use an open connection instead of the traditional star or delta closed connection. This structure gives them unique control flexibility and performance advantages, and has attracted widespread attention in various fields in recent years. Open-winding PMSMs can be applied to drive systems for electric and hybrid vehicles, as well as energy recovery systems for hybrid vehicles. They feature high power density, specifically provided by high magnetic field strength in the permanent magnets, combined with efficient control from two sets of inverters, making them suitable for space-constrained vehicle environments. They also feature optimized energy feedback, with dual inverters allowing for flexible control of energy flow. Furthermore, they offer improved regenerative braking efficiency and redundancy design; specifically, when one inverter fails, the other inverter can still drive the motor, enhancing vehicle safety.

[0003] An open-winding permanent magnet synchronous motor is equipped with two sets of inverters, so its voltage vector is the result of the combined output of the two inverters. The essence of decoupling angle is to achieve zero-sequence current suppression and efficient utilization of voltage resources by adjusting the phase difference of the output voltage vectors of the two inverters. In a dual-inverter power supply system, if the voltage vectors of the two inverters are not properly coordinated, it will lead to common-mode voltage superposition, triggering zero-sequence circulating current, increasing copper losses, reducing efficiency, and even threatening the safety of power devices. By introducing a dynamic decoupling angle, the voltage synthesis path can be optimized on the space vector plane, effectively canceling the zero-sequence voltage component, while expanding the linear modulation region of the system and improving the DC bus voltage utilization rate. For example, the traditional vector decoupling angle generally uses a fixed 180° decoupling angle control strategy to maximize the number of output levels, but it needs to be combined with closed-loop control to dynamically suppress zero-sequence current; while the adaptive decoupling angle strategy can adjust the phase difference in real time according to the load, balancing efficiency and dynamic response. The proper design of the decoupling angle not only affects the steady-state performance of the motor (such as torque ripple, total motor loss, and current harmonic distortion rate), but also directly determines the fault-tolerant operation capability of the system under fault conditions.

[0004] Although the theoretical value of decoupling angles is widely recognized, their practical application still faces many challenges. First, the coupling relationship between the decoupling angle and modulation strategies, as well as zero-sequence current closed-loop control, is complex, making it difficult to establish accurate mathematical models for multi-objective optimization. Second, different application scenarios have different requirements for decoupling angles. For example, electric vehicle drive systems require rapid dynamic response, while wind power grid-connected systems prioritize low harmonics and high stability, necessitating highly adaptable decoupling angle strategies. Current research largely focuses on static analysis of fixed decoupling angles, while exploration of adaptive adjustment of dynamic decoupling angles and the integration of intelligent algorithms remains insufficient.

[0005] To address the aforementioned issues, this invention focuses on the optimization design and control strategy of the vector decoupling angle of open-winding permanent magnet synchronous motors. It proposes a dynamic decoupling angle cooperative modulation strategy to achieve high-performance output by reducing total motor losses, total harmonic distortion, and torque ripple. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a vector decoupling angle design method for open-winding permanent magnet synchronous motors.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] A method for designing the vector decoupling angle of an open-winding permanent magnet synchronous motor includes the following steps:

[0009] Step 1: Construct a mathematical model of the open-winding permanent magnet synchronous motor, and based on the mathematical model of the open-winding permanent magnet synchronous motor, design a cost function for the vector decoupling angle of the open-winding permanent magnet synchronous motor according to the control target performance requirements; this cost function includes the motor output performance that the user needs to consider, such as torque error, total motor loss and total harmonic current distortion rate.

[0010] Step 2: Set the application driving scenario and operating conditions of the open-winding permanent magnet synchronous motor, and determine the range of values ​​for torque, speed and vector decoupling angle;

[0011] Step 3: Based on the determined range of torque, speed and vector decoupling angle, traverse the different torques and decoupling angles at different speeds one by one, scan the operating points, and calculate the torque error, total motor loss and total harmonic current distortion rate corresponding to different vector decoupling angles under different operating conditions.

[0012] Step 4: Based on the calculated torque error, total motor loss and total harmonic current distortion rate, and combined with the multi-objective Osprey optimization algorithm, a trained neural network model is obtained.

[0013] Step 5: Use the speed, torque and vector decoupling angle under different operating conditions as the second dataset, and input the second dataset into the trained neural network model to output the predicted performance index data;

[0014] Step 6: Substitute the predicted performance index data under different working conditions into the cost function to calculate multiple cost function values. Then, the minimum cost function value under different working conditions and the corresponding vector decoupling angle data can be obtained, and a database can be constructed.

[0015] Step 7: Based on the constructed database, the vector decoupling angle of the open-winding permanent magnet synchronous motor under different operating conditions is dynamically optimized by using a lookup table method.

[0016] Preferably, in step 1, the mathematical model of the open-winding permanent magnet synchronous motor is expressed as follows:

[0017]

[0018] In the formula, i d with i q U represents the stator current along the d-axis and q-axis, respectively. d with u q L represents the stator voltage along the d-axis and q-axis, respectively. d With L q These represent the equivalent inductances along the d-axis and q-axis, respectively, where R is the stator resistance and ω is the inductance. e Let ψ be the electric angular velocity. f L represents the flux linkage of the permanent magnet, and L represents the inductance of the filter.

[0019] Preferably, in step 1, the expression for the cost function is:

[0020] J = 10 × |T e * -T e |+0.1×P loss +100×THD i

[0021] In the formula, T e * T represents the actual load torque. e For the actual measured torque; P loss Total motor losses, THD i denoted as Total Harmonic Distortion (THD).

[0022] Preferably, in step 2, the torque ranges from 0 to 60 N·m, and the speed ranges from 0 to 1500 r·min. -1 The vector decoupling angle ranges from 60 to 300°.

[0023] Preferably, step 4 includes the following steps:

[0024] Step 41: Take the calculated torque error, total motor loss and total harmonic current distortion rate as the first dataset, normalize the first dataset, and then divide the first dataset into a training set and a validation set.

[0025] Step 42: Optimize the weights and bias coefficients of the feedforward neural network based on the multi-objective Osprey optimization algorithm to obtain the optimized feedforward neural network;

[0026] Step 43: Train the optimized feedforward neural network using the training set and validate it using the validation set to obtain the trained neural network model.

[0027] Preferably, in step 42, the optimized feedforward neural network includes an input layer, hidden layer 1, hidden layer 2 and an output layer connected in sequence;

[0028] Hidden layer 1 and hidden layer 2 are connected using the Sigmoid activation function, and hidden layer 2 and the output layer are connected using the ReLU activation function.

[0029] Preferably, the input layer has 3 neurons, the hidden layer 1 has 11 neurons, the hidden layer 2 has 3 neurons, and the output layer has 3 neurons.

[0030] Preferably, in step 5, the predicted performance index data includes torque error, total motor loss, and total harmonic distortion of current.

[0031] Compared with the prior art, the advantages of this invention are as follows:

[0032] (1) The present invention adopts the design of feedforward neural network based on multi-objective Osprey optimization algorithm. Compared with traditional feedforward neural network, it can make the neural network converge faster, predict more accurately, and improve the robustness and generalization of the neural network.

[0033] (2) The present invention can adjust the optimal vector decoupling angle in real time according to the actual working conditions to improve the output performance of the system. Attached Figure Description

[0034] Figure 1 This is a flowchart of the vector decoupling angle design method for open-winding permanent magnet synchronous motors proposed in this embodiment of the invention;

[0035] Figure 2 This is a schematic diagram illustrating the design concept of the predictive control weight factor for the mathematical model of the open-winding permanent magnet synchronous motor in this embodiment of the invention.

[0036] Figure 3 This is a diagram of the drive system of the open-winding permanent magnet synchronous motor in an embodiment of the present invention;

[0037] Figure 4 This is a loss model diagram of iron loss and copper loss of an open-winding permanent magnet synchronous motor in an embodiment of the present invention.

[0038] Figure 5 This is a structural diagram of the feedforward neural network in an embodiment of the present invention;

[0039] Figure 6 This is an evolutionary iteration diagram of the multi-objective Osprey optimization algorithm in this embodiment of the invention;

[0040] Figure 7 This is a comparison chart of the effects of the optimized feedforward neural network for multi-target ospreys and the traditional feedforward neural network in this embodiment of the invention;

[0041] Figure 8 This is a diagram showing the optimized feedforward neural network training, prediction, and verification in an embodiment of the present invention.

[0042] Figure 9 A comparison diagram of the cost function J under operating condition 1 provided in the embodiments of the present invention;

[0043] Figure 10 A comparison diagram of the cost function J under working condition 2 provided in the embodiments of the present invention;

[0044] Figure 11 The cost function J is shown in the comparison diagram for working condition 3 provided in this embodiment of the invention;

[0045] Figure 12 A comparison diagram of the cost function J under operating condition 4 provided in the embodiments of the present invention; Detailed Implementation

[0046] The following will be described in conjunction with embodiments of the present invention. Figures 1 to 12 The technical solutions in the embodiments of the present invention are clearly and completely described herein. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0047] like Figure 1 and Figure 2 As shown in the figure, this invention proposes a vector decoupling angle design method for an open-winding permanent magnet synchronous motor, which specifically includes the following steps:

[0048] Step 1: Construct a mathematical model of the open-winding permanent magnet synchronous motor, and based on the mathematical model of the open-winding permanent magnet synchronous motor, design a cost function for the vector decoupling angle of the open-winding permanent magnet synchronous motor according to the control target performance requirements. This cost function includes the motor output performance that the user needs to consider, such as torque error (Teerror), total motor loss (Ploss), and total harmonic distortion rate (THDi).

[0049] Table 1 Parameters of Open-Winding Permanent Magnet Synchronous Motor

[0050]

[0051] like Figure 3 As shown in Table 1, an open-winding permanent magnet synchronous motor is used as the controlled object in this embodiment of the invention. The mathematical model of the open-winding permanent magnet synchronous motor (i.e., the stator voltage equation under the dq axis) is constructed as follows:

[0052]

[0053] In the formula, i d with i q U represents the stator current along the d-axis and q-axis, respectively. d with u q L represents the stator voltage along the d-axis and q-axis, respectively. d With L q Let represent the equivalent inductances of the d-axis and q-axis, respectively, and L be the filter inductance value. Since the equivalent inductances of the d-axis and q-axis of a surface-mount motor are equal, let L be... d =L q =L, R is the stator resistance, ω e Let ψ be the electric angular velocity. f It is a permanent magnet flux linkage.

[0054] The torque expression in the dq coordinate system is:

[0055]

[0056] Among them, P n P is the number of pole pairs of the motor. dq This refers to the output power of the motor.

[0057] Based on the established mathematical model of the open-winding permanent magnet synchronous motor, the following is obtained: Figure 4 The loss models for iron and copper losses in a motor are shown below. Copper loss is equivalent to the power consumed when stator current flows through stator resistance, and iron loss is equivalent to the power consumed when iron loss branch current flows through equivalent iron resistance. The expressions for the loss models of iron and copper losses in a motor are as follows:

[0058]

[0059] Among them, i wd and i wq These are the useful work components of the stator current along the d-axis and q-axis, respectively.

[0060] When the system is stable, the speed and torque are constant, and the total motor loss can be expressed as:

[0061]

[0062] Among them, P loss P represents the total power consumption of the motor. Cu For motor copper losses, P Fe For the iron loss of the motor, R c For equivalent iron resistance, i cd and i cq These are the iron loss components of the stator current along the d-axis and q-axis, respectively.

[0063] Derivation of i using Kirchhoff's Current Law and Voltage Law wd i wq i cd i cq The relationship between these factors allows us to convert the expression for the total motor loss into:

[0064]

[0065] Among them, P loss T represents the total power consumption of the motor. e The torque is in the dq coordinate system.

[0066] The calculation method for the total harmonic distortion (THD) of current is as follows:

[0067]

[0068] In the formula, F1 is the fundamental frequency amplitude, and F is the total harmonic frequency amplitude.

[0069] A cost function is defined to evaluate the performance of the selected vector decoupling angle. To optimize the overall operating performance of the motor, torque error, total motor loss, and total harmonic distortion rate are included in the cost function. Therefore, the expression of the cost function is:

[0070] J = 10 × |T e * -T e |+0.1×P loss +100×THD i

[0071] In the formula, T e * T represents the actual load torque. e The actual measured torque; the difference is T.eerror T eerror For torque error, P loss Total motor losses, THD i The coefficients preceding the total harmonic distortion (THD) of the current, torque error, and total motor loss are used to unify the three terms in the expression to a range of 0-10, so as to comprehensively reflect the various output performances of the motor.

[0072] Step 2: Import the above model and motor parameters into Matlab, and follow the steps as follows: Figure 4 The simulation model shown is constructed to define the application driving scenario of the open-winding permanent magnet synchronous motor, specifically determining the range of torque, speed, and vector decoupling angle. In this embodiment, the torque range is 0-60 N·m with a step size of 0.1875; the speed range is 0-1500 r·min. -1 The step size is 10; the vector decoupling angle ranges from 60 to 300° with a step size of 1.

[0073] Step 3: Based on the determined range of torque, speed and vector decoupling angle, traverse the different torques and decoupling angles at different speeds one by one, scan the operating points, and calculate the torque error, total motor loss and total harmonic current distortion rate corresponding to different vector decoupling angles under different operating conditions.

[0074] Step 4: Based on the calculated torque error, total motor loss, and total harmonic current distortion rate, and combined with the multi-objective Osprey optimization algorithm, a trained neural network model is obtained, which specifically includes the following steps:

[0075] Step 41: Take the calculated torque error, total motor loss and total harmonic current distortion rate as the first dataset, normalize the first dataset, and then divide the first dataset into a training set and a validation set; specifically, divide the first dataset into a training set and a validation set in a 7:3 ratio.

[0076] Step 42: Optimize the weights and bias coefficients of the feedforward neural network based on the multi-objective Osprey optimization algorithm to enable it to converge quickly and reduce prediction error; use speed, torque, and decoupling angle as input layers, and torque error Teerror, total motor loss Ploss, and total harmonic distortion rate THDi as output layers, with minimizing root mean square error as the evaluation criterion, and finally use the combination of motor operating conditions and decoupling angle as a set of optimal values ​​to obtain the optimized feedforward neural network;

[0077] like Figure 5 and Figure 6As shown, this embodiment of the invention applies a multi-objective osprey optimization algorithm to optimize the weights and bias coefficients of a feedforward neural network (randomly initializing the osprey population, global exploration (first stage: location identification and fishing), and local exploitation (second stage: bringing fish to a suitable location) to optimize the parameters of the feedforward neural network). Specifically, the population size is set to 100, the dimension is set to 3, and the boundary values ​​are set to [-20, 20]. After setting, the multi-objective osprey optimization algorithm is used to optimize the weights and bias coefficients of the feedforward neural network. The training iterations are 1000, the learning rate is 0.01, and the minimum objective is 0.00001. The optimized weights and bias coefficients are substituted into the feedforward neural network to obtain the optimized feedforward neural network. The minimum root mean square error (RMSE) is used as the evaluation index of the feedforward neural network.

[0078] In this embodiment of the invention, the optimized feedforward neural network includes an input layer, a hidden layer 1, a hidden layer 2, and an output layer connected in sequence.

[0079] Hidden layer 1 and hidden layer 2 are connected using the Sigmoid activation function, and hidden layer 2 and the output layer are connected using the ReLU activation function.

[0080] Specifically, the input layer has 3 neurons, hidden layer 1 has 11 neurons, hidden layer 2 has 3 neurons, and the output layer has 3 neurons.

[0081] In this embodiment of the invention, the Sigmoid activation function is as follows:

[0082]

[0083] The ReLU activation function is shown below:

[0084] f(x) = max(0,x)

[0085] The RMSE expression is as follows:

[0086]

[0087] Where y is the actual value, This is the predicted value from the feedforward neural network.

[0088] Figure 7 and Figure 8 To optimize the process diagram, by Figure 7 and Figure 8 It can be seen that the feedforward neural network optimized by the multi-objective Osprey optimization algorithm has a smaller error and is more stable. The error of the feedforward neural network optimized by the multi-objective Osprey algorithm fluctuates within [-1, +1], which clearly demonstrates the superiority of this invention.

[0089] Step 43: Train the optimized feedforward neural network using the training set and validate it using the validation set. If the output network meets the requirements, i.e., the error is less than 0.00001, training can be stopped to obtain the trained neural network model. Otherwise, adjust the learning rate, network structure and other parameters and continue training until the output requirements are met.

[0090] Step 5: Take the speed, torque and vector decoupling angle under different operating conditions as a second dataset, and input the second dataset into the trained neural network model to output the predicted performance index data, namely torque error, total motor loss and total harmonic distortion rate of current.

[0091] Step 6: Substitute the predicted performance index data under different working conditions into the cost function to calculate multiple cost function values. Then, the minimum cost function value under different working conditions and the corresponding vector decoupling angle data can be obtained, and a database can be constructed.

[0092] Step 7: Dynamically optimize the vector decoupling angle of the open-winding permanent magnet synchronous motor under different operating conditions using a lookup table method. During operation, based on the constructed database, and according to the current operating conditions (i.e., different speeds and torques), the system queries the vector decoupling angle corresponding to the minimum cost function value and sets this angle as the vector decoupling angle for the open-winding permanent magnet synchronous motor, thus completing the dynamic optimization.

[0093] This invention uses a constructed database to predict the optimal vector decoupling angle for four different working conditions, and the results are shown in Table 2 below.

[0094] Table 24 shows the prediction results of the optimal vector decoupling angle for different working conditions.

[0095]

[0096] Figures 9 to 12 A comparison chart of the cost function J under different operating conditions, as shown in Table 2 and... Figures 9 to 12 The results show that this invention verifies that the decoupling angle should change under different operating conditions, rather than remaining fixed. Different vector decoupling angles should be designed for different operating conditions to achieve optimal performance of the open-winding permanent magnet synchronous motor drive system. This method is better than the traditional fixed vector decoupling angle of 180°, as it can adaptively adjust the decoupling angle after changes in operating conditions, maintaining optimal system performance. Figures 9 to 12As can be seen, under operating condition 1, the system performance of the present invention is improved by about 15% compared with the traditional decoupling angle control method; under operating condition 2, the system performance of the present invention is improved by about 13% compared with the traditional decoupling angle control method; under operating condition 3, the system performance of the present invention is improved by about 16% compared with the traditional decoupling angle control method; under operating condition 4, the system performance of the present invention is improved by about 8% compared with the traditional decoupling angle control method. In summary, the effectiveness of the vector decoupling angle design method for open-winding permanent magnet synchronous motors proposed in this invention can be seen.

[0097] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for designing the vector decoupling angle of an open-winding permanent magnet synchronous motor, characterized in that, Includes the following steps: Step 1: Construct a mathematical model of an open-winding permanent magnet synchronous motor, and design a cost function for the vector decoupling angle of the open-winding permanent magnet synchronous motor based on the mathematical model; Step 2: Set the application driving scenario and operating conditions of the open-winding permanent magnet synchronous motor, and determine the range of values ​​for torque, speed and vector decoupling angle; Step 3: Based on the determined range of torque, speed and vector decoupling angle, traverse the different torques and decoupling angles at different speeds one by one, and calculate the torque error, total motor loss and total harmonic current distortion rate corresponding to different vector decoupling angles under different operating conditions; Step 4: Based on the calculated torque error, total motor loss and total harmonic current distortion rate, and combined with the multi-objective Osprey optimization algorithm, a trained neural network model is obtained. Step 5: Use the speed, torque and vector decoupling angle under different operating conditions as the second dataset, and input the second dataset into the trained neural network model to output the predicted performance index data; Step 6: Substitute the predicted performance index data under different working conditions into the cost function to calculate multiple cost function values. Then, the minimum cost function value under different working conditions and the corresponding vector decoupling angle data can be obtained, and a database can be constructed. Step 7: Based on the constructed database, the vector decoupling angle of the open-winding permanent magnet synchronous motor under different operating conditions is dynamically optimized by using a lookup table method.

2. The vector decoupling angle design method for an open-winding permanent magnet synchronous motor according to claim 1, characterized in that, In step 1, the mathematical model of the open-winding permanent magnet synchronous motor is expressed as follows: ； In the formula, i d with i q U represents the stator current along the d-axis and q-axis, respectively. d with u q L represents the stator voltage along the d-axis and q-axis, respectively. d With L q Let represent the equivalent inductance along the d-axis and q-axis, respectively, and R be the stator resistance. Electric angular velocity, L represents the flux linkage of the permanent magnet, and L is the inductance of the filter.

3. The vector decoupling angle design method for an open-winding permanent magnet synchronous motor according to claim 1, characterized in that, In step 1, the expression for the cost function is: ； In the formula, T e * T represents the actual load torque. e For the actual measured torque; P loss Total motor losses, THD i denoted as Total Harmonic Distortion (THD).

4. The vector decoupling angle design method for an open-winding permanent magnet synchronous motor according to claim 1, characterized in that, In step 2, the torque ranges from 0 to 60 N·m, and the speed ranges from 0 to 1500 r·min. -1 The vector decoupling angle ranges from 60 to 300°.

5. The vector decoupling angle design method for an open-winding permanent magnet synchronous motor according to claim 1, characterized in that, Step 4 includes the following steps: Step 41: Take the calculated torque error, total motor loss and total harmonic current distortion rate as the first dataset, normalize the first dataset, and then divide the first dataset into a training set and a validation set. Step 42: Optimize the weights and bias coefficients of the feedforward neural network based on the multi-objective Osprey optimization algorithm to obtain the optimized feedforward neural network; Step 43: Train the optimized feedforward neural network using the training set and validate it using the validation set to obtain the trained neural network model.

6. The vector decoupling angle design method for an open-winding permanent magnet synchronous motor according to claim 5, characterized in that, In step 42, the optimized feedforward neural network includes an input layer, hidden layer 1, hidden layer 2 and an output layer connected in sequence; Hidden layer 1 and hidden layer 2 are connected using the Sigmoid activation function, and hidden layer 2 and the output layer are connected using the ReLU activation function.

7. The vector decoupling angle design method for an open-winding permanent magnet synchronous motor according to claim 6, characterized in that, The input layer has 3 neurons, hidden layer 1 has 11 neurons, hidden layer 2 has 3 neurons, and the output layer has 3 neurons.

8. The vector decoupling angle design method for an open-winding permanent magnet synchronous motor according to claim 1, characterized in that, In step 5, the predicted performance index data includes torque error, total motor loss, and total harmonic distortion of current.

Images