A method and device for modeling mutual inductance of coils in wireless power transmission system

Abstract

The present application relates to a method and device for modeling the mutual inductance of coils in a wireless power transmission system. The method includes obtaining multiple groups of measurement data of two coils, and dividing the multiple groups of measurement data into training samples and test samples; constructing a coil mutual inductance model; the model is a BP neural network, including an input layer, a hidden layer and an output layer; the model is trained using training samples to obtain a trained coil mutual inductance model. The model is used to estimate the relationship between the input parameters of the coil mutual inductance model that does not exist in the training samples and the test samples and the corresponding coil mutual inductance. The method can model the mutual inductance of coils in any relative position, and is applicable to coils with different parameters; using experimental data as the data source for modeling, the parameter items that cannot be quantified in the traditional theoretical derivation can be modeled as a whole, which can improve the model accuracy of parameters to coil mutual inductance, and has a certain improvement effect on the analysis and design of magnetically coupled resonant wireless power transmission systems.

Description

Technical Field

[0001] The present application relates to the technical field of magnetically coupled resonant wireless power transmission, and in particular to a method and device for modeling mutual inductance of coils in a wireless power transmission system. Background Art

[0002] Magnetic coupling resonant wireless power transfer technology is a new type of contactless energy transfer technology that can efficiently transmit electrical energy within a range of several meters and can pass through non-magnetic obstacles. It has good application prospects in wireless power supply for embedded sensors.

[0003] The mutual inductance between the transmitting and receiving coils (hereinafter referred to as coil mutual inductance) is an important parameter of the magnetic coupling resonant wireless power transmission system and has an important impact on the overall performance of the system. The size of the coil mutual inductance depends on the coil's own geometry, size, relative position, number of turns, and the conductivity, magnetic permeability, temperature, pressure and other parameters of the medium in the coil's environment. In the traditional modeling method, the coil mutual inductance is calculated by theoretical formulas, but in the derivation of the theoretical calculation formula, due to the lack of some internal mechanisms of the magnetic coupling resonant wireless power transmission system or the difficulty in accurately describing it with analytical expressions, there is a difference between the theoretical calculated value and the actual value of the coil mutual inductance, which affects the analysis and design of the entire wireless power transmission system. The data-driven coil mutual inductance modeling method can effectively overcome the above shortcomings. When the internal mechanism of the system cannot be obtained, the input and output data of the system can be used to train the data model of the coil mutual inductance. The coil mutual inductance value obtained according to the data model is closer to the actual value. Therefore, the data modeling method is of great value in the analysis, design and optimization of wireless power transmission systems. Summary of the invention

[0004] Based on this, it is necessary to provide a method and device for modeling mutual inductance of coils in a wireless power transmission system in response to the above technical problems.

[0005] A method for modeling mutual inductance of coils in a wireless power transmission system, the method comprising:

[0006] Acquire multiple groups of measurement data of two magnetic field coupling coils for wireless power transmission, and divide the multiple groups of measurement data into training samples and test samples; the measurement data include the coil's own parameters, the coil's environmental parameters, the relative position parameters of the two magnetic field coupling coils, and the corresponding coil mutual inductance between the two coils.

[0007] Construct a coil mutual inductance model; the coil mutual inductance model is a BP neural network, including an input layer, a hidden layer and an output layer; the number of nodes contained in the input layer is the same as the sum of the number of parameters of the coil's own parameters, the environmental parameters of the coil, and the relative position parameters of the two magnetic field coupling coils in the measurement data, the number of hidden layers is 2, and the number of nodes in each hidden layer is determined according to the total number of samples in the training sample, the number of input layer nodes, the number of output layer nodes and a preset ratio parameter; the output layer contains 1 node.

[0008] The coil mutual inductance model is trained using training samples. The model is trained by forward signal transmission, reverse error transmission, and cyclic adjustment of connection weights until the output layer error on the test data is less than the preset threshold error or the number of cycles reaches the preset training number, and the cycle ends to obtain a trained coil mutual inductance model.

[0009] The trained coil mutual inductance model is used to estimate the relationship between the coil's own parameters, the coil's environmental parameters, and the relative position parameters of the two magnetic field coupling coils and the coil mutual inductance that do not exist in the training samples and test samples.

[0010] In one embodiment, the measurement data includes the coil's own parameters, the coil's environmental parameters, the relative position parameters of the two magnetic field coupling coils, and the corresponding coil mutual inductance between the two coils; the coil's own parameters include the coil's shape, size, and number of turns; the environmental parameters include the medium's electrical conductivity, magnetic permeability, temperature, and pressure.

[0011] Acquiring multiple groups of measurement data of two coils, and dividing the multiple groups of measurement data into training samples and test samples, including:

[0012] When the coil's own parameters, the coil's environmental parameters, and the relative position parameters of the two magnetic field coupling coils are determined, an impedance tester is used to measure the self-inductance of the inductor composed of the two coils connected in series and the self-inductance of the inductor composed of the two coils connected in series, and the corresponding mutual inductance of the coils is obtained.

[0013] By changing the coil's own parameters, the coil's environmental parameters, and the relative position parameters of the two magnetic field coupling coils, and measuring the corresponding coil mutual inductance, multiple sets of measurement data are obtained.

[0014] In one embodiment, the coil mutual inductance model is trained using training samples. The model is trained by forward signal transmission, reverse error transmission, and cyclic adjustment of connection weights until the output layer error on the test data is less than a preset threshold error or the number of cycles reaches a preset training number, and the cycle is terminated to obtain a trained coil mutual inductance model, including:

[0015] A random function is used to initialize the connection weights between layers to obtain the initialization weights.

[0016] The coil's own parameters in the training sample, the coil's environmental parameters, and the relative position parameters of the two magnetic field coupling coils are used as the forward input of the coil mutual inductance model input layer, and directly used as the forward output of the input layer.

[0017] The forward output of the input layer is multiplied by the connection weight between the input layer and the first hidden layer, and the sum is taken as the forward input of the first hidden layer. After activation by the activation function, the forward output of the first hidden layer is obtained; and so on, until the forward output of the output layer is calculated and used as the calculation output of the model.

[0018] The output layer error is determined according to the difference between the calculated output of the model and the standard output, and is used as the reverse input of the output layer; the standard output is the mutual inductance of the coil in the training sample.

[0019] The reverse output of the output layer can be calculated according to the reverse input of the output layer and the inverse of the activation function; the reverse input of the second hidden layer can be calculated according to the reverse output of the output layer and the transpose of the connection weight matrix between the output layer and the second hidden layer; and so on, until the reverse output of the first hidden layer is calculated.

[0020] According to the reverse output of the i-th node in the next layer, the preset learning rate and the forward input of the j-th node in the previous layer, the adjustment amount of the connection weight between the previous and next layers is obtained.

[0021] The connection weights between the corresponding layers are updated according to the adjustment amount of the connection weights between the previous and next layers and the connection weights between the corresponding layers, and the next round of training is carried out until every training sample participates in the training and the training is completed.

[0022] The trained coil mutual inductance model is tested using test data until the output layer error on the test data is less than a preset threshold error or the number of cycles reaches a preset number of training rounds, and the cycle is terminated to obtain a trained coil mutual inductance model.

[0023] In one embodiment, the input parameters of the coil mutual inductance model include: the coil's own parameters, the coil's environment parameters, and the relative position parameters of the two magnetic field coupling coils; the method further includes:

[0024] The input parameters of the coil mutual inductance model other than the training samples and the test samples are input into the trained coil mutual inductance model to obtain the output result of the model.

[0025] The output error is obtained according to the output result and the corresponding coil mutual inductance obtained from the experiment.

[0026] When the output error is greater than a preset threshold error, the preset error threshold or the preset number of training times is adjusted to retrain until the verification result meets the error requirement, thereby obtaining a trained coil mutual inductance model.

[0027] A wireless power transmission system coil mutual inductance modeling device, the device comprising:

[0028] A data acquisition module is used to acquire multiple groups of measurement data of two magnetic field coupling coils used for wireless power transmission, and divide the multiple groups of measurement data into training samples and test samples; the measurement data includes the coil's own parameters, the environment parameters of the coil, the relative position parameters of the two magnetic field coupling coils, and the coil mutual inductance between the two corresponding coils.

[0029] A coil mutual inductance model construction module is used to construct a coil mutual inductance model; the coil mutual inductance model is a BP neural network, including an input layer, a hidden layer and an output layer; the number of nodes contained in the input layer is the same as the sum of the number of parameters of the coil's own parameters, the environmental parameters of the coil, and the relative position parameters of the two magnetic field coupling coils in the measurement data, the number of hidden layers is 2, and the number of nodes in each hidden layer is determined based on the total number of samples in the training sample, the number of input layer nodes, the number of output layer nodes and a preset ratio parameter; the output layer contains 1 node.

[0030] The coil mutual inductance model training module is used to train the coil mutual inductance model using training samples. The model is trained by forward signal transmission, reverse error transmission, and cyclic adjustment of connection weights until the output layer error on the test data is less than the preset threshold error or the number of cycles reaches the preset training number, and the cycle ends to obtain a trained coil mutual inductance model.

[0031] The coil mutual inductance module is used to estimate the relationship between the coil's own parameters that do not exist in the training samples and test samples, the coil's environmental parameters, and the relative position parameters of the two magnetic field coupling coils and the coil mutual inductance using the trained coil mutual inductance model.

[0032] In one embodiment, the measurement data includes the coil's own parameters, the coil's environmental parameters, the relative position parameters of the two magnetic field coupling coils, and the corresponding coil mutual inductance between the two coils; the coil's own parameters include the coil's shape, size, and number of turns; the environmental parameters include the medium's electrical conductivity, magnetic permeability, temperature, and pressure.

[0033] The data acquisition module is also used to measure the self-inductance of the inductor composed of two coils connected in series and the self-inductance of the inductor composed of two coils connected in series reversely, using an impedance tester, when the coil's own parameters, the coil's environmental parameters and the relative position parameters of the two magnetic field coupling coils are determined, to obtain the corresponding mutual inductance of the coils; by changing the coil's own parameters, the coil's environmental parameters and the relative position parameters of the two magnetic field coupling coils, and measuring the corresponding mutual inductance of the coils, multiple sets of measurement data are obtained.

[0034] In one of the embodiments, the coil mutual inductance model training module is further used to initialize the connection weights between the layers using a random function to obtain the initialization weights.

[0035] The coil's own parameters in the training sample, the environment parameters of the coil, and the relative position parameters of the two magnetic field coupling coils are used as the forward input of the coil mutual inductance model input layer, and directly used as the forward output of the input layer; the forward output of the input layer is multiplied by the connection weight between the input layer and the first hidden layer and the sum is used as the forward input of the first hidden layer, and after activation by the activation function, the forward output of the first hidden layer is obtained; and so on, until the forward output of the output layer is calculated and used as the calculated output of the model; according to the difference between the calculated output of the model and the standard output, the output layer error is determined and used as the reverse input of the output layer; the standard output is the mutual inductance of the coil in the training sample; according to the reverse input of the output layer and the inverse of the activation function, the reverse output of the output layer can be calculated; according to the output layer The reverse output of the second hidden layer and the transpose of the connection weight matrix between the output layer and the second hidden layer are used to calculate the reverse input of the second hidden layer; and so on until the reverse output of the first hidden layer is calculated; according to the reverse output of the ith node of the latter layer, the preset learning rate and the forward input of the jth node of the previous layer, the adjustment amount of the connection weight between the front and rear layers is obtained; according to the adjustment amount of the connection weight between the front and rear layers and the connection weight between the corresponding layers, the connection weight between the corresponding layers is updated, and the next round of training is carried out until every training sample participates in the training and the training is completed; the coil mutual inductance model obtained by training is tested using the test data until the output layer error on the test data is less than the preset threshold error or the number of cycles reaches the preset number of training rounds, and the cycle is terminated to obtain the trained coil mutual inductance model.

[0036] In one embodiment, the input parameters of the coil mutual inductance model include: the coil's own parameters, the environment parameters of the coil, and the relative position parameters of the two magnetic field coupling coils; the device also includes a verification module, which is used to input the input parameters of the coil mutual inductance model other than the training samples and the test samples into the trained coil mutual inductance model to obtain the output result of the model; according to the output result and the corresponding coil mutual inductance obtained by the experiment, the output error is obtained; when the output error is greater than the preset threshold error, the preset error threshold or the preset number of training times is adjusted to re-train until the verification result meets the error requirement to obtain the trained coil mutual inductance model.

[0037] The above-mentioned method and device for modeling the mutual inductance of coils in a wireless power transmission system include obtaining multiple groups of measurement data of two coils, and dividing the multiple groups of measurement data into training samples and test samples; constructing a coil mutual inductance model; the model is a BP neural network, including an input layer, a hidden layer and an output layer; the model is trained by forward signal transmission, reverse error transmission, and cyclic adjustment of connection weights using the training samples, until the output layer error on the test data is less than the preset threshold error or the number of cycles reaches the preset number of training times to end the cycle, and a trained coil mutual inductance model is obtained. The trained coil mutual inductance model is used to establish the relationship between the input parameters of the coil mutual inductance model that does not exist in the training samples and the test samples and the coil mutual inductance. This method can model the mutual inductance of coils at any relative position, and is applicable to coils with different parameters; using experimental data as the data source for modeling, the parameter items that cannot be quantified in the traditional theoretical derivation can be modeled as a whole, which can improve the model accuracy from parameters to coil mutual inductance, and has a certain improvement effect on the analysis and design of magnetically coupled resonant wireless power transmission systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A schematic diagram of a flow chart of a method for modeling mutual inductance of a coil of a wireless power transmission system in one embodiment;

[0039] Figure 2 A schematic flow chart of a method for modeling mutual inductance of coils in a wireless power transmission system in another embodiment;

[0040] Figure 3 is a spatial schematic diagram of two circular coils in one embodiment;

[0041] Figure 4 A BP neural network training flow chart in another embodiment;

[0042] Figure 5 It is a structural block diagram of a device for modeling mutual inductance of coils of a wireless power transmission system in one embodiment. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0044] Explanation of parameter symbols for the coil mutual inductance model:

[0045] (1) x is the forward input, v is the forward output, e is the reverse input, δ is the reverse output, w is the connection weight between nodes, and W is the inter-layer connection weight matrix;

[0046] For positive input and positive output, the input layer, the first hidden layer, the second hidden layer and the output layer have them;

[0047] For reverse input and forward output, only the first hidden layer, the second hidden layer and the output layer have;

[0048] For connection weights, there are only three types between two layers.

[0049] (2) The superscript of the parameter represents the corresponding network layer, which can be expressed in numbers or English. The subscript of the parameter represents the corresponding node in the layer. All parameters must have superscripts. Subscripts represent nodes, and no subscripts represent vectors.

[0050] The superscript of the input layer is 1 or in; the superscript of the first hidden layer is 2 or Hidden1; the superscript of the second hidden layer is 3 or Hidden2; the superscript of the output layer is 4 or out;

[0051] Example: Forward input of the jth node in the input layer: or Reverse output of the output layer: δ (out) or (4) ; Connection weight (matrix) from the second hidden layer to the output layer: W (Hidden2,out) or W (3,4) ; The weight adjustment of a single connection from the jth node in the kth layer to the ith node in the k+1th layer

[0052] In one embodiment, Figure 2 As shown, a method for modeling mutual inductance of coils in a wireless power transmission system is provided, the method comprising the following steps:

[0053] Step 100: Acquire multiple groups of measurement data of two magnetic field coupling coils for wireless power transmission, and divide the multiple groups of measurement data into training samples and test samples; the measurement data include the coil's own parameters, the coil's environmental parameters, the relative position parameters of the two magnetic field coupling coils, and the corresponding coil mutual inductance between the two coils.

[0054] The input parameters of the coil mutual inductance model include: the coil's own parameters, the coil's environmental parameters, and the relative position parameters of the two magnetic field coupling coils.

[0055] Step 102: construct a coil mutual inductance model; the coil mutual inductance model is a BP neural network, including an input layer, a hidden layer and an output layer; the number of nodes included in the input layer is the same as the sum of the number of parameters of the coil's own parameters, the environmental parameters of the coil, and the relative position parameters of the two magnetic field coupling coils in the measurement data, the number of hidden layers is 2, and the number of nodes in each hidden layer is determined according to the total number of samples in the training sample, the number of input layer nodes, the number of output layer nodes and a preset ratio parameter; the output layer contains 1 node.

[0056] The node activation function uses the ELU function, and the activation function only acts on the nodes of the hidden layer and the output layer;

[0057] Step 104: Use the training samples to train the coil mutual inductance model. The model is trained by forward signal transmission, reverse error transmission, and cyclic adjustment of connection weights until the output layer error on the test data is less than the preset threshold error or the number of cycles reaches the preset training number, and the cycle ends to obtain a trained coil mutual inductance model.

[0058] Step 106: Use the trained coil mutual inductance model to estimate the relationship between the coil's own parameters that do not exist in the training sample and the test sample, the coil's environmental parameters, and the relative position parameters of the two magnetic field coupling coils and the coil mutual inductance.

[0059] In the above-mentioned method for modeling the mutual inductance of coils in a wireless power transmission system, the method includes obtaining multiple groups of measurement data of two coils, and dividing the multiple groups of measurement data into training samples and test samples; constructing a coil mutual inductance model; the model is a BP neural network, including an input layer, a hidden layer and an output layer; using the training samples, the model is trained by forward signal transmission, reverse error transmission, and cyclic adjustment of connection weights, until the output layer error on the test data is less than the preset threshold error or the number of cycles reaches the preset number of training times to end the cycle, and obtain a trained coil mutual inductance model. The trained coil mutual inductance model is used to establish the relationship between the input parameters of the coil mutual inductance model that does not exist in the training samples and the test samples and the coil mutual inductance. This method can model the mutual inductance of coils at any relative position, and is applicable to coils with different parameters; using experimental data as the data source for modeling, the parameter items that cannot be quantified in the traditional theoretical derivation can be modeled as a whole, which can improve the model accuracy from parameters to coil mutual inductance, and has a certain improvement effect on the analysis and design of magnetically coupled resonant wireless power transmission systems.

[0060] In one of the embodiments, the measurement data includes the coil's own parameters, the coil's environmental parameters, the relative position parameters of the two magnetic field coupling coils, and the corresponding coil mutual inductance between the two coils; the coil's own parameters include the shape, size, and number of turns of the coil; the environmental parameters include the conductivity, magnetic permeability, temperature, and pressure of the medium; step 100 includes: when the coil's own parameters, the coil's environmental parameters, and the relative position parameters of the two magnetic field coupling coils are determined, using an impedance tester to measure the self-inductance of the inductance composed of the two coils connected in series and the self-inductance of the inductance composed of the two coils connected in series, and obtain the corresponding coil mutual inductance; by changing the coil's own parameters, the coil's environmental parameters, and the relative position parameters of the two magnetic field coupling coils, and measuring the corresponding coil mutual inductance, multiple sets of measurement data are obtained.

[0061] Specifically, the parameter definition of the two coils and the mutual inductance measurement method include:

[0062] Considering the two-coil model, the coil's own parameters such as geometric shape (such as circle or square), size (such as circle radius or square side length), number of turns, etc., and the coil's environmental parameters such as medium permeability, temperature, pressure, etc. can be set as λ1,λ2…λ m , where m is the sum of the coil's own parameters and the coil's environment parameters.

[0063] The geometric center point of the coil is regarded as the center of the coil, and the central axis of the coil is established through the center point. The central axis is perpendicular to the plane where the current in the coil is located, and the direction of the central axis is determined by the direction of the current according to the right-hand rule.

[0064] In three-dimensional space, the relative position parameters of the two coils can be expressed as the center point relative offset Coil center axis offset vector and around the vector The rotation angle ψ is uniquely determined.

[0065] The mutual inductance of the two coils is the parameters λ1, λ2…λ m ,s x ,s y ,s z ,n x ,n y ,n z , the function of ψ is denoted as M().

[0066] Assume the self-inductance of the two coils is L1 and L2 respectively, and the mutual inductance is M. Then the self-inductance of the inductor formed by the two coils connected in sequence is L 顺 =L1+L2+2M, the self-inductance of the two coils connected inversely is L 反 =L1+L2-2M, measure L 顺 and L 反 , we can get M=(L顺 -L 反 ) / 4.

[0067] By changing the parameter measurement corresponding to M(λ1,λ2…λ m ,s x ,s y ,s z ,n x ,n y ,n z ,ψ), record multiple groups of data as training data and test data for subsequent BP neural network.

[0068] In one embodiment, step 104 includes: using a random function to initialize the connection weights between each layer to obtain the initialization weights; using the coil's own parameters in the training sample, the environmental parameters of the coil, and the relative position parameters of the two magnetic field coupling coils as the forward input of the coil mutual inductance model input layer, and directly using it as the forward output of the input layer; multiplying the forward output of the input layer by the connection weights between the input layer and the first hidden layer and summing them as the forward input of the first hidden layer, and after activation by the activation function, obtaining the forward output of the first hidden layer; and so on, until the forward output of the output layer is calculated and used as the calculated output of the model; according to the difference between the calculated output of the model and the standard output, determining the output layer error and using it as the reverse input of the output layer; the standard output is the coil mutual inductance in the training sample; according to the reverse input of the output layer and the activation function, the output layer error is calculated and used as the reverse input of the output layer. The inverse of the function can be used to calculate the reverse output of the output layer; the reverse input of the second hidden layer is calculated according to the reverse output of the output layer and the transpose of the connection weight matrix between the output layer and the second hidden layer; and so on, until the reverse output of the first hidden layer is calculated; according to the reverse output of the ith node of the latter layer, the preset learning rate and the forward input of the jth node of the previous layer, the adjustment amount of the connection weight between the front and rear layers is obtained; according to the adjustment amount of the connection weight between the front and rear layers and the connection weight between the corresponding layers, the connection weight between the corresponding layers is updated, and the next round of training is carried out until every training sample participates in the training and the training is completed; the coil mutual inductance model obtained by training is tested using the test data, until the output layer error on the test data is less than the preset threshold error or the number of cycles reaches the preset number of training rounds, the cycle is terminated, and the trained coil mutual inductance model is obtained.

[0069] In one of the embodiments, the input parameters of the coil mutual inductance model include: the coil's own parameters, the environment parameters of the coil, and the relative position parameters of the two magnetic field coupling coils; the method also includes inputting the input parameters of the coil mutual inductance model other than the training samples and the test samples into the trained coil mutual inductance model to obtain the output result of the model; obtaining the output error based on the output result and the corresponding coil mutual inductance obtained from the experiment; when the output error is greater than the preset threshold error, adjusting the preset error threshold or the preset number of training times to re-train until the verification result meets the error requirement, thereby obtaining the trained coil mutual inductance model.

[0070] In one embodiment, Figure 2 The flow chart of the method for modeling the mutual inductance of coils in a wireless power transmission system is shown, and the method comprises the following steps:

[0071] S1. Definition of parameters of two circular coils and mutual inductance measurement method

[0072] Consider the two circular coil models, such as Figure 3 As shown, the radius is R P The center of the circle corresponds to the origin of the plane O(0,0,0), the axis corresponds to the z-axis, and the radius R S The center point of the circle is C(x C ,y C ,z C ) Axis direction At this time, the relative position parameters of the two circular coils themselves can be used P ,R S ,x C ,y C ,z C ,n x ,n y ,n z The remaining environmental parameters can be set to λ1,λ2…λ k , mutual inductance can be regarded as a function of the above parameters and is recorded as M.

[0073] Assume the self-inductance of the two coils is L1 and L2 respectively, and the mutual inductance is M. Then the self-inductance of the inductor formed by the two coils connected in sequence is L 顺 =L1+L2+2M, the self-inductance of the two coils connected inversely is L 反 =L1+L2-2M. Use the impedance tester to measure L 顺 and L 反 , we can get M=(L 顺 -L 反 ) / 4.

[0074] By changing the parameter measurement corresponding to M(λ1,λ2…λ k ,R P ,R S ,xC ,yC,z C ,n x ,n y ,n z ), record multiple sets of data as training data and test data for subsequent BP neural network. BP neural network is the coil mutual inductance model.

[0075] The input X of the BP neural network is the coil's own parameters, the coil's environmental parameters, and the relative position parameters of the two magnetic field coupling coils: λ1, λ2…λ m ,s x ,s y ,s z ,n x ,n y ,n z ,ψ, the standard output d is the coil mutual inductance M(λ1,λ2…λ m ,s x ,s y ,s z ,n x ,n y ,n z ,ψ).

[0076] Activation function of nodes in BP neural network Sure.

[0077] The input layer of the BP neural network contains n nodes, corresponding to the coil's own parameters, the coil's environmental parameters, and the relative position parameters of the two magnetic field coupling coils: λ1, λ2…λ k ,R P ,R S ,x C ,y C ,z C ,n x ,n y ,n z ;

[0078] The number of hidden layers of the BP neural network is 2, and the number of nodes in each layer S can be obtained by Determine, where N represents the total number of training samples, n is the number of input layer nodes, m is the number of output layer nodes, and α can be any positive number, usually in the range of 2-10.

[0079] The output layer of the BP neural network contains 1 node, and its node output y is defined, which is related to the standard output d, that is, the coil mutual inductance M(λ1,λ2…λ k ,R P ,R S ,x C ,y C ,z C ,n x ,ny ,n z ), the output node error is e=dy.

[0080] The forward signal propagates from front to back: the forward output v of the node in the previous layer (kth layer) (k) , multiplied by the corresponding connection weight w and summed to get the positive input x of the node in the next layer (k+1 layer) (k+1) ,Right now:

[0081]

[0082] where x j (k+1) represents the positive input of the jth node in the next layer (k+1th layer), v i (k) represents the positive output of the i-th node in the previous layer (k-th layer), w ij (k,k+1) Represents the connection weight between the i-th node in the previous layer (the k-th layer) and the j-th node in the next layer, and the matrix form is recorded as x (k+1) =W (k,k+1) v (k) , the connection weights are initialized by a random function.

[0083] The forward signal propagation starts with the forward input of the input layer. The coil's own parameters in the training sample, the environmental parameters of the coil, and the relative position parameters of the two magnetic field coupling coils are used as the forward input of the coil mutual inductance model input layer, and directly used as the forward output of the input layer. The forward output of the input layer is multiplied by the connection weight between the input layer and the first hidden layer and summed as the forward input of the first hidden layer. After activation by the activation function, the forward output of the first hidden layer is obtained; and so on, until the forward output of the output layer is calculated and used as the calculation output of the model.

[0084] The reverse error propagation is from back to front, starting from the reverse input of the output layer, and the reverse input of the output layer node is defined as the output node error e (out) =dy, calculate the reverse output of the output layer node in yes The inverse of the point y. According to the reverse output of the output layer δ out , the reverse input e of the second hidden layer can be calculated (Hidden2) =(W (Hidden2,out) ) T δ (out) and the reverse output of the second hidden layer Among them (W (Hidden2,out) ) T is the transpose of the connection weight matrix between the output layer and the second hidden layer, v (Hidden2)is the vector composed of the forward outputs of the nodes in the second hidden layer; and so on, the reverse input and reverse output of the first hidden layer are calculated. (Note: The superscript represents the number of layers; the parameters with subscripts represent the parameters of the nodes in the layer, and the parameters without subscripts represent the vector composed of the parameters of the nodes in the layer.)

[0085] The adjustment amount ΔW of the connection weight from the kth layer to the k+1th layer (k,k+1) =ηδ (k+1) x (k) , where η represents the learning rate, ΔW (k,k+1) is the adjustment amount of the connection weight from the kth layer to the k+1th layer, δ (k+1) is the reverse output of the k+1 layer, x (k) is the forward input of the kth layer, k=1,2,3.

[0086] S3, BP neural network training is carried out by forward signal transmission, reverse error transmission, and connection weight adjustment cycle until the output layer error on the test data is less than the threshold ε or the number of cycles reaches the preset maximum value Max, and the cycle ends to obtain a trained BP neural network.

[0087] S4. Use the trained BP neural network to establish the relationship between the coil's own parameters that do not exist in the S1 training data, the coil's environmental parameters, the relative position parameters of the two magnetic field coupling coils, and the coil mutual inductance. Compare the BP neural network results with the experimental results. If the prediction is not met, adjust ε or Max and retrain.

[0088] It should be understood that although Figure 1-2 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1-2 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0089] In one embodiment, Figure 5 As shown, a coil mutual inductance modeling device for a wireless power transmission system is provided, comprising: a data acquisition module, a coil mutual inductance model building module, a coil mutual inductance model training module and a coil mutual inductance modeling module, wherein:

[0090] The data acquisition module is used to acquire multiple sets of measurement data of two magnetic field coupling coils used for wireless power transmission, and divide the multiple sets of measurement data into training samples and test samples; the measurement data include the coil's own parameters, the environment parameters of the coil, the relative position parameters of the two magnetic field coupling coils, and the corresponding coil mutual inductance between the two coils.

[0091] The coil mutual inductance model construction module is used to construct the coil mutual inductance model; the coil mutual inductance model is a BP neural network, including an input layer, a hidden layer and an output layer; the number of nodes contained in the input layer is the same as the sum of the number of parameters of the coil's own parameters, the environmental parameters of the coil, and the relative position parameters of the two magnetic field coupling coils in the measurement data, the number of hidden layers is 2, and the number of nodes in each hidden layer is determined based on the total number of samples in the training samples, the number of input layer nodes, the number of output layer nodes, and the preset ratio parameters; the output layer contains 1 node.

[0092] The coil mutual inductance model training module is used to train the coil mutual inductance model using training samples. The model is trained by forward signal transmission, reverse error transmission, and cyclic adjustment of connection weights until the output layer error on the test data is less than the preset threshold error or the number of cycles reaches the preset training number, and the cycle ends to obtain a trained coil mutual inductance model.

[0093] The coil mutual inductance module is used to estimate the relationship between the coil's own parameters that do not exist in the training samples and test samples, the coil's environmental parameters, and the relative position parameters of the two magnetic field coupling coils and the coil mutual inductance using the trained coil mutual inductance model.

[0094] In one of the embodiments, the measurement data includes the coil's own parameters, the coil's environmental parameters, the relative position parameters of the two magnetic field coupling coils, and the corresponding coil mutual inductance between the two coils; the coil's own parameters include the shape, size, and number of turns of the coil; the environmental parameters include the conductivity, magnetic permeability, temperature, and pressure of the medium; the data acquisition module is also used to measure the self-inductance of the inductance composed of the two coils connected in series and the self-inductance of the inductance composed of the two coils connected in series with each other using an impedance tester when the coil's own parameters, the coil's environmental parameters, and the relative position parameters of the two magnetic field coupling coils are determined, so as to obtain the corresponding coil mutual inductance; by changing the coil's own parameters, the coil's environmental parameters, and the relative position parameters of the two magnetic field coupling coils, and measuring the corresponding coil mutual inductance, multiple sets of measurement data are obtained.

[0095] In one of the embodiments, the coil mutual inductance model training module is also used to initialize the connection weights between each layer using a random function to obtain the initialization weights; the coil's own parameters in the training sample, the environmental parameters of the coil, and the relative position parameters of the two magnetic field coupling coils are used as the forward input of the coil mutual inductance model input layer, and directly used as the forward output of the input layer; the forward output of the input layer is multiplied by the connection weight between the input layer and the first hidden layer and the sum is used as the forward input of the first hidden layer, and after activation by the activation function, the forward output of the first hidden layer is obtained; and so on, until the forward output of the output layer is calculated and used as the calculated output of the model; according to the difference between the calculated output of the model and the standard output, the output layer error is determined and used as the reverse input of the output layer; the standard output is the coil mutual inductance in the training sample; according to the reverse input of the output layer and the inverse of the activation function, the reverse output of the output layer can be calculated; the reverse input of the second hidden layer is calculated according to the reverse output of the output layer and the transpose of the connection weight matrix between the output layer and the second hidden layer; and so on, until the reverse output of the first hidden layer is calculated; according to the reverse output of the ith node of the latter layer, the preset learning rate and the forward input of the jth node of the previous layer, the adjustment amount of the connection weights between the front and rear layers is obtained; according to the adjustment amount of the connection weights between the front and rear layers and the connection weights between the corresponding layers, the connection weights between the corresponding layers are updated, and the next round of training is carried out until every training sample participates in the training and the training is completed; the trained coil mutual inductance model is tested using the test data until the output layer error on the test data is less than the preset threshold error or the number of cycles reaches the preset number of training rounds, and the cycle is terminated to obtain the trained coil mutual inductance model.

[0096] In one of the embodiments, the device also includes a verification module, which is used to input the input parameters of the coil mutual inductance model other than the training samples and the test samples into the trained coil mutual inductance model to obtain the output result of the model; obtain the output error based on the output result and the corresponding coil mutual inductance obtained from the experiment; when the output error is greater than the preset threshold error, adjust the preset error threshold or the preset number of training times to re-train until the verification result meets the error requirement to obtain the trained coil mutual inductance model.

[0097] In one of the embodiments, the device also includes a verification module, which is used to input the input parameters of the coil mutual inductance model other than the training samples and the test samples into the trained coil mutual inductance model to obtain an output result; obtain an output error based on the output result and the corresponding coil mutual inductance obtained from the experiment; when the output error is greater than a preset threshold error, adjust the preset error threshold or the preset number of training times to re-train until the verification result meets the error requirement to obtain a trained coil mutual inductance model.

[0098] For the specific definition of the mutual inductance modeling device of the wireless power transmission system coil, please refer to the definition of the mutual inductance modeling method of the wireless power transmission system coil mentioned above, which will not be repeated here. Each module in the above-mentioned mutual inductance modeling device of the wireless power transmission system coil can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0099] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0100] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for modeling mutual inductance of coils in a wireless power transmission system, characterized in that: The method comprises: Acquire multiple groups of measurement data of two magnetic field coupling coils for wireless power transmission, and divide the multiple groups of measurement data into training samples and test samples; the measurement data include the coil's own parameters, the coil's environment parameters, the two magnetic field coupling coils relative position parameters and the corresponding coil mutual inductance between the two coils; Constructing a coil mutual inductance model; the coil mutual inductance model is a BP neural network, including an input layer, a hidden layer and an output layer; the number of nodes included in the input layer is the same as the sum of the number of parameters of the coil itself, the environmental parameters of the coil, and the relative position parameters of the two magnetic field coupling coils; the number of hidden layers is 2, and the number of nodes in each hidden layer is determined according to the total number of samples in the training sample, the number of input layer nodes, the number of output layer nodes, and a preset ratio parameter; the output layer includes 1 node; The coil mutual inductance model is trained using training samples. The model is trained by forward signal transmission, reverse error transmission, and cyclic adjustment of connection weights until the output layer error on the test data is less than the preset threshold error or the number of cycles reaches the preset training number, and the cycle ends to obtain a trained coil mutual inductance model. The trained coil mutual inductance model is used to estimate the relationship between the coil's own parameters, the coil's environmental parameters, and the relative position parameters of the two magnetic field coupling coils and the coil mutual inductance that do not exist in the training samples and test samples.

2. The method according to claim 1, characterized in that The measurement data includes the coil's own parameters, the coil's environmental parameters, the relative position parameters of the two magnetic field coupling coils, and the corresponding coil mutual inductance between the two coils; the coil's own parameters include the coil's shape, size, and number of turns; the environmental parameters include the medium's electrical conductivity, magnetic permeability, temperature, and pressure; Acquiring multiple groups of measurement data of two coils, and dividing the multiple groups of measurement data into training samples and test samples, including: When the parameters of the coil itself, the parameters of the environment in which the coil is located, and the relative position parameters of the two magnetic field coupling coils are determined, the self-inductance of the inductor formed by the two coils connected in series and the self-inductance of the inductor formed by the two coils connected in series are measured using an impedance tester to obtain the corresponding coil mutual inductance; By changing the coil's own parameters, the coil's environmental parameters, and the relative position parameters of the two magnetic field coupling coils, and measuring the corresponding coil mutual inductance, multiple sets of measurement data are obtained.

3. The method according to claim 1, characterized in that The coil mutual inductance model is trained using training samples. The model is trained by forward signal transmission, reverse error transmission, and cyclic adjustment of connection weights until the output layer error on the test data is less than the preset threshold error or the number of cycles reaches the preset number of training times to end the cycle, and a trained coil mutual inductance model is obtained, including: Use a random function to initialize the connection weights between layers to obtain the initialization weights; The coil's own parameters in the training sample, the coil's environment parameters, and the relative position parameters of the two magnetic field coupling coils are used as the positive input of the coil mutual inductance model input layer, and directly used as the positive output of the input layer; The forward output of the input layer is multiplied by the connection weight between the input layer and the first hidden layer, and the sum is taken as the forward input of the first hidden layer. After activation by the activation function, the forward output of the first hidden layer is obtained. This process is repeated until the forward output of the output layer is calculated and used as the calculation output of the model. According to the difference between the calculated output and the standard output of the model, the output layer error is determined and used as the reverse input of the output layer; the standard output is the mutual inductance of the coil in the training sample; According to the reverse input of the output layer and the inverse of the activation function, the reverse output of the output layer can be calculated; according to the reverse output of the output layer and the transpose of the connection weight matrix between the output layer and the second hidden layer, the reverse input of the second hidden layer can be calculated; and so on, until the reverse output of the first hidden layer is calculated; According to the reverse output of the i-th node in the next layer, the preset learning rate and the forward input of the j-th node in the previous layer, the adjustment amount of the connection weight between the previous and next layers is obtained; Update the connection weights between the corresponding layers according to the adjustment amount of the connection weights between the previous and next layers and the connection weights between the corresponding layers, and conduct the next round of training until every training sample participates in the training and the training is completed; The trained coil mutual inductance model is tested using test data until the output layer error on the test data is less than a preset threshold error or the number of cycles reaches a preset number of training rounds, and the cycle is terminated to obtain a trained coil mutual inductance model.

4. The method according to claim 1, characterized in that: The input parameters of the coil mutual inductance model include: the coil's own parameters, the coil's environment parameters, and the relative position parameters of the two magnetic field coupling coils; the method also includes: Inputting the input parameters of the coil mutual inductance model other than the training samples and the test samples into the trained coil mutual inductance model to obtain the output result of the model; Obtaining an output error according to the output result and the corresponding coil mutual inductance obtained from the experiment; When the output error is greater than a preset threshold error, the preset error threshold or the preset number of training times is adjusted to retrain until the verification result meets the error requirement, thereby obtaining a trained coil mutual inductance model.

5. A wireless power transmission system coil mutual inductance modeling device, characterized in that: The device comprises: A data acquisition module, used to acquire multiple groups of measurement data of two magnetic field coupling coils used for wireless power transmission, and divide the multiple groups of measurement data into training samples and test samples; the measurement data include the coil's own parameters, the coil's environmental parameters, the relative position parameters of the two magnetic field coupling coils, and the corresponding coil mutual inductance between the two coils; A coil mutual inductance model construction module is used to construct a coil mutual inductance model; the coil mutual inductance model is a BP neural network, including an input layer, a hidden layer and an output layer; the number of nodes included in the input layer is the same as the sum of the number of parameters of the coil itself, the environmental parameters of the coil, and the relative position parameters of the two magnetic field coupling coils in the measurement data, the number of hidden layers is 2, and the number of nodes in each hidden layer is determined according to the total number of samples in the training sample, the number of input layer nodes, the number of output layer nodes and a preset ratio parameter; the output layer contains 1 node; The coil mutual inductance model training module is used to train the coil mutual inductance model using training samples. The model is trained by forward signal transmission, reverse error transmission, and cyclic adjustment of connection weights until the output layer error on the test data is less than the preset threshold error or the number of cycles reaches the preset training number, and the cycle ends to obtain a trained coil mutual inductance model; The coil mutual inductance module is used to estimate the relationship between the coil's own parameters that do not exist in the training samples and test samples, the coil's environmental parameters, and the relative position parameters of the two magnetic field coupling coils and the coil mutual inductance using the trained coil mutual inductance model.

6. The device according to claim 5, characterized in that The measurement data includes the coil's own parameters, the coil's environmental parameters, the relative position parameters of the two magnetic field coupling coils, and the corresponding coil mutual inductance between the two coils; the coil's own parameters include the coil's shape, size, and number of turns; the environmental parameters include the medium's electrical conductivity, magnetic permeability, temperature, and pressure; The data acquisition module is also used to measure the self-inductance of the inductor composed of two coils connected in series and the self-inductance of the inductor composed of two coils connected in series reversely, using an impedance tester, when the coil's own parameters, the coil's environmental parameters and the relative position parameters of the two magnetic field coupling coils are determined, to obtain the corresponding mutual inductance of the coils; by changing the coil's own parameters, the coil's environmental parameters and the relative position parameters of the two magnetic field coupling coils, and measuring the corresponding mutual inductance of the coils, multiple sets of measurement data are obtained.

7. The device according to claim 5, characterized in that The coil mutual inductance model training module is also used to initialize the connection weights between layers using a random function to obtain the initialization weights; The coil's own parameters in the training sample, the environment parameters of the coil, and the relative position parameters of the two magnetic field coupling coils are used as the forward input of the input layer of the coil mutual inductance model, and directly used as the forward output of the input layer; the forward output of the input layer is multiplied by the connection weight between the input layer and the first hidden layer and the sum is used as the forward input of the first hidden layer, and after activation by the activation function, the forward output of the first hidden layer is obtained; and so on, until the forward output of the output layer is calculated and used as the calculated output of the model; the output layer error is determined according to the difference between the calculated output and the standard output of the model, and used as the reverse input of the output layer; the standard output is the mutual inductance of the coil in the training sample; the reverse output of the output layer can be calculated according to the reverse input of the output layer and the inverse of the activation function; the reverse input of the second hidden layer is calculated according to the reverse output of the output layer and the transpose of the connection weight matrix between the output layer and the second hidden layer; And so on, until the reverse output of the first hidden layer is calculated; According to the reverse output of the i-th node in the next layer, the preset learning rate and the forward input of the j-th node in the previous layer, the adjustment amount of the connection weight between the previous and next layers is obtained; The connection weights between the corresponding layers are updated according to the adjustment amount of the connection weights between the front and rear layers and the connection weights between the corresponding layers, and the next round of training is carried out until every training sample participates in the training and the training is completed; the trained coil mutual inductance model is tested using the test data until the output layer error on the test data is less than the preset threshold error or the number of cycles reaches the preset number of training rounds, and the cycle is terminated to obtain the trained coil mutual inductance model.

8. The device according to claim 5, characterized in that The input parameters of the coil mutual inductance model include: the coil's own parameters, the environment parameters of the coil, and the relative position parameters of the two magnetic field coupling coils; the device also includes a verification module, which is used to input the input parameters of the coil mutual inductance model other than the training samples and the test samples into the trained coil mutual inductance model to obtain the output result of the model; according to the output result and the corresponding coil mutual inductance obtained by the experiment, the output error is obtained; when the output error is greater than the preset threshold error, the preset error threshold or the preset number of training times is adjusted to re-train until the verification result meets the error requirement, and the trained coil mutual inductance model is obtained.

Images