Method, system and equipment for calculating and evaluating space radiation field effect of long and straight control cable and medium

Abstract

The invention relates to a long straight control cable space radiation field effect calculation and evaluation method, system and device and a medium. The method comprises the following steps: constructing a partial differential equation constraint model based on a cable geometric structure and electromagnetic transient characteristics; fusing an actually measured magnetic field and an additional physical rule through a physical information neural network, and generating a complete electromagnetic data set to compensate for electric field data missing; a multi-dimensional interference factor is calculated by combining time-frequency distribution characteristics, equipment sensitive parameters and cable equivalent inductance, and the radiation field effect complexity is objectively quantified; establishing a probability mapping model of interference indexes and effect states by adopting multi-dimensional Gaussian process regression to realize risk probability distribution prediction; and finally, generating a visual probability curved surface of an electric field-magnetic field plane, and identifying a high-risk area to guide electromagnetic protection optimization. According to the method, the evaluation precision is improved in a data limited scene, and a technical closed loop from multi-dimensional feature fusion to probabilistic risk early warning is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of electromagnetic transient and calculation technology of power system, in particular to a long straight control cable spatial radiation field effect calculation and evaluation method, system, device and medium. BACKGROUND

[0002] In the field of key infrastructure such as power system and rail transit, long straight control cable as the core carrier of signal transmission, its spatial radiation field effect evaluation is directly related to the electromagnetic compatibility of equipment and the safety of system operation. This kind of evaluation method aims to quantify the interference risk of cable radiation electromagnetic field to adjacent sensitive equipment, and provides a theoretical basis for cable layout optimization and shielding design by establishing the mapping relationship between electromagnetic field distribution and equipment interference state. With the increasing integration of power electronic equipment, the coupling effect between cable radiation field and disturbed equipment is becoming more and more complex, especially in high-density cable channel or strong electromagnetic environment area, the traditional evaluation method is difficult to accurately capture the time-frequency characteristics and risk probability distribution of transient electromagnetic interference.

[0003] The existing technology mainly has three defects: first, in the scene where the necessary data measurement of electromagnetic interference evaluation is limited (such as electric field data is difficult to obtain directly), the existing model lacks data integrity guarantee mechanism, which leads to significant amplification of the calculation deviation of radiation field effect; second, the traditional complexity index only depends on single time domain or frequency domain feature, and cannot fuse multi-dimensional elements such as cable structure parameters, equipment sensitive characteristics and time-frequency energy distribution, so it is difficult to objectively represent the actual interference strength; finally, the existing method cannot establish a continuous mapping from interference index to risk probability, which makes the risk evaluation result lack of probabilistic support and visual warning ability, and it is difficult to guide accurate electromagnetic protection decision. SUMMARY

[0004] Based on this, the purpose of the present application is to provide a long straight control cable spatial radiation field effect calculation and evaluation method, system, device and medium which can guarantee the evaluation accuracy under the condition of data loss, fuse multi-dimensional interference characteristics and realize probabilistic risk visualization.

[0005] The purpose of the present application is realized by the following scheme:

[0006] In a first aspect, the present application provides a long straight control cable spatial radiation field effect calculation and evaluation method, comprising the following steps:

[0007] S1: Obtain the physical structure parameters of the long straight control cable, based on the geometric size of the cable and the preset current characteristics, call Maxwell equation to derive electromagnetic partial differential equation and define the initial conditions and boundary conditions of the spatial position of the cable, generate the electromagnetic transient model and electromagnetic calculation constraint conditions of the long straight control cable;

[0008] S2: Based on the electromagnetic transient model and the electromagnetic calculation constraint condition, the obtained field measurable magnetic field data is trained by a physical information neural network, the network is trained by combining the measurable magnetic field data and an additional physical information loss function constructed according to the physical rules, the network is used to solve the missing electric field data, and a complete electromagnetic data set containing the predicted electric field and the measured magnetic field is generated;

[0009] S3: Based on the complete electromagnetic data set, the predicted electric field and the measured magnetic field are processed by time-frequency transformation, the time-frequency distribution characteristics are extracted, and the interference factor is calculated by combining the characteristic frequency of the disturbed equipment and the equivalent inductance parameter of the cable, and the radiation field effect index representing the electromagnetic interference complexity is generated;

[0010] S4: The effect state of the field observation record is obtained, the radiation field effect index is used as the input feature, and the effect state is used as the output label, a nonlinear mapping relationship between the radiation field effect index and the effect state is established through a multi-dimensional Gaussian process regression, the trained Gaussian process regression model is used to predict the effect state probability distribution corresponding to any radiation field effect index, and a risk probability distribution model is generated;

[0011] S5: The risk probability distribution model is subjected to engineering risk assessment, the probability distribution of different effect states is calculated by inputting the radiation field effect index to be evaluated, and a visual risk probability distribution surface is generated.

[0012] In one of the embodiments, the long straight control cable space radiation field effect calculation and evaluation method provided by the application specifically comprises the following steps:

[0013] S11: Obtain the physical structure parameters of the long straight control cable, perform electromagnetic transient modeling processing based on the actual layout under the connection of two metal sections in the physical structure parameters, construct an electromagnetic coupling model of the finite length non-shielded cable according to the cable profile structure and the geometric characteristics along the z-axis, and combine the conduction characteristics of the z-direction only current density, and generate an electromagnetic transient model;

[0014] S12: Perform electromagnetic equation derivation on the electromagnetic transient model based on Maxwell's equations and continuity equations, introduce the uniform medium assumption and the constraint condition that the charge density is zero, combine the characteristics that the first derivative of the magnetic field intensity at the boundary is zero, derive the partial differential relationship between the space electric field and the magnetic field intensity, and generate the electromagnetic partial differential equation;

[0015] S13: Perform solving condition definition processing on the electromagnetic partial differential equation, set the electric field intensity zero value boundary condition according to the cutoff characteristics of the cable end current and potential in the physical structure parameters, and obtain the field intensity data at the initial time based on the physical structure parameters;

[0016] S14: Perform initial field distribution calculation on electric field intensity zero value boundary condition and field strength data, determine initial electric field and magnetic field intensity distribution through mathematical modeling of radiation principle of finite length straight conductor, and generate electromagnetic calculation constraint condition in combination with charge density continuity equation.

[0017] In one of the embodiments, the S2 of the long straight control cable spatial radiation field effect calculation evaluation method provided by the application specifically comprises the following steps:

[0018] S21: Perform neural network architecture construction processing on the electromagnetic transient model and the electromagnetic calculation constraint condition, construct a multi-objective loss function based on residual calculation of partial differential equation, initial condition matching error, boundary condition matching error and measurable magnetic field data deviation, and generate a physical information driven neural network optimization target by fusing additional physical information rules;

[0019] S22: Perform parameter iterative training processing on the neural network optimization target, calculate the gradient of the loss function to the network weight through automatic differentiation technology and update by back propagation until the total loss value converges to a preset convergence threshold, and generate a trained field quantity prediction model;

[0020] S23: Perform electric field calculation processing on the field quantity prediction model and the field measurable magnetic field data, input the magnetic field time sequence to calculate the electric field intensity distribution of the spatial target point through forward propagation, and generate a complete electromagnetic data set containing the predicted electric field and the measured magnetic field.

[0021] In one of the embodiments, the multi-objective loss function of the long straight control cable spatial radiation field effect calculation evaluation method provided by the application includes an initial condition loss function, a boundary condition loss function and a partial differential equation condition loss function, and the expression of the multi-objective loss function is:

[0022]

[0023]

[0024]

[0025] wherein, represents the boundary condition loss function, is the number of sampling points under the boundary condition, , represents the number of sampling points under the initial condition, are initial input magnetic field and electric field data respectively, are initial magnetic field data and electric field data respectively, is the nabla operator, represents the initial condition loss function, represents the partial differential equation condition loss function, to calculate the number of sampling points in the domain.

[0026] In one embodiment, the application provides a long straight control cable space radiation field effect calculation evaluation method S3 specifically comprising the following steps:

[0027] S31: Perform time-frequency joint analysis and processing on the predicted electric field and the measured magnetic field in the complete electromagnetic data set, call the short-time Fourier transform to calculate the energy distribution intensity of the signal in the time-frequency dimension, and generate a two-dimensional time-frequency spectrum matrix of the signal;

[0028] S32: Perform dominant frequency extraction processing on the two-dimensional time-frequency spectrum matrix, locate the frequency value corresponding to the maximum energy intensity in the time-frequency spectrum matrix, and generate key frequency characteristic parameters in combination with the equivalent inductance parameters of the cable and the preset characteristic frequency of the disturbed device;

[0029] S33: Perform interference factor synthesis processing on the key frequency characteristic parameters, calculate the radiation field effect index according to the relative deviation relationship between the dominant frequency and the device characteristic frequency through a preset objective complexity formula, and the expression of the objective complexity formula is:

[0030]

[0031]

[0032]

[0033] wherein, is the objective complexity index of the signal, i.e., the radiation field effect index, is the time-frequency energy distribution of the signal, is the average value of the time-frequency energy, is the frequency point corresponding to the maximum time-frequency distribution intensity of the disturbed signal, is the characteristic frequency of the disturbed sensitive device, M is the sequence number of the time direction in the time-frequency spectrum of the signal, N is the sequence number of the frequency direction in the time-frequency spectrum of the signal, A is the average weight coefficient, is the energy distribution value of the i th time sequence and the j th frequency sequence in the time-frequency spectrum matrix, and B is the starting energy threshold value, is the equivalent inductance of the cable, and c is the speed of light.

[0034] In one embodiment, the application provides a long straight control cable space radiation field effect calculation evaluation method S4 specifically comprising the following steps:

[0035] S41: Perform sample alignment processing on the acquired field observation record effect state and the corresponding radiation field effect indicator, take the complexity indicator EF value as the input feature vector, and the observed normal or interference state as the discrete output label, and generate a labeled model training sample set;

[0036] S42: Perform multi-dimensional Gaussian process regression modeling processing on the model training sample set, configure a square exponential covariance function and optimize hyperparameters and length scale matrix through maximum likelihood estimation, establish a probability mapping relationship from complexity indicator to effect state, and generate an initial probability mapping model;

[0037] S43: Perform posterior distribution derivation processing on the initial probability mapping model, update the conditional probability distribution through Bayesian inference, match the posterior probability distribution predicted by the model with the observation state of the training sample in KL divergence, and generate a risk probability distribution model.

[0038] In one embodiment, the long straight control cable space radiation field effect calculation evaluation method provided by the present application specifically comprises the following steps:

[0039] S51: Perform scene application processing on the risk probability distribution model, input the calculated radiation field effect indicators under the working conditions of the cable to be evaluated, derive the possible state distribution through Gaussian process regression, and generate an initial probability estimate under the working conditions of the cable to be evaluated;

[0040] S52: Perform state probability calculation processing on the initial probability estimate, compare the probability density function values corresponding to different effect states, calculate the specific probability values belonging to each state in proportion, and generate a quantitative effect state probability distribution;

[0041] S53: Perform visual rendering processing on the quantitative effect state probability distribution, draw probability contour lines or surface graphs of different states on the two-dimensional plane of electric field-magnetic field complexity, and generate a visual probability distribution surface for risk warning.

[0042] In a second aspect, the present application provides a long straight control cable space radiation field effect calculation evaluation system, which is configured with the following modules:

[0043] An electromagnetic transient model construction module is configured to acquire physical structure parameters of a long straight control cable, based on cable geometric dimensions and preset current characteristics, call Maxwell's equation to derive electromagnetic partial differential equations and define initial conditions and boundary conditions of cable space positions, and generate an electromagnetic transient model of the long straight control cable and electromagnetic calculation constraint conditions;

[0044] The physical information neural network training module is configured to train the acquired field measurable magnetic field data based on an electromagnetic transient model and electromagnetic calculation constraint conditions, train the network by combining the measurable magnetic field data and an additional physical information loss function constructed according to physical rules, solve the missing electric field data by using the network, and generate a complete electromagnetic data set containing predicted electric fields and measured magnetic fields.

[0045] The time-frequency feature extraction and interference factor calculation module is configured to perform time-frequency transformation processing on the predicted electric fields and the measured magnetic fields based on the complete electromagnetic data set, extract time-frequency distribution features, and calculate interference factors by combining characteristic frequencies of the disturbed equipment and cable equivalent inductance parameters, and generate radiation field effect indexes representing electromagnetic interference complexity.

[0046] The Gaussian process regression model training module is configured to obtain an effect state of a field observation record, take the radiation field effect indexes as input features and the effect state as an output label, establish a nonlinear mapping relationship between the radiation field effect indexes and the effect state by using a multidimensional Gaussian process regression, predict an effect state probability distribution corresponding to any radiation field effect index by using the trained Gaussian process regression model, and generate a risk probability distribution model.

[0047] The engineering risk assessment and visualization module is configured to perform engineering risk assessment on the risk probability distribution model, calculate probability distributions of different effect states by inputting the radiation field effect indexes to be evaluated, and generate a visual risk probability distribution surface.

[0048] In a third aspect, the present application provides a computer device, comprising a memory and a processor, the memory stores a computer program, and the processor implements any of the above long straight control cable space radiation field effect calculation and evaluation methods when executing the computer program.

[0049] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement any of the above long straight control cable space radiation field effect calculation and evaluation methods.

[0050] In summary, the long straight control cable space radiation field effect calculation evaluation method provided by the application can systematically solve the three core problems of insufficient evaluation accuracy, one-sided interference feature representation and weak risk decision support in the data missing scenario of the existing method by constructing a technology closed loop of physical rule driving and data intelligent fusion. Specifically, based on the cooperative mechanism of the electromagnetic transient model and the physical information neural network, the dynamic compensation and integrity reconstruction of the electric field data can be realized, and the problem of missing key electromagnetic parameters caused by limited measurement conditions can be effectively overcome; by fusing the interference factor calculation model of the time-frequency distribution feature, the equipment sensitive characteristic and the cable structure parameter, the limitation of the traditional single time domain or frequency domain index can be broken through, and multi-dimensional objective quantization of the radiation field effect complexity can be realized; with the help of the probability mapping mechanism of the multi-dimensional Gaussian process regression, a continuous nonlinear relationship from the interference index to the risk state can be established, and the accurate prediction of the effect occurrence probability can be realized; and finally, the generated electric field-magnetic field two-dimensional probability distribution surface can realize the gradientized visual identification of the high-risk area, so as to guide the accurate electromagnetic protection strategy formulation.

[0051] For better understanding and implementation, the application will be described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 A flowchart of a long straight control cable space radiation field effect calculation evaluation method provided by an embodiment of the application is shown in the figure.

[0053] Figure 2 A flowchart of generating a radiation field effect index representing electromagnetic interference complexity provided by the application is shown in the figure.

[0054] Figure 3 A structural diagram of a long straight control cable space radiation field effect calculation evaluation device provided by another embodiment of the application is shown in the figure. DETAILED DESCRIPTION

[0055] In order to facilitate the understanding of the application, the application will be described in more detail below with reference to the related drawings. The preferred embodiments of the application are shown in the drawings. However, the application can be implemented in many different forms, and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the application more thorough and comprehensive.

[0056] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the application belongs. The terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the application. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0057] In one embodiment, as shown in Figure 1 A method for evaluating the spatial radiation field effect of long straight control cable is provided. The method can be applied to a terminal, a server, or a system including a terminal and a server, and can be implemented through the interaction of the terminal and the server. The method includes the following steps:

[0058] S1: Obtain the physical structure parameters of the long straight control cable. Based on the cable geometric size and the preset current characteristics, call Maxwell equations to derive electromagnetic partial differential equations and define the initial conditions and boundary conditions of the cable spatial position, generate the electromagnetic transient model and electromagnetic calculation constraint conditions of the long straight control cable.

[0059] Specifically, the system obtains the physical structure parameters of the long straight control cable, which include the core material, the insulating layer material, the laying length, the equivalent inductance, etc. Based on the cable geometric size and the preset current characteristics, the system calls Maxwell equations to expand and derive. During the derivation process, the system sets assumptions in combination with the laying characteristics of the long straight cable, i.e., the cable is laid along a fixed coordinate axis, the structure is uniform and has no shielding layer, the magnetic field is only distributed along the cable loop, the electric field and the magnetic field have dominant directions and are only related to a single spatial coordinate, and there is no free charge accumulation in the cable laying area.

[0060] Further, the system simplifies Maxwell equations based on the above assumptions and further derives in combination with the continuity equation to obtain partial differential equations describing the electromagnetic transient process of the cable. Then the system defines the initial conditions of the cable spatial position, which are determined based on the quasi-steady field theory, i.e., at the initial moment, the electric field strength and the magnetic field strength are related through the conductive characteristics of the cable medium when there is no external excitation. The system also defines the boundary conditions, which are set around the spatial solution domain formed by the cable laying. The solution domain covers the cable transverse, longitudinal and time dimension ranges, and the derivatives of the electric field strength and the magnetic field strength at the boundary comply with the ideal conductor boundary characteristics. The system integrates the partial differential equations, the initial conditions and the boundary conditions to form the electromagnetic transient model and the electromagnetic calculation constraint conditions of the long straight control cable.

[0061] S2: Based on the electromagnetic transient model and the electromagnetic calculation constraint conditions, train the obtained field measurable magnetic field data with physical information neural network, train the network with the measurable magnetic field data and the additional physical information loss function constructed according to the physical rules, solve the missing electric field data with the network, and generate a complete electromagnetic data set including the predicted electric field and the measured magnetic field.

[0062] Specifically, the system invokes the electromagnetic transient model and electromagnetic calculation constraints as prior physical information, embedding them into the training process of the physical information neural network. Further, the system acquires measurable magnetic field data from the field and preprocesses this data, including removing power frequency interference and smoothing noise to reduce the impact of data noise on training. The system divides the preprocessed measurable magnetic field data into training and validation sets. The training set is used for neural network parameter optimization, while the validation set is used to monitor the network's generalization ability. The system constructs an additional physical information loss function, which sets constraints based on the electric field tolerance characteristics of the cable insulation layer to prevent the network output from exceeding the physically reasonable range.

[0063] Furthermore, the system inputs the measurable magnetic field data and electromagnetic calculation constraints into the neural network, and initiates training by combining an additional physical information loss function. During training, an optimizer dynamically adjusts the learning rate until the network loss meets the preset convergence threshold and the validation set loss shows no upward trend. After training, the system inputs all field-measurable magnetic field data into the neural network to solve for the missing electric field data. The system extracts the amplitude parameters of each set of magnetic field data and the corresponding electric field data, and combines them with the original time history data to generate a complete electromagnetic dataset containing both the predicted electric field and the measured magnetic field.

[0064] S3: Based on a complete electromagnetic dataset, the predicted electric field and the measured magnetic field are processed by time-frequency transformation to extract time-frequency distribution features. The interference factor is calculated by combining the characteristic frequency of the disturbed equipment and the equivalent inductance parameter of the cable, and a radiation field effect index characterizing the complexity of electromagnetic interference is generated.

[0065] Specifically, the system calls upon a complete electromagnetic dataset to extract predicted electric field data and measured magnetic field data. The system performs time-frequency transformation on both the predicted and measured magnetic field data, employing a short-time Fourier transform (SFT) method. During the transformation, the window function, window length, and overlap rate are configured to balance time-frequency resolution. The system obtains the time-frequency distribution matrix for each data set through the time-frequency transformation, and then extracts the time-frequency distribution features. These features are the frequencies corresponding to the maximum intensity in the matrix, representing the frequency range where electromagnetic interference energy is concentrated. The system acquires the characteristic frequency of the disturbed device and the equivalent inductance parameters of the cable. The characteristic frequency of the disturbed device is calculated based on the characteristics of the cable's equivalent half-wave antenna, with the calculation process relating the cable's equivalent inductance to the speed of light. The system calculates the interference factor using the frequency corresponding to the time-frequency distribution features as the numerator and the characteristic frequency of the disturbed device as the denominator. The system defines the interference factor as a radiation field effect index characterizing the complexity of electromagnetic interference, reflecting the correlation between the frequency of concentrated electromagnetic interference energy and the sensitive frequency of the disturbed device.

[0066] S4: Obtain the effect state recorded by field observation, take the radiation field effect index as the input feature and the effect state as the output label, establish the nonlinear mapping relationship between the radiation field effect index and the effect state through multidimensional Gaussian process regression, use the trained Gaussian process regression model to predict the probability distribution of the effect state corresponding to any radiation field effect index, and generate a risk probability distribution model.

[0067] Specifically, the system acquires the effect states recorded in the field observations. These effect states are determined based on equipment operation logs and sampling data quality assessment results, and are categorized into normal and disturbance states. The system calls upon radiation field effect indicators, combining them with electric and magnetic field amplitudes extracted from the complete electromagnetic dataset to construct a three-dimensional input feature matrix. Each set of data in the input feature matrix corresponds to a set of effect states observed in the field. The system uses the effect states as output labels, combining them with the input feature matrix to form a sample set. This sample set is then proportionally divided into a training set and a test set.

[0068] Preferably, the system employs a multidimensional Gaussian process regression algorithm, configuring a squared exponential covariance function. The shape parameter and input length scale matrix in the covariance function are determined through optimization using a training set, with the optimization objective being to minimize the negative log-likelihood function. The system trains the multidimensional Gaussian process regression model using the training set and verifies the model's prediction accuracy using a test set. After training, the system inputs any radiation field effect index into the model, and the model outputs the probability distribution of the corresponding effect state. The probability distribution includes the probability of the normal state and the probability of the disturbance state, and the sum of the two probabilities satisfies the basic axioms of probability theory. The system integrates the model and the probability distribution calculation logic to generate a risk probability distribution model.

[0069] S5: Perform engineering risk assessment on the risk probability distribution model, input the radiation field effect index to be assessed to calculate the probability distribution of different effect states, and generate a visualized risk probability distribution surface.

[0070] Specifically, the system acquires the radiation field effect indicators to be evaluated, including electric field amplitude, magnetic field amplitude, and radiation field effect indicators. The system inputs these indicators into a risk probability distribution model, which calculates the probability distributions for different effect states. Based on engineering application requirements, the system sets risk level classification standards, categorizing risk levels into low, medium, and high risk based on the probability interval of the interference state. The system determines the risk level corresponding to each indicator based on the calculated probability distribution. The system then activates a visualization generation module, using electric field amplitude, magnetic field amplitude, and radiation field effect indicators as three-dimensional coordinate axes, and interference state probability as the color mapping basis to construct a visualized risk probability distribution surface. This visualized surface intuitively presents the risk distribution patterns under different combinations of electromagnetic field parameters. The system uses this visualized surface to identify high-risk parameter intervals, providing a quantitative reference for cable layout optimization and shielding design.

[0071] In summary, the spatial radiation field effect calculation and evaluation method for long straight control cables provided in this application systematically solves three core problems of existing methods: insufficient evaluation accuracy in scenarios with missing data, one-sided representation of interference characteristics, and weak support for risk decision-making by constructing a technical closed loop that integrates physical rule-driven and data intelligence. Specifically: based on the synergistic mechanism of electromagnetic transient model and physical information neural network, dynamic compensation and complete reconstruction of electric field data can be achieved, effectively overcoming the problem of missing key electromagnetic parameters due to limited measurement conditions; by integrating interference factor calculation model of time-frequency distribution characteristics, equipment sensitivity characteristics, and cable structural parameters, the limitations of traditional single time-domain or frequency-domain indicators can be overcome, realizing multi-dimensional objective quantification of the complexity of radiation field effect; by using the probability mapping mechanism of multi-dimensional Gaussian process regression, a continuous nonlinear relationship from interference indicators to risk state can be established, realizing accurate prediction of the probability of effect occurrence; finally, the generated two-dimensional probability distribution surface of electric field and magnetic field can realize gradient visual identification of high-risk areas to guide the formulation of precise electromagnetic protection strategies.

[0072] In one embodiment, S1 of the method for calculating and evaluating the spatial radiation field effect of a long straight control cable provided by the present invention specifically includes the following steps:

[0073] S11: Obtain the physical structure parameters of the long straight control cable, perform electromagnetic transient modeling based on the actual layout of the two metal connections in the physical structure parameters, and construct an electromagnetic coupling model of the finite-length unshielded cable by combining the cable cross-sectional structure and geometric characteristics extending along the z-axis with the conduction characteristics of the current density only in the z-direction component, and generate an electromagnetic transient model.

[0074] Specifically, the system acquires the physical structural parameters of the long straight control cable. These parameters include the connection method of the two metal sections, the cable cross-sectional structure, the finite length value, the material of the unshielded layer, and the geometric parameters extending along the z-axis. Based on the actual layout of the two metal sections, the system initiates the electromagnetic transient modeling process. The system analyzes the interlayer distribution characteristics of the cable cross-sectional structure, determines the spatial proportion of the inner core and insulation layer, and, combined with the linear geometric characteristics of the cable extending along the z-axis, defines the spatial dimension range of the model as x∈[0,L], y∈[0,W], z∈[0,H], where L is the finite length of the cable, W is the cross-sectional width, and H is the cross-sectional height. Based on the conduction characteristics of the current density containing only the z-axis component, the system sets the current density vector J=(0,0, To eliminate interference from the current components in the x and y directions and focus... Conduction behavior along the z-axis. The system integrates layout, geometry, and conduction characteristics to construct an electromagnetic coupling model of a finite-length unshielded cable. This model can represent the connection between two metal sections. By studying the variation law and radiation characteristics at the lower boundary with a finite length, an electromagnetic transient model that can support the derivation of subsequent equations is finally generated.

[0075] S12: Based on Maxwell's equations and the continuity equation, the electromagnetic equations of the electromagnetic transient model are derived. The assumption of a homogeneous medium and the constraint of zero charge density are introduced. Combined with the characteristic that the first derivative of the magnetic field strength is zero at the boundary, the partial differential relationship between the spatial electric field and the magnetic field strength is derived, and the electromagnetic partial differential equation is generated.

[0076] Specifically, the system calls the generated electromagnetic transient model and initiates the electromagnetic equation derivation process based on Maxwell's equations, with the following expression:

[0077]

[0078] in The magnetic field strength, For electric field strength, For current density, It is the electric displacement vector. It represents the magnetic flux density. Let be the charge density. The system also introduces a continuity equation. Furthermore, the derivation introduces the assumption of a homogeneous medium, setting:

[0079]

[0080]

[0081] in, The vacuum permittivity, Relative permittivity of the medium The permeability of free space, Let be the relative permeability of the medium; and simultaneously introduce the constraint that the charge density is zero. At this point, the continuity equation simplifies to The system takes into account the characteristic that the first derivative of the magnetic field strength is zero at the model boundary. , where n is the boundary normal vector, clearly defines the variation of the magnetic field in the boundary region. The system is based on the current density vector. and The derivation yields Then, by simultaneously applying Maxwell's equations, the partial differential relationships between the electric and magnetic field strengths in space are gradually established, ultimately generating the electromagnetic partial differential equations describing the electromagnetic transient process of the cable. The expression for these electromagnetic partial differential equations is as follows:

[0082]

[0083] S13: Define the solution conditions for the electromagnetic partial differential equations. Based on the cutoff characteristics of the cable end current and potential in the physical structure parameters, set the boundary condition of zero electric field strength, and obtain the field strength data at the initial moment based on the physical structure parameters.

[0084] Specifically, the system analyzes the current and potential characteristics at the cable end in the physical structural parameters based on the electromagnetic partial differential equations. It finds that current conduction at the cable end exhibits a cutoff characteristic, meaning that at the end... , combined It can be known And then based on The first derivative of the magnetic field strength at the terminal boundary is determined to be zero, i.e. Based on the termination characteristics of the potential distribution at the cable end, the system determines the value pattern of the electric field intensity at the cable end boundary and sets a zero-value boundary condition for the electric field intensity. and Place This ensures that the electric field at the boundary conforms to actual physical phenomena. Subsequently, based on the acquired physical structure parameters, the system extracts information related to the initial electric field strength, such as the initial excitation state of the cable and the initial electromagnetic properties of the medium. Through processing and analysis of this information, the initial electric field strength is obtained. Electric field strength data at time With magnetic field strength data .

[0085] S14: Calculate the initial field distribution based on the zero-value boundary conditions of the electric field strength and the field strength data. Determine the initial electric and magnetic field strength distributions through mathematical modeling based on the radiation principle of a finite-length straight conductor, and generate electromagnetic calculation constraints by combining the charge density continuity equation.

[0086] Specifically, the system invokes the determined boundary conditions of zero electric field strength and the initial field strength data to initiate the initial field distribution calculation process. The system uses the radiation principle of a finite-length straight conductor for mathematical modeling. This model quantifies the radiation field distribution of the finite-length conductor in the initial state, thereby determining the initial electric field strength distribution. With magnetic field strength distribution After obtaining the initial field distribution, the system combines the charge density continuity equation with the curl equation in Maxwell's equations, namely:

[0087]

[0088]

[0089] in, The current density at the initial moment, The system verifies whether the change in charge density in the initial field distribution satisfies the continuity requirement by simultaneously solving the above equations, and adjusts the initial field distribution parameters accordingly. and This ensures that it conforms to the fundamental laws of electromagnetism. The system integrates and verifies the initial field distribution, the zero-value boundary condition of the electric field intensity, and the continuity constraint of the charge density, ultimately generating electromagnetic calculation constraints that can be used for subsequent neural network training and electromagnetic calculations.

[0090] In one embodiment, S2 of the method for calculating and evaluating the spatial radiation field effect of a long straight control cable provided by the present invention specifically includes the following steps:

[0091] S21: The electromagnetic transient model and electromagnetic calculation constraints are processed by constructing a neural network architecture. Based on the residual calculation of partial differential equations, initial condition matching error, boundary condition matching error and measurable magnetic field data deviation, additional physical information rules are integrated to construct a multi-objective loss function and generate a physical information-driven neural network optimization objective.

[0092] Specifically, the system constructs a neural network architecture to process the electromagnetic transient model and electromagnetic computational constraints. First, the input layer, hidden layer, and output layer structure are determined. The input layer corresponds to the time and space coordinate parameters in the electromagnetic transient model, and the output layer corresponds to the predicted values ​​of electric and magnetic field strengths. The number of hidden layers and neurons is determined based on the data dimensions in the electromagnetic computational constraints. The system constructs a multi-objective loss function based on residual calculations of partial differential equations, initial condition matching errors, boundary condition matching errors, and deviations in measurable magnetic field data, incorporating additional physical information rules. The multi-objective loss function includes an initial condition loss function, a boundary condition loss function, and a partial differential equation condition loss function. The expression for the multi-objective loss function is:

[0093]

[0094]

[0095]

[0096] in, Represents the boundary condition loss function. The number of sampling points under boundary conditions. , This indicates the number of sampling points under the initial conditions. These are the initial input magnetic field and electric field data, respectively. These are the initial magnetic field data and electric field data, respectively. For the nabla operator, Represents the initial condition loss function. This represents the conditional loss function of a partial differential equation. This represents the number of sampling points in the computational domain. The system integrates three loss functions to generate a physically-driven neural network optimization objective.

[0097] In one embodiment, the loss function Loss of the physical information neural network can also be derived from the initial conditional loss. Boundary condition loss Residual loss of partial differential equations and losses under special conditions The loss of additional physical information (i.e., the loss of physical information) consists of four parts, and its expression is as follows:

[0098]

[0099] The data for initial conditions, boundary conditions, and special conditions can be obtained through measurement, simulation, or calculation based on physical formulas. The introduction of this method effectively overcomes the common problem of divergence in physical information neural networks when the initial value is 0. The data-driven part of the loss function... , , and the physical model-driven part of the loss function The expression is as follows:

[0100]

[0101]

[0102]

[0103] in, Represents the initial condition loss function. Represents the boundary condition loss function. and represent the number of samples given for initial and boundary conditions training, respectively. and The training data is obtained based on the initial and boundary conditions, or it can be obtained through experimental measurement or simulation generation. Represents the boundary condition loss function. For training points of partial differential equations, the residuals of the equations can be obtained efficiently through automatic differentiation techniques. .

[0104] S22: Perform iterative parameter training on the neural network optimization objective, calculate the gradient of the loss function with respect to the network weights using automatic differentiation technology, and backpropagate to update it until the total loss value converges to the preset convergence threshold, generating a fully trained field prediction model.

[0105] Specifically, the system performs iterative parameter training on the neural network optimization objective. First, it initializes the network weight parameters, setting their initial values ​​according to a uniform distribution rule. The system then uses automatic differentiation to calculate the gradient of the loss function with respect to the network weights. This automatic differentiation technique, based on the chain rule, derives the loss function from the output layer back to the input layer, progressively calculating the contribution of each weight parameter to the total loss value, i.e., the gradient value. After obtaining the gradient, the system updates the network weights using a backpropagation algorithm. During the update process, the weight values ​​are adjusted according to the gradient direction to reduce the total loss value.

[0106] Furthermore, the system repeatedly performs the iterative process of gradient calculation and weight update, calculating the current total loss value after each iteration. The total loss value is... , and The system iterates until the total loss value converges to a preset convergence threshold. The convergence criterion is that the change in the total loss value across multiple consecutive iterations is less than the preset threshold range. When the convergence condition is met, the system stops iterating and generates a fully trained field prediction model. This model can output predicted electric and magnetic field values ​​that meet the electromagnetic calculation constraints based on the input parameters.

[0107] S23: Perform electric field estimation processing on the field quantity prediction model and the field measurable magnetic field data. Input the magnetic field time series and calculate the electric field intensity distribution of the spatial target point through forward propagation to generate a complete electromagnetic dataset containing the predicted electric field and the measured magnetic field.

[0108] Specifically, the system performs electric field estimation on the field quantity prediction model and the field-measurable magnetic field data. First, the field-measurable magnetic field data is preprocessed, including data format conversion and outlier removal, to ensure the magnetic field data meets the input requirements of the field quantity prediction model. The system inputs the preprocessed magnetic field time series into the field quantity prediction model. Through the model's forward propagation process, the electric field intensity distribution at spatial target points is calculated. During forward propagation, the magnetic field time series passes through the input layer and hidden layer sequentially. The hidden layer extracts and transforms features from the input data based on trained weight parameters. Finally, the output layer outputs the predicted electric field intensity values ​​for each target point at different times. After acquiring the predicted electric field data output by the field quantity prediction model, the system integrates it with the original field-measurable magnetic field data. The integration is based on the data acquisition timestamp and spatial coordinates, ensuring that each set of predicted electric field data corresponds to the measured magnetic field data under the same time and spatial conditions. Through this integration, the system generates a complete electromagnetic dataset containing both predicted and measured electric and magnetic fields. This dataset covers electric and magnetic field information at different times and spatial locations, providing data support for subsequent calculations of radiation field effect indicators.

[0109] In one embodiment, S3 of the method for calculating and evaluating the spatial radiation field effect of a long straight control cable provided by the present invention specifically includes the following steps:

[0110] S31: Perform time-frequency joint analysis on the predicted electric field and measured magnetic field in the complete electromagnetic dataset, call the short-time Fourier transform to calculate the energy distribution intensity of the signal in the time-frequency dimension, and generate the two-dimensional time-frequency matrix of the signal.

[0111] Specifically, the system performs joint time-frequency analysis on the predicted electric field and measured magnetic field in the complete electromagnetic dataset, covering electric and magnetic field strength data corresponding to all time series in the dataset. The system calls the Short-Time Fourier Transform (SFT) algorithm, which segments the time-domain signal through a sliding time window, performs a Fourier transform on the signal within each time window, mapping the time-domain signal to a two-dimensional time-frequency space, and calculates the energy distribution intensity at different frequencies for each time point. During the SFT, the length of the time window is determined according to the signal's temporal resolution requirements, ensuring that the details of signal changes in the time dimension and the distribution characteristics in the frequency dimension are captured. The system stores the energy distribution intensity at each time point and corresponding frequency in numerical form, arranging them in time series and frequency series order to generate a two-dimensional time-frequency spectrum matrix. The row dimension of this matrix corresponds to the time direction sequence, and the column dimension corresponds to the frequency direction sequence. The value of each element in the matrix represents the signal energy distribution intensity at the corresponding time-frequency point.

[0112] S32: Perform dominant frequency extraction processing on the two-dimensional time spectrum matrix, locate the frequency value corresponding to the maximum energy intensity in the time spectrum matrix, and generate key frequency characteristic parameters by combining the cable equivalent inductance parameters and the preset characteristic frequency of the disturbed equipment.

[0113] Specifically, the system extracts the dominant frequency from the two-dimensional time-spectrum matrix. A matrix traversal algorithm is used to read the energy distribution intensity values ​​of each element in the time-spectrum matrix one by one. The matrix position corresponding to the maximum energy intensity is determined by numerical comparison. The column index of this position corresponds to a specific frequency value in the frequency sequence; this frequency value is the dominant frequency of the signal. The system calls the cable equivalent inductance parameter from the cable physical structure parameter library. This parameter is an inherent parameter obtained based on the cable structure characteristics and electromagnetic simulation. Simultaneously, it reads the preset characteristic frequency of the disturbed equipment, which is a reference frequency pre-set according to the electromagnetic sensitivity characteristics of the disturbed equipment. The system integrates the extracted dominant frequency, cable equivalent inductance parameter, and the preset characteristic frequency of the disturbed equipment, clarifying the numerical correlation between the three. The dominant frequency reflects the energy concentration frequency of the interference signal; the cable equivalent inductance parameter is used as an objective value for subsequent calculation of the characteristic frequency of the disturbed equipment; and the preset characteristic frequency of the disturbed equipment serves as a reference benchmark for interference matching. These three together constitute key frequency characteristic parameters, providing data support for subsequent interference factor synthesis.

[0114] S33: Perform interference factor synthesis processing on key frequency characteristic parameters. Based on the relative deviation between the dominant frequency and the equipment characteristic frequency, calculate the radiation field effect index using a preset objective complexity formula. The expression for the objective complexity formula is:

[0115]

[0116]

[0117]

[0118] in, This is an objective indicator of signal complexity, specifically a radiation field effect indicator. This represents the time-frequency energy distribution of the signal. This represents the average value of the time-frequency energy. The frequency point corresponding to the maximum time-frequency distribution intensity of the disturbance signal. For objectively calculating the characteristic frequencies of the disturbance-sensitive device, M is the number of sequences in the time direction of the signal's time spectrum, N is the number of sequences in the frequency direction of the signal's time spectrum, and A is the average weighting coefficient. Let be the energy distribution value of the i-th time series and the j-th frequency series in the time-spectrum matrix, and B be the start-up energy threshold. Let be the equivalent inductance of the cable, and c be the speed of light.

[0119] In one embodiment, S4 of the method for calculating and evaluating the spatial radiation field effect of a long straight control cable provided by the present invention specifically includes the following steps:

[0120] S41: Perform sample alignment processing on the effect states and corresponding radiation field effect indices of the acquired field observation records, use the complexity index EF value as the input feature vector, and use the observed normal or disturbance states as discrete output labels to generate a labeled model training sample set.

[0121] Specifically, the system performs sample alignment processing on the acquired field observation records of effect states and corresponding radiation field effect indices. First, it reads the effect state data from the field observation records, which includes the operating status of the disturbed equipment at different time points, categorized into normal and disturbed states. Simultaneously, it reads the radiation field effect indices corresponding to the time points, i.e., the complexity index EF value. Each EF value corresponds one-to-one with the effect state at a specific time point. The system achieves sample alignment between effect states and EF values ​​through timestamp matching, eliminating invalid data with mismatched timestamps to ensure that each EF value has a unique corresponding effect state label. The system uses the complexity index EF value as the input feature vector. A single EF value constitutes a one-dimensional input feature vector, and the EF values ​​of multiple samples are arranged sequentially to form an input feature matrix. The observed normal or disturbed states are used as discrete output labels, labeled using numerical encoding methods, such as using a specific numerical value to represent the normal state and another specific numerical value to represent the disturbed state. The system integrates the input feature vector and discrete output labels, stores them according to a preset sample format, and generates a labeled model training sample set. This sample set contains the input feature matrix and the corresponding output label vector, providing a data foundation for subsequent modeling.

[0122] S42: Perform multidimensional Gaussian process regression modeling on the model training sample set, configure the squared exponential covariance function and optimize the hyperparameters and length scale matrix through maximum likelihood estimation, establish the probability mapping relationship from complexity index to effect state, and generate the initial probability mapping model.

[0123] Specifically, the system performs multidimensional Gaussian process regression modeling on the model training sample set. First, the basic framework of the multidimensional Gaussian process is defined, using the input feature vector (EF value) in the model training sample set as the input variable. The output label (effect state) is the output variable. Construct a Gaussian process:

[0124]

[0125] in, It is a mean function. This refers to the covariance function. The system is configured with a squared exponential covariance function, the expression of which is:

[0126]

[0127] in, This is the amplitude hyperparameter of the covariance function, used to adjust the overall scale of the covariance; and For any two samples in the input feature vector; It is a length-scale matrix, and it is a diagonal matrix with diagonal elements. This is used to adjust the degree to which input features (EF values) affect the output. The system optimizes hyperparameters through maximum likelihood estimation. With length scale matrix The goal is to maximize the log-likelihood function of the model training sample set to find the optimal values ​​of the hyperparameters. Through the above modeling and optimization, the system establishes a probabilistic mapping relationship between complexity indices and effect states. That is, given any EF value, it can output the probability of the corresponding normal or disturbance state, generating an initial probability mapping model.

[0128] S43: Perform posterior distribution derivation on the initial probability mapping model, update the conditional probability distribution through Bayesian inference, and perform KL divergence matching between the model-predicted posterior probability distribution and the observed state of the training samples to generate a risk probability distribution model.

[0129] Specifically, the system derives the posterior distribution of the initial probability mapping model. Based on the Bayesian inference framework, it combines the prior distribution of the initial probability mapping model with the likelihood function of the model training sample set to update and obtain the posterior probability distribution. The core formula of Bayesian inference is:

[0130]

[0131] in, Hyperparameters of the initial probability mapping model and , For model training sample set, Let be the prior distribution of the hyperparameters. Let be the likelihood function. Let be the posterior distribution of the hyperparameters. This is an evidence factor.

[0132] Furthermore, the system calculates the posterior distribution of the hyperparameters using this formula, and then updates the conditional probability distribution of the initial probability mapping model to obtain a probability prediction model based on the posterior distribution. The system performs KL divergence matching between the model's predicted posterior probability distribution and the observed states of the training samples. The formula for calculating KL divergence is:

[0133]

[0134] in, The probability distribution of the observed states of the training samples. This represents the posterior probability distribution predicted by the model. The system adjusts the model parameters by minimizing the KL divergence to minimize the difference between the model's predicted distribution and the observed distribution, ultimately generating a risk probability distribution model that can output the probabilities of normal and disturbed states corresponding to any EF value.

[0135] In one embodiment, S5 of the method for calculating and evaluating the spatial radiation field effect of a long straight control cable provided by the present invention specifically includes the following steps:

[0136] S51: Apply scenario processing to the risk probability distribution model, input the radiation field effect index calculated under the cable condition to be evaluated, and deduce its possible state distribution through Gaussian process regression to generate the initial probability estimate under the cable condition to be evaluated.

[0137] Specifically, the system performs scenario application processing on the risk probability distribution model. First, it obtains the radiation field effect index under the cable's operating condition to be evaluated. This index is calculated using an objective complexity formula, specifically the EF value under the evaluated condition. Simultaneously, it associates the electric field strength characteristic values ​​and magnetic field strength characteristic values ​​corresponding to the cable under evaluation, collectively forming the model's input feature vector. The system imports the input feature vector into the risk probability distribution model and calls the model's built-in multidimensional Gaussian process regression algorithm. Based on the squared exponential covariance function determined during model training and the optimized hyperparameters, it deduces the posterior distribution of the effect state corresponding to the input feature vector. During the Gaussian process regression deduction, the posterior probability distribution formula is used to calculate:

[0138]

[0139] in, The state of the effect to be evaluated. The input feature vector to be evaluated. Input matrix for training samples, Output labels for the training samples. The posterior mean is... The autocovariance of the input feature vector. This represents the model noise variance. The system uses this formula to output the probability distribution range of the effect state under the operating conditions of the cable to be evaluated, generating an initial probability estimate that includes the probability intervals for both the normal and disturbance states.

[0140] S52: Perform state probability calculation processing on the initial probability estimate, compare the probability density function values ​​corresponding to different effect states, calculate the specific probability values ​​belonging to each state proportionally, and generate a quantitative effect state probability distribution.

[0141] Specifically, the system performs state probability calculations on the initial probability estimate. First, it extracts the probability density functions corresponding to the normal state and the disturbance state from the initial probability estimate. The probability density functions are determined based on the Gaussian distribution probability density formula:

[0142]

[0143] in, For the effect state dimension (here) (corresponding to normal and interference states). This is the mean vector of state probabilities. Let be the state probability covariance matrix. The system calculates the probability density function values ​​of the two effect states within the probability distribution range one by one, and determines the peak position and distribution range of the probability density of each state by numerical comparison. Then, the system calculates the specific probability value belonging to each state according to the probability density integral ratio. The integration range is the effective interval corresponding to the probability density function of each state. The integration result satisfies the constraint that the sum of the probability of the normal state and the probability of the disturbance state is 1. Through the above calculations, the system generates a quantitative effect state probability distribution, which clearly gives the specific probability values ​​of the normal state and the disturbance state under the cable condition to be evaluated.

[0144] S53: Visualize and render the quantitative probability distribution of effect states, draw probability contour lines or surface plots of different states on the two-dimensional plane of electric field-magnetic field complexity, and generate a visualized probability distribution surface for risk warning.

[0145] Specifically, the system performs visualization rendering on the quantitative probability distribution of effect states. First, a two-dimensional plane of electric-magnetic field complexity is constructed. The horizontal axis of the plane is set to the characteristic value of the magnetic field strength under the cable condition to be evaluated, and the vertical axis is set to the characteristic value of the electric field strength. Each coordinate point in the plane corresponds to a set of electric-magnetic field characteristic combinations and the corresponding radiation field effect index EF value. The system calls a visualization drawing tool to draw probability contour lines or surface plots of different states on the two-dimensional plane based on the quantitative probability distribution of effect states. During the drawing process, the probability values ​​of normal and interference states are mapped to different color gradients or surface heights; for example, higher probability values ​​correspond to darker colors or higher surface heights. The probability value ranges are also marked, clearly defining the ranges of low risk (high probability of normal state), medium risk (probabilities of the two states are close), and high risk (high probability of interference state). Through the above rendering process, the system generates a visualized probability distribution surface for risk warning. This surface can intuitively display the probability distribution of effect states under different electric-magnetic field combinations, providing engineers with an intuitive visual reference for judging the interference risk level of the cable condition to be evaluated and formulating protective measures.

[0146] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0147] Based on the same inventive concept, this application also provides a device for calculating and evaluating the spatial radiation field effect of a long straight control cable, used to implement the aforementioned method for calculating and evaluating the spatial radiation field effect of a long straight control cable. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the device for calculating and evaluating the spatial radiation field effect of a long straight control cable provided below can be found in the limitations of the method for calculating and evaluating the spatial radiation field effect of a long straight control cable described above, and will not be repeated here.

[0148] Preferably, such as Figure 3 As shown, the present invention provides a spatial radiation field effect calculation and evaluation system 600 for long straight control cables, which is configured with the following modules:

[0149] The electromagnetic transient model construction module 610 is used to obtain the physical structure parameters of the long straight control cable. Based on the cable geometry and preset current characteristics, it calls Maxwell's equations to derive the electromagnetic partial differential equations and defines the initial and boundary conditions of the cable's spatial position, generating the electromagnetic transient model and electromagnetic calculation constraints of the long straight control cable.

[0150] The Physical Information Neural Network Training Module 620 is used to train a physical information neural network based on the acquired field measurable magnetic field data according to the electromagnetic transient model and electromagnetic calculation constraints. It combines the measurable magnetic field data and the additional physical information loss function constructed according to physical rules to train the network. The network is used to solve for the missing electric field data and generate a complete electromagnetic dataset containing the predicted electric field and the measured magnetic field.

[0151] The time-frequency feature extraction and interference factor calculation module 630 is used to perform time-frequency transformation processing on the predicted electric field and the measured magnetic field based on the complete electromagnetic dataset, extract the time-frequency distribution features, and calculate the interference factor by combining the characteristic frequency of the disturbed equipment and the equivalent inductance parameter of the cable, and generate a radiation field effect index that characterizes the complexity of electromagnetic interference.

[0152] The Gaussian process regression model training module 640 is used to obtain the effect state recorded by field observation. It takes the radiation field effect index as the input feature and the effect state as the output label. It establishes a nonlinear mapping relationship between the radiation field effect index and the effect state through multidimensional Gaussian process regression. It uses the trained Gaussian process regression model to predict the probability distribution of the effect state corresponding to any radiation field effect index and generate a risk probability distribution model.

[0153] The Engineering Risk Assessment and Visualization Module 650 is used to assess engineering risks using a risk probability distribution model. It inputs the radiation field effect index to be assessed, calculates the probability distribution of different effect states, and generates a visualized risk probability distribution surface.

[0154] Preferably, the electromagnetic transient model construction module 610 provided in this application is configured with the following units:

[0155] The electromagnetic coupling model building unit is used to obtain the physical structure parameters of a long straight control cable. Based on the actual layout of the two metal connections in the physical structure parameters, electromagnetic transient modeling is performed. According to the cable cross-sectional structure and geometric characteristics extending along the z-axis, combined with the conduction characteristics of the current density only in the z-direction component, an electromagnetic coupling model of a finite-length unshielded cable is constructed to generate an electromagnetic transient model.

[0156] The electromagnetic partial differential equation derivation unit is used to derive electromagnetic equations for electromagnetic transient models based on Maxwell's equations and the continuity equation. It introduces the assumption of a homogeneous medium and the constraint of zero charge density, and combines the characteristic that the first derivative of the magnetic field strength is zero at the boundary to derive the partial differential relationship between the spatial electric field and the magnetic field strength, thereby generating electromagnetic partial differential equations.

[0157] The solution condition definition unit is used to define the solution conditions for electromagnetic partial differential equations. Based on the cutoff characteristics of the cable end current and potential in the physical structure parameters, it sets the boundary condition of zero electric field strength and obtains the field strength data at the initial moment based on the physical structure parameters.

[0158] The initial field distribution calculation unit is used to calculate the initial field distribution based on the zero-value boundary conditions of the electric field strength and the field strength data. It determines the initial electric and magnetic field strength distribution through mathematical modeling based on the radiation principle of a finite-length straight conductor, and generates electromagnetic calculation constraints by combining the charge density continuity equation.

[0159] Preferably, the physical information neural network training module 620 provided in this application is configured with the following units:

[0160] The multi-objective loss function construction unit is used to construct a neural network architecture for electromagnetic transient models and electromagnetic calculation constraints. Based on the residual calculation of partial differential equations, initial condition matching error, boundary condition matching error and measurable magnetic field data deviation, it integrates additional physical information rules to construct a multi-objective loss function and generate a physical information-driven neural network optimization objective.

[0161] The parameter iteration training unit is used to perform parameter iteration training on the neural network optimization target. It calculates the gradient of the loss function with respect to the network weights through automatic differentiation and backpropagates the update until the total loss value converges to the preset convergence threshold, generating a fully trained field prediction model.

[0162] The electric field data extrapolation unit is used to perform electric field extrapolation processing on the field quantity prediction model and the field measurable magnetic field data. The input magnetic field time series is used to calculate the electric field intensity distribution of the spatial target point through forward propagation, and generate a complete electromagnetic dataset containing the predicted electric field and the measured magnetic field.

[0163] Preferably, the time-frequency feature extraction and interference factor calculation module 630 provided in this application is configured with the following units:

[0164] The time-frequency joint analysis unit is used to perform time-frequency joint analysis on the predicted electric field and measured magnetic field in the complete electromagnetic dataset. It calls the short-time Fourier transform to calculate the energy distribution intensity of the signal in the time-frequency dimension and generates the two-dimensional time-frequency spectrum matrix of the signal.

[0165] The dominant frequency extraction unit is used to extract the dominant frequency from the two-dimensional time spectrum matrix, locate the frequency value corresponding to the maximum energy intensity in the time spectrum matrix, and generate key frequency characteristic parameters by combining the cable equivalent inductance parameters and the preset characteristic frequency of the disturbed equipment.

[0166] The interference factor synthesis and index calculation unit is used to synthesize interference factors for key frequency characteristic parameters and calculate the radiation field effect index based on the relative deviation between the dominant frequency and the equipment characteristic frequency using a preset objective complexity formula.

[0167] Preferably, the Gaussian process regression model training module 640 provided in this application is configured with the following units:

[0168] The training sample alignment building unit is used to perform sample alignment processing on the effect state of the acquired field observation records and the corresponding radiation field effect index. The complexity index EF value is used as the input feature vector, and the observed normal or disturbance state is used as the discrete output label to generate a labeled model training sample set.

[0169] The multidimensional Gaussian process modeling unit is used to perform multidimensional Gaussian process regression modeling on the model training sample set, configure the squared exponential covariance function and optimize the hyperparameters and length scale matrix through maximum likelihood estimation, establish the probability mapping relationship between complexity index and effect state, and generate the initial probability mapping model.

[0170] The posterior distribution derivation optimization unit is used to derive the posterior distribution of the initial probability mapping model, update the conditional probability distribution through Bayesian inference, and perform KL divergence matching between the posterior probability distribution predicted by the model and the observed state of the training samples to generate a risk probability distribution model.

[0171] Preferably, the engineering risk assessment and visualization module 650 provided in this application is configured with the following units:

[0172] The working condition probability extrapolation unit is used to process the risk probability distribution model in a scenario application. It takes the radiation field effect index calculated under the working condition of the cable to be evaluated as input, and extrapolates its possible state distribution through Gaussian process regression to generate the initial probability estimate under the working condition of the cable to be evaluated.

[0173] The state probability quantitative calculation unit is used to perform state probability calculation processing on the initial probability estimate, compare the probability density function values ​​corresponding to different effect states, calculate the specific probability values ​​belonging to each state according to the proportion, and generate a quantitative effect state probability distribution.

[0174] The probability distribution visualization rendering unit is used to visualize and render the quantitative probability distribution of effect states. It draws probability contour lines or surface plots of different states on the two-dimensional plane of electric field-magnetic field complexity, and generates a visualized probability distribution surface for risk warning.

[0175] In one embodiment, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described method for calculating and evaluating the spatial radiation field effect of long straight control cables.

[0176] In one embodiment, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for calculating and evaluating the spatial radiation field effect of long straight control cables.

[0177] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0178] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0179] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for calculating and evaluating the spatial radiation field effect of a long straight control cable, characterized in that, Includes the following steps: S1: Obtain the physical structure parameters of the long straight control cable, derive the electromagnetic partial differential equations based on the cable geometry and preset current characteristics, call Maxwell's equations to define the initial and boundary conditions of the cable's spatial position, and generate the electromagnetic transient model and electromagnetic calculation constraints of the long straight control cable. S2: Based on the electromagnetic transient model and the electromagnetic calculation constraints, the acquired on-site measurable magnetic field data is used to train a physical information neural network. The network is trained by combining the measurable magnetic field data and the additional physical information loss function constructed according to physical rules. The missing electric field data is solved by the network to generate a complete electromagnetic dataset containing the predicted electric field and the measured magnetic field. S3: Based on the complete electromagnetic dataset, perform time-frequency transformation processing on the predicted electric field and the measured magnetic field, extract the time-frequency distribution characteristics, and calculate the interference factor by combining the characteristic frequency of the disturbed equipment and the equivalent inductance parameter of the cable, and generate a radiation field effect index that characterizes the complexity of electromagnetic interference. S4: Obtain the effect state recorded in the field observation, take the radiation field effect index as the input feature and the effect state as the output label, establish the nonlinear mapping relationship between the radiation field effect index and the effect state through multidimensional Gaussian process regression, use the trained Gaussian process regression model to predict the probability distribution of the effect state corresponding to any radiation field effect index, and generate a risk probability distribution model. S5: Perform engineering risk assessment on the risk probability distribution model, input the radiation field effect index to be assessed to calculate the probability distribution of different effect states, and generate a visualized risk probability distribution surface.

2. The method according to claim 1, characterized in that, S1 includes: S11: Obtain the physical structure parameters of the long straight control cable, perform electromagnetic transient modeling based on the actual layout of the two metal connections in the physical structure parameters, and construct an electromagnetic coupling model of the finite length unshielded cable based on the cable cross-sectional structure and geometric characteristics extending along the z-axis, combined with the conduction characteristics of the current density only in the z-direction component, and generate an electromagnetic transient model. S12: Based on Maxwell's equations and the continuity equation, the electromagnetic transient model is derived by introducing the assumption of a homogeneous medium and the constraint of zero charge density. Combining the characteristic that the first derivative of the magnetic field strength is zero at the boundary, the partial differential relationship between the spatial electric field and the magnetic field strength is derived, and the electromagnetic partial differential equation is generated. S13: Define the solution conditions for the electromagnetic partial differential equation, set the zero-value boundary condition for electric field strength according to the cutoff characteristics of the cable end current and potential in the physical structure parameters, and obtain the field strength data at the initial moment based on the physical structure parameters. S14: Calculate the initial field distribution based on the zero-value boundary condition of the electric field strength and the field strength data. Determine the initial electric and magnetic field strength distributions through mathematical modeling based on the radiation principle of a finite-length straight conductor, and generate electromagnetic calculation constraints by combining the charge density continuity equation.

3. The method according to claim 1, characterized in that, S2 includes: S21: The electromagnetic transient model and electromagnetic calculation constraints are processed by constructing a neural network architecture. Based on the residual calculation of partial differential equations, initial condition matching error, boundary condition matching error and measurable magnetic field data deviation, additional physical information rules are fused to construct a multi-objective loss function and generate a physical information-driven neural network optimization objective. S22: Perform parameter iterative training on the neural network optimization target, calculate the gradient of the loss function with respect to the network weights through automatic differentiation technology and backpropagate to update it until the total loss value converges to the preset convergence threshold, and generate a fully trained field quantity prediction model. S23: Perform electric field estimation processing on the field quantity prediction model and the field measurable magnetic field data. Input the magnetic field time series and calculate the electric field intensity distribution of the spatial target point through forward propagation to generate a complete electromagnetic dataset containing the predicted electric field and the measured magnetic field.

4. The method according to claim 3, characterized in that, The multi-objective loss function includes an initial condition loss function, a boundary condition loss function, and a partial differential equation condition loss function. The expression of the multi-objective loss function is as follows: in, Represents the boundary condition loss function. The number of sampling points under boundary conditions. , This indicates the number of sampling points under the initial conditions. These are the initial input magnetic field and electric field data, respectively. These are the initial magnetic field data and electric field data, respectively. For the nabla operator, Represents the initial condition loss function. This represents the conditional loss function of a partial differential equation. This represents the number of sampling points in the computational domain.

5. The method according to claim 1, characterized in that, S3 includes: S31: Perform time-frequency joint analysis on the predicted electric field and measured magnetic field in the complete electromagnetic dataset, call the short-time Fourier transform to calculate the energy distribution intensity of the signal in the time-frequency dimension, and generate the two-dimensional time-frequency matrix of the signal. S32: Perform dominant frequency extraction processing on the two-dimensional time spectrum matrix, locate the frequency value corresponding to the maximum energy intensity in the time spectrum matrix, and generate key frequency characteristic parameters by combining the cable equivalent inductance parameters and the preset characteristic frequency of the disturbed equipment. S33: The key frequency characteristic parameters are subjected to interference factor synthesis processing. Based on the relative deviation between the dominant frequency and the equipment characteristic frequency, the radiation field effect index is calculated using a preset objective complexity formula. The expression of the objective complexity formula is as follows: in, This is an objective indicator of signal complexity, specifically a radiation field effect indicator. This represents the time-frequency energy distribution of the signal. This represents the average value of the time-frequency energy. The frequency point corresponding to the maximum time-frequency distribution intensity of the disturbance signal. For objectively calculating the characteristic frequencies of the disturbance-sensitive device, M is the number of sequences in the time direction of the signal's time spectrum, N is the number of sequences in the frequency direction of the signal's time spectrum, and A is the average weighting coefficient. Let be the energy distribution value of the i-th time series and the j-th frequency series in the time-spectrum matrix, and B be the start-up energy threshold. Let be the equivalent inductance of the cable, and c be the speed of light.

6. The method according to claim 1, characterized in that, S4 includes: S41: Perform sample alignment processing on the effect state of the acquired field observation records and the corresponding radiation field effect index, take the complexity index EF value as the input feature vector, and take the observed normal or disturbance state as the discrete output label to generate a labeled model training sample set. S42: Perform multidimensional Gaussian process regression modeling on the training sample set of the model, configure the squared exponential covariance function and optimize the hyperparameters and length scale matrix through maximum likelihood estimation, establish the probability mapping relationship from complexity index to effect state, and generate the initial probability mapping model. S43: Perform posterior distribution derivation on the initial probability mapping model, update the conditional probability distribution through Bayesian inference, and perform KL divergence matching between the posterior probability distribution predicted by the model and the observation state of the training samples to generate a risk probability distribution model.

7. The method according to any one of claims 1-6, characterized in that, S5 includes: S51: Perform scenario application processing on the risk probability distribution model, input the radiation field effect index calculated under the cable condition to be evaluated, and deduce its possible state distribution through Gaussian process regression to generate the initial probability estimate under the cable condition to be evaluated. S52: Perform state probability calculation processing on the initial probability estimate, compare the probability density function values ​​corresponding to different effect states, calculate the specific probability values ​​belonging to each state according to the proportion, and generate a quantitative effect state probability distribution. S53: Perform visualization rendering on the quantitative effect state probability distribution, draw probability contour lines or surface diagrams of different states on the electric field-magnetic field complexity two-dimensional plane, and generate a visualization probability distribution surface for risk warning.

8. A system for calculating and evaluating the spatial radiation field effect of a long straight control cable, characterized in that, The system includes: The electromagnetic transient model construction module is used to obtain the physical structure parameters of long straight control cables. Based on the cable geometry and preset current characteristics, it calls Maxwell's equations to derive electromagnetic partial differential equations and defines the initial and boundary conditions of the cable's spatial position, generating the electromagnetic transient model and electromagnetic calculation constraints of the long straight control cable. The physical information neural network training module is used to train the physical information neural network based on the electromagnetic transient model and the electromagnetic calculation constraints, and to train the network by combining the measurable magnetic field data and the additional physical information loss function constructed according to physical rules. The network is then used to solve for the missing electric field data to generate a complete electromagnetic dataset containing the predicted electric field and the measured magnetic field. The time-frequency feature extraction and interference factor calculation module is used to perform time-frequency transformation processing on the predicted electric field and the measured magnetic field based on the complete electromagnetic dataset, extract the time-frequency distribution features, and calculate the interference factor by combining the characteristic frequency of the disturbed equipment and the equivalent inductance parameter of the cable, and generate a radiation field effect index characterizing the complexity of electromagnetic interference. The Gaussian process regression model training module is used to obtain the effect state recorded by field observation. The radiation field effect index is used as the input feature and the effect state is used as the output label. The nonlinear mapping relationship between the radiation field effect index and the effect state is established through multidimensional Gaussian process regression. The trained Gaussian process regression model is used to predict the probability distribution of the effect state corresponding to any radiation field effect index and generate a risk probability distribution model. The engineering risk assessment and visualization module is used to assess engineering risks using the risk probability distribution model. It inputs the radiation field effect index to be assessed, calculates the probability distribution of different effect states, and generates a visualized risk probability distribution surface.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

Images