A hybrid permanent magnet variable flux memory motor multi-mode identification and coordinated control method

Abstract

The application discloses a hybrid permanent magnet variable flux memory motor multi-mode identification and coordinated control method, comprising: training an offline probabilistic neural network based on the operating parameters of the hybrid permanent magnet variable flux memory motor, and obtaining a smoothing factor value; identifying the motor working condition based on the set parameters and actual parameters of the hybrid permanent magnet variable flux memory motor; and obtaining the magnetization current value of the operating area classification result based on the motor working condition to control the hybrid permanent magnet variable flux memory motor. The application solves the problems of too complex operation process identification classification and slow convergence in the coordinated control process of the existing hybrid permanent magnet variable flux memory motor control, so that the motor control converges faster and good control quality is obtained based on the operating mode classification.

Description

Technical Field

[0001] This invention belongs to the field of electrical drive technology, and in particular relates to a multi-mode identification and coordinated control method for a hybrid permanent magnet variable flux memory motor. Background Technology

[0002] Variable Flux Memory Machines (VFMMs) change the magnetization state (MS) of low-coercive force (LCF) permanent magnet materials by applying instantaneous demagnetizing or magnetizing pulses, achieving flexible adjustment of the air gap flux linkage with almost no loss. Due to this advantage, VFMMs are highly competitive in applications requiring a wide constant power speed range (CPSR), such as electric vehicles. Hybrid permanent magnet VFMMs employ both high- and low-coercive force permanent magnet materials, achieving high torque density while maintaining adjustable permanent magnet flux linkage. Their operating trajectory and the current solution at the stable operating point vary across different operating regions. Therefore, how to quickly determine the motor's operating region and obtain the current solution at its stable operating point based on observed parameters such as motor speed, torque, and permanent magnet flux linkage, thereby improving the speed and accuracy of motor control, is an important research problem for VFMMs. Probabilistic Neural Networks (PNNs) are a class of neural network algorithms highly suitable for pattern classification. Existing technologies do not cover the classification of operating modes of hybrid permanent magnet variable flux memory motors based on probabilistic neural networks, nor the corresponding magnetization current coordinated control technology. Summary of the Invention

[0003] The purpose of this application is to provide a multi-mode identification and coordinated control method for a hybrid permanent magnet variable flux memory motor, thereby solving the problems of operating mode identification and magnetization current coordinated control of the hybrid permanent magnet variable flux memory motor within a certain speed range, achieving simpler operating mode classification and faster convergence speed. The specific technical solution of this application is as follows:

[0004] A multi-mode identification and coordinated control method for a hybrid permanent magnet variable flux memory motor includes:

[0005] Based on the operating parameters of the hybrid permanent magnet variable flux memory motor, an offline probabilistic neural network was trained, and the smoothing factor value was obtained.

[0006] Based on the set parameters and actual parameters of the hybrid permanent magnet variable flux memory motor, the motor operating condition is identified;

[0007] Based on the motor operating conditions, the magnetization current value of the operating area classification result is obtained to control the hybrid permanent magnet variable flux memory motor.

[0008] Optionally, the offline probabilistic neural network is trained based on the operating parameters of the hybrid permanent magnet variable flux memory motor, and the smoothing factor value is obtained, including:

[0009] Obtain the motor parameters of the offline probabilistic neural network and establish a motor parameter vector;

[0010] Obtain the operating mode of the offline probabilistic neural network and establish a motor mode vector;

[0011] Obtain samples of the operating modes of the offline probabilistic neural network to form a training set;

[0012] Based on the classification results of the running regions of the training set, the value of the smoothing factor is adjusted.

[0013] Optionally, obtaining the motor parameters of the offline probabilistic neural network and establishing the motor parameter vector includes:

[0014] The actual speed, given speed, given torque, and permanent magnet flux linkage observations of the hybrid permanent magnet variable flux memory motor are obtained to obtain the motor parameters of the offline probabilistic neural network.

[0015] An input feature vector is constructed based on the motor parameters of the offline probabilistic neural network. The input feature vector is:

[0016] X=[x1,x2,x3,x4] (1)

[0017] Where X represents the feature vector, and x1 to x4 represent the values ​​within the feature vector.

[0018] Optionally, obtaining the operating mode of the offline probabilistic neural network and establishing the motor mode vector includes:

[0019] The operating mode of the offline probabilistic neural network is obtained by acquiring the field adjustment current, current limit, heavy load voltage limit, maximum torque-current ratio, maximum torque-voltage ratio, and light load voltage limit of the hybrid permanent magnet variable flux memory motor.

[0020] A probability feature vector is constructed based on the operating mode of the offline probabilistic neural network, and the probability feature vector is:

[0021] F=[f1, f2,···,f9] (2)

[0022] Where F represents the probability feature vector, and f1 to f9 represent the values ​​within the probability feature vector.

[0023] Optionally, obtaining samples of the operating modes of the offline probabilistic neural network to form a training set includes:

[0024] Training samples are obtained from the operating mode of the offline probabilistic neural network.

[0025] The training samples are input into the same set to form the training set;

[0026] The training samples in the training set are set as neuron nodes to form a pattern layer neural network.

[0027] Optionally, correcting the smoothing factor value based on the region classification results of the training set includes:

[0028] Based on the pattern unit output equation of the training set, the pattern unit output quantity of the training set is obtained, and the pattern unit output equation is:

[0029]

[0030] Where i = 1, 2, ..., n, n represents the total number of classes in the training samples, d represents the dimension of the sample space, and x ij Let represent the j-th center of the i-th sample, and δ represent the smoothing factor;

[0031] Based on the output of the mode units in the training set, the estimated probability density function of the working condition mode is obtained, and the estimated probability density function is:

[0032]

[0033] Among them, f i (x) represents the output of the i-th sample, and m represents the number of neurons in the i-th class;

[0034] Based on the operating mode and the motor parameters of the offline probabilistic neural network, the error of the smoothing factor is obtained, and the smoothing factor is corrected.

[0035] Optionally, the identification of motor operating conditions based on the set parameters and actual parameters of the hybrid permanent magnet variable flux memory motor includes:

[0036] The set parameters and actual parameters of the hybrid permanent magnet variable flux memory motor are obtained, input into the offline probabilistic neural network, and calculations are performed.

[0037] Based on the calculation results of the offline probabilistic neural network, the operating area classification results of the permanent magnet variable flux memory motor are obtained;

[0038] Based on the classification results of the operating area, the motor operating condition is obtained.

[0039] Optionally, the step of obtaining the magnetizing current value based on the motor operating conditions and the classification result of the operating area to control the hybrid permanent magnet variable flux memory motor includes:

[0040] Based on the motor operating conditions, obtain the solution for the steady-state operating point;

[0041] Based on the solution at the steady-state operating point, adjust the current type;

[0042] Based on the current type, magnetization is performed.

[0043] Optionally, obtaining the solution for the steady-state operating point based on the motor operating conditions includes:

[0044] The voltage limit and maximum torque-current ratio equations at the steady-state operating point are confirmed by combining them.

[0045] The equation for the maximum torque-current ratio is:

[0046]

[0047] Where, ψ pm_MS =0% represents the permanent magnet flux linkage parameter when the permanent magnet is magnetized to 0, L d L represents the d-axis torque. q Represents the q-axis torque. The x-coordinate represents the steady-state operating point. The ordinate represents the steady-state operating point;

[0048] The equation of the ordinate of the working point is:

[0049]

[0050] Among them, T ess ω represents the maximum torque, p represents the q-axis momentum, and ω represents the maximum torque. ss This indicates the speed when the permanent magnet is magnetized to 0.

[0051] Optionally, adjusting the current type based on the solution at the steady-state operating point includes:

[0052] When the speed of the hybrid permanent magnet variable flux memory motor is lower than the speed when the permanent magnet is magnetized to 0, the current parameter is a short-time pulse current;

[0053] When the speed of the hybrid permanent magnet variable flux memory motor is not lower than the speed when the permanent magnet is magnetized to 0, the current parameter is a continuous current.

[0054] The beneficial effects of this application are as follows:

[0055] 1. Achieved accurate and efficient identification. The magnetization current control method for hybrid permanent magnet variable flux memory motor based on probabilistic neural network achieves fast and accurate identification of various operating regions of the memory motor through a probabilistic neural network algorithm trained offline and executed online, avoiding the slow convergence speed of traditional serial conditional control statements.

[0056] 2. Rapid control is achieved. Based on the classification results of the operating area and the current value of the steady-state operating point in each area, a short-time or continuous magnetizing current is applied to the motor in a coordinated manner. On the basis of determining the operating load, the motor is further controlled, which can significantly improve the adjustment speed while ensuring the accuracy of motor control. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. The drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. They are used together with the following detailed description to explain the present disclosure, but do not constitute a limitation of the present disclosure. In the drawings:

[0058] Figure 1 This is a block diagram illustrating the principle of a hybrid permanent magnet variable flux memory motor control system based on a probabilistic neural network in a specific embodiment of the present invention.

[0059] Figure 2 This is a schematic diagram of the running trajectory of the hybrid permanent magnet variable flux memory motor under different operating modes in a specific embodiment of the present invention;

[0060] Figure 3 This is a schematic diagram of the probabilistic neural network structure in a specific embodiment of the present invention;

[0061] Figure 4 This is a schematic diagram of the offline training results of the probabilistic neural network in a specific embodiment of the present invention;

[0062] Figure 5 This is a schematic diagram of the d-axis current and speed waveforms in the control system permanent magnet flux linkage adjustment (MS=100% to MS=50%) and current limit mode in a specific embodiment of the present invention. Detailed Implementation

[0063] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Furthermore, in the embodiments of the present invention, terms such as "first" and "second" are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0064] like Figure 1 The diagram shown is a block diagram of a hybrid permanent magnet variable flux memory motor control system based on a probabilistic neural network according to a specific embodiment of the present invention, including:

[0065] A multi-mode identification and coordinated control method for a hybrid permanent magnet variable flux memory motor includes:

[0066] Based on the operating parameters of the hybrid permanent magnet variable flux memory motor, an offline probabilistic neural network was trained, and the smoothing factor value was obtained.

[0067] Based on the set parameters and actual parameters of the hybrid permanent magnet variable flux memory motor, the motor operating condition is identified;

[0068] Based on the motor operating conditions, the magnetization current value of the operating area classification result is obtained to control the hybrid permanent magnet variable flux memory motor.

[0069] Optionally, the offline probabilistic neural network is trained based on the operating parameters of the hybrid permanent magnet variable flux memory motor, and the smoothing factor value is obtained, including:

[0070] Obtain the motor parameters of the offline probabilistic neural network and establish a motor parameter vector;

[0071] Obtain the operating mode of the offline probabilistic neural network and establish a motor mode vector;

[0072] Obtain samples of the operating modes of the offline probabilistic neural network to form a training set;

[0073] Based on the classification results of the running regions of the training set, the value of the smoothing factor is adjusted.

[0074] Optionally, obtaining the motor parameters of the offline probabilistic neural network and establishing the motor parameter vector includes:

[0075] The actual speed, given speed, given torque, and permanent magnet flux linkage observations of the hybrid permanent magnet variable flux memory motor are obtained to obtain the motor parameters of the offline probabilistic neural network.

[0076] An input feature vector is constructed based on the motor parameters of the offline probabilistic neural network. The input feature vector is:

[0077] X=[x1,x2,x3,x4] (1)

[0078] Where X represents the feature vector, and x1 to x4 represent the values ​​within the feature vector.

[0079] Optionally, obtaining the operating mode of the offline probabilistic neural network and establishing the motor mode vector includes:

[0080] The operating mode of the offline probabilistic neural network is obtained by acquiring the field adjustment current, current limit, heavy load voltage limit, maximum torque-current ratio, maximum torque-voltage ratio, and light load voltage limit of the hybrid permanent magnet variable flux memory motor.

[0081] A probability feature vector is constructed based on the operating mode of the offline probabilistic neural network, and the probability feature vector is:

[0082] F=[f1, f2,···,f9] (2)

[0083] Where F represents the probability feature vector, and f1 to f9 represent the values ​​within the probability feature vector.

[0084] Optionally, obtaining samples of the operating modes of the offline probabilistic neural network to form a training set includes:

[0085] Training samples are obtained from the operating mode of the offline probabilistic neural network.

[0086] The training samples are input into the same set to form the training set;

[0087] The training samples in the training set are set as neuron nodes to form a pattern layer neural network.

[0088] Optionally, correcting the smoothing factor value based on the region classification results of the training set includes:

[0089] Based on the pattern unit output equation of the training set, the pattern unit output quantity of the training set is obtained, and the pattern unit output equation is:

[0090]

[0091] Where i = 1, 2, ..., n, n represents the total number of classes in the training samples, d represents the dimension of the sample space, and x ij Let represent the j-th center of the i-th sample, and δ represent the smoothing factor;

[0092] Based on the output of the mode units in the training set, the estimated probability density function of the working condition mode is obtained, and the estimated probability density function is:

[0093]

[0094] Among them, f i (x) represents the output of the i-th sample, and m represents the number of neurons in the i-th class;

[0095] Based on the operating mode and the motor parameters of the offline probabilistic neural network, the error of the smoothing factor is obtained, and the smoothing factor is corrected.

[0096] Optionally, the identification of motor operating conditions based on the set parameters and actual parameters of the hybrid permanent magnet variable flux memory motor includes:

[0097] The set parameters and actual parameters of the hybrid permanent magnet variable flux memory motor are obtained, input into the offline probabilistic neural network, and calculations are performed.

[0098] Based on the calculation results of the offline probabilistic neural network, the operating area classification results of the permanent magnet variable flux memory motor are obtained;

[0099] Based on the classification results of the operating area, the motor operating condition is obtained.

[0100] Optionally, the step of obtaining the magnetizing current value based on the motor operating conditions and the classification result of the operating area to control the hybrid permanent magnet variable flux memory motor includes:

[0101] Based on the motor operating conditions, obtain the solution for the steady-state operating point;

[0102] Based on the solution at the steady-state operating point, adjust the current type;

[0103] Based on the current type, magnetization is performed.

[0104] Optionally, obtaining the solution for the steady-state operating point based on the motor operating conditions includes:

[0105] The voltage limit and maximum torque-current ratio equations at the steady-state operating point are confirmed by combining them.

[0106] The equation for the maximum torque-current ratio is:

[0107]

[0108] Where, ψ pm_MS =0% represents the permanent magnet flux linkage parameter when the permanent magnet is magnetized to 0, L d L represents the d-axis torque. q Represents the q-axis torque. The x-coordinate represents the steady-state operating point. The ordinate represents the steady-state operating point;

[0109] The equation of the ordinate of the working point is:

[0110]

[0111] Among them, T ess ω represents the maximum torque, p represents the q-axis momentum, and ω represents the maximum torque. ss This indicates the speed when the permanent magnet is magnetized to 0.

[0112] Optionally, adjusting the current type based on the solution at the steady-state operating point includes:

[0113] When the speed of the hybrid permanent magnet variable flux memory motor is lower than the speed when the permanent magnet is magnetized to 0, the current parameter is a short-time pulse current;

[0114] When the speed of the hybrid permanent magnet variable flux memory motor is not lower than the speed when the permanent magnet is magnetized to 0, the current parameter is a continuous current.

[0115] For hybrid permanent magnet variable flux memory motors, there are six operating conditions: Magnetization Current (MC), Current Limit (CL), Voltage Limit with Heavy-load VL_H, Maximum Torque per Ampere (MTPA), Maximum Torque per Voltage (MTPV), and Voltage Limit with Light-load VL_L. In MC mode, it can be further refined into multiple magnetization states according to actual operating requirements. This paper subdivides these into four types, denoted as MC4, MC3, MC2, and MC1, and defines these conditions sequentially as numbers 1 to 9 in the PNN program. The variable determining the motor's operating condition can be summarized as the actual operating speed ω of the motor. r Given running speed Given electromagnetic torque Observations of permanent magnet flux This forms a 4-dimensional feature vector X = [x1, x2, x3, x4]. Let f k The output operating condition mode is represented by k = 1…9, which represent operating condition modes MC4, MC3, MC2, MC1, CL, VL_H, MTPA, MTPV, and VL_L, respectively. Three samples are collected for each operating condition mode, for a total of 27 training samples. The specific training samples for operating condition judgment are shown in the table below:

[0116] Table 1 Training Sample Table for Judging the Operating Conditions of Hybrid Permanent Magnet Memory Motors

[0117]

[0118] The established PNN algorithm model and process are as follows: Figure 1 As shown, the PNN model includes an input layer, a pattern layer, a summation layer, and an output layer. The input layer contains 4 neurons, with the same dimension as the input feature vector, used to pass the feature vector to the network. The pattern layer contains 27 neuron nodes, each corresponding to a training sample, used to calculate the matching relationship between the feature vector and each pattern in the training set. The output of each pattern unit in this layer is equation (3), that is:

[0119]

[0120] Where i = 1, 2, ..., n, n represents the total number of classes in the training samples, d represents the dimension of the sample space, and x ij Let δ be the j-th center of the i-th class of samples, and let δ be the smoothing factor. This value plays a very important role in classification, and its value determines the width of the bell curve centered on the sample.

[0121] By using a summation layer to perform a weighted average of the outputs of neurons belonging to a certain working condition in the pattern layer, the estimated probability density function of the working condition pattern is obtained. This equation is equation (4), i.e.:

[0122]

[0123] Among them, f i (x) represents the output of the i-th sample, and m represents the number of neurons in the i-th class.

[0124] Figure 2 The diagram shows the regional magnetic adjustment trajectory of a hybrid permanent magnet memory motor, with T in the figure. L This represents the load torque in steady state. ω b This is the base speed. To reduce copper losses and prevent uncontrollable magnetic weakening, speeds below the base speed are i... d =0 control method. With MS = 100% magnetization, the motor speed reaches ω. b When the electromagnetic torque reaches the rated value T eN ,like Figure 2 As shown at point F in (a). This point is obtained by using i. d The point at which the =0 control method can achieve the maximum torque. When the motor speed exceeds the base speed, the motor operating range can be divided into the following modes:

[0125] (a) 0 ≤ ω r ≤ωb

[0126] During the operation of the motor, a starting process is obviously required. Therefore, when processing this period, it is also necessary to analyze the adjustment scheme for the motor to reach its base speed from 0 speed, where ω b This represents the base velocity. In specific control procedures, to prevent uncontrollable demagnetization, a [method / mechanism] is employed. Control method, where, in ω b The permanent magnets of the motor remain fully magnetized, i.e., MS = 100%. At this point, in... Under the given conditions, the equation is obtained:

[0127]

[0128] When MS = 100% and the motor speed reaches ω b When the electromagnetic torque of the motor reaches the rated value T, eN ,like Figure 2 Point F in (a) represents the maximum torque point achievable using this control method. Based on the voltage limit and torque expression equation, point F can be expressed as:

[0129]

[0130] (b)ω b <ω r ≤ω ss

[0131] When the motor speed exceeds ω b To further increase speed, a negative d-axis pulse current can be used for demagnetization. When the motor is in the MS = 0% state, the speed is ω. ss .like Figure 2 (a) Describe the process of demagnetizing from MS=100% state to MS=0% state, ω b Voltage limit at speed and The positive semi-axis intersects at point A, and... The positive semi-axis intersects at point F. During the weakening process, this is equivalent to the voltage limit expanding to the same level as... The intersection point G of the positive semi-axis is shown by the dark dashed line in 2(a). After that, the velocity increases, the voltage limit decreases inward with equal curvature, and finally stabilizes at the point passing through F and parallel to i. r ds The positive semi-axis intersects at point C, i.e., it remains... Under the condition that the value remains unchanged, make the permanent magnet flux from ψ pm_MS＝100％ Reduce to ψ pm_MS＝0％ When MS = 0%, the motor speed increases to ω. ss The voltage limit's positive and negative half-axis both recover to a standard elliptical trajectory, such as... Figure 2(a) is shown by the light-colored dashed line. During the magnetization weakening process, due to the decrease in permanent magnet flux linkage, the center point of the voltage limit shifts to the right, i.e., from point D in the MS=100% state to point E in the MS=0% state. During this process, the change in permanent magnet flux linkage causes a change in the constant torque trajectory; in the MS=0% state, the... The maximum torque that can be obtained by the control method is denoted as T. ess At this point, the motor speed and torque reach equilibrium simultaneously at point F. According to the voltage limit and torque expression, point F can be expressed as equation (6), which represents:

[0132]

[0133] This is also the steady-state operating point within the working area. untie.

[0134] Among them, T ess The slope must be greater than T at point F. eN This magnetization weakening process can be divided into several segments. In each segment, the amplitude of the negative d-axis pulse current is gradually increased to perform stepwise demagnetization. This application divides this process into 4 segments, namely MS4, MS3, MS2, and MS1.

[0135] (c)(ω ss <ω r ≤ω v )

[0136] ω v It is the velocity corresponding to the voltage limit at the intersection point L of the MTPV trajectory and the current limit. When the velocity equals ω ss At this point, the motor has reached MS=0%. To further increase the speed, the pulsed current field weakening method is no longer applicable. Instead, a continuous negative d-axis current can be used for field weakening. Due to the presence of the continuous negative d-axis current, the reluctance torque component will be superimposed on the permanent magnet torque, resulting in the motor's electromagnetic torque being higher than T. ess When using a continuous negative d-axis current, the MTPA trajectory and ω under the MS=0% state... ss Speed ​​corresponds to voltage limit and T ess The constant torque trajectory intersects at points I and J, respectively. The velocity corresponding to the voltage limit at point J is ω. J It is important to note that when MS > 0%, the permanent magnet flux linkage is variable, and the voltage limit is not a standard ellipse. Therefore, the intersection of its MTPA trajectory and the current limit circle will move from point H and stabilize at point I as the magnetic field weakens. When MS = 0%, the MTPA trajectory can be described as a problem of solving for the extreme points of a given bivariate function (torque equation) under current circle constraints. The specific equation is as follows:

[0137]

[0138] Using Lagrange's extremum theorem, we introduce an auxiliary function H:

[0139]

[0140] Where ξ is the Lagrange factor and H is the Lagrange auxiliary function, taking the partial derivative of the auxiliary function and setting it to 0, we get:

[0141]

[0142] In the formula, and Let be the derivatives of the auxiliary function with respect to the direct and quadrature axis currents. Simplifying and dividing the two equations yields the d / q axis current equations. Since the motor's control region is in the second quadrant, we take... The solution is a negative real number, and the expression for the MTPA trajectory is as follows:

[0143]

[0144] The solutions at each point on the MTPA trajectory can be obtained by combining equation (13) and the voltage limit at that point. When the velocity is higher than ω ss Depending on the magnitude of the load torque and the given electromagnetic torque, there are three possible scenarios:

[0145] 1) The stable operating point moves along the current limiting circle to the point where it coincides with T. ecl Intersection point.

[0146] Where T ecl It is the constant torque trajectory passing through the intersection of the voltage limit and the current limit at a certain speed, which can be obtained by the following formula:

[0147]

[0148] It can be seen that different ω r The value corresponds to a unique T ecl .like Figure 2 (b) shows point K as T eclK Constant torque, ω J The point where the three trajectories corresponding to the voltage limit, current limit, and circular intersections intersect simultaneously.

[0149] Figure 2 (b) gives In this case, the motor's trajectory is from point F to point I at MTPA, and then along the current limit circle to point K, where the speed increases to ω. J Simultaneously, torque and T L Achieving balance. If To maintain a constant speed, the motor torque can only reach point T at point K.eclK Of course, point K is determined by ω. r The value changes and it can move along the current-limiting circle on the IL segment. Point L is ω. v Speed ​​T ecl The point where the trajectories of the constant torque, voltage limit, and current limit circles intersect simultaneously. From Figure 2 (b) It can also be seen that the MTPA control method can handle the situation when MS=0%. The operating condition is equivalent to the motor speed remaining constant but the torque increasing during the transition from point F to point I. This increases output power at the cost of increased losses. The maximum output torque under this condition is the constant torque T passing through point I. essmax The torque can be obtained by equation (14) at ω. ss The solution obtained below can also be derived by simultaneously solving the MTPA expression and the current limiting circle:

[0150]

[0151] 2) The stable operating point reaches the given velocity ω along the current limiting circle. MTPA Then move along the voltage limit to T e * At the intersection of the trajectories.

[0152] T eMTPA This represents the constant torque trajectories that intersect the MTPA trajectory, and the velocity corresponding to the voltage limit at these intersection points is ω. MTPA ,like Figure 2 T shown in (c) eMTPA =T ess ω MTPA =ω J At this point, the motor's trajectory jumps from point F to point I at MTPA, and then moves along the current limit circle to point K, where the speed increases to ω. J However, the torque is not yet balanced. The torque T at this point... e =T eclK Greater than the given torque value Therefore, in order to achieve torque balance while maintaining constant speed, the operating point is along ω J The corresponding voltage limit ellipse shifts to the point where... The intersection point M. The coordinates of point M can be obtained by simultaneously solving the equations ω... J The voltage limit ellipse at speed and the given torque expression are obtained as follows:

[0153]

[0154] 3) The stable operating point reaches the given velocity ω along the current limiting circle. MTPAThen move along the voltage limit to the MTPA trajectory and T eMTPA The intersection point, and then move along the MTPA trajectory to... The intersection of the trajectory and the MTPA trajectory.

[0155] like Figure 2 T shown in (d) eMTPA =T ess ω MTPA =ω J At this point, the motor's trajectory jumps from point F to point I at MTPA, and then moves along the current limit circle to point K, where the speed increases to ω. J However, the torque is not yet balanced. The torque at this point... Less than T eMTPA Therefore, in order to achieve torque balance while maintaining constant speed, the operating point is along ω. J The corresponding voltage limit ellipse shifts. But T e * Constant torque trajectory and ω J The corresponding voltage limit ellipse intersects the right side of the MTPA trajectory. Therefore, the operating point cannot cross the MTPA trajectory when moving along the voltage limit ellipse. That is, the operating point moves along the voltage limit ellipse to point J, then moves along the MTPA trajectory and stabilizes at point N. The coordinates of point N can be obtained by simultaneously determining the torque T. e * The MTPA trajectory expression is obtained as follows:

[0156]

[0157] (d)ω v <ω r ≤ω max

[0158] ω v It is the velocity corresponding to the voltage limit at the intersection point L of the MTPV trajectory and the current limit, T. eclL With ω v Corresponding to the limiting ellipse tangent, ω max This is the set maximum operating speed of the motor. MTPV trajectory and ω r The corresponding limiting ellipses intersect, and the constant torque trajectory passing through the intersection point is T. evl For example, the MTPV trajectory and ω max The corresponding limiting ellipses intersect at point P, and the constant torque trajectory passing through point P is T. evlP ,like Figure 2 As shown in (e).

[0159] 1) The stable operating point is reached along the current limiting circle at a speed ω v Then, it moves along the maximum torque / voltage ratio trajectory to the point where... Points where the trajectories intersect.

[0160] At this point, the motor is in a deep field weakening state and uses MTPV control, which is the minimum stator voltage control method that achieves the maximum speed under the same output torque conditions. Its trajectory is a curve connecting the minimum voltage points required for different torques, i.e., the line connecting the tangent point of the torque hyperbola and the voltage limit circle. Since the torque trajectory and the voltage limit ellipse are tangent (i.e., the tangent directions are consistent), the slopes of the two curves are equal at their intersection point.

[0161]

[0162] Its equation can be expressed as:

[0163]

[0164] By differentiating the torque and voltage expressions and substituting them into equation (19), we obtain the trajectory equation of the MTPV as follows:

[0165]

[0166] The MTPV trajectory intersects the current limiting circle at point L, therefore the coordinates of point L can be obtained by equation (13):

[0167]

[0168] For T e * The intersection of the constant torque and the MTPV trajectory can be expressed as:

[0169]

[0170] For example Figure 2 (e) After the motor moves along the current limiting circle to point L, it then moves along the MTPV trajectory to point T. evlP Point P where the trajectories intersect. It can be seen that under this operating condition, the stable operating point of the motor can move along the LP segment on the MTPV trajectory.

[0171] 2) The stable operating point is along the current limiting circle ω v Then move along the MTPV trajectory to ω vl At the velocity point, then along ω vl The corresponding voltage limit ellipse moves to the point where The intersection of the trajectories. Where ω vl T represents the velocity corresponding to the voltage limit ellipse passing through each point on the MTPV trajectory. elim This indicates the ultimate cut-off torque, meaning the motor will stop if the torque remains below this value after normal startup. For example... Figure 2In (f), ω vl =ω max T evl =T evlP T * e =T elim At this point, the motor's operating point moves from point L to point P and reaches speed equilibrium, but the torque is unbalanced. Therefore, the motor's operating point moves from point P along the voltage limit ellipse to point Q, i.e., ω... max Corresponding voltage limit ellipse and T elim The intersection of the constant torque trajectory. The coordinates of point Q can be obtained by simultaneously solving the equations T... e * Constant torque and ω vl Solving for the corresponding voltage limit elliptic expression:

[0172]

[0173] Figure 3 The probabilistic neural network structure in the method of this invention is given, including an input layer, a pattern layer, a summation layer, and an output layer.

[0174] The input layer contains four neurons, with the same dimension as the input feature vector, used to pass the feature vector to the network. The input feature vector is chosen as a 4-dimensional vector, X = [x1, x2, x3, x4], where x1 to x4 represent the values ​​within the feature vector, including the actual speed, given speed, given torque, and observed permanent magnet flux linkage of the hybrid permanent magnet variable flux memory motor. The probability feature vector is chosen as a 9-dimensional vector F = [f1, f2, ..., f9], corresponding to the nine operating modes of the hybrid permanent magnet variable flux memory motor: Magnetic Current Adjustment Mode 4 (MC4), Magnetic Current Adjustment Mode 3 (MC3), Magnetic Current Adjustment Mode 2 (MC2), Magnetic Current Adjustment Mode 1 (MC1), Current Limit Mode (CL), Voltage Limit Mode under Heavy Load (VL_H), Maximum Torque-to-Current Ratio Mode (MTPA), Maximum Torque-to-Voltage Ratio Mode (MTPV), and Voltage Limit Mode under Light Load (VL_L).

[0175] Each operating mode collects 3 samples, for a total of 21 training samples, forming 21 neuron nodes in the mode layer. Each node corresponds to one training sample, used to calculate the matching relationship between the tested feature vector and each mode in the training set.

[0176]

[0177] In the formula, x is the measured feature vector, i represents the i-th operating mode (i = 1, 2, ..., 7), d is the dimension of the sample space (d = 4), and x ij Let δ be the j-th training sample of the i-th running mode, and let δ be the smoothing factor (default value is spread = 0.1).

[0178] The summation layer performs a weighted average of the outputs of neurons belonging to a certain operating mode in the pattern layer, thereby obtaining the estimated probability density function for each operating mode:

[0179]

[0180] In the formula f i (x) represents the probability that the measured feature vector is in the i-th type of operation mode, and m represents the number of neurons in the i-th type of operation mode, where m = 3.

[0181] The output layer consists of competing neurons, with the same number of neurons as the summation layer, which is also equal to the number of operating modes. Each neuron corresponds to one operating mode. The output layer receives the estimated probability density of each operating mode from the summation layer and uses a threshold to select the neuron with the highest posterior probability density, outputting 1. This neuron represents the operating mode of the sample to be identified, while the outputs of other neurons are all 0. This can be represented by the following function:

[0182] y = argmax(f i ) (twenty three)

[0183] After constructing the offline probabilistic neural network, considering that different numbers of training samples will affect the subsequent response speed of the model, it is necessary to analyze the specific impact of the number of samples on the offline probabilistic neural network. Specifically, for the identification of the training set, this application adopted training schemes with 21 and 54 states in the training set. The identification results of the obtained training set are as follows: Figure 4 The diagram shown illustrates the offline training results of the probabilistic neural network in a specific embodiment of the present invention. With 21 samples, a relatively good recognition result was achieved. However, for some samples, a significant deviation was observed, resulting in an accuracy rate of approximately 90.5%. After further training with an increased number of samples, it was found that the increased number of training samples and the high data similarity among some samples allowed for the integration and simplification of the large amount of data involved. Furthermore, this method can further improve recognition accuracy.

[0184] like Figure 5 As shown, this is a specific embodiment of the present invention in T LThe waveforms of d / q axis current, speed, and permanent magnet flux linkage of the hybrid permanent magnet variable flux memory motor control system under the magnetization current adjustment mode (MS=100% to MS=50% and then to MS=75%) are shown. It can be seen that the motor exhibits four operating states: demagnetization acceleration, stable speed, deceleration, and magnetization. During the demagnetization acceleration process, the initial motor speed is 2000 r / min, and other indicators remain at steady-state values ​​under this condition. During the actual demagnetization acceleration operation, the motor performs demagnetization, and the d / q axis currents change according to the set magnetization current trajectory. The permanent magnet flux linkage of the motor decreases to 0.072 Wb, and the speed increases to 2900 r / min. Simultaneously, dynamic adjustments are made based on the q-axis current, while the d-axis current undergoes minimal changes based on the actual control of the current controller to avoid misoperation. After the motor enters a stable state and after 6 seconds, it enters the deceleration operating range, where the motor speed is reduced to 2200 r / min due to the magnetization boundary limitations. Considering the dynamic fluctuations in the q-axis current during deceleration, a sprint operation is performed according to the predetermined magnetization current trajectory after entering steady state. The permanent magnet flux increases to 0.099Wb. After magnetization is completed, the motor enters the magnetization state, and the d-axis power remains at 0A, thereby maintaining the current magnetization state.

[0185] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any transformations or substitutions that can be conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A multi-mode identification and coordinated control method for a hybrid permanent magnet variable flux memory motor, characterized in that, include: Based on the operating parameters of the hybrid permanent magnet variable flux memory motor, an offline probabilistic neural network was trained, and the smoothing factor value was obtained. Based on the set parameters and actual parameters of the hybrid permanent magnet variable flux memory motor, the motor operating condition is identified; Based on the motor operating conditions, the magnetization current value of the operating area classification result is obtained to control the hybrid permanent magnet variable flux memory motor.

2. The multi-mode identification and coordinated control method for a hybrid permanent magnet variable flux memory motor according to claim 1, characterized in that, The operating parameters of the hybrid permanent magnet variable flux memory motor are used to train an offline probabilistic neural network and obtain the smoothing factor value, including: Obtain the motor parameters of the offline probabilistic neural network and establish a motor parameter vector; Obtain the operating mode of the offline probabilistic neural network and establish a motor mode vector; Obtain samples of the operating modes of the offline probabilistic neural network to form a training set; Based on the classification results of the running regions of the training set, the value of the smoothing factor is adjusted.

3. The multi-mode identification and coordinated control method for a hybrid permanent magnet variable flux memory motor according to claim 2, characterized in that, The step of obtaining the motor parameters of the offline probabilistic neural network and establishing the motor parameter vector includes: The actual speed, given speed, given torque, and permanent magnet flux linkage observations of the hybrid permanent magnet variable flux memory motor are obtained to obtain the motor parameters of the offline probabilistic neural network. An input feature vector is constructed based on the motor parameters of the offline probabilistic neural network. The input feature vector is: X=[x1,x2,x3,x4] (1) Where X represents the feature vector, and x1 to x4 represent the values ​​within the feature vector.

4. The multi-mode identification and coordinated control method for a hybrid permanent magnet variable flux memory motor according to claim 2, characterized in that, The step of obtaining the operating mode of the offline probabilistic neural network and establishing the motor mode vector includes: The operating mode of the offline probabilistic neural network is obtained by acquiring the field adjustment current, current limit, heavy load voltage limit, maximum torque-current ratio, maximum torque-voltage ratio, and light load voltage limit of the hybrid permanent magnet variable flux memory motor. A probability feature vector is constructed based on the operating mode of the offline probabilistic neural network, and the probability feature vector is: F = [f1, f2, ..., f9] (2) where F represents the probability feature vector and f1 to f9 represent the values ​​within the probability feature vector.

5. The multi-mode identification and coordinated control method for a hybrid permanent magnet variable flux memory motor according to claim 2, characterized in that, The step of obtaining samples of the operating modes of the offline probabilistic neural network to form a training set includes: Training samples are obtained from the operating mode of the offline probabilistic neural network. The training samples are input into the same set to form the training set; The training samples in the training set are set as neuron nodes to form a pattern layer neural network.

6. The multi-mode identification and coordinated control method for a hybrid permanent magnet variable flux memory motor according to claim 2, characterized in that, The method of correcting the smoothing factor value based on the region classification results of the training set includes: Based on the pattern unit output equation of the training set, the pattern unit output quantity of the training set is obtained, and the pattern unit output equation is: Where i = 1, 2, ..., n, n represents the total number of classes in the training samples, d represents the dimension of the sample space, and x ij Let represent the j-th center of the i-th sample, and δ represent the smoothing factor; Based on the output of the mode units in the training set, the estimated probability density function of the working condition mode is obtained, and the estimated probability density function is: Among them, f i (x) represents the output of the i-th sample, and m represents the number of neurons in the i-th class; Based on the operating mode and the motor parameters of the offline probabilistic neural network, the error of the smoothing factor is obtained, and the smoothing factor is corrected.

7. The multi-mode identification and coordinated control method for a hybrid permanent magnet variable flux memory motor according to claim 1, characterized in that, The method of identifying motor operating conditions based on the set parameters and actual parameters of the hybrid permanent magnet variable flux memory motor includes: The set parameters and actual parameters of the hybrid permanent magnet variable flux memory motor are obtained, input into the offline probabilistic neural network, and calculations are performed. Based on the calculation results of the offline probabilistic neural network, the operating area classification results of the permanent magnet variable flux memory motor are obtained; Based on the classification results of the operating area, the motor operating condition is obtained.

8. The multi-mode identification and coordinated control method for a hybrid permanent magnet variable flux memory motor according to claim 1, characterized in that, The step of obtaining the magnetizing current value based on the motor operating conditions and the classification result of the operating area to control the hybrid permanent magnet variable flux memory motor includes: Based on the motor operating conditions, obtain the solution for the steady-state operating point; Based on the solution at the steady-state operating point, adjust the current type; Based on the current type, magnetization is performed.

9. The multi-mode identification and coordinated control method for a hybrid permanent magnet variable flux memory motor according to claim 8, characterized in that, The process of obtaining the solution for the steady-state operating point based on the motor operating conditions includes: The voltage limit and maximum torque-current ratio equations at the steady-state operating point are confirmed by combining them. The equation for the maximum torque-current ratio is: Where, ψ pm_MS =0% represents the permanent magnet flux linkage parameter when the permanent magnet is magnetized to 0, L d L represents the d-axis torque. q Represents the q-axis torque. The x-coordinate represents the steady-state operating point. The ordinate represents the steady-state operating point; The equation of the ordinate of the working point is: Among them, T ess ω represents the maximum torque, p represents the number of pole pairs, and ω represents the maximum torque. ss This indicates the speed when the permanent magnet is magnetized to 0.

10. The multi-mode identification and coordinated control method for a hybrid permanent magnet variable flux memory motor according to claim 8, characterized in that, The adjustment of the current type based on the solution at the steady-state operating point includes: When the speed of the hybrid permanent magnet variable flux memory motor is lower than the speed when the permanent magnet is magnetized to 0, the current parameter is the short-time magnetizing current; When the speed of the hybrid permanent magnet variable flux memory motor is not lower than the speed when the permanent magnet is magnetized to 0, the current parameter is the continuously adjusting magnetic current.

Images