Cable channel state sensing method and system, storage medium and equipment

Abstract

The embodiment of the invention discloses a cable channel state sensing method. The method comprises the following steps: acquiring a plurality of characteristic parameters of the running state of a cable channel in real time; carrying out dimension reduction processing on the plurality of feature parameters to obtain a fusion feature vector; constructing a neural network model, and determining a corresponding partial differential equation and a loss function; training the neural network model according to the fusion feature vector until the corresponding loss function meets a threshold condition; and determining the operation state of the to-be-detected cable channel based on the trained neural network model. According to the invention, by combining the technical means of real-time data acquisition, data dimension reduction and fusion, and neural network model construction and training, efficient and accurate sensing of the operation state of the cable channel is realized, and a powerful guarantee is provided for safe and stable operation of a cable system.

Description

Technical Field

[0001] The present invention relates to the technical field of power distribution cable network status perception, and in particular to a cable channel status perception method, system, storage medium and device. Background Art

[0002] With the acceleration of urbanization, power cables, as the "blood vessels" of urban power transmission, are directly related to the normal operation of the city and the quality of life of residents due to their safety and reliability. However, since power cables are generally laid underground, their operating environment is complex and changeable and difficult to observe directly, resulting in huge challenges in the discovery and repair of cable faults. The cables in underground cable channels are densely packed and complexly arranged. In addition, the power supply distance is long and the channel environment is harsh (such as the presence of toxic and harmful gases, high temperature and humidity, water accumulation, etc.). These factors greatly increase the risk of cable operation.

[0003] The traditional cable inspection method mainly relies on manual inspection, but this method is inefficient, time-consuming and labor-intensive, and is easily limited by the inspectors' experience, physical strength and environmental factors. It is difficult to fully and accurately grasp the actual conditions of the cable channel. More importantly, manual inspections in harsh channel environments also pose serious threats to the personal safety of inspectors. Therefore, with the development of smart grids, higher requirements are placed on the operation and maintenance of cable channels, and an efficient, intelligent and safe inspection method is urgently needed.

[0004] In recent years, the evolution of cable channel risks and digital operation and maintenance technology have become research hotspots. This technology aims to establish a cable channel risk assessment model and simulation system by comprehensively considering multiple factors such as cable body factors (such as sheath current, core current, intermediate joint temperature, etc.), channel environmental factors (such as temperature, humidity, gas concentration, etc.) and external force damage, so as to accurately identify, evaluate and locate the operation risks of cable channels. This technology can not only improve inspection efficiency and reduce operation and maintenance costs, but also effectively prevent cable failures and ensure the safe and stable operation of the power grid. Summary of the invention

[0005] Based on this, it is necessary to propose a cable channel status perception method to address the above problems.

[0006] A method for sensing cable channel status, the method comprising the following steps:

[0007] Collect several characteristic parameters of the cable channel operation status in real time;

[0008] Performing dimensionality reduction processing on the plurality of feature parameters to obtain a fused feature vector;

[0009] Build a neural network model and determine the corresponding partial differential equation and loss function;

[0010] Training the neural network model according to the fused feature vector until the corresponding loss function meets a threshold condition;

[0011] The operating status of the cable channel to be detected is determined based on the trained neural network model.

[0012] In the above scheme, the real-time acquisition of several characteristic parameters of the cable channel operation status specifically includes:

[0013] Obtain 16 characteristic parameters under different laying modes of cable channels: cable sheath current, cable core current, cable intermediate joint temperature, cable partial discharge current, cable current carrying capacity, grounding current, vibration information, ambient temperature, humidity, carbon monoxide concentration, hydrogen sulfide concentration, oxygen concentration, water level, smoke concentration, vibration parameters, and external damage parameters;

[0014] Construct a vector sequence based on the 16 characteristic parameters:

[0015] X={x1,x2,…,x n} m×n

[0016] Among them, n is 16, which represents the number of samples, m represents the vector dimension, and x i represents the i-th feature parameter.

[0017] In the above solution, the dimensionality reduction process is performed on the plurality of feature parameters to obtain a fused feature vector, specifically including:

[0018] Decentralize the feature parameters and obtain the covariance matrix:

[0019]

[0020] Among them, A represents an M×M dimensional symmetric matrix, X represents a vector sequence, and X T Represents the transposed matrix of a vector sequence;

[0021] Perform eigendecomposition on the matrix A:

[0022] A=Q∑Q -1

[0023] Among them, the vectors contained in the matrix Q are the eigenvectors of the square matrix A, ∑ represents an M×M dimensional diagonal matrix, and the values ​​on the diagonal are the eigenvalues ​​λ i , where i = 1, 2, ... m;

[0024] Preset λ1≥λ2≥…≥λ m , select the eigenvector a where the first k eigenvalues ​​are located i Construct the projection matrix:

[0025] W=[a1,a2,…,a k ]

[0026] Among them, a i represents an m-dimensional column vector;

[0027] Select the first k characteristic roots to determine the contribution coefficient K t ;

[0028] The feature space is determined by k feature vectors, and the original data is projected into the feature space to obtain a sample of the merged features:

[0029] Y=W T X

[0030] Among them, Y represents the fused feature vector, W T represents the transposed matrix of the projection matrix, and X represents a vector sequence.

[0031] In the above scheme, the construction of the neural network model and the determination of the corresponding partial differential equation and loss function specifically include:

[0032]

[0033] in, and Ω represent the boundary and spatial domain respectively, u(Y,t) represents the spatial variable Y(y1,y2,…,y k ) is the solution of the domain Ω and the time variable t in the time domain, I(Y) and B(Y,t) represent the initial conditions and boundary conditions respectively, N represents the partial differential operator with respect to the spatial variable, and λ represents the coefficient related to the physical information;

[0034] The loss function consists of the initial conditions, boundary conditions, and the residuals of the governing equations:

[0035] L(θ)=ω i L i (θ)+ω b L b (θ)+ω r L r (θ)

[0036] Among them, ω i ,ω b ,ω r They represent the weights of the initial conditions, boundary conditions and loss function of the control equation respectively.

[0037] In the above scheme, the loss function of the initial conditions, boundary conditions and control equations is:

[0038]

[0039] Among them, θ represents all parameters in the neural network model, Y and t represent the input, and u θ Represents the output value of the neural network model under the parameter θ, L i (θ) represents the loss function of the initial condition, L b (θ) Loss function of boundary conditions, L r (θ) The loss function governing the equation.

[0040] In the above scheme, the neural network model is trained according to the fused feature vector until the corresponding loss function meets the threshold condition, specifically including:

[0041] The feature parameters are mapped using the CORAL transfer learning algorithm to obtain the covariance distance between the source domain and target domain features of the CORAL function:

[0042]

[0043] Among them, C S and C T Represent the covariance matrices of the source domain and the target domain respectively, and d represents the number of neural network outputs;

[0044] When the covariance distance is within a preset range, the training of the neural network model is completed.

[0045] In the above scheme, the operation status of the cable channel to be detected is determined based on the trained neural network model, which specifically includes:

[0046] The data of the cable channel to be detected is input into the trained neural network model to determine the operation status of the cable channel, wherein the operation status of the cable channel includes a normal state, a warning state and an alarm state.

[0047] The present application also proposes a cable channel state perception system, the system comprising: a fusion feature vector acquisition unit, a neural network model construction unit, a neural network model training unit and a cable channel state perception unit;

[0048] The fused feature vector acquisition unit is used to collect several feature parameters of the cable channel operation status in real time; perform dimensionality reduction processing on the several feature parameters to obtain a fused feature vector;

[0049] The neural network model building unit is used to determine the corresponding partial differential equation and loss function;

[0050] The neural network model training unit is used to train the constructed neural network model based on the fused feature vector until the corresponding loss function meets a threshold condition;

[0051] The cable channel state sensing unit is used to determine the operating state of the cable channel to be detected based on the trained neural network model.

[0052] The present application also proposes a readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the following steps: collecting several characteristic parameters of the operating status of the cable channel in real time;

[0053] Performing dimensionality reduction processing on the plurality of feature parameters to obtain a fused feature vector;

[0054] Build a neural network model and determine the corresponding partial differential equation and loss function;

[0055] Training the neural network model according to the fused feature vector until the corresponding loss function meets a threshold condition;

[0056] The operating status of the cable channel to be detected is determined based on the trained neural network model.

[0057] The present application also proposes a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the following steps of the computer program:

[0058] Collect several characteristic parameters of the cable channel operation status in real time;

[0059] Performing dimensionality reduction processing on the plurality of feature parameters to obtain a fused feature vector;

[0060] Build a neural network model and determine the corresponding partial differential equation and loss function;

[0061] Training the neural network model according to the fused feature vector until the corresponding loss function meets a threshold condition;

[0062] The operating status of the cable channel to be detected is determined based on the trained neural network model.

[0063] The use of the embodiments of the present invention has the following beneficial effects: first, several characteristic parameters of the operating status of the cable channel are collected in real time; the several characteristic parameters are subjected to dimensionality reduction processing to obtain a fused characteristic vector; a neural network model is constructed to determine the corresponding partial differential equation and loss function; the neural network model is trained according to the fused characteristic vector until the corresponding loss function meets the threshold condition; the operating status of the cable channel to be detected is determined based on the trained neural network model. The present invention realizes efficient and accurate perception of the operating status of the cable channel by combining the technical means of real-time data collection, data dimensionality reduction and fusion, and neural network model construction and training, providing a strong guarantee for the safe and stable operation of the cable system. This method combines a deep learning model and can take into account more dimensional information, thereby more accurately and intelligently confirming and identifying the status of the cable system. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0065] in:

[0066] Figure 1 A schematic diagram of a method for sensing cable channel status in an embodiment;

[0067] Figure 2 The cable channel status perception result determined based on the test set. DETAILED DESCRIPTION

[0068] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0069] In the following description, a large number of specific details are given in order to provide a more thorough understanding of the present invention; however, it is obvious to those skilled in the art that the present invention can be implemented without one or more of these details; in other examples, in order to avoid confusion with the present invention, some technical features known in the art are not described, and it should be understood that the present invention can be implemented in different forms and should not be interpreted as limited to the embodiments presented here; on the contrary, providing these embodiments will make the disclosure thorough and complete and fully convey the scope of the present invention to those skilled in the art.

[0070] The purpose of the terms used herein is only to describe specific embodiments and is not intended to be limiting of the present invention. When used herein, the singular forms "a", "an" and "said / the" are also intended to include the plural forms, unless the context clearly indicates otherwise, and it should also be understood that the terms "consisting of" and / or "comprising", when used in this specification, determine the presence of the features, integers, steps, operations, elements and / or parts, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, parts and / or groups. When used herein, the term "and / or" includes any and all combinations of the relevant listed items.

[0071] To facilitate understanding, the relevant terms involved in this application are first introduced below.

[0072] (1) Least squares method, which seeks the best function matching the data by minimizing the sum of squared errors. In statistics, it is often used to estimate the parameters of linear regression models.

[0073] (2) Principal component analysis is a commonly used data dimensionality reduction technique that can reduce the dimensions of the data while retaining the most important features in the data set, thereby simplifying the complexity of data analysis. PCA transforms the original data into a new coordinate system through linear transformation. The coordinate axes of this new coordinate system (i.e., principal components) are linear combinations of the dimensions of the original data, and these coordinate axes are arranged from large to small according to the data variance.

[0074] The cable channel risk assessment model established by the existing technology still has some shortcomings in practical application, such as incomplete data collection, low data processing accuracy, and inaccurate risk assessment model. To this end, this paper proposes a cable channel status perception method and system based on an improved physical information neural network. This method integrates 16 key parameters such as cable sheath current, cable core current, intermediate joint temperature, partial discharge current, current carrying capacity, grounding current, vibration information, ambient temperature, humidity, carbon monoxide concentration, hydrogen sulfide concentration, oxygen concentration, water level, smoke concentration, vibration and external force damage, and uses principal component analysis to extract features and reduce dimensions. It is then combined with the improved physical information neural network for deep learning and intelligent analysis, thereby achieving comprehensive and accurate perception of the status of dense urban cable channels; this not only helps to improve the level of intelligence in cable channel operation and maintenance management, but also provides strong support for ensuring the safe and stable operation of urban power grids.

[0075] In order to thoroughly understand the present invention, a detailed structure will be presented in the following description to illustrate the technical solution proposed by the present invention; optional embodiments of the present invention are described in detail as follows, but in addition to these detailed descriptions, the present invention may also have other implementation methods.

[0076] like Figure 1 As shown, in one embodiment, a cable channel state sensing method is provided, and the cable channel state sensing method includes steps S101 to S105, which are described in detail as follows:

[0077] S101, collecting several characteristic parameters of the cable channel operation status in real time;

[0078] By collecting several characteristic parameters of the cable channel operation status in real time, the current status of the cable channel can be reflected immediately, which helps to promptly discover potential problems or faults and improve the safety and stability of the cable system. Real-time monitoring is particularly important for key infrastructure such as electricity.

[0079] The variation characteristics of various parameters under different laying modes and operating conditions of cable channels are studied. This application describes the same object and the same process from different angles for urban dense cable channels from the aspects of theoretical construction, evolution model, evolution path and system complexity, quantifies the corresponding risk development degree, analyzes the impact of risk evolution, explores the parameter variation mechanism caused by risk factors, obtains the risk characteristics and variation laws of urban dense cable channels, and first obtains several (16) characteristic parameters of cable channels under different laying modes and operating conditions.

[0080] In some embodiments, several characteristic parameters of the cable channel operation status are collected in real time, specifically including:

[0081] Obtain 16 characteristic parameters under different laying modes of cable channels: cable sheath current, cable core current, cable intermediate joint temperature, cable partial discharge current, cable current carrying capacity, grounding current, vibration information, ambient temperature, humidity, carbon monoxide concentration, hydrogen sulfide concentration, oxygen concentration, water level, smoke concentration, vibration parameters, and external damage parameters;

[0082] Construct a vector sequence based on 16 characteristic parameters:

[0083] X={x1,x2,…,x n} m×n

[0084] Among them, n is 16, which represents the number of samples, m represents the vector dimension, and x i represents the i-th feature parameter.

[0085] The operation status of the cable channel is monitored, and a total of 16-dimensional feature data is obtained. The different characteristic properties of each group of signals lead to low accuracy of the state perception results and a large amount of calculation. Therefore, the principal component analysis (PCA) method is used to extract the main components of all eigenvalues. PCA can compress the feature quantities of all signals. It is a fusion method that uses the idea of ​​dimensionality reduction and is adaptive. The original data is converted into a new set of sample data through transformation and reconstruction. Although the data obtained has a low dimension, it can also reflect most of the characteristic information of the original data, which is conducive to the subsequent state perception calculation.

[0086] S102, performing dimensionality reduction processing on a number of feature parameters to obtain a fused feature vector;

[0087] Dimensionality reduction processing of the multiple feature parameters collected can remove redundant information and reduce the complexity and computational cost of data processing. At the same time, the fused feature vector obtained through dimensionality reduction can more comprehensively reflect the status of the cable channel and improve the accuracy and efficiency of subsequent analysis.

[0088] Preferably, all vector feature data are formed into a vector sequence X = {x1, x2, ..., x n} m×n , n represents the number of samples, m represents the vector dimension, and in this application, there are 16 types of feature quantities, so n is taken as 16, x i is the i-th feature quantity, specifically including:

[0089] First, the mean-variance calculation method is used to normalize the dimensions of all signals to solve the problem of large dimensional differences between different signals, and then the data is decentralized: i =x i -μ; where the mean i=1,2,3…n.

[0090] In some embodiments, a plurality of feature parameters are subjected to dimensionality reduction processing to obtain a fused feature vector, specifically including:

[0091] Decentralize the feature parameters and obtain the covariance matrix:

[0092]

[0093] Among them, A represents an M×M dimensional symmetric matrix, X represents a vector sequence, and X T Represents the transposed matrix of a vector sequence;

[0094] Perform eigendecomposition on the matrix A:

[0095] A=Q∑Q -1

[0096] Among them, the vectors contained in the matrix Q are the eigenvectors of the square matrix A, ∑ represents an M×M dimensional diagonal matrix, and the values ​​on the diagonal are the eigenvalues ​​λ i , where i = 1, 2, ... m;

[0097] Preset λ1≥λ2≥…≥λ m , select the eigenvector a where the first k eigenvalues ​​are located i Construct the projection matrix:

[0098] W=[a1,a2,…,a k ]

[0099] Among them, a i represents an m-dimensional column vector;

[0100] Select the first k characteristic roots to determine the contribution coefficient K t ;

[0101] The feature space is determined by k eigenvectors, and the original data is projected into the feature space to obtain a sample of the merged features:

[0102] Y=W T X

[0103] Among them, Y represents the fused feature vector, W T represents the transposed matrix of the projection matrix, and X represents a vector sequence.

[0104] Preferably, the number of principal components is determined in the order of the eigenvalues. The larger the eigenvalue, the greater the contribution coefficient. The first k eigenvalues ​​are selected, and the contribution coefficient K t Defined as:

[0105] In the formula, λ i is the i-th characteristic root.

[0106] S103, constructing a neural network model, and determining the corresponding partial differential equation and loss function;

[0107] By utilizing the ability of neural network models to handle complex nonlinear problems and combining the diversity and uncertainty of cable channel states, a high-precision and high-robustness prediction model can be constructed. By determining the corresponding partial differential equations and loss functions and training based on fused feature vectors, it can be ensured that the model can accurately learn the mapping relationship between cable channel states and characteristic parameters. During the training process, the model parameters are continuously optimized so that the loss function meets the threshold conditions, thereby ensuring the prediction accuracy of the model.

[0108] The improved physical information neural network algorithm is used to realize the state perception of dense cable channels in cities. The physical information neural network is a machine learning model that combines deep learning and physics knowledge. Unlike traditional data-driven neural networks, the physical information neural network uses physical laws to guide the model during the learning process, thereby improving the generalization ability of the model.

[0109] In some embodiments, constructing a neural network model and determining the corresponding partial differential equation and loss function specifically include:

[0110]

[0111] in, and Ω represent the boundary and spatial domain respectively, u(Y,t) represents the spatial variable Y(y1,y2,…,y k ) is the solution of the domain Ω and the time variable t in the time domain, I(Y) and B(Y,t) represent the initial conditions and boundary conditions respectively, N represents the partial differential operator with respect to the spatial variable, and λ represents the coefficient related to the physical information;

[0112] The loss function consists of the initial conditions, boundary conditions, and the residuals of the governing equations:

[0113] L(θ)=ω i L i (θ)+ω b L b (θ)+ω r L r (θ)

[0114] Among them, ω i ,ω b ,ω r They represent the weights of the initial conditions, boundary conditions and loss function of the control equation respectively.

[0115] Specifically, in the physical information neural network structure, θ represents all parameters in the neural network model, Y(y1, y2,…, yk) and t represent the input quantity, uθ is the output value of the neural network under the parameter θ, and the physical information neural network obtains the optimal θ by minimizing the loss function. The loss function is composed of the initial conditions, boundary conditions and residuals of the control equation.

[0116] In some embodiments, the initial conditions, boundary conditions, and loss functions of the governing equations are:

[0117]

[0118] Among them, θ represents all parameters in the neural network model, Y and t represent the input, and u θ Represents the output value of the neural network model under the parameter θ, L i (θ) represents the loss function of the initial condition, L b (θ) Loss function of boundary conditions, L r (θ) The loss function governing the equation.

[0119] For physical information neural networks, inappropriate setting of θ will greatly affect the accuracy and solution speed of the algorithm. In this application, the transfer learning algorithm is used to minimize L(θ) and obtain the optimal parameter θ to achieve the improvement of the physical information neural network. Transfer learning is a training strategy in machine learning. It has broad application prospects when combined with the use of physical information neural networks. The use of transfer learning can make the model more robust. For transient problems with low-fidelity data, the use of transfer learning can greatly improve the training rate through pre-training of a certain transient.

[0120] S104, training the neural network model according to the fused feature vector until the corresponding loss function meets the threshold condition;

[0121] Based on the trained neural network model, the operating status of the cable channel to be tested can be determined quickly and accurately. This method not only improves the efficiency of status determination, but also reduces the misjudgment rate caused by human factors, providing strong technical support for the maintenance and management of the cable channel.

[0122] Preferably, transfer learning improves the accuracy of the physical information neural network when the number of samples is insufficient by means of knowledge transfer, feature transfer, parameter transfer, etc. The present application uses the CORAL transfer learning algorithm to map the feature parameters. The basic principle of CORAL is to maximize the correlation between the source domain and the target domain, enhance the similarity distribution of the feature space, and thus reduce the distance between the two domains.

[0123] In some embodiments, the neural network model is trained according to the fused feature vector until the corresponding loss function meets the threshold condition, specifically including:

[0124] The CORAL transfer learning algorithm is used to map the feature parameters and obtain the covariance distance between the source domain and target domain features of the CORAL function:

[0125]

[0126] Among them, C S and C T Represent the covariance matrices of the source domain and the target domain respectively, and d represents the number of neural network outputs;

[0127] When the covariance distance is within the preset range, the training of the neural network model is completed.

[0128] 16 characteristic parameters of the known cable operation channel state are obtained, and they are divided into three operation states: normal, warning and alarm according to the actual operation state. The three operation state evaluation criteria are obtained from the experience of operation and maintenance personnel. The obtained data is divided into training set and test set according to a certain ratio, and input into the improved physical information neural network model to realize the state perception of dense urban cable channels.

[0129] S105: Determine the operating status of the cable channel to be detected based on the trained neural network model.

[0130] The entire perception process has achieved intelligence and automation from data collection, processing to status determination, reducing manual intervention and improving work efficiency. At the same time, it also provides possibilities for remote monitoring and intelligent operation and maintenance of cable channels, helping to improve the intelligence level of the entire power system.

[0131] like Figure 2 As shown, in some embodiments, under different laying modes and operating conditions of cable channels, 100 sets of data containing 16 parameters are obtained, 70 sets of data are set as training sets, and 30 sets are set as test sets to perform state perception of dense urban cable channels. The perception results are shown in FIG. Figure 2 As shown in the figure, the horizontal axis represents the test set sample number, and the vertical axis represents the classification label. It can be seen that the proposed method can accurately perceive the three operating states of normal, warning and alarm. The three operating states are represented by classification labels 1, 2, and 3 respectively. The accuracy of urban dense cable channel state perception is 96.7%, so the proposed method has good effect. In addition, the method has good scalability and flexibility. It can adjust the collected feature parameters and optimize the structure and parameters of the neural network model according to actual needs to adapt to the cable channel state perception tasks of different scenarios and needs.

[0132] In some embodiments, determining the operating state of the cable channel to be detected based on the trained neural network model specifically includes:

[0133] The data of the cable channel to be detected is input into the trained neural network model to determine the operating status of the cable channel. The operating status of the cable channel includes normal status, warning status and alarm status.

[0134] Preferably, the determination conditions of the three states: normal state, warning state and alarm state can be preset manually.

[0135] In summary, the present application proposes a method for sensing the state of a cable channel, which takes into account 16 parameters, including cable sheath current, cable core current, cable intermediate joint temperature, cable partial discharge current, cable current carrying capacity, grounding current, vibration information, ambient temperature, humidity, carbon monoxide concentration, hydrogen sulfide concentration, oxygen concentration, water level, smoke concentration, vibration parameters, and external damage parameters. It combines PCA and an improved physical information neural network algorithm to realize the state perception of dense urban cable channels and accurately perceive the three operating states of normal, warning, and alarm. Among them, PCA can compress the feature quantities of all signals. It is a fusion method that uses the idea of ​​dimensionality reduction and is adaptive. The original data is converted into a set of new sample data through transformation and reconstruction. Although the obtained data has a low dimension, it can also reflect most of the characteristic information of the original data, which is conducive to subsequent state perception calculations. It can process new monitoring data in real time and accurately to realize the state perception of dense urban cable channels.

[0136] The present application also proposes a cable channel state perception system, the system comprising: a fusion feature vector acquisition unit, a neural network model construction unit, a neural network model training unit and a cable channel state perception unit;

[0137] A fusion feature vector acquisition unit is used to collect several feature parameters of the cable channel operation status in real time; perform dimensionality reduction processing on several feature parameters to obtain a fusion feature vector;

[0138] A neural network model building unit, used to determine the corresponding partial differential equations and loss functions;

[0139] A neural network model training unit, used to train the constructed neural network model based on the fused feature vector until the corresponding loss function meets the threshold condition;

[0140] The cable channel state sensing unit is used to determine the operating state of the cable channel to be detected based on the trained neural network model.

[0141] The present application also proposes a readable storage medium storing a computer program, which, when executed by a processor, causes the processor to execute the following steps: collecting several characteristic parameters of the operating status of the cable channel in real time;

[0142] Perform dimensionality reduction processing on several feature parameters to obtain a fused feature vector;

[0143] Build a neural network model and determine the corresponding partial differential equation and loss function;

[0144] The neural network model is trained according to the fused feature vector until the corresponding loss function meets the threshold condition;

[0145] The operating status of the cable channel to be detected is determined based on the trained neural network model.

[0146] The present application also proposes a computer device, including a memory and a processor, wherein the memory stores a computer program, and the computer program is executed by the processor to perform the following steps:

[0147] Collect several characteristic parameters of the cable channel operation status in real time;

[0148] Perform dimensionality reduction processing on several feature parameters to obtain a fused feature vector;

[0149] Build a neural network model and determine the corresponding partial differential equation and loss function;

[0150] The neural network model is trained according to the fused feature vector until the corresponding loss function meets the threshold condition;

[0151] The operating status of the cable channel to be detected is determined based on the trained neural network model.

[0152] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in the present application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0153] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0154] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the patent of the present application. It should be pointed out that, for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. The above disclosure is only a preferred embodiment of the present invention, and of course it cannot be used to limit the scope of the rights of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope covered by the present invention.

Claims

1. A method for sensing cable channel status, characterized in that: The method comprises: Collect several characteristic parameters of the cable channel operation status in real time; Performing dimensionality reduction processing on the plurality of feature parameters to obtain a fused feature vector; Build a neural network model and determine the corresponding partial differential equation and loss function; Training the neural network model according to the fused feature vector until the corresponding loss function meets a threshold condition; The operating status of the cable channel to be detected is determined based on the trained neural network model.

2. The cable channel status perception method according to claim 1, characterized in that: The real-time acquisition of several characteristic parameters of the cable channel operation status specifically includes: Obtain 16 characteristic parameters under different laying modes of cable channels: cable sheath current, cable core current, cable intermediate joint temperature, cable partial discharge current, cable current carrying capacity, grounding current, vibration information, ambient temperature, humidity, carbon monoxide concentration, hydrogen sulfide concentration, oxygen concentration, water level, smoke concentration, vibration parameters, and external damage parameters; Construct a vector sequence based on the 16 characteristic parameters: X={x1,x2,…,x n } m×n Among them, n is 16, which represents the number of samples, m represents the vector dimension, and x i represents the i-th feature parameter.

3. The cable channel status perception method according to claim 2, characterized in that: The performing dimensionality reduction processing on the plurality of feature parameters to obtain a fused feature vector specifically includes: Decentralize the feature parameters and obtain the covariance matrix: Among them, A represents an M×M dimensional symmetric matrix, X represents a vector sequence, and X T Represents the transposed matrix of a vector sequence; Perform eigendecomposition on the matrix A: A=Q∑Q -1 Among them, the vectors contained in the matrix Q are the eigenvectors of the square matrix A, ∑ represents an M×M dimensional diagonal matrix, and the values ​​on the diagonal are the eigenvalues ​​λ i , where i = 1, 2, ... m; Preset λ1≥λ2≥…≥λ m , select the eigenvector a where the first k eigenvalues ​​are located i Construct the projection matrix: <h2 style=";text-align:left;direction:ltr">W=[a1,a2,…,a<h2 style=";text-align:left;direction:ltr"> k <h2 style=";text-align:left;direction:ltr"> ] Among them, a i represents an m-dimensional column vector; Select the first k characteristic roots to determine the contribution coefficient K t ; The feature space is determined by k feature vectors, and the original data is projected into the feature space to obtain a sample of the merged features: Y=W T X Among them, Y represents the fused feature vector, W T represents the transposed matrix of the projection matrix, and X represents a vector sequence.

4. The cable channel status perception method according to claim 3, characterized in that: The said constructing a neural network model and determining the corresponding partial differential equation and loss function specifically include: in, and Ω represent the boundary and spatial domain respectively, u(Y,t) represents the spatial variable Y(y1,y2,…,y k ) is the solution of the domain Ω and the time variable t in the time domain, I(Y) and B(Y,t) represent the initial conditions and boundary conditions respectively, N represents the partial differential operator with respect to the spatial variable, and λ represents the coefficient related to the physical information; The loss function consists of the initial conditions, boundary conditions, and the residuals of the governing equations: L(θ)=ω i L i (θ)+ω b L b (θ)+ω r L r (i) Among them, ω i ,ω b ,ω r They represent the weights of the initial conditions, boundary conditions and loss function of the control equation respectively.

5. The cable channel status perception method according to claim 4, characterized in that: The loss function of the initial conditions, boundary conditions and control equations is: Among them, θ represents all parameters in the neural network model, Y and t represent the input, and u θ Represents the output value of the neural network model under the parameter θ, L i (θ) represents the loss function of the initial condition, L b (θ) Loss function of boundary conditions, L r (θ) The loss function governing the equation.

6. The cable channel status perception method according to claim 5, characterized in that: The training of the neural network model according to the fused feature vector until the corresponding loss function meets a threshold condition specifically includes: The feature parameters are mapped using the CORAL transfer learning algorithm to obtain the covariance distance between the source domain and target domain features of the CORAL function: Among them, C S and C T Represent the covariance matrices of the source domain and the target domain respectively, and d represents the number of neural network outputs; When the covariance distance is within a preset range, the training of the neural network model is completed.

7. The cable channel status perception method according to claim 6, characterized in that: The step of determining the operating state of the cable channel to be detected based on the trained neural network model specifically includes: The data of the cable channel to be detected is input into the trained neural network model to determine the operation status of the cable channel, wherein the operation status of the cable channel includes a normal state, a warning state and an alarm state.

8. A cable channel status perception system, characterized in that: The system comprises: a fusion feature vector acquisition unit, a neural network model construction unit, a neural network model training unit and a cable channel state perception unit; The fused feature vector acquisition unit is used to collect several feature parameters of the cable channel operation status in real time; perform dimensionality reduction processing on the several feature parameters to obtain a fused feature vector; The neural network model building unit is used to determine the corresponding partial differential equation and loss function; The neural network model training unit is used to train the constructed neural network model based on the fused feature vector until the corresponding loss function meets a threshold condition; The cable channel state sensing unit is used to determine the operating state of the cable channel to be detected based on the trained neural network model.

9. A readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor is caused to execute the steps of the method according to any one of claims 1 to 7.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method according to any one of claims 1 to 7.

Images