Low-power-consumption current monitoring device and method for optimizing magnetoelectric composite material based on hybrid model

Abstract

The invention relates to the technical field of current monitoring, in particular to a low-power-consumption current monitoring device and method for optimizing a magnetoelectric composite material based on a hybrid model. According to the low-power-consumption current monitoring device and method for optimizing the magnetoelectric composite material based on the hybrid model, through material-structure-circuit collaborative optimization, the defects of a traditional current monitoring method in the aspects of performance, power consumption and adaptability are overcome; and an efficient and reliable current monitoring solution is provided for smart power grid and Internet of Things scenes.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of current monitoring technology, and in particular to a low-power current monitoring device and method based on hybrid model optimization of magneto-electric composite materials. BACKGROUND

[0002] With the rapid development of smart grid and Internet of Things technology, current monitoring technology needs to meet the demands of low power consumption, high precision and strong adaptability. Traditional current monitoring methods (such as electromagnetic current transformers) have problems such as large size, narrow frequency band, high power consumption, etc., and are difficult to adapt to complex scenarios. Magneto-electric composite materials provide a new direction for current monitoring due to their high magneto-electric conversion coefficient and pT-level magnetic field detection capability. However, the magnetic-electric coupling characteristics of magneto-electric composite materials are affected by the nonlinear coupling of multiple parameters (such as composition ratio, bias magnetic field, etc.), and traditional optimization methods rely on experimental trial and error, which is inefficient and prone to overfitting.

[0003] In addition, the existing sensor structure is difficult to adapt to different diameter wires, has weak common mode noise resistance, and has high signal chain power consumption; traditional sensors rely on external power supply, and the static power consumption is difficult to approach zero, which restricts their application in low-power scenarios.

[0004] Therefore, how to design a low-power current monitoring device and method based on hybrid model optimization of magneto-electric composite materials to solve the above technical problems is a technical problem to be solved. SUMMARY

[0005] To solve the above problems, the purpose of the present application is to provide a low-power current monitoring device and method based on hybrid model optimization of magneto-electric composite materials, which solves the deficiencies of traditional current monitoring methods in performance, power consumption and adaptability through material-structure-circuit collaborative optimization, and provides an efficient and reliable current monitoring solution for smart grid and Internet of Things scenarios.

[0006] In a first aspect, the present application provides a low-power current monitoring method based on hybrid model optimization of magneto-electric composite materials, comprising the following steps:

[0007] Step S1: Collecting magneto-electric composite material data and performing standardization processing;

[0008] Step S2: According to the data processing result, adopting PCA dimension reduction to perform original data standardization processing on the original data to eliminate dimensional differences;

[0009] Step S3: Combining the data standardization processing result, adopting random forest regression modeling to select Bootstrap sampling to generate multiple sub-training sets, constructing multiple regression trees and integrating each tree prediction result through ensemble learning, and outputting the accurate prediction value of the magneto-electric coefficient;

[0010] Step S4: According to the representation result of the output magneto-electric coefficient accurate prediction value, the influence of the principal component on the magneto-electric coefficient is quantified based on the Gini impurity reduction amount and / or the replacement importance;

[0011] Step S5: The performance of the above regression model is evaluated in combination with the determination coefficient (R2), the mean square error (MSE), the root mean square error (RMSE) and the absolute mean error (MAE).

[0012] The standardization process is as follows: the magneto-electric composite experimental data is input into the data matrix:

[0013]

[0014] Wherein, n is the sample number, p is the original parameter of the bias magnetic field value, the aspect ratio, the piezomagnetic phase ratio value, the piezomagnetic coefficient, the piezoelectric phase flexibility coefficient, the temperature and the compression load 7 parameters; the magneto-electric coefficient is used as the output parameter; the standardization formula is as follows:

[0015]

[0016] The step S2 is as follows:

[0017] Firstly, the covariance matrix is calculated, and the formula is as follows:

[0018]

[0019] Wherein, Z is the standardized matrix;

[0020] Then, the eigenvalue decomposition is carried out on the above C representation result, and the first k principal components are selected according to the size of the eigenvalue;

[0021] Finally, according to the selected k principal components, the principal component matrix is constructed, and the formula is as follows:

[0022] T = Z·W K

[0023] Wherein, W K is the projection matrix of the first k eigenvectors.

[0024] In step S3, the accurate prediction value of the output electromagnetic coefficient is output, and the formula is as follows:

[0025]

[0026] Wherein, y i，pred is the prediction value of the ith tree.

[0027] In step S4, the formula is as follows:

[0028]

[0029] The step S5 is specifically formulated as follows:

[0030]

[0031] Wherein, y i，pred The PCA-RF model prediction value is represented by y i，real The actual value is represented by y, and y represents the average value of all actual values.

[0032] In the second aspect, the application provides a low-power current monitoring device for optimizing magnetoelectric composite materials based on a hybrid model, which comprises a box body, core sensitive elements symmetrically arranged on the upper and lower sides of the box body and connected in series at the output ends, and U-shaped elastic clamps symmetrically arranged between the core sensitive elements on the upper and lower sides.

[0033] Further, the inner side of the clamp is also symmetrically provided with elastic long pieces for fixing wires with different diameters.

[0034] The application has the following beneficial effects:

[0035] 1. The PCA-RF-based magnetoelectric composite material optimization model of the application eliminates multi-parameter redundancy through principal component analysis dimension reduction, accurately predicts the magnetoelectric coefficient by combining random forest nonlinear regression, reveals the synergistic optimization mechanism of material composite ratio and bias magnetic field, and significantly improves the magnetoelectric conversion efficiency.

[0036] 2. The application designs a self-powered low-power circuit: the energy conversion characteristics of the magnetoelectric composite material are used to realize self-power supply, and the ultra-low power signal processing circuit and the direct current bias magnetic optimization technology are combined to make the static power consumption close to zero.

[0037] 3. The application realizes high-precision non-destructive current detection by equivalent modeling of distributed parasitic resistance, integrates the parasitic resistance of the power line into a "virtual" detection resistance, and further reduces the system power consumption.

[0038] 4. The application adopts a double-magnetoelectric sensitive unit split design, combines a U-shaped elastic clamp, realizes self-adaptive fixing of wire diameter and common mode noise suppression, and enhances signal stability. DETAILED DESCRIPTION

[0039] Figure 1 It is a schematic diagram of the first embodiment of the application;

[0040] Figure 2 It is a schematic diagram of the second embodiment of the application;

[0041] Figure 3 It is a schematic diagram of the application;

[0042] Figure 4 It is a schematic diagram of the application.

[0043] Description of reference numerals: DETAILED DESCRIPTION

[0044] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0045] Example 1

[0046] See also Figure 1 As shown, this embodiment provides a low-power current monitoring method based on hybrid model optimization of magnetoelectric composite materials, including the following steps:

[0047] Step S1: collecting magnetoelectric composite material data and performing standardization processing;

[0048] Step S2: Based on the data processing results, PCA dimensionality reduction is used to standardize the original data to eliminate dimensional differences;

[0049] Step S3: Combined with the data normalization results, random forest regression modeling is used to select Bootstrap sampling to generate multiple sub-training sets, multiple regression trees are constructed, and the prediction results of each tree are integrated through ensemble learning to output the accurate prediction value of the magnetoelectric coefficient;

[0050] Step S4: quantifying the influence of the principal component on the magnetoelectric coefficient based on the Gini impurity reduction and / or substitution importance according to the representation result of the output magnetoelectric coefficient accurate prediction value;

[0051] Step S5: The performance of the above regression model is evaluated by combining the coefficient of determination (R2), mean square error (MSE), root mean square error (RMSE) and mean absolute error (MAE).

[0052] The standardization process is as follows: the experimental data of the magnetoelectric composite material is input into the data matrix:

[0053]

[0054] Where n is the number of samples, p is the original parameters of the seven parameters: bias magnetic field value, aspect ratio, piezomagnetic phase volume fraction, piezomagnetic coefficient, piezoelectric phase compliance coefficient, temperature, and compression load; the magnetoelectric coefficient is used as the output parameter; the normalization formula is as follows:

[0055]

[0056] The step S2 is specifically as follows:

[0057] First, calculate the covariance matrix, which is as follows:

[0058]

[0059] Where Z is the normalization matrix;

[0060] Then, the C result is subjected to eigenvalue decomposition, sorted by eigenvalue size and select the first k principal components;

[0061] Finally, according to the selected k principal components, the principal component matrix is constructed, and the formula is as follows:

[0062] T = Z·W K

[0063] Where W K is the projection matrix of the first k eigenvectors.

[0064] In step S3, the accurate prediction value of electromagnetic coefficient is output, and the formula is as follows:

[0065]

[0066] Where y i,pred is the prediction value of the ith tree.

[0067] In step S4, the formula is as follows:

[0068]

[0069] In step S5, the specific formula is as follows:

[0070]

[0071] Where y i，pred represents the PCA-RF model prediction value, y i，real represents the actual value, and y represents the average of all actual values. R2 shows the fitting degree between the model prediction value and the actual value. MSE, RMSE, MAE represent the error value between the prediction value and the true value. When the model has higher R 2 and lower error, it means that the prediction performance of the model is good.

[0072] In the experimental part, the parameters of RF are as follows: the number of trees m∈[50, 150], determined by the convergence test of out-of-bag error (OOB error); the maximum depth d∈[3, 7], used to limit the model complexity to prevent overfitting; the minimum leaf sample number d min ∈[3, 7], used to avoid noise interference. The data set has 120 samples, which is divided into training set (96 groups) and test set (24 groups) in the ratio of 8:2. The comparative experiment is genetic algorithm optimized BP network, and the network structure is input layer (10 nodes), hidden layer (8 nodes), output layer (1 node); the population size is 50, the iteration number is 100, the crossover probability is 0.8, and the mutation probability is 0.1.

[0073] As Figure 3 shown, the performance comparison of PCA-RF and BP models is shown, and the analysis shows that the prediction results of the two models of magnetoelectric coefficient all exhibit high fitting degree, wherein the R2 of the PCA-RF model is higher than that of the BP model, reaching 97.43%, which is 2.01% higher than the R2 of the BP model without optimization, and the three error indicators of the optimized PCA-RF model are also lower than the indicators of the model without optimization. This shows that in the prediction of the magnetoelectric coefficient, the difference between the prediction result and the actual result of the optimized PCA-RF model is smaller, and the prediction value is closer to the actual value, which shows that the model has good prediction ability. Therefore, the prediction accuracy of the magnetoelectric coefficient can be improved by using the model, and the optimization and design of the magnetoelectric composite material can be promoted.

[0074] Figure 4 To use the proposed low-power current monitoring technology in current-mode switching power supply chips, and based on the 0.8 μm BiCMOS process design, the circuit is simulated and the static power consumption is optimized, and the reference experiments for comparison are LEEC-F and LEUNG-Y. Figure 4 It is shown that when the number of MOS tubes is 25-40, the static power consumption of the low-power current monitoring circuit is significantly better than that of LEEC-F and LEUNG-Y. When the number of MOS tubes is 30, the static power consumption of the embodiment is 70 (unit: μW), which is 22.2% and 17.6% lower than that of LEEC-F (90 μW) and LEUNG-Y (85 μW), respectively.

[0075] Example two

[0076] Referring to Figure 2 As shown, the embodiment provides a low-power current monitoring device for optimizing magnetoelectric composite materials based on a hybrid model, which comprises a box body, core sensitive elements symmetrically arranged on the upper and lower sides of the box body and connected in series at the output ends, and U-shaped elastic clamps symmetrically arranged between the core sensitive elements on the upper and lower sides. The box body is composed of a ceramic substrate and an acrylic cover, and the core sensitive elements are made of magnetoelectric composite materials.

[0077] The double-unit split differential current sensor utilizes the product effect of the magnetoelectric effect and the Ampere loop theorem, treats the alternating magnetic field as an intermediate quantity, and the measured current causes the change of the surrounding magnetic field, and the change of the magnetic field causes the deformation of the material to cause the change of the output voltage, thereby achieving the purpose of measuring the current. From the perspective of energy transfer, the energy conversion of the whole measurement process is: electric → magnetic → machine → electric. When there is no ferromagnetic material around the current-carrying wire, the magnetic lines of force generated by the current-carrying wire around it are concentric circles, and the magnetic field size is: Wherein, I is the current size of the wire, R is the distance from the wire center to the magnetic field Hi. When the ferromagnetic material is placed around the wire, the magnetic lines are attracted by the ferromagnetic material, and the distribution of the magnetic lines changes. However, due to the symmetry of the structure in the box, the two magnetostrictive units still receive magnetic fields with equal size and opposite direction.

[0078] Further, the inside of the clamp is also symmetrically provided with an elastic long piece for fixing the wire with different diameters. In order to realize the differential effect, the permanent magnet is designed to apply a bias magnetic field with equal size and same direction to the two magnetostrictive units. At this time, under the action of the magnetic field generated by the current-carrying wire, when the magnetostrictive unit 1 generates relative elongation, the magnetostrictive unit 2 generates relative shortening, and vice versa. When the current in the wire is alternating current, the output of the composite magneto-sensitive unit 1 at the same time is -U, and the output of the composite magneto-sensitive unit 2 is +U. Connecting the two outputs in series can realize the difference, and get the output voltage of 2U. Under the action of temperature, the units 1 and 2 expand or contract at the same time, and the differential output is zero.

[0079] According to the low-frequency magneto-electric effect, after the L-T type composite magneto-electric material senses the low-frequency magnetic field around the current-carrying wire, the magneto-electric coupling between the magnetostrictive layer and the piezoelectric layer occurs due to the magnetic elastic strain. The theoretical formula of the magneto-electric voltage coefficient αd of the two differential output composite magneto-sensitive units at low frequency is:

[0080]

[0081] Through the above analysis, the main cause of the magneto-electric response of the composite magneto-electric material is the alternating magnetic field Hi generated around the current-carrying wire. The voltage-current conversion coefficient αV-I, that is, the sensitivity of the sensor, can be represented as: Wherein, represents the change of the magnetic field strength caused by the change of the unit current, and represent the piezoelectric effect and the magnetostrictive effect respectively, and the mechanical stress o of the intermediate layer acts as an intermediate variable in this conversion process. Based on the above formula, the sensitivity of the sensor can be represented as:

[0082] It can be seen that under the condition of determined material parameters and size, the voltage value 2U output by the sensor and the current value I to be measured have a linear increasing relationship.

[0083] According to the above analysis and summary, a self-powered low-power circuit system is constructed, including M power tube units on the same row. Rs, Rc and Rds represent the parasitic resistance of the wire connected to the single power tube unit, the metal hole resistance and the power tube channel resistance. Since R ds +R c >>R s , the current on each power tube is affected by Rs The impact is very small, use I uniformly ds The total voltage drop V across adjacent parasitic resistors is sm =V sm =∑ i R s (k+i)I ds Since there are M resistors that are considered as a centralized "virtual" resistor,

[0084] V sm It can be expressed as: Output current I o =MNI ds , then V sm It can be further expressed as: Among them, R sm =(M+1)R s / 2N, represents the collective representation of all adjacent parasitic sense resistors, the “virtual” resistor R sm It can be regarded as a lossless current detection resistor directly connected in series with the power tube M

[0085] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0086] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0087] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 the function specified in the one or more blocks.

[0088] These computer program instructions can also be loaded into computer or other programmable data processing devices, so that a series of operation steps are performed on the computer or other programmable data processing devices to generate computer-implemented processes, so that the instructions executed on the computer or other programmable data processing devices provide processes for implementing the flow Figure 1 the flow or flows and / or blocks Figure 1 the steps of the function specified in the one or more blocks.

[0089] The above description is only the preferred embodiment of the present application, not other forms of the present application, any skilled in the art can use the above disclosed technical content to change or modify as equivalent embodiments of equivalent changes. But any simple modification, equivalent change and modification of the above embodiments without departing from the technical solution of the present application, according to the technical essence of the present application, still belongs to the protection scope of the technical solution of the present application.

Claims

1. A low-power current monitoring method based on hybrid model optimization of magnetoelectric composite materials, characterized by: The following steps are involved: Step S1: collecting magnetoelectric composite material data and performing standardization processing; Step S2: Based on the data processing results, PCA dimensionality reduction is used to standardize the original data to eliminate dimensional differences; Step S3: Combined with the data normalization results, random forest regression modeling is used to select Bootstrap sampling to generate multiple sub-training sets, multiple regression trees are constructed, and the prediction results of each tree are integrated through ensemble learning to output the accurate prediction value of the magnetoelectric coefficient; Step S4: quantifying the influence of the principal component on the magnetoelectric coefficient based on the Gini impurity reduction and / or substitution importance according to the representation result of the output magnetoelectric coefficient accurate prediction value; Step S5: The performance of the above regression model is evaluated by combining the coefficient of determination (R2), mean square error (MSE), root mean square error (RMSE) and mean absolute error (MAE).

2. The low-power current monitoring method based on hybrid model optimization of magnetoelectric composite materials according to claim 1, characterized in that: The standardization process is as follows: the experimental data of the magnetoelectric composite material is input into the data matrix: Where n is the number of samples, p is the original parameters of the seven parameters: bias magnetic field value, aspect ratio, piezomagnetic phase volume fraction, piezomagnetic coefficient, piezoelectric phase compliance coefficient, temperature, and compression load; the magnetoelectric coefficient is used as the output parameter; the normalization formula is as follows:

3. The low-power current monitoring method based on hybrid model optimization of magnetoelectric composite materials according to claim 1, characterized in that: The step S2 is specifically as follows: First, calculate the covariance matrix, which is as follows: Where Z is the normalization matrix; Then, perform eigenvalue decomposition on the above C representation results, sort them by eigenvalue size and select the first k principal components; Finally, based on the selected k principal components, the principal component matrix is ​​constructed, and its formula is as follows: T=Z·W K Among them, W K is the projection matrix of the first k eigenvectors.

4. The low-power current monitoring method based on hybrid model optimization of magnetoelectric composite materials according to claim 1, characterized in that: In step S3, the accurate predicted value of the electromagnetic coefficient is output, and the formula is as follows: Among them, y i，pred is the predicted value of the i-th tree.

5. The low-power current monitoring method based on hybrid model optimization of magnetoelectric composite materials according to claim 1, characterized in that: In step S4, the formula is as follows:

6. The low-power current monitoring method based on hybrid model optimization of magnetoelectric composite materials according to claim 1, characterized in that: The specific formula of step S5 is as follows: Among them, y i，pred Represents the PCA-RF model prediction value, y i，real Represents the actual value, Represents the average of all actual values.

7. A low-power current monitoring device based on hybrid model optimization of magnetoelectric composite materials, characterized by: The invention comprises a box body, core sensitive elements symmetrically fixed on the upper and lower sides of the box body and with output ends connected in series, and a U-shaped elastic clamp symmetrically arranged between the core sensitive elements on the upper and lower sides.

8. The low-power current monitoring device based on hybrid model optimization of magnetoelectric composite materials according to claim 7, characterized in that: Elastic long pieces for fixing wires of different diameters are symmetrically provided on the inner side of the clamp.

Images