A high-efficiency torque distribution control method for a modular multiphase motor and application thereof

Abstract

This invention discloses a high-efficiency torque distribution control method for a modular multiphase motor and its application. It provides n torque distribution conditions, specifically including power output from a single motor, power output from two motors (average), ..., power output from n-1 motors (average), and power output from n motors (average). The efficiency values ​​corresponding to the n torque distribution conditions are calculated using a speed-torque-efficiency relationship fitted by a BP neural network. The torque distribution condition that consistently maintains the highest efficiency value is used as the torque distribution strategy executed by the modular multiphase motor. Obtaining the speed-torque-torque distribution condition relationship helps to expand the high-efficiency range and significantly improves the efficiency in the low-torque range. Furthermore, the torque distribution strategy is obtained through sample training using a BP neural network, making it efficient, accurate, and reliable, significantly improving the control accuracy and efficiency level of the modular multiphase motor.

Description

Technical Field

[0001] This invention belongs to the field of motor control, and specifically relates to a high-efficiency torque distribution control method for a modular multiphase motor. This invention also relates to the application of this high-efficiency torque distribution control method for a modular multiphase motor. Background Technology

[0002] Modular multiphase motor drive control systems are highly favored due to their flexible control, high power factor, and strong fault tolerance, and have become an important direction for the development of high-power propulsion motor systems both domestically and internationally.

[0003] In practical applications, when dealing with multiple independent three-phase motor units in a multi-module multiphase motor, it is necessary to allocate and manage each independent three-phase motor unit. Otherwise, problems such as inconsistent states of each independent three-phase motor unit or low operating efficiency may occur. The applicant believes that it is necessary to introduce an optimal allocation strategy (see CN115416496A).

[0004] As the applicant further conducted in-depth research and experiments, it was found that in the optimal allocation calculation of multiple independent three-phase motor units, it is necessary to deal with complex data, which makes it difficult for the system to make fast and accurate allocation decisions, thereby affecting the system's sensitivity and clarity.

[0005] Therefore, the applicant seeks a technical solution to address the above-mentioned technical problems. Summary of the Invention

[0006] In view of this, the purpose of this invention is to provide a high-efficiency torque distribution control method for modular multiphase motors and its application, which can help expand the high-efficiency range and significantly improve the efficiency in the low-torque range; at the same time, the torque distribution strategy is obtained by training samples using a BP neural network, which is efficient, accurate and reliable, and significantly improves the control accuracy and efficiency level of modular multiphase motors.

[0007] Before proposing the technical solution of this invention, the applicant conducted extensive analysis and thought, and found that when distributing torque among n working motors with the same parameters, under certain speed conditions, when the relationship between the loss of a single multiphase motor unit and its torque is a concave function, the efficiency of distributing torque evenly among n multiphase motor units is higher than that of distributing torque unevenly among n working motors.

[0008] To verify the above points of this application, this application takes the distribution of the total required torque T between two working motors (one working motor is equivalent to one multiphase motor unit) as an example. Under the condition that the loss value of each motor is a concave function of its torque, The constraint split torque T, The allocation strategy is optimal. Please see further details.Figure 1 As shown, at a certain speed, the total motor loss is a concave function of its torque:

[0009] Let point A be the torque-bisector operating point, i.e. The loss is .

[0010] The torque at point B is The loss is Then the torque at point C is The loss is The total loss at points B and C is ,

[0011] The unevenly distributed torque loss value is .

[0012] right Differentiation yields .because For a concave function of T, therefore when From time to time .

[0013] Therefore, when hour, ,at this time There is a minimum value, which confirms that when the loss value is a concave function of torque, the total required torque T is distributed between the two working motors, and the average distribution between the two motors results in the lowest loss and the highest efficiency.

[0014] Similarly, it can be confirmed that when the total required torque T is distributed among 3, 4 or more working motors, the optimal distribution scheme is still to distribute the torque evenly among the working motors.

[0015] Therefore, the technical solution adopted by the present invention is as follows:

[0016] A high-efficiency torque distribution control method for a modular multiphase motor is disclosed. The modular multiphase motor comprises n independent multiphase motor units, each driven and controlled by its corresponding independent driver, where n ≥ 2. All independent multiphase motor units are identical. The high-efficiency torque distribution control method provides n torque distribution conditions, specifically including power output from a single motor, power output from two motors (average), ..., power output from n-1 motors (average), and power output from n motors (average). The efficiency value corresponding to each of the n torque distribution conditions is calculated using a speed-torque-efficiency relationship fitted by a BP neural network. The torque distribution condition where the efficiency value consistently remains at its highest value is used as the torque distribution strategy executed by the modular multiphase motor, thus obtaining the speed-torque-torque distribution condition relationship.

[0017] This application further tested the changes in the relationship curves between motor losses and torque at different speeds: please refer to [link to relevant documentation]. Figure 2 , Figure 3 , Figure 4 and Figure 5 The curves showing the relationship between measured motor losses and torque at different speeds are presented. It can be seen that at low and medium speeds, the motor losses and torque have a concave function relationship; while at medium and high speeds and medium torque, the losses of each independent multiphase motor unit have a slightly convex function property with respect to torque, and the overall losses of the independent multiphase motor unit are approximately a linear function of torque.

[0018] Therefore, preferably, when the efficient torque distribution control method is running, the speed of the motor is not higher than 4500 r / min, so that the motor loss and its torque have a concave function relationship or are close to a concave function relationship; more preferably, when the efficient torque distribution control method is running, the speed of the motor is not higher than 2000 r / min, so that the motor loss and its torque have a concave function relationship or are close to a concave function relationship.

[0019] Preferably, in the speed-torque-torque distribution condition relationship, the dividing line between the different n torque distribution conditions is a non-linear curve; wherein, several coordinate points corresponding to the dividing line can be found by exhaustive search, and then the non-linear curve is fitted to determine the dividing line.

[0020] Preferably, in the speed-torque-torque distribution condition relationship, the boundary line between the different n torque distribution conditions is a non-linear curve; wherein, by generating multiple data of speed, torque and their corresponding torque distribution conditions, based on the data as sample points, the trained BP neural network is used to classify the sample points, and the boundary line is directly determined through the classification.

[0021] Preferably, the BP neural network includes an input layer, a hidden layer, and an output layer; wherein the input data of the input layer is rotational speed and torque, the output layer is a softmax function with n output neuron nodes, and the hidden layer has multiple neuron nodes.

[0022] Preferably, the number of neurons in the hidden layer is greater than n.

[0023] Preferably, the accuracy of classifying the sample points using the neural network is not less than 99%.

[0024] Preferably, the application of the high-efficiency torque distribution control method described above is as a motor torque distribution control method for a modular multiphase motor used in vehicles.

[0025] Preferably, the modular multiphase motor for vehicles includes n independent multiphase motor units, each driven and controlled by its corresponding independent driver, and each independent driver powered by its corresponding independent power supply. Each independent multiphase motor unit, each independent driver, and each independent power supply are connected to an on-board computer. The control method includes: the on-board computer has a preset optimal efficiency algorithm, and based on the power requirements of the vehicle and with satisfying the optimal efficiency algorithm as the driving target, it calculates the number of independent multiphase motor units to be operated and their corresponding output power.

[0026] It should be noted that in this application, "BP" is an abbreviation for "back propagation," and "r / min" refers to "revolutions per minute."

[0027] This application calculates the efficiency values ​​corresponding to the n torque distribution conditions by fitting the speed-torque-efficiency relationship through a BP neural network. The torque distribution condition that consistently maintains the highest efficiency value is used as the torque distribution strategy executed by the modular multiphase motor. In other words, the speed-torque-torque distribution condition relationship is obtained. Practical verification shows that the torque distribution strategy provided by this application can help expand the high-efficiency range and significantly improve the efficiency in the low-torque range. At the same time, the torque distribution strategy is obtained by training the BP neural network with samples, which is efficient, accurate and reliable, and significantly improves the control accuracy and efficiency level of the modular multiphase motor. Attached Figure Description

[0028] Figure 1 This is a graph showing the concave function of motor loss value and torque in this application;

[0029] Figure 2 This is a graph showing the relationship between motor loss and torque when the motor speed is 1875 r / min in this application.

[0030] Figure 3 This is a graph showing the relationship between motor loss and torque when the motor speed is 3750 r / min in this application.

[0031] Figure 4 This is a graph showing the relationship between motor loss and torque when the motor speed is 5625 r / min in this application.

[0032] Figure 5 This is a graph showing the relationship between motor loss and torque when the motor speed is 7500 r / min in this application.

[0033] Figure 6 This is a diagram showing the relationship between speed, torque, and torque distribution under specific embodiments of this application;

[0034] Figure 7a It is the classification confusion matrix of the BP neural network training set;

[0035] Figure 7b It is the classification confusion matrix of the validation set;

[0036] Figure 7c It is the classification confusion matrix of the test set;

[0037] Figure 7d It is the classification confusion matrix for all sets (including training set, validation set, and test set);

[0038] Figure 8 It is a map of the total motor system when the four motors have the average output under all operating conditions;

[0039] Figure 9 This is the overall motor system map corresponding to the decision-making through a BP neural network in the specific implementation of this application;

[0040] Figure 10 This is a schematic diagram of the structure of the BP neural network according to a specific embodiment of this application;

[0041] Figure 11 This is a graph showing the error change of the BP neural network during the training process under a specific embodiment of this application;

[0042] Figure 12 It is the error value between the actual motor efficiency value and the fitted motor efficiency value under the specific implementation of this application;

[0043] Figure 13 This is a motor MAP diagram obtained based on experimental test data in a specific embodiment of this application;

[0044] Figure 14 This is a motor MAP diagram obtained based on BP neural network fitting in a specific embodiment of this application. Detailed Implementation

[0045] This invention discloses an efficient torque distribution control method for a modular multiphase motor. The modular multiphase motor includes n independent multiphase motor units, each driven and controlled by its corresponding independent driver, where n≥2. All independent multiphase motor units are identical. The efficient torque distribution control method provides n torque distribution conditions, specifically including power output from a single motor, power output from two motors (average), ..., power output from n-1 motors (average), and power output from n motors (average). The efficiency value corresponding to each of the n torque distribution conditions is calculated using a speed-torque-efficiency relationship (which can be represented by a motor MAP diagram) fitted by a BP neural network. The torque distribution condition where the efficiency value consistently remains at its highest value is used as the torque distribution strategy executed by the modular multiphase motor, thus obtaining the speed-torque-torque distribution condition relationship.

[0046] Preferably, in this embodiment, when the high-efficiency torque distribution control method is running, the motor speed is not higher than 4500 r / min, so that the motor loss and its torque have a concave function relationship or are close to a concave function relationship; preferably, in this embodiment, when the high-efficiency torque distribution control method is running, the motor speed is not higher than 2000 r / min, more preferably the motor speed is not higher than 1875 r / min, so that the motor loss and its torque have a concave function relationship or are even closer to a concave function relationship.

[0047] In this embodiment, in the speed-torque-torque distribution working condition relationship, the dividing line between different n torque distribution working conditions is a non-linear curve; preferably, several coordinate points corresponding to the dividing line can be found by exhaustive search, and then the non-linear curve is fitted to determine the dividing line.

[0048] Further preferably, to achieve simpler and more efficient decision-making, in this embodiment, the boundary line between the different n torque distribution conditions in the speed-torque-torque distribution relationship is a non-linear curve; wherein, by generating multiple data points of speed, torque and their corresponding torque distribution conditions, and using this data as sample points, a trained BP neural network is used to classify the sample points, and the boundary line is directly determined through this classification. Further preferably, the BP neural network includes an input layer, a hidden layer and an output layer; wherein, the input data of the input layer is speed and torque, the output layer is a softmax function with n output neuron nodes, and the hidden layer has multiple neuron nodes; to improve the classification accuracy, through a large number of experiments, specifically preferably, in this embodiment, the number of neuron nodes in the hidden layer is greater than n; preferably, in this embodiment, the accuracy of classifying sample points by the neural network is not less than 99%.

[0049] Preferably, the application of the high-efficiency torque distribution control method described above is as a motor torque distribution control method for a modular multiphase motor used in vehicles.

[0050] Preferably, the modular multiphase motor for vehicles includes n independent multiphase motor units, each driven and controlled by its corresponding independent driver, and each independent driver powered by its corresponding independent power supply. Each independent multiphase motor unit, each independent driver, and each independent power supply are connected to an on-board computer. The control method includes: the on-board computer has a preset optimal efficiency algorithm, and based on the vehicle's power requirements and with meeting the optimal efficiency algorithm as the driving target, it calculates the number of independent multiphase motor units to be operated and their corresponding output power.

[0051] Since the torque-speed-efficiency diagram of a motor is a two-dimensional strongly nonlinear function, the prediction accuracy of multidimensional nonlinear functions fitted by existing conventional methods is often low. Therefore, control based on the motor MAP diagram is usually difficult to achieve the optimal control strategy for the motor. Preferably, this embodiment also proposes a high-precision fitting method for the motor MAP diagram, which uses the motor efficiency fitting value output by the trained BP neural network to create the motor MAP diagram. Preferably, in this embodiment, the horizontal axis of the motor MAP diagram is the motor speed, and the vertical axis is the motor torque, based on the motor efficiency fitting value corresponding to the motor speed and motor torque. Preferably, in this embodiment, the motor MAP diagram can be specifically created using MATLAB software, Python language, or similar software tools.

[0052] In this embodiment, the training method for the BP neural network includes: training the BP neural network using actual motor efficiency values ​​under different speeds and torques as samples, so that the error between the fitted motor efficiency value obtained by the BP neural network and the actual motor efficiency value is within a preset error range. Preferably, in this embodiment, the preset error range is set as follows: the difference between the fitted motor efficiency value and the actual motor efficiency value is ±1.5 × 10⁻⁶. -3 .

[0053] Preferably, please refer to [reference needed]. Figure 1 The BP neural network structure shown in this embodiment includes an input layer and an output layer. The input of the input layer is the motor speed value and the motor torque value. There is one or more hidden layers between the input layer and the output layer. The output of the output layer is the motor efficiency fitting value.

[0054] After multiple training tests by the applicant, the following preferred solution was obtained: In this embodiment, the number of hidden layers is 2, including a first hidden layer and a second hidden layer; more preferably, the number of neurons in the first hidden layer is 10 and the number of neurons in the second hidden layer is 5; based on the structure of this BP neural network, it has been confirmed that a high-precision motor efficiency fitting effect can be obtained.

[0055] It should be noted that, in this embodiment, the neuron parameters in the BP neural network are determined by training the neural network based on the data obtained from the hidden layer after the input layer inputs data to the hidden layer. The parameters of specific nodes do not have a clear physical meaning, therefore, this embodiment will not elaborate on them.

[0056] Preferably, through multiple experimental verifications, in this embodiment, the number of training iterations for training the BP neural network is at least 300, more preferably 380-480, ensuring that the error function value of the BP neural network training set decreases to 10. -7 The magnitude ensures a high-precision fitting effect for the motor MAP diagram.

[0057] This embodiment trains a BP neural network using actual motor efficiency values ​​under different speeds and torques as samples, resulting in a BP neural network with high-precision motor efficiency fitting. Actual testing has verified that the motor efficiency fitting values ​​output by this BP neural network can be used as the data basis to obtain a high-precision motor MAP, providing a reliable and stable data foundation for the optimal efficiency algorithm of modular multiphase motors for vehicles.

[0058] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this invention.

[0059] Please refer to CN115416496A, the earlier application of this application, which discloses a drive system for a modular three-phase motor for vehicles. Specifically, it includes a modular three-phase motor (specifically a permanent magnet synchronous motor) with n independent three-phase motor units, n independent drivers, n independent power supplies, and a single on-board computer. Each independent three-phase motor unit is driven and controlled by its corresponding independent driver, and each independent driver is powered by its corresponding independent power supply. Each independent three-phase motor unit, each independent driver, and each independent power supply are connected to the on-board computer. In this embodiment, the modular three-phase motor for vehicles adopts a fractional-slot concentrated winding type motor. The operation of the k-th independent three-phase motor unit involves its three-phase windings being set within its corresponding fixed k-th stator sector. The k-th independent driver is only used to drive the k-th independent three-phase motor unit, and the k-th independent power supply is only used to power the k-th independent driver. Here, 1 ≤ k ≤ n, and n and k are positive integers greater than 1.

[0060] In this embodiment, the on-board computer collects the real-time temperature of n independent three-phase motor units and the real-time charge of n independent power supplies. The on-board computer has a motor efficiency cloud map (specifically including a motor speed-torque efficiency map) and a driver efficiency cloud map pre-installed. In practical applications, AI algorithms can be used to obtain the motor efficiency cloud map and the driver efficiency cloud map.

[0061] In this embodiment, the remaining technical solutions for the modular three-phase motor and its drive system for vehicles can be directly referred to the description in CN115416496A, and will not be repeated in this embodiment.

[0062] Specifically, in this embodiment, n=4, meaning the motor specifically includes 4 independent three-phase motor units, modular multi-phase motors (hereinafter referred to as motors), and the motors adopt the high-efficiency torque distribution control method described below:

[0063] In this embodiment, the motor's operating speed is controlled within a range where its loss value and torque exhibit a concave functional relationship. The optimal torque distribution strategy includes four torque distribution conditions: power output from a single motor, power output from two motors (average), power output from three motors (average), and power output from four motors (average). The efficiency values ​​corresponding to these four torque distribution conditions are calculated using the speed-torque-efficiency relationship fitted by a BP neural network. The torque distribution condition that consistently maintains the highest efficiency value is selected as the torque distribution strategy executed by the modular multiphase motor, resulting in the speed-torque-torque distribution condition relationship:

[0064] Please see Figure 6As shown, the red dotted operating condition area B1 represents the highest efficiency when powered by a single motor; the blue dotted operating condition area B2 represents the highest efficiency when powered by an average of two motors; the green dotted operating condition area B3 represents the highest efficiency when powered by an average of three motors; and the yellow dotted operating condition area B4 represents the highest efficiency when powered by an average of four motors. Through this... Figure 6 It can be seen that different operating conditions have their corresponding optimal torque distribution methods. Therefore, when determining the boundary between different torque distribution conditions, it can be regarded as a classification problem, that is, the input is the speed and torque, and the output is the torque distribution method corresponding to that speed and torque. This classification problem can be handled by a BP neural network, and the following scheme can be adopted:

[0065] A BP neural network consists of an input layer, a hidden layer, and an output layer. The input layer has two neurons for inputting rotational speed and torque. The hidden layer has twelve neurons. The output layer uses a softmax function, where the sum of the output values ​​of its four neurons is 1. The output value of each neuron represents the probability that the neural network determines the class to which it belongs.

[0066] In practical implementation, multiple speed-torque-torque distribution mode data are generated, that is, multiple data points for speed, torque, and their corresponding torque distribution conditions are generated. The mode with a single motor output is encoded as [1,0,0,0], the mode with two motors averaging output is encoded as [0,1,0,0], the mode with three motors averaging output is encoded as [0,0,1,0], and the mode with four motors averaging output is encoded as [0,0,0,1]. The generated data is used to train a BP neural network. After training, the following can be obtained: Figure 7a , Figure 7b , Figure 7c as well as Figure 7d The results shown are obtained using either MATLAB software or Python.

[0067] Please refer to the following for details. Figure 7a , Figure 7b , Figure 7c as well as Figure 7d As shown, where, Figure 7a The classification confusion matrix of the BP neural network training set; Figure 7b The classification confusion matrix for the validation set; Figure 7c The classification confusion matrix for the test set; Figure 7d This is the classification confusion matrix for all sets (including training, validation, and test sets); it should be noted that... Figure 7aThe squares "99.3%, 0.7%" shown in the lower right corner mean that 99.3% of the sample points were successfully classified into the corresponding class, and 0.7% represents the proportion of sample points that were not successfully classified. In other words, the trained BP neural network provided in this application embodiment has a classification accuracy of 99.3% for the number of sample points, and can accurately output the corresponding torque distribution method after inputting the speed and torque.

[0068] Furthermore, the very few sample points that were successfully classified (i.e., the sample points that were misclassified by the BP neural network) mainly exist at the switching points between different torque distribution methods. Since the difference in motor efficiency at the switching points between the two torque distribution methods is small, even if there are a very small proportion of misclassified sample points, it can be considered that the misclassification of the BP neural network decision has a small impact on motor efficiency and will not affect the optimal efficiency control of the motor in this application. At the same time, the classification method provided in this embodiment is simple, convenient and efficient.

[0069] Please refer to the following: Figure 8 and Figure 9 As shown, where, Figure 8 To create a map of the total motor system corresponding to the average output of the four motors under all operating conditions. Figure 9 This is the total motor system map diagram corresponding to the decision made by the BP neural network in the above-described embodiments of this application regarding the torque distribution condition where the efficiency value is always kept at the highest value.

[0070] pass Figure 8 and Figure 9 By comparing the two, we can see that differentiating the applicable torque distribution conditions based on speed and torque is beneficial to expanding the high-efficiency range of the motor system and significantly improving the operating efficiency of the motor system in the low torque range.

[0071] Preferably, in this embodiment, the motor efficiency cloud map is generated using the high-precision fitting method for the motor MAP map described below:

[0072] Please refer to the above. Figure 10 As shown, the BP neural network includes an input layer and an output layer. The input of the input layer is the motor speed value and the motor torque value. There is a first hidden layer and a second hidden layer between the input layer and the output layer. The first hidden layer has 10 neurons and the second hidden layer has 5 neurons. The output of the output layer is the motor efficiency fitting value.

[0073] During BP neural network training, the actual motor efficiency values ​​of permanent magnet synchronous motors under different speeds and torques are used as samples to train the BP neural network, so that the error between the motor efficiency fitting value obtained by the BP neural network and the actual motor efficiency value is within the preset error range.

[0074] Please see Figure 11 The graph shown depicts the error change during the training process of the BP neural network. The horizontal axis represents the number of "Epochs" (i.e., a single training iteration), which ran for a total of 442 Epochs. The vertical axis represents "Mean Squared Error," abbreviated as mse, which means mean squared error. Specifically, it is a measure that reflects the degree of difference between the estimator and the estimated quantity. Figure 2 The graph shows the error function decrease curves a1 (blue, Train), a2 (green, Validation), and a3 (red, Test) for the training, validation, and test sets, respectively, combined with the Goal curve a4 and the Best curve a5. We can see that after more than 300 iterations, the error function values ​​for the test, validation, and training sets all decrease to 10. -7 The magnitude indicates that the BP neural network in this embodiment not only provides good fitting accuracy for the training data, but also has good generalization ability.

[0075] Please refer to the following for details. Figure 12 As shown, the difference between the fitted motor efficiency value output by the trained BP neural network and the actual motor efficiency value remains within ±1.5 × 10⁻⁶. -3 Within.

[0076] Please refer to the following for further details. Figure 13 and Figure 14 As shown, this application uses experimental test data to generate a MAP diagram of the motor using MATLAB software (see [link]). Figure 13 (As shown); simultaneously, based on the motor efficiency fitting value output by the trained BP neural network, a MAP plot of the fitted motor is generated using MATLAB software (see...). Figure 14 ), through the Figure 13 and Figure 14 A comparison shows that Figure 13 and Figure 14 The previous differences were very small, which indicates that the BP neural network provided in this application has a high-precision fitting capability for generating motor MAP diagrams, and obtains a high-precision motor efficiency fitting effect, thereby providing a reliable and stable data foundation for the optimal efficiency algorithm of modular multiphase motors for vehicles.

[0077] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0078] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A high efficiency torque sharing control method for a modular multiphase electric machine, the modular multiphase electric machine comprising n independent multiphase electric machine units, each independent multiphase electric machine unit being driven and controlled by its corresponding independent drive, n≥2; characterized in that, Each independent multiphase motor unit is the same multiphase motor unit. The high-efficiency torque distribution control method has n torque distribution conditions, specifically including power output by a single motor, power output by two motors on average, ..., power output by n-1 motors on average, and power output by n motors on average. The efficiency values ​​corresponding to the n torque distribution conditions are calculated by fitting the speed-torque-efficiency relationship through a BP neural network. The torque distribution condition that always maintains the highest efficiency value is taken as the torque distribution strategy executed by the modular multiphase motor, thus obtaining the speed-torque-torque distribution condition relationship. In the speed-torque-torque distribution condition relationship, the boundary line between the different n torque distribution conditions is a non-linear curve. Specifically, several coordinate points corresponding to the boundary line are found through an exhaustive method, and then the non-linear curve is fitted to determine the boundary line. Alternatively, in the speed-torque-torque distribution condition relationship, the boundary line between the different n torque distribution conditions is a non-linear curve. Specifically, multiple data points of speed, torque, and their corresponding torque distribution conditions are generated, and based on these data as sample points, a trained BP neural network is used to classify the sample points, and the boundary line is directly determined through this classification.

2. The high-efficiency torque distribution control method according to claim 1, characterized in that, When the high-efficiency torque distribution control method is running, the speed of the motor is no higher than 4500 r / min, so that the motor loss and its torque have a concave function relationship or are close to a concave function relationship.

3. The high-efficiency torque distribution control method according to claim 1, characterized in that, When the high-efficiency torque distribution control method is running, the speed of the motor is no higher than 2000 r / min, so that the motor loss and its torque have a concave function relationship or are close to a concave function relationship.

4. The high-efficiency torque distribution control method according to claim 1, characterized in that, The BP neural network includes an input layer, a hidden layer, and an output layer; wherein, the input data of the input layer is rotational speed and torque, the output layer is a softmax function with n output neuron nodes, and the hidden layer has multiple neuron nodes.

5. The high-efficiency torque distribution control method according to claim 4, characterized in that, The number of neurons in the hidden layer is greater than n.

6. The high-efficiency torque distribution control method according to claim 4, characterized in that, The accuracy of classifying the sample points using the neural network is no less than 99%.

7. An application of the high-efficiency torque distribution control method according to any one of claims 1-6, characterized in that, It is applied as a motor torque distribution control method for modular multiphase motors used in vehicles.

8. The application of the high-efficiency torque distribution control method according to claim 7, characterized in that, The modular multiphase motor for vehicles includes n independent multiphase motor units, each driven and controlled by its corresponding independent driver, and each independent driver powered by its corresponding independent power supply. Each independent multiphase motor unit, each independent driver, and each independent power supply are connected to an on-board computer. The control method includes: the on-board computer has a preset optimal efficiency algorithm, and based on the power requirements of the vehicle and with satisfying the optimal efficiency algorithm as the driving target, it calculates the number of independent multiphase motor units to be operated and their corresponding output power.

Images