Excitation brushless motor accurate calculation method and system based on equivalent magnetic circuit method

Abstract

The invention discloses an excitation brushless motor accurate calculation method based on an equivalent magnetic circuit method. The method comprises the following steps: acquiring parameter information of a target excitation brushless motor; constructing an initial equivalent magnetic circuit as a current equivalent magnetic circuit; calculating yoke plate magnetic flux and tooth magnetic flux; constructing an air gap magnetic flux expression and expressing the rest of magnetic fluxes as air gap magnetic fluxes; carrying out loop calculation by adopting a KVL law; calculating to obtain an intermediate matrix differential equation; decomposing and processing the intermediate matrix differential equation to obtain an air gap magnetic flux calculation result based on the current equivalent magnetic circuit; updating the current equivalent magnetic circuit of the target excitation brushless motor at the next moment based on the pre-trained graph structure neural network; and repeating the steps to realize real-time calculation of the air gap magnetic flux of the target excitation brushless motor. And completing calculation of the excitation brushless motor based on the equivalent magnetic circuit method according to the real-time air gap magnetic flux. The invention also discloses a system for realizing the accurate calculation method of the excitation brushless motor based on the equivalent magnetic circuit method. The method is high in reliability, good in precision and high in dynamic precision.

Description

Technical Field

[0001] The present invention belongs to the field of motor simulation calculation, and in particular relates to a precise calculation method and system for an excitation brushless motor based on an equivalent magnetic circuit method. Background Art

[0002] The magnetic equivalent circuit method (MEC) is an important computational approach for calculating magnetic fields. Compared to finite element methods, MEC significantly reduces computational costs while maintaining accuracy and offers greater theoretical interpretability. Therefore, the MEC is widely used in calculating magnetic fields in electromagnetic devices such as magnetic bearings and transformers. By incorporating principles such as virtual work and Faraday induction, MEC can be further applied to calculate electromagnetic parameters such as electromagnetic force and voltage.

[0003] MEC equates the magnetic field of the object of study to a circuit, where magnetic flux lines are equivalent to wires, magnetic materials are equivalent to resistors, and excitation sources such as coils / permanent magnets are equivalent to power sources, thereby obtaining corresponding parameter equivalence. Taking an induction motor as an example, MEC equates the original physical field to a magnetic circuit model. The location of the energized coil is represented as the power source, and the magnitude of the magnetomotive force is represented as the voltage generated by the power source. The stator and rotor themselves appear as magnetic resistance, whose magnitude is related to the relative magnetic permeability of the material. The current flowing in the wire is the magnetic flux. The calculation is then completed by solving the matrix equation based on Kirchhoff's laws (KCL and KVL).

[0004] However, the accuracy of MEC depends largely on the accuracy of the magnetic circuit model itself and the accuracy of the equivalent reluctance value. Currently, when applied to motor calculations, MEC solutions do not implement dynamic updates during motor operation. This makes it impossible for existing solutions to accurately simulate the changes in air gap reluctance caused by the cogging effect in real time, thus affecting the relative permeability of the magnetic material and the accuracy of the final results. Summary of the Invention

[0005] One of the purposes of the present invention is to provide a precise calculation method for an excitation brushless motor based on an equivalent magnetic circuit method with high reliability, good accuracy and high dynamic precision.

[0006] A second object of the present invention is to provide a system for realizing the accurate calculation method of the excitation brushless motor based on the equivalent magnetic circuit method.

[0007] The present invention provides a method for accurately calculating the excitation of a brushless motor based on the equivalent magnetic circuit method, comprising the following steps:

[0008] S1. Obtain parameter information of the target excitation brushless motor;

[0009] S2. Based on the parameter information obtained in step S1, construct the initial equivalent magnetic circuit of the target excitation brushless motor and use it as the current equivalent magnetic circuit;

[0010] S3. Calculate the target excitation brushless motor yoke plate magnetic flux and tooth magnetic flux according to the current equivalent magnetic circuit;

[0011] S4 constructs the target excitation brushless motor air gap flux representation, and the calculation result obtained in step S3 is expressed in terms of air gap flux;

[0012] S5. Based on the calculation results obtained in step S4, the KVL law is used for loop calculation;

[0013] S6. Based on the loop calculation results obtained in step S5, combined with the circuit control equation, the intermediate matrix differential equation of the target excitation brushless motor is calculated;

[0014] S7. Based on the singular value decomposition scheme, the intermediate matrix differential equation obtained in step S6 is decomposed and processed to obtain the air gap flux calculation result based on the current equivalent magnetic circuit;

[0015] S8. Based on the pre-trained graph structure neural network, update the current equivalent magnetic circuit of the target excitation brushless motor at the next moment;

[0016] S9 repeats steps S3 to S8 to achieve real-time calculation of the air gap flux of the target excitation brushless motor;

[0017] S10. According to the real-time air gap magnetic flux of the target excitation brushless motor obtained in step S9, the calculation of the excitation brushless motor based on the equivalent magnetic circuit method is completed.

[0018] The target excitation brushless motor described in step S1 specifically includes a three-phase six-pole excitation brushless motor.

[0019] The step S3 of calculating the yoke magnetic flux and the tooth magnetic flux of the target excitation brushless motor according to the current equivalent magnetic circuit specifically includes the following steps:

[0020] Set the magnetic flux of the rotor yoke Φ r for Stator yoke magnetic flux Φ s for

[0021] According to the magnetic flux of the rotor yoke plate Φ r and stator yoke plate magnetic flux Φ s , the magnetic flux of the rotor teeth Φ rt Expressed as

[0022]

[0023] The stator tooth magnetic flux Φst Expressed as

[0024]

[0025] In the formula is the magnetic flux flowing through the i-th mover yoke plate; is the magnetic flux flowing through the i-th stator yoke plate.

[0026] The step S4 of constructing the air gap flux representation of the target excitation brushless motor and representing the calculation result obtained in step S3 in terms of the air gap flux specifically includes the following steps:

[0027] During the rotation of the motor, the magnetic flux lines emitted by each pole enter at most two adjacent teeth at the same time; therefore, the air gap flux vector Φ with a maximum length of 2p is constructed. a , p is the number of magnetic poles;

[0028] Air gap flux vector Φ a Expressed as in is the jth component separated from the magnetic flux flowing through the i-th rotor tooth and entering the air gap; i ranges from 1 to 6, and j ranges from 1 to 2;

[0029] According to the KCL law, Φ is calculated a , Φ rt and Φ st The relationship is

[0030] Φ st (i) = Φ a (2i-1)+Φ a (2i-2)

[0031] Φ rt (i) = Φ a (2i-1)+Φ a (2i)

[0032] Where Φ st (i) is Φ st The i-th element in Φ rt (i) is Φ rt The i-th element in Φ a (2i-1) is Φ a The 2i-1th element in Φ a (2i) is Φ a The 2ith element in Φ a (2i-2) is Φ a The 2i-2th element in Φ, and if 2i-2=0, then directly a (2i-2) is replaced by Φ a (12).

[0033] Step S5, based on the calculation result obtained in step S4, uses the KVL law to perform loop calculation, which specifically includes the following steps:

[0034] The KVL law is used to calculate the two circuits formed by the two mover poles; the mover poles are numbered i-1 and i respectively;

[0035] Assume that the motor rotor rotates to the stator position j, then the stator teeth are numbered j-1 and j respectively, so the equation group is:

[0036]

[0037] Where R st (j-1) is the magnetic resistance of stator tooth j-1 in the magnetic circuit direction; R s (j) is the magnetic resistance of stator yoke plate j in the magnetic circuit direction; R a (2i-1) is the air gap magnetic resistance corresponding to the magnetic circuit in the air gap 2i-1; R rt (i) is the magnetic resistance of the i-th rotor tooth in the magnetic circuit direction; R r (i) is the magnetic resistance of the yoke plate of the i-th rotor in the direction of the magnetic circuit; θ s (j) is the magnetomotive force generated by the circuit at the position of stator tooth j; θ r (i) is the magnetomotive force generated by the circuit at the position of the tooth i of the rotor;

[0038] Traverse the values ​​of i, i takes values ​​from 1 to 6, and if i-1=0, directly replace i-1 with 6, and solve all the equations corresponding to all the values ​​of i simultaneously, and replace Φ in the simultaneous equations st , Φ s , Φ rt and Φ r Use Φ a To express;

[0039] Arrange the simultaneous equations into matrix form RΦ a =Tθ; where R is the value including R a 、R s 、R r 、R st and R rt The magnetoresistance matrix; T is the conversion matrix used to convert the dimension of θ to Φ a Align; θ is the magnetomotive force matrix, and θ includes the magnetomotive force information generated by the AC coil control circuit and the magnetomotive force information generated by the DC coil control circuit.

[0040] Step S6, based on the loop calculation result obtained in step S5 and combined with the circuit control equation, calculates the intermediate matrix differential equation of the target excitation brushless motor, which specifically includes the following steps:

[0041] The magnetomotive force matrix θ is decomposed into the form of nI, where n is the number of coil turns and I is the vector composed of the currents of each coil; thus RΦ a =Tθ is expressed as RΦ a =TnI;

[0042] Combined with the circuit control equation Where ε is the voltage of the coil wound on a single tooth, R is the resistance of the coil wound on a single tooth, is the rate of change of magnetic flux at the corresponding coil position;

[0043] Will Substitute into the equation RΦ a =TnI, expressed as where R c is the total resistance of a single control loop; is the vector of the rate of change of magnetic flux at all coil positions. Since coil windings are arranged on both stator teeth and rotor teeth, It is related to the rate of change of magnetic flux at both the stator teeth and the rotor teeth, and

[0044] Will pass Indicates that the intermediate matrix differential equation is finally obtained as Where Q is Expressed as The coefficient matrix of .

[0045] The singular value decomposition scheme described in step S7 decomposes and processes the intermediate matrix differential equation obtained in step S6 to obtain the air gap flux calculation result based on the current equivalent magnetic circuit, which specifically includes the following steps:

[0046] Since TQ is a singular matrix, TQ is decomposed into singular values, which is expressed as TQ = USV T , U is the first positive definite matrix, V is the second positive definite matrix, S is the expansion matrix and Σ is the singular value of TQ and λ1~λ n are the n singular values ​​of TQ;

[0047] TQ=USV T Substituting into the intermediate matrix differential equation, we get And sorted into

[0048] Since U and V are positive definite matrices, there exists U -1 =U T and V -1 =V T , and substitute get

[0049] Since the expanded matrix S has only one non-zero sub-block, According to the dimension of Σ, it can be divided into ordinary differential part and algebraic part:

[0050] Assume S∈R m×n and Σ∈R k×k , m is the dimension of the row space of the T matrix, n is the dimension of the column space of the Q matrix, and k is the dimension of the non-zero block in S; set U = [U1 U2], U1 is the first sub-block of U and U1∈R m×k , U2 is the second sub-block of U and U2∈R m×(m-k) ; Set V = [V1 V2], V1 is the first sub-block of V and V1∈R n×k , V2 is the second sub-block of V and V2∈R n×(n-k) ;set up y is an intermediate variable satisfying Φ a =Vy, y1 is the first sub-block of y and satisfies y2 is the second sub-block of y; after splitting the ordinary differential equation and algebraic equation, we get

[0051]

[0052] Since Σ is a reversible matrix, the ordinary differential equation is expressed in explicit form as

[0053] Φ a Represented as Φ a =V1y1+V2y2, thus we get

[0054] is a full rank matrix. Using the full rank reversible property of square matrix, set And solve the algebraic equation to get Substitute into the ordinary differential equation to obtain the solution equation

[0055] Solve the equation to get y1; solve y2 based on y1; get y based on y1 and y2, and according to Φ a =Vy solves to obtain the air gap flux calculation result Φ based on the current equivalent magnetic circuit a .

[0056] The pre-trained graph structure neural network described in step S8 updates the current equivalent magnetic circuit of the target excitation brushless motor at the next moment, specifically including the following steps:

[0057] According to the BH curve of the material, the updated magnetic permeability μ is Where B is the magnetic induction intensity, H is the magnetic field intensity; the calculation formula is Update the partition resistance, where l is the length of the magnetic circuit in the corresponding partition, and A is the cross-sectional area of ​​the material perpendicular to the magnetic circuit;

[0058] Use graph structure neural network GNN for link prediction;

[0059] The stator tooth position information and the magnetic pole tooth position angle information are used as the node feature vector input of GNN; the feature vector input of node i is set to h i ;

[0060] Construct a single GNN neuron:

[0061] Based on the attention mechanism, the weight coefficient α between node i and node j is calculated using the following formula ij :

[0062]

[0063] Where σ() is the Leaky ReLU activation function; α T is the attention weight vector; W is the mapping matrix; || is the vector concatenation operation; is the number of neighbor nodes;

[0064] Using formula Update the embedding vector of each node at the next moment;

[0065] The update process is abstractly represented as h'=Attention(h,A), where Attention() represents the attention mechanism processing process and A represents the adjacency matrix;

[0066] Since the change in motor position will cause the magnetic flux lines emitted by the same rotor tooth to enter different stator teeth, corresponding to the edges between nodes in the GNN appearing or disappearing over time, the update gate and reset gate are introduced in the gated recurrent unit to determine how much historical information to retain at the current moment and whether to ignore part of the historical state information;

[0067] Considering the current time t and the previous n-1 time steps, it can be expressed as

[0068]

[0069] In the formula The reset gate information of the i-th node at the j-th time step is used to indicate the proportion of the hidden state retained in the previous time step; Reset gate weight matrix for the i-th node at the j-th time step; The output of the gated recurrent unit of the previous time step before the time step calculated by the i-th node; is the node feature vector updated by the attention mechanism at the calculated time step; is the reset gate bias matrix of the i-th node at the j-th time step; is the update gate information of the i-th node at the j-th time step, which is used to represent the mixing ratio of the hidden state of the previous time step and the candidate hidden state of the current time step; is the updated gate weight matrix of the i-th node at the j-th time step; is the update gate bias matrix of the i-th node at the j-th time step;

[0070] The following formula is used to calculate the node i at t j The node embedding vector updated by the gated recurrent unit at every moment and will As the input of the gated recurrent unit at the next time step:

[0071]

[0072] In the formula It is a standby hidden state; is the state weight matrix of the i-th node at the j-th time step; is the state bias matrix of the i-th node at the j-th time step; * is the Hadamard product;

[0073] Complete the construction of a single GNN neuron;

[0074] There are n GNN neurons in a layer, corresponding to n time steps;

[0075] Establish a two-layer GNN, obtain n time-step embedding vectors of node i in the last layer, and concatenate them in sequence to obtain an i-node embedding vector H containing timing information i ;

[0076] Use the following formula to calculate H i The elements in the matrix are weightedly fused to obtain the node embedding Z with spatiotemporal characteristics. i :

[0077] Z i =Softmax((H i W q )(H i W k ) T +M)(H i W v )

[0078] Where Softmax() is the normalized exponential function; W q 、Wk 、W v are all mapping matrices; M is a shielding matrix, and M ij is the element in row i and column j in M;

[0079] Prediction by logistic regression: set x=<Z i ,Z j >, x is the variable that integrates the information of nodes i and j; < > is the vector inner product;

[0080] Set p = Sigmoid(W p x+b p ), where Sigmoid() is the Sigmoid activation function, p is the prediction result, W p is the weight vector, b p is the bias vector;

[0081] Prediction results according to rules We get , where existence represents the prediction result, and existence = 1 means the edge exists, and existence = 0 means it does not exist;

[0082] Update the adjacency matrix A based on the prediction results between pairs of nodes of different categories;

[0083] Multilayer perceptron is used to predict the magnetic resistance value, expressed as R=MLP R (x), R is the predicted magnetic resistance value, MLP R () is the multi-layer perceptron processing function;

[0084] During pre-training, the following formula is used as the loss function of the GNN model:

[0085] L total =λL1+(1-λ)L2

[0086] Where λ is the loss weight value; L1 is the first loss function, and N is the number of nodes, is the true value, p i is the predicted value; L2 is the second loss function, and is the true value of magnetic resistance, R i is the predicted value of magnetic resistance;

[0087] According to the predicted adjacency matrix A and magnetic resistance value, the current equivalent magnetic circuit of the target excitation brushless motor at the next moment is updated.

[0088] The present invention also provides a system for realizing the accurate calculation method of the excitation brushless motor based on the equivalent magnetic circuit method, comprising a data acquisition module, a magnetic circuit equivalent module, a magnetic flux calculation module, an air gap representation module, a loop calculation module, an intermediate calculation module, an air gap calculation module, a magnetic circuit update module, a circulation module and a motor calculation module; the data acquisition module, the magnetic circuit equivalent module, the magnetic flux calculation module, the air gap representation module, the loop calculation module, the intermediate calculation module, the air gap calculation module, the magnetic circuit update module and the circulation module are connected in series in sequence; the output of the circulation module is simultaneously connected to the motor calculation module and the magnetic flux calculation module; the data acquisition module is used to obtain parameter information of the target excitation brushless motor, and the data is stored in the system. Information is uploaded to the magnetic circuit equivalent module; the magnetic circuit equivalent module is used to construct the initial equivalent magnetic circuit of the target excitation brushless motor according to the received data information and the obtained parameter information, and use it as the current equivalent magnetic circuit, and upload the data information to the magnetic flux calculation module; the magnetic flux calculation module is used to construct the air gap magnetic flux representation of the target excitation brushless motor according to the received data information, and represent the calculated result in air gap magnetic flux, and upload the data information to the air gap representation module; the air gap representation module is used to calculate the yoke plate magnetic flux and tooth magnetic flux of the target excitation brushless motor according to the received data information and the current equivalent magnetic circuit, and upload the data information to the loop calculation module; loop calculation module It is used to perform loop calculation using the KVL law according to the received data information and the obtained calculation results, and upload the data information to the intermediate calculation module; the intermediate calculation module is used to calculate the intermediate matrix differential equation of the target excitation brushless motor according to the received data information and the obtained loop calculation results in combination with the circuit control equation, and upload the data information to the air gap calculation module; the air gap calculation module is used to decompose and process the obtained intermediate matrix differential equation based on the received data information and the singular value decomposition scheme, obtain the air gap magnetic flux calculation result based on the current equivalent magnetic circuit, and upload the data information to the magnetic circuit update module; the magnetic circuit update module is used to calculate the air gap magnetic flux based on the received data information based on the received data information. Data information, based on the pre-trained graph structure neural network, updates the current equivalent magnetic circuit of the target excitation brushless motor at the next moment, and uploads the data information to the loop module; the loop module is used to repeat the work of the magnetic flux calculation module, the air gap representation module, the loop calculation module, the intermediate calculation module, the air gap calculation module and the magnetic circuit update module according to the received data information, to realize the real-time calculation of the air gap magnetic flux of the target excitation brushless motor, and upload the data information to the motor calculation module and the magnetic flux calculation module; the motor calculation module is used to complete the calculation of the excitation brushless motor based on the equivalent magnetic circuit method according to the received data information and the real-time air gap magnetic flux of the target excitation brushless motor.

[0089] The present invention provides a method and system for accurately calculating the excitation brushless motor based on the equivalent magnetic circuit method. By constructing and calculating the equivalent magnetic circuit model of the target excitation brushless motor and updating the equivalent magnetic circuit model in real time based on a graph structure neural network, the method and system not only achieve accurate calculation of the excitation brushless motor based on the equivalent magnetic circuit method, but also have higher reliability, better accuracy, and higher dynamic precision. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] Figure 1 Schematic diagram of the process flow of the present invention.

[0091] Figure 2 Schematic diagram of a motor in the method of the present invention.

[0092] Figure 3 Schematic diagram of the equivalent magnetic circuit model in the method of the present invention.

[0093] Figure 4 Schematic diagram of the functional modules of the system of the present invention. DETAILED DESCRIPTION

[0094] like Figure 1 The method flow chart of the present invention is shown as follows: The method for accurately calculating the excitation of a brushless motor based on the equivalent magnetic circuit method disclosed in the present invention comprises the following steps:

[0095] S1. Obtain parameter information of the target excitation brushless motor. The target excitation brushless motor specifically includes a three-phase six-pole excitation brushless motor. The following processing steps are all for a three-phase six-pole excitation brushless motor. At the same time, other types of excitation brushless motors can refer to the following steps for processing.

[0096] S2. According to the parameter information obtained in step S1, the initial equivalent magnetic circuit of the target excitation brushless motor is constructed and used as the current equivalent magnetic circuit; the initial state of the target excitation brushless motor is as follows Figure 2 As shown, its initial equivalent magnetic circuit is as follows Figure 3 As shown;

[0097] S3. Calculate the yoke magnetic flux and tooth magnetic flux of the target excitation brushless motor according to the current equivalent magnetic circuit; specifically comprising the following steps:

[0098] Set the magnetic flux of the rotor yoke Φ r for Stator yoke magnetic flux Φ s for

[0099] According to the magnetic flux of the rotor yoke plate Φ r and stator yoke plate magnetic flux Φ s , the magnetic flux of the rotor teeth Φ rt Expressed as

[0100]

[0101] The stator tooth magnetic flux Φ st Expressed as

[0102]

[0103] In the formula is the magnetic flux flowing through the i-th mover yoke plate; is the magnetic flux flowing through the i-th stator yoke plate

[0104] S4. Constructing the air gap flux representation of the target excitation brushless motor and expressing the calculation result obtained in step S3 in terms of the air gap flux; specifically comprising the following steps:

[0105] During the rotation of the motor, the magnetic flux lines emitted by each pole enter at most two adjacent teeth at the same time; therefore, the air gap flux vector Φ with a maximum length of 2p is constructed. a , p is the number of magnetic poles;

[0106] Air gap flux vector Φ a Expressed as in is the jth component separated from the magnetic flux flowing through the i-th rotor tooth and entering the air gap; i ranges from 1 to 6, and j ranges from 1 to 2;

[0107] According to the KCL law, Φ is calculated a , Φ rt and Φ st The relationship is

[0108] Φ st (i) = Φ a (2i-1)+Φ a (2i-2)

[0109] Φ rt (i) = Φ a (2i-1)+Φ a (2i)

[0110] Where Φ st (i) is Φ st The i-th element in Φ rt (i) is Φ rt The i-th element in Φ a (2i-1) is Φ a The 2i-1th element in Φ a (2i) is Φ a The 2ith element in Φ a (2i-2) is Φ a The 2i-2th element in Φ, and if 2i-2=0, then directlya (2i-2) is replaced by Φ a (12);

[0111] S5. Based on the calculation results obtained in step S4, the KVL law is used to perform loop calculation; specifically, the steps include:

[0112] The KVL law is used to calculate the two circuits formed by the two mover poles; the mover poles are numbered i-1 and i respectively;

[0113] Assuming that the motor rotor rotates to the jth stator position, the stator teeth are numbered j-1 and j, respectively, so the equations are:

[0114]

[0115] Where R st (j-1) is the magnetic resistance of stator tooth j-1 in the magnetic circuit direction; R s (j) is the magnetic resistance of stator yoke plate j in the magnetic circuit direction; R a (2i-1) is the air gap magnetic resistance corresponding to the magnetic circuit in the air gap 2i-1; R rt (i) is the magnetic resistance of the i-th rotor tooth in the magnetic circuit direction; R r (i) is the magnetic resistance of the yoke plate of the i-th rotor in the direction of the magnetic circuit; θ s (j) is the magnetomotive force generated by the circuit at the position of stator tooth j; θ r (i) is the magnetomotive force generated by the circuit at the position of the tooth i of the rotor;

[0116] Traverse the values ​​of i, i takes values ​​from 1 to 6, and if i-1=0, directly replace i-1 with 6, and solve all the equations corresponding to all the values ​​of i simultaneously, and replace Φ in the simultaneous equations st , Φ s , Φ rt and Φ r Use Φ a To express;

[0117] Arrange the simultaneous equations into matrix form RΦ a =Tθ; where R is the value including R a 、R s 、R r 、R st and R rt The magnetoresistance matrix; T is the conversion matrix used to convert the dimension of θ to Φ a Alignment; θ is the magnetomotive force matrix, and θ includes the magnetomotive force information generated by the AC coil control circuit and the magnetomotive force information generated by the DC coil control circuit;

[0118] S6. Based on the loop calculation results obtained in step S5, combined with the circuit control equation, the intermediate matrix differential equation of the target excitation brushless motor is calculated; specifically comprising the following steps:

[0119] The magnetomotive force matrix θ is decomposed into the form of nI, where n is the number of coil turns and I is the vector composed of the currents of each coil; thus RΦ a =Tθ is expressed as RΦ a =TnI;

[0120] Combined with the circuit control equation Where ε is the voltage of the coil wound on a single tooth, R is the resistance of the coil wound on a single tooth, is the rate of change of magnetic flux at the corresponding coil position;

[0121] Will Substitute into the equation RΦ a =TnI, expressed as where R c is the total resistance of a single control loop; is the vector of the rate of change of magnetic flux at all coil positions. Since coil windings are arranged on both stator teeth and rotor teeth, It is related to the rate of change of magnetic flux at both the stator teeth and the rotor teeth, and

[0122] Since Φ st and Φ rt Can be used a Therefore, Φ c Through Φ a Indicates that the intermediate matrix differential equation is finally obtained as Where Q is Expressed as The coefficient matrix of

[0123] S7. Based on the singular value decomposition scheme, the intermediate matrix differential equation obtained in step S6 is decomposed and processed to obtain the air gap flux calculation result based on the current equivalent magnetic circuit; specifically comprising the following steps:

[0124] Since TQ is a singular matrix, TQ is decomposed into singular values, which is expressed as TQ = USV T , U is the first positive definite matrix, V is the second positive definite matrix, S is the expansion matrix and Σ is the singular value of TQ and λ1~λ n are the n singular values ​​of TQ;

[0125] TQ = USV T Substituting into the intermediate matrix differential equation, we get And sorted into

[0126] Since U and V are positive definite matrices, there exists U -1 =U T and V -1 =V T , and substitute get

[0127] Since the expanded matrix S has only one non-zero sub-block, According to the dimension of Σ, it can be divided into ordinary differential part and algebraic part:

[0128] Assume S∈R m×n and Σ∈R k×k , m is the dimension of the row space of the T matrix, n is the dimension of the column space of the Q matrix, and k is the dimension of the non-zero block in S; set U = [U1 U2], U1 is the first sub-block of U and U1∈R m×k , U2 is the second sub-block of U and U2∈R m×(m-k) ; Set V = [V1 V2], V1 is the first sub-block of V and V1∈R n×k , V2 is the second sub-block of V and V2∈R n×(n-k) ;set up y is an intermediate variable satisfying Φ a =Vy, y1 is the first sub-block of y and satisfies y2 is the second sub-block of y; after splitting the ordinary differential equation and algebraic equation, we get

[0129]

[0130] Since Σ is a reversible matrix, the ordinary differential equation is expressed in explicit form as

[0131] Φ a Represented as Φ a =V1y1+V2y2, thus we get

[0132] is a full rank matrix. Using the full rank reversible property of square matrix, set And solve the algebraic equation to get Substitute into the ordinary differential equation to obtain the solution equation

[0133] Solve the equation to get y1; solve y2 based on y1; get y based on y1 and y2, and according to Φ a =Vy solves to obtain the air gap flux calculation result Φ based on the current equivalent magnetic circuit a ;

[0134] S8. Based on the pre-trained graph structure neural network, update the current equivalent magnetic circuit of the target excitation brushless motor at the next moment; specifically, the steps include:

[0135] According to the BH curve of the material, the updated magnetic permeability μ is Where B is the magnetic induction intensity, H is the magnetic field intensity; the calculation formula is Update the partition resistance, where l is the length of the magnetic circuit in the corresponding partition, and A is the cross-sectional area of ​​the material perpendicular to the magnetic circuit;

[0136] Use graph structure neural network GNN for link prediction;

[0137] The stator tooth position information and the magnetic pole tooth position angle information are used as the node feature vector input of GNN; the feature vector input of node i is set to h i ;

[0138] Construct a single GNN neuron:

[0139] Based on the attention mechanism, the weight coefficient α between node i and node j is calculated using the following formula ij :

[0140]

[0141] Where σ() is the Leaky ReLU activation function; α T is the attention weight vector; W is the mapping matrix; || is the vector concatenation operation; is the number of neighbor nodes;

[0142] Using formula Update the embedding vector of each node at the next moment;

[0143] The update process is abstractly represented as h'=Attention(h,A), where Attention() represents the attention mechanism processing process and A represents the adjacency matrix;

[0144] Since the change in motor position will cause the magnetic flux lines emitted by the same rotor tooth to enter different stator teeth, corresponding to the edges between nodes in the GNN appearing or disappearing over time, the update gate and reset gate are introduced in the gated recurrent unit to determine how much historical information to retain at the current moment and whether to ignore part of the historical state information;

[0145] Considering the current time t and the previous n-1 time steps, it can be expressed as

[0146]

[0147] In the formula The reset gate information of the i-th node at the j-th time step is used to indicate the proportion of the hidden state retained in the previous time step; Reset gate weight matrix for the i-th node at the j-th time step; The output of the gated recurrent unit of the previous time step before the time step calculated by the i-th node; is the node feature vector updated by the attention mechanism at the calculated time step; is the reset gate bias matrix of the i-th node at the j-th time step; is the update gate information of the i-th node at the j-th time step, which is used to represent the mixing ratio of the hidden state of the previous time step and the candidate hidden state of the current time step; is the updated gate weight matrix of the i-th node at the j-th time step; is the update gate bias matrix of the i-th node at the j-th time step;

[0148] The following formula is used to calculate the node i at t j The node embedding vector updated by the gated recurrent unit at every moment and will As the input of the gated recurrent unit at the next time step:

[0149]

[0150] In the formula It is a standby hidden state; is the state weight matrix of the i-th node at the j-th time step; is the state bias matrix of the i-th node at the j-th time step; * is the Hadamard product;

[0151] Complete the construction of a single GNN neuron;

[0152] There are n GNN neurons in a layer, corresponding to n time steps;

[0153] Establish a two-layer GNN, obtain n time-step embedding vectors of node i in the last layer, and concatenate them in sequence to obtain an i-node embedding vector H containing timing information i ;

[0154] Use the following formula to calculate H i The elements in the matrix are weightedly fused to obtain the node embedding Z with spatiotemporal characteristics. i :

[0155] Z i =Softmax((H i W q )(H i W k ) T +M)(Hi W v )

[0156] Where Softmax() is the normalized exponential function; W q 、W k 、W v are all mapping matrices; M is a shielding matrix, and M ij is the element in row i and column j of M; the shielding matrix is ​​used to prevent the previous time step from obtaining the information of the next time step;

[0157] Prediction by logistic regression: set x=<Z i ,Z j >, x is the variable that integrates the information of nodes i and j; < > is the vector inner product;

[0158] Set p = Sigmoid(W p x+b p ), where Sigmoid() is the Sigmoid activation function, p is the prediction result, W p is the weight vector, b p is the bias vector;

[0159] Prediction results according to rules We get: existence represents the prediction result, and existence = 1 means the edge exists, and existence = 0 means it does not exist.

[0160] Update the adjacency matrix A based on the prediction results between pairs of nodes of different categories;

[0161] Multilayer perceptron is used to predict the magnetic resistance value, expressed as R=MLP R (x), R is the predicted magnetic resistance value, MLP R () is the multi-layer perceptron processing function;

[0162] During pre-training, the following formula is used as the loss function of the GNN model:

[0163] L total =λL1+(1-λ)L2

[0164] Where λ is the loss weight value; L1 is the first loss function, and N is the number of nodes, is the true value, p i is the predicted value; L2 is the second loss function, and is the true value of magnetic resistance, R i is the predicted value of magnetic resistance;

[0165] According to the predicted adjacency matrix A and magnetic resistance value, the current equivalent magnetic circuit of the target excitation brushless motor at the next moment is updated; in specific implementation, Figure 3 R between the annular regions in a 、R a With R st The connection relationship, R a With R rt Update the connection relationship;

[0166] S9 repeats steps S3 to S8 to achieve real-time calculation of the air gap flux of the target excitation brushless motor;

[0167] S10. According to the real-time air gap magnetic flux of the target excitation brushless motor obtained in step S9, the calculation of the excitation brushless motor based on the equivalent magnetic circuit method is completed;

[0168] At this time, the electromagnetic force F based on the virtual work method can be further developed according to the air gap magnetic flux obtained in step S9. M , Power P based on energy conservation M Calculation of other parameters.

[0169] like Figure 4The figure shows a schematic diagram of the functional modules of the system of the present invention: the system disclosed in the present invention for realizing the accurate calculation method of the excitation brushless motor based on the equivalent magnetic circuit method comprises a data acquisition module, a magnetic circuit equivalent module, a magnetic flux calculation module, an air gap representation module, a loop calculation module, an intermediate calculation module, an air gap calculation module, a magnetic circuit update module, a circulation module and a motor calculation module; the data acquisition module, the magnetic circuit equivalent module, the magnetic flux calculation module, the air gap representation module, the loop calculation module, the intermediate calculation module, the air gap calculation module, the magnetic circuit update module and the circulation module are connected in series in sequence; the output of the circulation module is simultaneously connected to the motor calculation module and the magnetic flux calculation module; the data acquisition module is used to obtain the target excitation brushless motor. The parameter information of the brushless motor is collected and the data information is uploaded to the magnetic circuit equivalent module; the magnetic circuit equivalent module is used to construct the initial equivalent magnetic circuit of the target excitation brushless motor according to the received data information and the acquired parameter information, and use it as the current equivalent magnetic circuit, and upload the data information to the magnetic flux calculation module; the magnetic flux calculation module is used to construct the air gap magnetic flux representation of the target excitation brushless motor according to the received data information, and express the calculated result in air gap magnetic flux, and upload the data information to the air gap representation module; the air gap representation module is used to calculate the yoke plate magnetic flux and tooth magnetic flux of the target excitation brushless motor according to the received data information and the current equivalent magnetic circuit, and upload the data information to the circuit calculation module. Module; the loop calculation module is used to perform loop calculation using the KVL law based on the received data information and the obtained calculation results, and upload the data information to the intermediate calculation module; the intermediate calculation module is used to calculate the intermediate matrix differential equation of the target excitation brushless motor based on the received data information and the obtained loop calculation results in combination with the circuit control equation, and upload the data information to the air gap calculation module; the air gap calculation module is used to decompose and process the obtained intermediate matrix differential equation based on the received data information and the singular value decomposition scheme to obtain the air gap magnetic flux calculation result based on the current equivalent magnetic circuit, and upload the data information to the magnetic circuit update module; the magnetic circuit update module is used to calculate the air gap magnetic flux ... The received data information is used to update the current equivalent magnetic circuit of the target excitation brushless motor at the next moment based on the pre-trained graph structure neural network, and the data information is uploaded to the loop module; the loop module is used to repeat the work of the magnetic flux calculation module, the air gap representation module, the loop calculation module, the intermediate calculation module, the air gap calculation module and the magnetic circuit update module according to the received data information, so as to realize the real-time calculation of the air gap magnetic flux of the target excitation brushless motor, and upload the data information to the motor calculation module and the magnetic flux calculation module; the motor calculation module is used to complete the calculation of the excitation brushless motor based on the equivalent magnetic circuit method according to the received data information and the obtained real-time air gap magnetic flux of the target excitation brushless motor.

Claims

1. A precise calculation method for the excitation of a brushless motor based on the equivalent magnetic circuit method, comprising the following steps: S1. Obtain parameter information of the target excitation brushless motor; S2. Based on the parameter information obtained in step S1, construct the initial equivalent magnetic circuit of the target excitation brushless motor and use it as the current equivalent magnetic circuit; S3. Calculate the target excitation brushless motor yoke plate magnetic flux and tooth magnetic flux according to the current equivalent magnetic circuit; S4 constructs the target excitation brushless motor air gap flux representation, and the calculation result obtained in step S3 is expressed in terms of air gap flux; S5. Based on the calculation results obtained in step S4, the KVL law is used for loop calculation; S6. Based on the loop calculation results obtained in step S5, combined with the circuit control equation, the intermediate matrix differential equation of the target excitation brushless motor is calculated; S7. Based on the singular value decomposition scheme, the intermediate matrix differential equation obtained in step S6 is decomposed and processed to obtain the air gap flux calculation result based on the current equivalent magnetic circuit; S8. Based on the pre-trained graph structure neural network, update the current equivalent magnetic circuit of the target excitation brushless motor at the next moment; S9 repeats steps S3 to S8 to achieve real-time calculation of the air gap flux of the target excitation brushless motor; S10. According to the real-time air gap magnetic flux of the target excitation brushless motor obtained in step S9, the calculation of the excitation brushless motor based on the equivalent magnetic circuit method is completed.

2. The accurate calculation method of the excitation brushless motor based on the equivalent magnetic circuit method according to claim 1 is characterized in that The target excitation brushless motor described in step S1 specifically includes a three-phase six-pole excitation brushless motor.

3. The accurate calculation method of the excitation brushless motor based on the equivalent magnetic circuit method according to claim 2 is characterized in that The step S3 of calculating the yoke magnetic flux and the tooth magnetic flux of the target excitation brushless motor according to the current equivalent magnetic circuit specifically includes the following steps: Set the magnetic flux of the rotor yoke Φ r for Stator yoke magnetic flux Φ s for According to the magnetic flux of the rotor yoke plate Φ r and stator yoke plate magnetic flux Φ s , the magnetic flux of the rotor teeth Φ rt Expressed as The stator tooth magnetic flux Φ st Expressed as In the formula is the magnetic flux flowing through the i-th mover yoke plate; is the magnetic flux flowing through the i-th stator yoke plate.

4. The accurate calculation method of the excitation brushless motor based on the equivalent magnetic circuit method according to claim 3 is characterized in that The step S4 of constructing the air gap flux representation of the target excitation brushless motor and representing the calculation result obtained in step S3 in terms of the air gap flux specifically includes the following steps: During the rotation of the motor, the magnetic flux lines emitted by each pole enter at most two adjacent teeth at the same time; therefore, the air gap flux vector Φ with a maximum length of 2p is constructed. a , p is the number of magnetic poles; Air gap flux vector Φ a Expressed as in is the jth component separated from the magnetic flux flowing through the i-th rotor tooth and entering the air gap; i ranges from 1 to 6, and j ranges from 1 to 2; According to the KCL law, Φ is calculated a , Φ rt and Φ st The relationship is Φ st (i)=Φ a (2i-1)+Φ a (2i-2) Φ rt (i)=Φ a (2i-1)+Φ a (2i) Where Φ st (i) is Φ st The i-th element in Φ rt (i) is Φ rt The i-th element in Φ a (2i-1) is Φ a The 2i-1th element in Φ a (2i) is Φ a The 2ith element in Φ a (2i-2) is Φ a The 2i-2th element in Φ, and if 2i-2=0, then directly a (2i-2) is replaced by Φ a (12).

5. The accurate calculation method of the excitation brushless motor based on the equivalent magnetic circuit method according to claim 4 is characterized in that Step S5, based on the calculation result obtained in step S4, uses the KVL law to perform loop calculation, which specifically includes the following steps: The KVL law is used to calculate the two circuits formed by the two mover poles; the mover poles are numbered i-1 and i respectively; Assuming that the motor rotor rotates to the jth stator position, the stator teeth are numbered j-1 and j, respectively, so the equations are: Where R st (j-1) is the magnetic resistance of stator tooth j-1 in the magnetic circuit direction; R s (j) is the magnetic resistance of stator yoke plate j in the magnetic circuit direction; R a (2i-1) is the air gap magnetic resistance corresponding to the magnetic circuit in the air gap 2i-1; R rt (i) is the magnetic resistance of the i-th rotor tooth in the magnetic circuit direction; R r (i) is the magnetic resistance of the yoke plate of the i-th rotor in the direction of the magnetic circuit; θ s (j) is the magnetomotive force generated by the circuit at the position of stator tooth j; θ r (i) is the magnetomotive force generated by the circuit at the position of the tooth i of the rotor; Traverse the values ​​of i, i takes values ​​from 1 to 6, and if i-1=0, directly replace i-1 with 6, and solve all the equations corresponding to all the values ​​of i simultaneously, and replace Φ in the simultaneous equations st , Φ s , Φ rt and Φ r Use Φ a To express; Arrange the simultaneous equations into matrix form RΦ a =Tθ; where R is the a 、R s 、R r 、R st and R rt The magnetoresistance matrix; T is the transformation matrix used to transform the dimensions of θ to Φ a Alignment; θ is a magnetomotive force matrix, and θ includes magnetomotive force information generated by the AC coil control circuit and magnetomotive force information generated by the DC coil control circuit.

6. The accurate calculation method for the excitation brushless motor based on the equivalent magnetic circuit method according to claim 5 is characterized in that Step S6, based on the loop calculation result obtained in step S5 and combined with the circuit control equation, calculates the intermediate matrix differential equation of the target excitation brushless motor, which specifically includes the following steps: The magnetomotive force matrix θ is decomposed into the form of nI, where n is the number of coil turns and I is the vector composed of the currents of each coil; thus RΦ a =Tθ is expressed as RΦ a =TnI; Combined with the circuit control equation Where ε is the voltage of the coil wound on a single tooth, R is the resistance of the coil wound on a single tooth, is the rate of change of magnetic flux at the corresponding coil position; Will Substitute into the equation RΦ a =TnI, expressed as where R c is the total resistance of a single control loop; is the vector of the rate of change of magnetic flux at all coil positions, and Will pass Indicates that the intermediate matrix differential equation is finally obtained as Where Q is Expressed as The coefficient matrix of .

7. The accurate calculation method for the excitation brushless motor based on the equivalent magnetic circuit method according to claim 6 is characterized in that The singular value decomposition scheme described in step S7 decomposes and processes the intermediate matrix differential equation obtained in step S6 to obtain the air gap flux calculation result based on the current equivalent magnetic circuit, which specifically includes the following steps: Since TQ is a singular matrix, TQ is decomposed into singular values, which is expressed as TQ = USV T , U is the first positive definite matrix, V is the second positive definite matrix, S is the expansion matrix and Σ is the singular value of TQ and λ1~λ n are the n singular values ​​of TQ; TQ=USV T Substituting into the intermediate matrix differential equation, we get And sorted into Since U and V are positive definite matrices, there exists U -1 =U T and V -1 =V T , and substitute get Since the expanded matrix S has only one non-zero sub-block, According to the dimension of Σ, it can be divided into ordinary differential part and algebraic part: Assume S∈R m×n and Σ∈R k×k , m is the dimension of the row space of the T matrix, n is the dimension of the column space of the Q matrix, and k is the dimension of the non-zero block in S; set U = [U1 U2], U1 is the first sub-block of U and U1∈R m×k , U2 is the second sub-block of U and U2∈R m ×(m-k) ; Set V = [V1 V2], V1 is the first sub-block of V and V1∈R n×k , V2 is the second sub-block of V and V2∈R n×(n-k) ;set up y is an intermediate variable satisfying Φ a =Vy, y1 is the first sub-block of y and satisfies y2 is the second sub-block of y; after splitting the ordinary differential equation and algebraic equation, we get Since Σ is a reversible matrix, the ordinary differential equation is expressed in explicit form as Φ a Represented as Φ a =V1y1+V2y2, thus we get is a full rank matrix. Using the full rank reversible property of square matrix, set And solve the algebraic equation to get Substitute into the ordinary differential equation to obtain the solution equation Solve the equation to get y1; solve y2 based on y1; get y based on y1 and y2, and according to Φ a =Vy solves to obtain the air gap flux calculation result Φ based on the current equivalent magnetic circuit a .

8. The accurate calculation method for the excitation brushless motor based on the equivalent magnetic circuit method according to claim 7 is characterized in that The pre-trained graph structure neural network described in step S8 updates the current equivalent magnetic circuit of the target excitation brushless motor at the next moment, specifically including the following steps: According to the BH curve of the material, the updated magnetic permeability μ is Where B is the magnetic induction intensity, H is the magnetic field intensity; the calculation formula is Update the partition resistance, where l is the length of the magnetic circuit in the corresponding partition, and A is the cross-sectional area of ​​the material perpendicular to the magnetic circuit; Use graph structure neural network GNN for link prediction; The stator tooth position information and magnetic pole tooth position angle information are used as the node feature vector input of GNN; Set the feature vector input of node i to h i ; Construct a single GNN neuron: Based on the attention mechanism, the weight coefficient α between node i and node j is calculated using the following formula ij : Where σ() is the Leaky ReLU activation function; α T is the attention weight vector; W is the mapping matrix; || is the vector concatenation operation; is the number of neighbor nodes; Using formula Update the embedding vector of each node at the next moment; The update process is abstractly represented as h'=Attention(h,A), where Attention() represents the attention mechanism processing process and A represents the adjacency matrix; Since the change in motor position will cause the magnetic flux lines emitted by the same rotor tooth to enter different stator teeth, corresponding to the edges between nodes in the GNN appearing or disappearing over time, the update gate and reset gate are introduced in the gated recurrent unit to determine how much historical information to retain at the current moment and whether to ignore part of the historical state information; Considering the current time t and the previous n-1 time steps, it can be expressed as In the formula The reset gate information of the i-th node at the j-th time step is used to indicate the proportion of the hidden state retained in the previous time step; Reset gate weight matrix for the i-th node at the j-th time step; The output of the gated recurrent unit of the previous time step before the time step calculated by the i-th node; is the node feature vector updated by the attention mechanism at the calculated time step; is the reset gate bias matrix of the i-th node at the j-th time step; is the update gate information of the i-th node at the j-th time step, which is used to represent the mixing ratio of the hidden state of the previous time step and the candidate hidden state of the current time step; is the updated gate weight matrix of the i-th node at the j-th time step; is the update gate bias matrix of the i-th node at the j-th time step; The following formula is used to calculate the node i at t j The node embedding vector updated by the gated recurrent unit at every moment and will As the input of the gated recurrent unit at the next time step: In the formula It is a standby hidden state; is the state weight matrix of the i-th node at the j-th time step; is the state bias matrix of the i-th node at the j-th time step; * is the Hadamard product; Complete the construction of a single GNN neuron; There are n GNN neurons in a layer, corresponding to n time steps; Establish a two-layer GNN, obtain n time-step embedding vectors of node i in the last layer, and concatenate them in sequence to obtain an i-node embedding vector H containing timing information i ; Use the following formula to calculate H i The elements in the matrix are weightedly fused to obtain the node embedding Z with spatiotemporal characteristics. i : Z i =Softmax((H i W q )(H i W k ) T +M)(H i W v ) Where Softmax() is the normalized exponential function; W q 、W k 、W v are all mapping matrices; M is a shielding matrix, and M ij is the element in row i and column j in M; Prediction by logistic regression: set x=<Z i ,Z j >, x is the variable that integrates the information of nodes i and j; < > is the vector inner product; Set p = Sigmoid(W p x+b p ), where Sigmoid() is the Sigmoid activation function, p is the prediction result, W p is the weight vector, b p is the bias vector; Prediction results according to rules We get , where existence represents the prediction result, and existence = 1 means the edge exists, and existence = 0 means it does not exist; Update the adjacency matrix A based on the prediction results between pairs of nodes of different categories; Multilayer perceptron is used to predict the magnetic resistance value, expressed as R=MLP R (x), R is the predicted magnetic resistance value, MLP R () is the multi-layer perceptron processing function; During pre-training, the following formula is used as the loss function of the GNN model: L total =λL1+(1-λ)L2 Where λ is the loss weight value; L1 is the first loss function, and N is the number of nodes, is the true value, p i is the predicted value; L2 is the second loss function, and is the true value of magnetic resistance, R i is the predicted value of magnetic resistance; According to the predicted adjacency matrix A and magnetic resistance value, the current equivalent magnetic circuit of the target excitation brushless motor at the next moment is updated.

9. A system for realizing the accurate calculation method of the excitation brushless motor based on the equivalent magnetic circuit method as claimed in any one of claims 1 to 8, characterized in that It includes a data acquisition module, a magnetic circuit equivalent module, a magnetic flux calculation module, an air gap representation module, a loop calculation module, an intermediate calculation module, an air gap calculation module, a magnetic circuit update module, a circulation module and a motor calculation module; the data acquisition module, the magnetic circuit equivalent module, the magnetic flux calculation module, the air gap representation module, the loop calculation module, the intermediate calculation module, the air gap calculation module, the magnetic circuit update module and the circulation module are connected in series in sequence; the output of the circulation module is simultaneously connected to the motor calculation module and the magnetic flux calculation module; the data acquisition module is used to obtain parameter information of the target excitation brushless motor and upload the data information to the magnetic circuit equivalent module; the magnetic circuit equivalent module is used to construct the initial equivalent magnetic circuit of the target excitation brushless motor according to the received data information and the acquired parameter information, and use it as the current equivalent magnetic circuit, and upload the data information to the magnetic flux calculation module; The magnetic flux calculation module is used to construct the air gap magnetic flux representation of the target excitation brushless motor based on the received data information, express the calculated result in air gap magnetic flux, and upload the data information to the air gap representation module; The air gap representation module is used to calculate the yoke magnetic flux and tooth magnetic flux of the target excitation brushless motor based on the received data information and the current equivalent magnetic circuit, and upload the data information to the loop calculation module; the loop calculation module is used to perform loop calculation using the KVL law based on the received data information and the obtained calculation results, and upload the data information to the intermediate calculation module; The intermediate calculation module is used to calculate the intermediate matrix differential equation of the target excitation brushless motor based on the received data information, the obtained loop calculation results, and the circuit control equation, and upload the data information to the air gap calculation module; The air gap calculation module is used to decompose and process the intermediate matrix differential equation obtained based on the received data information and the singular value decomposition scheme, obtain the air gap magnetic flux calculation result based on the current equivalent magnetic circuit, and upload the data information to the magnetic circuit update module; The magnetic circuit update module is used to update the current equivalent magnetic circuit of the target excitation brushless motor at the next moment based on the received data information and the pre-trained graph structure neural network, and upload the data information to the loop module; The loop module is used to repeat the work of the magnetic flux calculation module, the air gap representation module, the loop calculation module, the intermediate calculation module, the air gap calculation module and the magnetic circuit update module according to the received data information, so as to realize the real-time calculation of the air gap magnetic flux of the target excitation brushless motor and upload the data information to the motor calculation module and the magnetic flux calculation module; the motor calculation module is used to complete the calculation of the excitation brushless motor based on the equivalent magnetic circuit method according to the received data information and the real-time air gap magnetic flux of the target excitation brushless motor.

Images