A method for identifying a medium of a power device based on a digital-analog drive

Abstract

The application relates to the technical field of power detection, and discloses a power equipment medium identification method based on digital-analog driving, which comprises the following steps: constructing an equipment geometric field and arranging boundary measuring points; obtaining physical quantity and normal derivative data of the measuring points through non-invasive measurement and fitting; based on a physical control equation, unknown medium parameters are converted into equivalent source terms through standardization and an equivalent source method, and then the boundary integral equation is converted into the boundary integral equation through a boundary element method, so that the measurement data and the internal source terms are associated; a surface quadrature neural network is constructed, the field point coordinates and the boundary data are taken as inputs, the internal equivalent source term distribution is output, and a loss function is constructed based on the boundary integral equation to train the network; real-time data are calculated forwardly by using the trained network, the equivalent source terms are obtained, and the medium distribution is inversed, the application provides the power equipment medium identification method based on the digital-analog driving, which can quickly judge the cause of the internal condition of the equipment and perform tracing and positioning.

Description

Technical Field

[0001] This invention relates to the field of power detection technology, and in particular to a method for identifying the medium of power equipment based on digital-analog driving. Background Technology

[0002] With the advancement of my country's power system construction, more electrical equipment is being put into use. As a core component of the power system, the operational reliability of electrical equipment directly affects the dynamic stability and power supply continuity of the power system. Therefore, the inspection of electrical equipment is crucial. Direct detection inside electrical equipment often requires invasive operations, which may cause irreversible damage to the equipment. Furthermore, traditional detection methods struggle to balance real-time performance and reliability, failing to achieve precise location and analysis of internal conditions (such as faults). Therefore, conducting internal media tracing based on limited surface information of the equipment has significant engineering application value.

[0003] A team from State Grid Sichuan Electric Power Company established a model using the finite element multiphysics analysis method to study the characteristics of the internal temperature field of power equipment, providing a reference for condition monitoring. A team from State Grid Liaoning Electric Power Company verified the feasibility of a method for diagnosing localized overheating faults in power equipment based on infrared images in experiments.

[0004] In existing technologies, the finite element analysis method is often limited by the accuracy of mesh discretization and the assumption of simplified boundary conditions. The experimental verification method also has large errors due to its reliance on subjective experience and experimental errors. Moreover, existing technologies are often studied for a specific power equipment (such as GIS, converter valve, etc.), and due to the characteristics of different equipment, the technology cannot be extended to multiple equipment. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a method for identifying the medium of power equipment based on digital-analog driving, which can quickly determine the cause of the internal condition of the equipment and perform source tracing and location.

[0006] This invention provides a method for identifying the medium of power equipment based on analog-digital driving, comprising the following steps: Construct the geometric field of the target power equipment and arrange measurement points on the boundary of the geometric field; The physical quantity data and normal derivative data related to the medium to be identified at all measurement points are obtained through non-invasive measurement, and the boundary data are fitted. Based on the governing equations describing the physical processes within the geometric field, a fusion of numerical and model-driven equations is constructed. The construction process includes: standardizing the governing equations, using the equivalent source method to transform unknown medium distribution parameters into unknown equivalent source terms, and applying the boundary element method to transform the governing equations into boundary integral equations. The boundary integral equations correlate boundary measurement data with equivalent source terms within the field. Construct and train a surface quadrature neural network; the input of the surface quadrature neural network includes the field point coordinates and the obtained boundary data, and the output is the equivalent source term distribution inside the field; during training, construct the core loss function based on the boundary integral equation, and adjust the neural network parameters by optimizing the loss function so that the output of the neural network satisfies the boundary integral equation; Based on a trained surface quadrature neural network, forward computation is performed on the input real-time measurement data to output the equivalent source term distribution at the current moment; according to the relationship between the equivalent source term and the medium distribution in the control equation, the medium distribution inside the geometric field is obtained by solving the problem. Based on the comparison between the medium distribution and the preset threshold, the internal status of the equipment is determined and an early warning message is output.

[0007] Furthermore, fitting is performed on the boundary data, including: The boundary parameterization method is determined based on the geometric shape characteristics of the boundary of the geometric field. Based on the parameterization method, select the interpolation or fitting algorithm that matches the geometric features; Based on the parameterization method and the selected algorithm, discrete measurement point data are reconstructed into continuous boundary functions.

[0008] Furthermore, the physical quantity includes at least one of temperature, electric potential, magnetic induction intensity, or sound pressure.

[0009] Furthermore, the boundary integral equation obtained by applying the boundary element method is: In the formula, As the boundary of the field, For a complete field. Represents the source point on the boundary. Representing field points on the boundary, Represents the field points within the field area. This represents the basic solution form of the physical quantity to be measured. The fundamental solution form of its directional derivative is given by, where This parameter specifies the distance between the source point and the field point. The properties of the field boundary determine this: if the boundary is smooth, then... If the boundary is not smooth, then .

[0010] Furthermore, the surface quadrature neural network is a multilayer perceptron containing at least one hidden layer; the core loss function is constructed based on the residuals of the boundary integral equation, and is expressed as: .

[0011] Furthermore, according to the equation relating the equivalent source term to the medium distribution parameters: Solving for the medium distribution ; in Let the input point be the position vector. For the distribution of media within the field, The distribution of physical quantities within the field. For source terms, is the unknown source term of the equation.

[0012] Furthermore, the dielectric distribution parameters include at least one of electrical conductivity, thermal conductivity, magnetic permeability, or acoustic impedance.

[0013] Furthermore, the obtained medium distribution parameters are compared with the reference parameters under normal operating conditions. If the parameter difference in a specific region exceeds a set threshold, the region is determined to be an abnormal region, and the abnormal operating condition type is identified.

[0014] A power equipment condition monitoring system, comprising: The measurement module is used to non-invasively acquire physical quantity data and their normal derivatives from measurement points on the surface of power equipment; The data processing and recognition module includes a power equipment media recognition method based on digital analog driving. And a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a digital-analog driven method for identifying power equipment media.

[0015] The technical solution provided by this invention has the following advantages compared with the prior art: This invention only requires obtaining the surface information of the power equipment, solving the problem of difficulty in obtaining the internal physical parameters of the equipment. Through neural network training, nonlinear relationships in the data can be captured, reflecting the internal physical model of the equipment, improving detection accuracy. Simultaneously, its rapid inference also improves the real-time performance of detection. The model-driven approach integrates the prior knowledge of the physical model with the fitting and generalization ability of the data-driven approach, avoiding the assumption bias and parameter errors of pure model-driven approaches, and overcoming the shortcomings of pure data-driven approaches such as dependence on large amounts of data and lack of interpretability, achieving higher accuracy and engineering practicality. By analyzing changes in the medium parameters within the equipment, the cause of the internal condition can be quickly determined, and the source can be traced and located. The technology has good scalability and can be applied to the detection of different equipment and different media. Attached Figure Description

[0016] Figure 1 A flowchart illustrating a power equipment dielectric identification method based on digital-analog driving, provided for an embodiment of the present invention; Figure 2 The above diagram illustrates the measurement points and boundary fitting on the surface of common power equipment (taking gas-insulated switches and reactors as examples) under different geometric fields, as provided in the embodiments of the present invention. Figure 3 This is a schematic diagram of the analog-digital fusion driving method provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of a surface quadrature neural network (SINN) provided in an embodiment of the present invention. Detailed Implementation

[0017] The following detailed description of a specific embodiment of the present invention is provided in conjunction with the accompanying drawings. However, it should be understood that the scope of protection of the present invention is not limited to the specific embodiment.

[0018] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the technical solution of this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0019] The present invention will be described below through several specific embodiments. To keep the following description of the embodiments clear and concise, detailed descriptions of known functions and components may be omitted. When any component of an embodiment of the present invention appears in more than one drawing, the component may be represented by the same reference numerals in each drawing.

[0020] Figure 1 A flowchart illustrating a power equipment media identification method based on analog-digital driving, provided in an embodiment of the present invention. Figure 2 This invention provides schematic diagrams of measuring points and boundary fitting on the surfaces of common power equipment (taking gas-insulated switches and reactors as examples) under different geometric fields. Figure 3 This is a schematic diagram of the analog-digital fusion driving method provided in an embodiment of the present invention. Figure 4 This is a schematic diagram of a surface quadrature neural network (SINN) provided in an embodiment of the present invention.

[0021] like Figure 1 and Figure 2As shown, this invention provides a method for identifying the medium of power equipment based on numerical simulation, which mainly includes the following steps: constructing the geometric field of the target power equipment and arranging measurement points on the boundary of the geometric field; acquiring physical quantity data and their normal derivative data related to the medium to be identified at all measurement points through non-invasive measurement, and fitting the boundary data; constructing a numerical simulation fusion driving equation based on the control equation describing the physical process inside the geometric field; the construction process includes: standardizing the control equation, using the equivalent source method to transform the unknown medium distribution parameters into unknown equivalent source terms, and applying the boundary element method to transform the control equation into a boundary integral equation; the boundary integral equation integrates the boundary measurement data with the field. The internal equivalent source terms are correlated; a surface quadrature neural network is constructed and trained; the input of the surface quadrature neural network includes the field point coordinates and the obtained boundary data, and the output is the distribution of equivalent source terms inside the field; during training, a core loss function is constructed based on the boundary integral equation, and the neural network parameters are adjusted by optimizing the loss function so that the output of the neural network satisfies the boundary integral equation; based on the trained surface quadrature neural network, forward calculation is performed on the input real-time measurement data to output the distribution of equivalent source terms at the current time; according to the relationship between the equivalent source terms and the medium distribution in the control equation, the medium distribution inside the geometric field is obtained; based on the comparison result of the medium distribution and the preset threshold, the internal state of the equipment is judged and a warning information is output.

[0022] Furthermore, the boundary data is fitted, including: determining the boundary parameterization method based on the geometric shape characteristics of the geometric field boundary; selecting an interpolation or fitting algorithm that matches the geometric shape characteristics based on the parameterization method; and reconstructing the discrete measurement point data into a continuous boundary function based on the parameterization method and the selected algorithm.

[0023] Specifically, the data on the complete boundary is obtained by fitting the measurement points. The fitting method is determined based on the geometry of the boundary. Essentially, this fitting method uses discrete point cloud data to reconstruct continuous, complete boundary data. Several typical boundary processing methods are as follows: Smooth boundary: First, the center coordinates and major and minor axis parameters are inferred from the measurement points, and the continuous geometric path is determined using polar coordinate equations. Next, the functional relationship between the polar angle at each point and the original data is constructed, and a data mapping model is established using piecewise cubic Hermitian polynomial interpolation. Finally, a small angle variation is used as the step size across the entire boundary range, and the data on the boundary is derived using the mapping model.

[0024] Polygon Boundary: First, vertices are identified based on the slope changes of the lines connecting adjacent points. Then, a dense network of boundary point coordinates is constructed between the vertices using linear interpolation. Next, the cumulative path length along the boundary contour is calculated for all measured points, mapping the discrete data at different points to a function of the path length. Finally, piecewise quadratic interpolation is used to calculate the data for each newly added boundary point, thus reconstructing the polygon boundary data.

[0025] Irregular Boundaries: First, the measurement points are sorted, and a smooth spline curve is fitted using cubic spline interpolation. Next, an arc length parameter is introduced, and the displacement of each point on the curve is recorded. Moving least squares method is used for local fitting of the data. Finally, high-density sampling is performed within the total arc length range, mapping each tiny arc length increment back to the coordinates of the points on the spline curve and their corresponding interpolated data.

[0026] Furthermore, the physical quantity includes at least one of temperature, electric potential, magnetic induction intensity, or sound pressure.

[0027] Furthermore, the boundary integral equation obtained by applying the boundary element method is: In the formula, As the boundary of the field, For a complete field. Represents the source point on the boundary. Representing field points on the boundary, Represents the field points within the field area. This represents the basic solution form of the physical quantity to be measured. The fundamental solution form of its directional derivative is given by, where This parameter specifies the distance between the source point and the field point. The properties of the field boundary determine this: if the boundary is smooth, then... If the boundary is not smooth, then .

[0028] Furthermore, the surface quadrature neural network is a multilayer perceptron containing at least one hidden layer; the core loss function is constructed based on the residuals of the boundary integral equation, and is expressed as: .

[0029] Furthermore, according to the equation relating the equivalent source term to the medium distribution parameters: Solving for the medium distribution ;in Let the input point be the position vector. For the distribution of media within the field, The distribution of physical quantities within the field. For source terms, is the unknown source term of the equation.

[0030] Furthermore, the dielectric distribution parameters include at least one of electrical conductivity, thermal conductivity, magnetic permeability, or acoustic impedance.

[0031] Furthermore, the obtained medium distribution parameters are compared with the reference parameters under normal operating conditions. If the parameter difference in a specific region exceeds a set threshold, the region is determined to be an abnormal region, and the abnormal operating condition type is identified.

[0032] Specifically, online identification is performed using a pre-trained method. The field and boundaries of the power equipment under test are defined, measurement point information on the equipment surface is acquired and fitted, physical equations are defined, and a surface integral neural network (SINN) driven by a numerical-analog fusion method is constructed. Measurement data is then input into the pre-trained neural network system for real-time online identification.

[0033] If the distribution difference of the neural network output medium in a certain area exceeds a set threshold, the system will alarm and indicate the abnormal area, abnormal operating condition type, and fault severity. The materials and media filling a device are known, and the normal operating range of the medium's physical properties is clearly defined (for example, for a certain material, the manufacturer will provide its relevant physical parameters and the normal fluctuation range, such as 100±5%). Therefore, if the deduced medium parameters exceed the normal parameters of the internal medium, then the set threshold has been exceeded.

[0034] A power equipment condition monitoring system includes: a measurement module for non-invasively acquiring physical quantity data and their normal derivatives of measurement points on the surface of the power equipment; a data processing and identification module including a power equipment medium identification method based on digital-analog driven method; and a computer-readable storage medium storing a computer program that, when executed by a processor, implements the power equipment medium identification method based on digital-analog driven method.

[0035] Specifically, the target equipment and parameter properties are determined, and the geometric field is constructed. After identifying the electrical equipment to be tested, determine the area to be measured based on the testing target, i.e., construct the geometric field, such as the cross-section of the equipment. On the boundary of the geometric field, arrange measurement points at equal or equal angular intervals. This means that different arrangement methods should be used on the boundaries of different geometric objects, but the requirement is the same: the boundary must be uniformly covered. For example, for a circle, the spacing cannot be equal because the boundary is not a straight line; therefore, the measurement points should be distributed at equal angles (radial angles).

[0036] Acquire relevant data about the device surface and perform fitting processing. Electrical information about a device's operation can be obtained through sensors or other measurement technologies. For example, current data can be measured using an ammeter, and temperature data can be measured using a thermocouple. Simultaneous measurement of the physical quantity at all measurement points is possible. and its derivative in the normal direction ,Right now:

[0037] (1) In the formula This represents the boundary of the geometric field. The data is grouped into... ,in For the number of measurement points, This step involves measuring the number of times the data will be measured. This non-invasive measurement method obtains input data for subsequent steps, acquiring system operational data while avoiding interference with the system.

[0038] To construct a data-analog fusion-driven method, the first step is to define the physical equations within the field based on the measurement target, and then standardize these equations, expressing them as follows: Form. Among them... Let the input point be the position vector. For the distribution of media within the field, The distribution of physical quantities within the field. This is the source term. At this point... This is an unknown quantity, and also a parameter to be determined in this invention.

[0039] For the standardized equation form, the equation can be decomposed and transformed into an equation containing unknown source terms using the equivalent source method, as follows: (2) In the formula, The unknown source term in the equation represents the energy distribution within the field. Using the equivalent source method, all unknown parameters in the equation are represented using equivalent sources.

[0040] Using the boundary element method, we can transform the field physical equations into boundary equations, and obtain: (3) In the formula, As the boundary of the field, For a complete field. Represents the source point on the boundary. Representing field points on the boundary, Represents the field points within the field area. This represents the basic solution form of the physical quantity to be measured. The fundamental solution form of its directional derivative is given by, where This parameter specifies the distance between the source point and the field point. The properties of the field boundary determine this: if the boundary is smooth, then... If the boundary is not smooth, then .

[0041] From equation (3), it can be seen that the incomplete measurement data of the equipment surface in equation (1) and the physical model with unknown medium parameters in equation (2) are successfully correlated, achieving the purpose of data-model fusion driving. Under transient problems (such as considering time variables), multiple standardized equation forms can be obtained by using the separation of variables method, and solved one by one according to the above method.

[0042] Take the following two-dimensional transient equation as an example: The first step is to break down the solution according to the variables. Taking this equation as an example, if we want to solve the equation without regard to the time variable, we can break down the solution into parts that are related to the time variable and parts that are not: The second step is to substitute the solution back into the original equation, and after simplification, we get: Finally, extract the terms that do not change over time and rearrange them into the first term in equation (2) of this invention, namely: In simple terms: decompose the problem according to variable requirements - substitute back into the equation - extract the required terms. Set training points, construct a surface quadrature neural network, and define the core loss function. First, set the training points for the neural network: training points on the boundary, i.e., the measurement points in step 1. The training points within the region need to uniformly cover the field. Define... This represents the total number of training points.

[0043] Construct the following neural network: Input: Coordinates of each source point and field point within the field. Data on the boundary and Function: To invert the unknown energy distribution within the field, providing support for subsequent steps. Structure: Employs a fully connected multilayer perceptron (MLP) architecture with two hidden layers, each containing 64 neurons. The activation function is the sigmoid function. The output layer has the same number of neurons as the input layer, with each neuron representing the energy inversion of one input point. Output: Equivalent source distribution. Training algorithm: Adam, learning rate lr = 0.001. Reason: The Adam optimizer automatically adjusts the learning rate by calculating the cumulative gradient and cumulative squared gradient for each parameter, eliminating the need for manual setting. This adaptive adjustment allows the model to update accordingly based on the gradient of different parameters during training, thereby improving convergence speed and stability. During network design, it was found that a learning rate of 0.001 was chosen because it strikes a balance between convergence speed and training stability, achieving fast convergence without oscillations or divergence in the results.

[0044] Iteration Termination Criteria: The total number of iterations (called epochs in neural networks) is 1000. Iteration terminates after 1000 epochs. The rationale for choosing 1000 epochs: This MLP size can achieve millisecond-level response times when trained on a GPU (such as an RTX 3060; training speed will be even faster if you choose a better GPU), with one epoch controlled to approximately 12-18ms. Testing shows that with 1000 epochs, the prediction accuracy guarantees a relative L2 error within 0.6%-1.8%. Therefore, in summary, this number of epochs ensures both accuracy and fast prediction response.

[0045] Of course, the number of training points also affects the training time. During the network design process, we conducted multiple tests, and the number of training points M should not exceed 1000. In fact, 1000 training points can fully cover typical power equipment of various sizes, such as GIS, smoothing reactors, and converter valves.

[0046] The core loss function is defined as follows using the surface quadrature equation (i.e., equation (3)): (4) Each training session will involve continuously adjusting the source terms. The value of is used to reduce the loss function. This step aims to make the values ​​calculated on the left and right sides of the quadrature equation in equation (3) equal, so as to satisfy the laws of the physical model when the data are input into the neural network.

[0047] The relationship between the equivalent source terms and the medium distribution is established, reflecting the physical conditions within the equipment. The distribution of equivalent sources within the equipment can be obtained through the preceding steps. The relationship between the equivalent source terms and the medium distribution can be established through equation (2): At this point, the unknowns in the equation only include the unknown medium distribution. Solving this equation will suffice. The distribution of the medium can be equated to the physical conditions occurring inside the equipment. For example, when an electrical fault occurs inside the equipment, the potential in the fault area increases, and electrical energy transmission is obstructed. In this case, the physical condition of the area can be equated to a decrease in conductivity. Other operating conditions such as thermal, magnetic, and mechanical conditions can also be analyzed using this equivalent method.

[0048] To verify that the present invention has higher accuracy compared with traditional detection methods, an experiment was conducted using a reactor as a typical power device to detect its internal thermal conditions. Multiple methods were employed and comparative experiments were performed as follows: Method A: Traditional detection methods rely on subjective experience to deduce the internal condition of the equipment.

[0049] Method B: Existing technical methods, using finite element multiphysics analysis to deduce the internal conditions of the equipment.

[0050] Method C: The present invention proposes a method that uses digital-analog fusion driving and SINN method to deduce the internal situation of the device.

[0051] After conducting 50 experiments, the experimental results are as follows: The error in identifying abnormal parameters is defined as the relative error between the measured value and the true parameter. The accuracy of identifying abnormal regions is defined as the percentage of the measured region within the true abnormal region. Real-time identification is determined by whether an identification result can be obtained within 0.5 minutes.

[0052] Experimental results show that, compared with existing methods, the proposed method reduces the error in anomaly parameter identification, improves the accuracy in anomaly region identification, and is real-time, outperforming existing methods in all aspects. This is because SINNs train rapidly and have strong nonlinear fitting capabilities, allowing for effective fusion of sparse measured data with incomplete models containing unknown parameters. This eliminates reliance on pure data or pure models while improving identification accuracy.

[0053] The above inventions are merely a few specific embodiments of the present invention. However, the embodiments of the present invention are not limited thereto, and any variations that can be conceived by those skilled in the art should fall within the protection scope of the present invention.

Claims

1. A method for identifying the medium of power equipment based on analog-digital driving, characterized in that, Includes the following steps: Construct the geometric field of the target power equipment, and arrange measuring points on the boundary of the geometric field; The physical quantity data and normal derivative data related to the medium to be identified at all measurement points are obtained through non-invasive measurement, and the boundary data are fitted. Based on the control equations describing the physical processes within the geometric field, a numerical-model fusion driving equation is constructed. The construction process includes: standardizing the control equations, using the equivalent source method to transform the unknown medium distribution parameters into unknown equivalent source terms, and applying the boundary element method to transform the control equations into boundary integral equations; the boundary integral equations correlate the boundary measurement data with the equivalent source terms inside the field. A surface quadrature neural network is constructed and trained. The input of the surface quadrature neural network includes the coordinates of the field points and the obtained boundary data, and the output is the equivalent source term distribution inside the field. During training, a core loss function is constructed based on the boundary integral equation, and the neural network parameters are adjusted by optimizing the loss function so that the output of the neural network satisfies the boundary integral equation. Based on the trained surface quadrature neural network, forward computation is performed on the input real-time measurement data to output the equivalent source term distribution at the current moment; according to the relationship between the equivalent source term and the medium distribution in the control equation, the medium distribution inside the geometric field is obtained by solving. Based on the comparison between the medium distribution and the preset threshold, the internal state of the device is determined and a warning message is output.

2. The power equipment medium identification method based on digital-analog driving as described in claim 1, characterized in that, The fitting of the boundary data includes: Based on the geometric shape characteristics of the boundary of the geometric field, the boundary parameterization method is determined; Based on the parameterization method, select an interpolation or fitting algorithm that matches the geometric shape feature; Based on the parameterization method and the selected algorithm, discrete measurement point data are reconstructed into continuous boundary functions.

3. The power equipment medium identification method based on digital-analog driving as described in claim 1, characterized in that, The physical quantity includes at least one of temperature, electric potential, magnetic induction intensity, or sound pressure.

4. The power equipment medium identification method based on digital-analog driving as described in claim 1, characterized in that, The boundary integral equation obtained by applying the boundary element method is: In the formula, As the boundary of the field, For a complete field. Represents the source point on the boundary. Representing field points on the boundary, Represents the field points within the field area. This represents the basic solution form of the physical quantity to be measured. The fundamental solution form of its directional derivative is given by, where To input the distance between the source point and the field point, the parameters are... The properties of the field boundary determine this: if the boundary is smooth, then... If the boundary is not smooth, then .

5. The power equipment medium identification method based on digital-analog driving as described in claim 1, characterized in that, The surface quadrature neural network is a multilayer perceptron containing at least one hidden layer; the core loss function is constructed based on the residuals of the boundary integral equation and is expressed as: 。 6. The power equipment medium identification method based on digital-analog driving as described in claim 1, characterized in that, Based on the equation relating the equivalent source term to the medium distribution parameters: Solving for the medium distribution ; in Let the input point be the position vector. For the distribution of media within the field, The distribution of physical quantities within the field. For source terms, is the unknown source term of the equation.

7. The power equipment medium identification method based on digital-analog driving as described in claim 5, characterized in that, The dielectric distribution parameters include at least one of electrical conductivity, thermal conductivity, magnetic permeability, or acoustic impedance.

8. The power equipment medium identification method based on digital-analog driving as described in claim 1, characterized in that, The obtained medium distribution parameters are compared with the reference parameters under normal operating conditions. If the parameter difference in a specific region exceeds the set threshold, the region is determined to be an abnormal region, and the abnormal operating condition type is identified.

9. A power equipment condition monitoring system, characterized in that, include: The measurement module is used to non-invasively acquire physical quantity data and their normal derivatives from measurement points on the surface of power equipment; The data processing and identification module includes the power equipment medium identification method based on digital-analog driving according to any one of claims 1 to 8; And a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the digital-analog driven power equipment media identification method as described in any one of claims 1 to 8.

Images