Fault early warning model construction method, fault early warning method and device

Abstract

The invention discloses a fault early warning model construction method and a fault early warning method and device, and belongs to the technical field of power cable joint temperature fault prediction, the fault early warning model construction method is used for cable joint temperature fault early warning, and comprises the following steps: obtaining a plurality of cable environment data and cable joint core temperatures corresponding to the cable environment data; performing dimension reduction on the cable environment data to obtain two-dimensional environment parameters; training a support vector machine according to the two-dimensional environment parameters; and optimizing a C parameter and a gamma parameter of the support vector machine through a whale optimization algorithm to obtain a cable joint temperature fault early warning model. According to the construction method of the fault early warning model, the fault early warning model used for cable joint temperature fault early warning can be constructed, and the problem that early warning of power cable joint temperature abnormity is not accurate enough is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power cable joint temperature fault prediction, and in particular relates to a fault warning model construction method, a fault warning method and a device. Background Art

[0002] With the acceleration of my country's urbanization process and the expansion of the scale of power grid construction, power cables, as a key component of the power grid, undertake important communication and power transmission functions.

[0003] At the same time, with the continuous expansion of urban construction, some distribution network cable lines have excessive load or fault current, resulting in excessive cable temperature. According to statistics, 10kV distribution network cable joints and terminals are the weakest parts of the cable. About 80% of cable failures are caused by joint or terminal failures. The main reason is that the manufacturing process is not sophisticated, and the finished products produced have loose joints, excessive contact resistance and uneven products. At the same time, when the city is flooded, various acidic and alkaline liquids are poured into the cable trench, which will also corrode the intermediate joints of the cable, causing insulation breakdown, explosion and fire and other safety accidents.

[0004] In this regard, a Chinese patent with publication number CN118408646B discloses a method and system for warning of abnormal temperature of power cable joints. Based on the historical temperature data of the internal core of the power cable joint, the system uses a trained EEMD-BiLSTM temperature prediction model to more accurately obtain the temperature prediction value at the next time point. By considering the surface temperature of the power cable joint and the heat transfer and heat radiation between the surface temperature and core temperature of the power cables connected on both sides and the internal core temperature of the power cable joint, a core temperature correction formula for the internal core temperature of the power cable is constructed to correct the temperature prediction value at the next time point, thereby obtaining an accurate predicted temperature, and providing an accurate warning based on the predicted temperature combined with a preset threshold and a deviation formula.

[0005] However, the current warning for abnormal temperature of power cable joints is not accurate enough, so there is an urgent need to develop a fault warning model construction method, a fault warning method and a device to solve the problems in the prior art. Summary of the invention

[0006] The purpose of the present invention is to provide a fault warning model construction method, a fault warning method and a device with good accuracy to solve the problem that the current warning of abnormal temperature of power cable joints is not accurate enough as mentioned in the above background technology.

[0007] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows:

[0008] A method for constructing a fault warning model for cable joint temperature fault warning includes the following steps:

[0009] Obtaining a number of cable environmental data and cable joint core temperatures corresponding to the cable environmental data;

[0010] Reduce the dimension of cable environmental data to obtain two-dimensional environmental parameters;

[0011] Optimize the C parameter and gamma parameter of the support vector machine through the whale optimization algorithm;

[0012] According to the two-dimensional environmental parameters, the support vector machine is trained to obtain the cable joint temperature fault warning model.

[0013] Furthermore, the cable environmental data includes cable load, ambient temperature and cable outer sheath temperature.

[0014] Furthermore, the cable joint core temperature corresponding to the cable environment data is the cable joint core simulation temperature, and the acquisition of the cable joint core simulation temperature includes the following steps:

[0015] Acquire cable model data, and construct a cable connector core model according to the cable model data;

[0016] According to the cable environment data and the cable joint core model, the cable joint core simulation temperature is obtained;

[0017] Among them, the cable joint core model is constructed by finite element simulation software; the cable model data includes cable joint structural parameters and cable joint core loss parameters, and the cable joint core loss parameters include conductor core loss, insulation layer dielectric loss, metal shielding layer loss and armor layer loss.

[0018] Furthermore, before obtaining the cable joint wire core simulation temperature according to the cable environment data and the cable joint wire core model, the method further includes: preprocessing the cable environment data;

[0019] Wherein, the preprocessing of the cable environment data comprises the following steps:

[0020] Impute missing values ​​in cable environment data;

[0021] Detect and handle outliers in cable environmental data;

[0022] Normalized cable environment data;

[0023] The missing value is estimated by the following formula:

[0024]

[0025] Where x(t) is the estimated missing value, t1 and t2 are two known data points adjacent to the missing value, and t1 and t2 are the times corresponding to the data points;

[0026] The detection and processing of outliers in cable environment data comprises the following steps:

[0027] Calculate the first quartile and the third quartile, and calculate the interquartile range based on the first quartile and the third quartile;

[0028] According to the interquartile range, define the outlier range and obtain the outlier value;

[0029] Replace the resulting outliers with the median.

[0030] Furthermore, the dimensionality reduction of the cable environment data to obtain two-dimensional environment parameters comprises the following steps:

[0031] The cable load, ambient temperature and cable outer sheath temperature are reduced in dimension by principal component analysis, so that the cable load, ambient temperature and cable outer sheath temperature are converted into two-dimensional environmental parameters, wherein the two-dimensional environmental parameters include two types of data.

[0032] Further, the optimization of the C parameter and the gamma parameter of the support vector machine by the whale optimization algorithm comprises the following steps:

[0033] Set algorithm parameters and randomly generate initial individuals;

[0034] Calculate individual fitness and record the best individual;

[0035] Determine whether the algorithm has reached the maximum number of iterations. If so, output the optimal solution. If not, the following steps are included:

[0036] The generation strategy controls the random number p;

[0037] Update the step length control parameter a, and update the step length control parameter a and parameter C according to the formula;

[0038] Determine whether p is less than 0.5; when p is less than 0.5, the spiral mechanism updates the individual position; when p is not less than 0.5, determine whether |A| is less than 1; when |A| is less than 1, the search mechanism updates the individual position; when |A| is not less than 1, the encirclement mechanism updates the individual position;

[0039] Determine again whether the algorithm has reached the maximum number of iterations. If not, perform the above steps again.

[0040] Further, the optimization of the C parameter and the gamma parameter of the support vector machine by the whale optimization algorithm comprises the following steps:

[0041] Set algorithm parameters and randomly generate multiple initial individuals of the population;

[0042] Calculate the fitness of each individual in the population and record the best individual in the best population;

[0043] Determine whether the algorithm has reached the maximum number of iterations. If so, output the optimal solution. If not, perform the following steps:

[0044] Adjust the search step size;

[0045] The generation strategy controls the random number p;

[0046] Update the step length control parameter a, and update the step length control parameter a and parameter C according to the formula;

[0047] Determine whether p is less than 0.5; when p is less than 0.5, the spiral mechanism updates the individual position; when p is not less than 0.5, determine whether |A| is less than 1; when |A| is less than 1, the search mechanism updates the individual position; when |A| is not less than 1, the encirclement mechanism updates the individual position;

[0048] Share the best individual population information;

[0049] Determine again whether the algorithm has reached the maximum number of iterations. If not, perform the above steps again.

[0050] A fault warning method for cable joint temperature fault warning, based on a cable joint temperature fault warning model, comprises the following steps:

[0051] Acquire measured data of cable environment; wherein the measured data of cable environment includes measured cable load, ambient temperature and cable outer sheath temperature;

[0052] According to the constructed cable joint temperature fault warning model and the measured data of the cable environment, the cable joint temperature warning data is obtained.

[0053] A computer device comprises a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the fault warning model construction method or the fault warning method.

[0054] A computer program product comprises a computer program, which implements the steps of the fault warning model construction method or the fault warning method when executed by a processor.

[0055] The present invention has the following advantages:

[0056] The support vector machine has a good judgment ability for cable joint temperature fault warning. The present application reduces the dimension of the cable environment data to obtain two-dimensional environmental parameters to adapt the support vector machine, so that the cable joint temperature fault warning model of the present application has good accuracy.

[0057] Furthermore, the C parameter and gamma parameter of the support vector machine are optimized through the whale optimization algorithm to further improve the accuracy of the cable joint temperature fault warning model.

[0058] At the same time, the cable environment data is preprocessed to improve the accuracy of subsequent training steps to further improve the accuracy of the cable joint temperature fault warning model.

[0059] Furthermore, the cable joint structure parameters and the cable joint core loss parameters are used to construct a cable joint core model, so as to obtain the cable joint core temperature corresponding to the cable environment data.

[0060] At the same time, the common whale optimization algorithm is used to optimize the C parameter and gamma parameter of the support vector machine to further improve the accuracy of the cable joint temperature fault warning model.

[0061] Furthermore, the present application can further improve the accuracy of the model and the iteration speed of the model through the improved multi-group whale optimization algorithm based on the ordinary whale optimization algorithm.

[0062] Other features and advantages of the present invention will be disclosed in detail in the following specific embodiments and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a schematic diagram of the overall process of the present invention;

[0064] Figure 2 The cable structure and property table of the first embodiment;

[0065] Figure 3 This is a graph of the ambient temperature data after the completion of Example 1;

[0066] Figure 4 This is a cable load data diagram after completion of Example 1;

[0067] Figure 5 This is a graph of the outer sheath temperature data after the completion of Example 1;

[0068] Figure 6 It is a two-dimensional geometric model diagram of a three-core cable of Embodiment 1;

[0069] Figure 7 This is a three-core cable grid division diagram of Example 1;

[0070] Figure 8 This is a three-core cable operation loss diagram of Example 1;

[0071] Fig. 9 This is a simulation diagram of the steady-state temperature field of a three-core cable in Example 1;

[0072] Fig.10 This is a classification diagram of the SVM classifier of Example 1;

[0073] Fig.11 This is a diagram of the classification accuracy of the SVM classifier before parameter adjustment in Example 1;

[0074] Fig.12 This is a schematic diagram of the whale optimization process of Example 1;

[0075] Fig.13 The search process and fitness relationship diagram of parameters C and gamma in Example 1;

[0076] Fig.14 This is a graph showing the relationship between the number of iterations after whale optimization in Example 1 and time;

[0077] Fig.15 This is a diagram showing the classification accuracy of the prediction model after whale optimization in Example 1;

[0078] Fig.16 This is a schematic diagram of the multi-group whale optimization process of Example 2;

[0079] Fig.17 This is a comparison chart of classification accuracy after whale optimization in Example 1 and Example 2;

[0080] Fig.18 This is a diagram showing the classification accuracy of the model after optimization of multiple groups of whales in Example 2;

[0081] Fig.19 The eigenvalue, variance contribution rate and cumulative contribution rate diagram of the first embodiment;

[0082] Fig. 20 It is a feature vector graph extracted by principal component analysis in Example 1;

[0083] Fig.21 This is the principal component biplot of Example 1;

[0084] Fig. 22 This is a relationship diagram between the ambient temperature and the cable temperature when the cable load is 320A in Example 1;

[0085] Fig.23 Graph showing the relationship between cable load and cable temperature when the ambient temperature is 40° C. in Example 1. DETAILED DESCRIPTION

[0086] In order to better understand the purpose, structure and function of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings.

[0087] Embodiment 1

[0088] A cable joint temperature fault early warning method, such as Figure 1 As shown, the following steps are included:

[0089] S1: Acquire a number of cable environmental data and the cable joint core temperature corresponding to the cable environmental data;

[0090] In this embodiment, the cable environmental data includes cable load, ambient temperature and cable outer sheath temperature.

[0091] S2: Reduce the dimension of cable environmental data to obtain two-dimensional environmental parameters.

[0092] S3: Optimize the C parameter and gamma parameter of the support vector machine through the whale optimization algorithm.

[0093] S4: According to the two-dimensional environmental parameters, a support vector machine is trained to obtain a cable joint temperature fault warning model.

[0094] Specifically, in this embodiment, the cable joint core temperature corresponding to the cable environment data in S1 is a cable joint core simulation temperature, and S1 obtains a number of cable environment data and the cable joint core temperature corresponding to the cable environment data, including the following steps:

[0095] S11: Obtain some cable environment data;

[0096] S12: Acquire cable model data, and construct a cable connector core model according to the cable model data;

[0097] S13: Obtain the cable joint core simulation temperature according to the cable environment data and the cable joint core model.

[0098] In this embodiment, the cable environmental data in S11 is obtained from actual measurement. Specifically, the cable environmental data comes from the actual measured data of a cable well in a certain place in July 2023. The test time is 28 days. The data includes cable load, ambient temperature and cable outer sheath temperature. The cable is a three-core cross-linked polyethylene cable with a rated current carrying capacity of Imax=640A and a YJV22-8.7 / 15kV-3x500mm2. The metal shielding layer is a copper tape shielding layer and the armor layer is a steel tape armor layer.

[0099] Optionally, since there are missing and outliers in the cable environmental data during the actual measurement and collection process, before S13 obtains the simulated temperature of the cable joint core according to the cable environmental data and the cable joint core model, it also includes: preprocessing the cable environmental data.

[0100] The preprocessing of the cable environment data comprises the following steps: completing the cable environment data.

[0101] Specifically, the method of completing the cable environment data comprises the following steps:

[0102] Impute missing values ​​in cable environment data;

[0103] Detect and handle outliers in cable environmental data.

[0104] The missing values ​​in the cable environment data can be estimated by the following formula:

[0105]

[0106] Where x(t) is the estimated missing value, x(t1) and x(t2) are two known data points adjacent to the missing value, and t1 and t2 are the times corresponding to the data points.

[0107] For outliers in the data, the quartile method (IQR) is used for detection and processing. First, the first quartile Q1 and the third quartile Q3 of the data are calculated, and then the interquartile range IQR is calculated.

[0108] IQR=Q3-Q1.

[0109] The outlier range is defined according to the interquartile range. Any data point less than Q1-1.5IQR or greater than Q3+1.5IQR is considered an outlier. For detected outliers, the median is used for replacement to reduce its impact on the overall trend of the data. The specific processing formula is as follows:

[0110] x ”new” =”median”(X);

[0111] Among them, x ”new” is the value of the outlier after replacement; median is the median; X is the cable environment data.

[0112] like Figure 3-Figure 5 The following table shows the ambient temperature data, cable load data and outer sheath temperature data after completion of this application. The horizontal axis is the acquisition sequence, and the vertical axis is the corresponding ambient temperature, cable load and outer sheath temperature data, in units of ℃, A and ℃ respectively.

[0113] Optionally, the preprocessing of the cable environment data further includes the following steps:

[0114] Normalized cable environment data: In this embodiment, the normalized cable environment data is obtained after the cable environment data is completed.

[0115] This application normalizes the data to avoid the influence of different dimensions for subsequent processing. Specifically, the normalization formula is as follows:

[0116]

[0117] Among them, x′ is the normalized eigenvalue, x is the original eigenvalue, max(x) and min(x) are the minimum and maximum eigenvalues, respectively. Through normalization, the eigenvalues ​​of the cable environment data are scaled to the interval [0, 1], so that different features have the same dimension, which helps to improve the prediction performance of the model.

[0118] The cable connector core model in S12 is constructed by finite element simulation software. In this embodiment, COMSOL Multiphysics multi-physics simulation software is used to construct the cable connector core model. Specifically, the step of using COMSOL Multiphysics to construct the cable connector core model according to the cable model data is prior art and will not be repeated in this application.

[0119] The cable model data in S12 are the parameters of various components required to construct the cable joint core model through COMSOL Multiphysics multi-physics field simulation software, including cable joint structure parameters and cable joint core loss parameters. In this embodiment, the cable joint structure parameters are as follows: Figure 2 The cable connector core model is as shown. Figure 6-9 shown.

[0120] Specifically, the cable joint core loss parameters include conductor core loss, insulation layer dielectric loss, metal shielding layer loss and armor layer loss. The conductor core loss, metal shielding loss and metal shielding layer loss are mainly calculated in the eddy current field. The conductor loss is mainly related to the current applied to the conductor. The shielding loss and armor loss are the eddy current losses induced in the metal shielding during the process of the magnetic field changing with the current. The insulation layer dielectric loss is related to the parameters such as the AC voltage applied to both ends of the cable insulation layer, the dielectric constant of the insulation layer and the dielectric loss tangent, and is obtained by solving the AC electric field.

[0121] In this embodiment, the calculation formula of the conductor core loss per unit length is as follows:

[0122] Q c =I 2 R;

[0123] Among them, Q c is the cable conductor core loss per unit length, in W / m; I is the load current flowing through the cable conductor core, in A; R is the AC resistance value of the conductor core per unit length, in Ω / m.

[0124] At working temperature, the AC resistance R per unit length of conductor core can be calculated by the following formula:

[0125] R=R′(1+Y s +Y P );

[0126] Where R is the AC resistance of the conductor core per unit length, R′ is the DC resistance of the conductor core per unit length, in Ω / m; Y s is the skin effect factor; Y P is the proximity effect factor.

[0127] In this embodiment, the DC resistance value R′ per unit length of the conductor core can be calculated by the following formula:

[0128]

[0129] Where R′ is the DC resistance per unit length of the conductor core; ρ 20 is the resistivity of the core material at 20°C. For standard soft copper, ρ 20 =0.17241×10 -6 Ω·m; A is the cross-sectional area of ​​the conductor core, in m 2 ; α is the resistance temperature coefficient of the core material at 20°C, for copper conductor α=0.00393°C -1 ; θ is the temperature of the conductor core, in °C; k1 is used for the core formed by twisting multiple conductors, generally 1.02 (250mm 2 Below)~1.03(250mm 2 and above); k2 is the conductor cabling coefficient, which is generally taken as 1.01; k3 is the coefficient introduced by the conductor hardening and increasing resistivity during the process of pressing the core, which is generally taken as 1.01.

[0130] In this embodiment, the skin effect factor Y s It can be calculated by the following formula:

[0131]

[0132] Among them, Y s is the percentage of resistance increase due to skin effect; f is the frequency of AC power supply, in Hz; R′ is the DC resistance value of the conductor core per unit length; k sis a constant. The conductor cores of different structures have different values. For round, compacted, non-dry impregnated conductors, k s =1.

[0133] In this embodiment, the proximity effect factor Y P It can be calculated by the following formula:

[0134]

[0135] Among them, Y P is the percentage of resistance increase due to proximity effect; d c is the diameter of the cable conductor core, in mm; s is the distance between the cable conductor core axes, in mm; f is the AC power frequency, in Hz; R′ is the DC resistance value of the conductor core per unit length; k p is a constant. The conductor cores of different structures have different values. For round, compacted, non-dry impregnated conductors, k p =1.

[0136] The dielectric loss of the cable insulation layer mainly refers to the energy loss caused by dielectric conductivity and dielectric hysteresis effect under the action of alternating voltage. The dielectric loss of the insulation layer of the cable per unit length (per phase) is:

[0137] Q d =ωcU 2 tanδ;

[0138] Among them, Q d is the dielectric loss of the insulation layer of the cable (per phase), ω=2πf, in this embodiment, f is 50Hz; c is the cable capacitance per unit length, in F / m; U is the cable phase voltage, in V; tanδ is the loss factor of the insulation material, which is related to the voltage level and the insulation material, and is generally taken as 0.004.

[0139] The calculation formula for cable capacitance per unit length is as follows:

[0140]

[0141] Where c is the cable capacitance per unit length, in F / m; ε is the dielectric constant of the insulating material, which is generally 2.5 F / m for cross-linked polyethylene material; D i is the diameter of the cable insulation layer, in mm; d C It is the cable core diameter in mm.

[0142] For a three-core cable in normal operation, the metal shielding layer loss mainly refers to the eddy current loss. In this embodiment, the calculation formula for the metal shielding layer (three-phase) loss per unit length is as follows:

[0143] Q e =3λ1qC ;

[0144] Among them, Q e is the metal shielding layer loss per unit length; λ1 is the loss factor of the metal shielding layer of a single core; Q c It is the core loss of the cable conductor per unit length.

[0145] The calculation formula for the loss factor λ1 of a single core metal shield is as follows:

[0146]

[0147] Among them, R′ cu is the resistance of the metal shielding layer per unit length, in Ω / m; R is the AC resistance of the conductor core per unit length; X is the reactance of the metal shielding layer per unit length, in Ω / m.

[0148] The calculation formula for the reactance X of the metal shielding layer per unit length is as follows:

[0149]

[0150] Wherein, X is the reactance of the metal shielding layer per unit length, in Ω / m; ω=2πf, in this embodiment, f is 50 Hz; d is the average diameter of the metal shielding layer, in mm; s is the distance between the core axes of the cable conductors.

[0151] The armor layer loss mainly includes two parts, hysteresis loss and eddy current loss, which can be obtained by calculating the ratio of armor layer loss to conductor loss. In this embodiment, the armor layer loss per unit length is calculated by the following formula:

[0152] Q f =3λ2Q c ;

[0153] Among them, Q f is the armor loss per unit length; λ2 is the loss factor of the armor layer of a single core; Q c It is the core loss of the cable conductor per unit length.

[0154] The calculation formula for the loss factor of a single core armor layer is as follows:

[0155] λ2=λ′2+λ″2;

[0156] Among them, λ2 is the loss factor of a single core armor layer; λ′2 is the hysteresis loss factor; λ″2 is the eddy current loss factor.

[0157] Under the condition of AC 50Hz, the calculation formulas of λ′2 and λ″2 are as follows:

[0158]

[0159] Where, λ′2 is the hysteresis loss factor; s is the distance between the core axes of the cable conductors, in mm; K is the coefficient; R is the AC resistance per unit length of the conductor core; d A is the average diameter of the armor layer, in mm; σ is the equivalent thickness of the armor layer, in mm.

[0160]

[0161] Among them, λ″2 is the eddy current loss factor; s is the distance between the core axes of the cable conductor; K is the coefficient; σ is the equivalent thickness of the armor layer; R is the AC resistance value of the conductor core per unit length; d A is the average diameter of the armor layer;

[0162] In this embodiment, the coefficient K is calculated by the following formula:

[0163]

[0164] Among them, K is the coefficient; d A is the average diameter of the armor layer, in mm; μ is the relative magnetic permeability of the steel belt, usually taken as 300; σ is the equivalent thickness of the armor layer, in mm.

[0165] In this embodiment, the boundary condition of the model is a constant temperature boundary condition, that is, it is assumed that the temperature of the air flowing near the cable remains constant.

[0166] The present application constructs a cable joint core model to obtain the simulated temperature of the cable joint core, and constructs the cable joint core model through the cable joint structure parameters and the cable joint core loss parameters. The cable joint core loss parameters include conductor core loss, insulation layer dielectric loss, metal shielding layer loss and armor layer loss, so as to correspond to the cable load, ambient temperature and cable outer sheath temperature, thereby enhancing the accuracy of the cable joint core model and improving the accuracy of the early warning model.

[0167] In the S13, the cable joint core simulation temperature is obtained according to the cable environment data and the cable joint core model, including the following steps:

[0168] The cable environment data is input into the cable joint core model, and the cable joint core model outputs the cable joint core simulated temperature according to the cable environment data. Specifically, the cable joint core simulated temperature obtained after inputting the actual data into the cable joint core model is a prior art used by finite element simulation software, and this application will not repeat it.

[0169] After obtaining the simulated temperature of the cable joint core in S13, the following steps are also included:

[0170] S14: Verify the simulated temperature of the cable connector core.

[0171] In this embodiment, the simulated temperature of the cable joint core is verified by the control equation of temperature field heat transfer.

[0172] The expression of temperature field is as follows:

[0173] T = f(xy,z,t);

[0174] Wherein, T represents the temperature of the point with coordinates (x, y, z) at time t, (x, y, z) are the spatial rectangular coordinates, and t is the time.

[0175] According to the relationship between temperature field and time, when the temperature distribution does not change with time, it is called a steady-state temperature field, and when the temperature distribution changes with time, it is called a transient temperature field. According to the relationship between temperature field and space, the temperature field can be divided into one-dimensional, two-dimensional and three-dimensional temperature fields.

[0176] Since the length of the cable body in actual engineering is much larger than the cable cross-section, and the cable materials are all isotropic homogeneous media, the heat transfer problem of the cable body in this embodiment is represented by a two-dimensional model. According to the energy conservation and Fourier's law, the heat conduction control equation in the plane can be obtained:

[0177]

[0178] Among them, Qv is the heat generation rate of the heat source per unit area, in W / m2; λ is the thermal conductivity of the material, in W / (m·K); T is the temperature value in the area, in K; x and y are coordinates; p is the mass per unit area of ​​the material, in kg / m2; w is the specific heat capacity of the material in the area, in J / (kg·K); t is the time, in s.

[0179] Specifically, the control equation of the steady-state temperature field of the area containing the internal heat source of the cable (conductor, insulation layer, shielding layer, etc.) is:

[0180]

[0181] Among them, Qv is the heat generation rate of the heat source per unit area, the unit is W / m2; λ is the thermal conductivity of the material, the unit is W / (m·K); T is the temperature value in the area, the unit is K; x, y are coordinates.

[0182] The control equation of the steady-state temperature field in the area without heat source is:

[0183]

[0184] Where T is the temperature value in the area, in K; x and y are coordinates.

[0185] When the calculated temperature value is close to the obtained cable joint wire core simulation temperature, it is verified that the obtained cable joint wire core simulation temperature is correct.

[0186] The principal component analysis method uses normalization processing and related matrix calculation to obtain a set of eigenvectors and corresponding eigenvalues. First, the number of principal components is determined based on the size of the eigenvalues ​​of the normalized data. At the same time, the number of principal components and the degree of interpretation of the data information are further determined by calculating the cumulative variance contribution rate of each feature. In this embodiment, the eigenvalues, variance contribution rates and cumulative contribution rates obtained are as follows: Fig.19 The eigenvectors extracted by principal component analysis are shown in Fig. 20 The principal component biplot is shown in Fig.21 , the sum of PC1 and PC2 is the cumulative variance contribution rate of the first two principal components, and the ambient temperature, outer sheath temperature, and cable load are the extracted eigenvectors. This application extracts feature data through principal component analysis (PCA), decomposes eigenvalues, compresses data, and denoises to further improve the information extraction rate of data and the effectiveness of data processing.

[0187] In this embodiment, the cable environment data includes three types, which are three-dimensional data, which are converted into two types through principal component analysis and reduced to two-dimensional data. Specifically, the principal component analysis method is a prior art and will not be described in detail in this application. Optionally, other methods can also be used to reduce the dimension to two-dimensional data.

[0188] Optionally, the principal component analysis method in this embodiment includes the following steps:

[0189] S21: Calculate the eigenvalues ​​and eigenvectors of the correlation matrix R of the characteristic parameters. The calculation formula is as follows:

[0190]

[0191] Among them, A is the sample matrix, A * is the standardized matrix after the original sample data is standardized; n is the total number of samples.

[0192] The sample matrix is ​​expressed by the following formula:

[0193] A=(a ij ) n×p ;

[0194] Among them, A is the sample matrix; a ij is the jth feature of the i-th sample; n is the total number of samples; p is the total number of features.

[0195] The standardized matrix is ​​expressed as follows:

[0196] A* =(a ij * ) n×p ;

[0197] Among them, A * It is the standardized matrix after the original sample data is standardized; and S j Represent the mean and variance of each feature respectively; n is the total number of samples; p is the total number of features.

[0198] In this embodiment, the calculated eigenvalues ​​of R are λ1≥λ2≥…≥λp, and the corresponding eigenvectors are μ1, μ2, …, μp.

[0199] S22: Calculate the contribution rate μi of the characteristic parameter variance and the contribution rate μ∑(m) of the cumulative variance.

[0200] Among them, the calculation formula of the contribution rate of characteristic parameter variance is as follows:

[0201]

[0202] The calculation formula for the contribution rate of cumulative variance is as follows:

[0203]

[0204] The cumulative variance contribution rate is used to determine the number of principal components. ∑ When (m) ≥ 80%, it can be considered that the first m principal components contain most of the information.

[0205] S23: Calculate the eigenvectors corresponding to each principal component, as follows:

[0206] U P×m =[U1,U2,…,U m ].

[0207] The principal component matrix can be expressed as follows:

[0208]

[0209] The support vector machine in S4 is a classification algorithm based on statistical learning theory and has been widely used in fault diagnosis. SVM uses a mapping function to map training data from input space to a high-dimensional feature space, and then finds the best classification hyperplane as the decision plane by maximizing the margin between different classes. SVM is suitable for classification and regression tasks, can process high-dimensional data, and is used to solve nonlinear classification problems.

[0210] For the binary classification problem, given a training sample data set D = {(x1, y1), (x2, y2), ..., (xm ,y m )}, where y i ={-1, +1}, a hyperplane can be used for classification, and its representation is:

[0211] g(x)=ω T x+b=0;

[0212] Among them, ω is the normal vector of the hyperplane, which determines its direction; b is the distance between the hyperplane and the origin.

[0213] If this hyperplane can completely separate the training samples correctly, and the distance between the sample closest to the hyperplane and the hyperplane is the largest, then this hyperplane is the optimal hyperplane. The decision function of the optimal hyperplane is:

[0214] f(x)=sgn(g(x))=sgn(ω T x+b);

[0215] Among them, sgn() is the sign function, ω is the normal vector of the hyperplane, b is the distance between the hyperplane and the origin, and it is 1 when the independent variable is positive and -1 when it is negative.

[0216] The distance from vector x to the classification surface g(x)=0 is |g(x)| / ||ω||. According to the above formula, making positive scale adjustments to the parameters ω and b will not affect the decision result. The above formula can be transformed into the following constraint conditions:

[0217] y i (ω T x i +b)≥1;

[0218] The samples from the hyperplane to both sides meet the boundary conditions, that is, the points where the above equation is true are support vectors. The sum of the distances from the hyperplane to the support vectors on both sides, that is, the classification interval, is Therefore, the solution process of the optimal hyperplane can be transformed into the following optimization problem:

[0219]

[0220] sty i [(ω·x i )+b]-1≥0, i=1, 2,...,N.

[0221] On this basis, a slack variable is introduced and reflected in the objective function in the form of a penalty term to solve the linear inseparable situation:

[0222]

[0223] sty i [(ω·xi )+b]-1+ξ i ≥0;

[0224] ξ i ≥0,i=1,2,…,N;

[0225] Among them, C is the regularization parameter, which indicates the tolerance to errors, ξ i is the relaxation factor. The greater the sample error, the greater the i The bigger.

[0226] The optimization problem in the above formula can be transformed into the saddle point problem of the Lagrangian functional. After derivation, the dual optimization problem of the generalized optimal classification surface can be obtained:

[0227]

[0228] 0≤α i ≤C,i=1,2,…,N;

[0229] Among them, α i α j is the Lagrange coefficient, y i y j is the label and category of the sample. Then the decision function of the generalized optimal classification surface is:

[0230]

[0231] When there is no classification hyperplane in the input sample space, further improvement is needed. The kernel method is introduced to map the original space to a high-dimensional space to make the samples separable:

[0232]

[0233] K represents the calculation result of the kernel function, and φ(·) represents the mapping function.

[0234] Then the decision function of the optimal hyperplane can be expressed as:

[0235]

[0236] where α is the solution to the following optimization problem:

[0237]

[0238] 0≤α i ≤C,i=1,2,…,N;

[0239] b can be obtained by satisfying the following support vector:

[0240]

[0241] In this embodiment, the two-dimensional environment parameters are input into the support vector machine SVM, and the support vector machine SVM is trained, including the following steps:

[0242] S41: obtaining a training set and a test set according to the two-dimensional environmental parameters and the cable connector core temperature corresponding to the two-dimensional environmental parameters.

[0243] Because the cables actually tested were all working cables and had no abnormal core temperature data, we artificially set the temperature of 100, 200 and 200 cables to be increased by 0.1-0.2 times, 0.2-0.3 times and 0.4-0.5 times as abnormal data, and divided them into three levels of warning. Correspondingly, from the comparison between the cable core temperature and the outer sheath temperature, we can see that due to the increase in the heat of the cable core, the outer sheath will also heat up accordingly. Fig. 22 The influence of ambient temperature on the cable temperature field shows that, when the cable structure is determined, the longitudinal temperature difference of the cable has nothing to do with the ambient temperature. Fig.23 The influence of load current on cable temperature field shows that the greater the load current, the faster the cable temperature rises and the greater the temperature rise. According to the above conclusion, the corresponding cable outer sheath temperature in the data set is also adjusted accordingly, thereby forming a fault data set. Due to the specific cable, this embodiment divides a total of 1459 data, 1021 of which are used as training sets and 438 as test sets.

[0244] S42: Train a support vector machine based on the training set and the test set.

[0245] Specifically, the steps of training the support vector machine according to the training set and the test set are prior art and will not be described in detail in this application. Fig.10 shown.

[0246] Among them, the decision function of the optimal hyperplane of the support vector machine is:

[0247] f(x)=sgn(g(x))=sgn(ω T x+b);

[0248] The kernel function of the support vector machine is:

[0249]

[0250] K(x, y) represents the calculation result of the kernel function, x and y are sample points in the original feature space, and φ(·) represents the mapping function. Because the cable joint core temperature has a linear relationship with the actual measured cable outer sheath and ambient temperature, the linear kernel (LinearKernel) is used in the kernel function selection of SVM, which has the best effect.

[0251] In the present invention, the search range of parameter C is: [0.01, 20], the step length is 0.01, and the search range of parameter gamma is: [0.001, 5], the step length is 0.001. If the parameter C and parameter gamma with the highest fitness are selected only by the grid search method, the parameter C and parameter gamma are 10 and 1 respectively.

[0252] The larger the penalty parameter C is, the more it is equivalent to penalizing the slack variable. We hope that the slack variable is close to 0, that is, the penalty for misclassification increases, tending to the situation where all the training sets are classified correctly. In this way, the accuracy is very high when testing the training set, but the generalization ability is weak. The smaller the C value is, the smaller the penalty for misclassification is, allowing fault tolerance, treating them as noise points, and the generalization ability is stronger. After derivation, we can get the dual optimization problem of the generalized optimal classification surface:

[0253]

[0254]

[0255] 0≤α i ≤C, i=1, 2, …, N;

[0256] Among them, α i is the Lagrange coefficient.

[0257] The red, blue and yellow labels are 1, 2 and 3 respectively. The classification report is as follows: Fig.11 As shown in the figure, f1 is the F1 score, which is an indicator used in statistics to measure the accuracy of a binary classification model. From the classification report, we can see that the accuracy of SVM classification success is only 87% after simple grid search optimization.

[0258] In order to improve the prediction success probability of the SVM classification network, the present invention uses the whale optimization algorithm to optimize the C and gamma parameters of the support vector machine. The whale optimization algorithm (WOA) specifically includes the following:

[0259] In WOA, the calculation formula for initializing the position of the whale population is as follows:

[0260] X i =lb+rand×(ub-lb);

[0261] Where: X i is the position of individual i, lb and ub are the lower and upper bounds of the search space, and rand is a random number between 0 and 1.

[0262] The WOA algorithm mainly goes through the prey search stage, the prey surround stage and the bubble net attack stage:

[0263] 1) Searching for prey

[0264] In this phase, the whale optimization algorithm simulates the behavior of whales swimming freely in the ocean, exploring a vast search space to find potential optimal solutions. The key to this phase is randomness, and the solutions (whales) in the algorithm move in a random manner within the search space to ensure that no possible areas are missed. This approach is similar to whales swimming randomly in the ocean in search of food sources. The position update of each solution is unpredictable, but the goal is to explore as many areas as possible to increase the chance of finding a high-quality solution. Random search is a method used in WOA for global search and jumping out of local optimality. Its mathematical representation is:

[0265] D=|C×X rand (t)-X(t)|;

[0266] X(t+1)=X rand (t)-A×D;

[0267] Among them, X rand (t) is a randomly selected whale (solution). It can be seen that the formula for random search is very similar to the formula for encircling prey. The only difference is that the direction is adjusted from the current optimal candidate solution to a random whale. Whether encircling prey and random search behaviors occur depends on the value of A. When |A|<1, encircling prey behavior is used; when |A|>1, random search behavior is used.

[0268] 2) Surrounding the prey

[0269] When the algorithm detects a promising solution, it enters the Surround Prey phase. In this phase, other solutions (whales) begin to focus on this potential best solution, performing a more purposeful search. This is similar to how a whale discovers a school of fish and begins to swim around them, preparing to prey. The position updates of the solutions become more precise and focused in this phase, and they will adjust their direction and position based on the location of the current best solution. The purpose of this phase is to narrow the search area and focus on finding the optimal solution in the most promising area.

[0270] In reality, humpback whales can identify the location of prey and surround it in a circle. However, since the optimal position in the search space is unknown, the WOA algorithm assumes that the current best candidate solution position is the target prey position. After defining the target prey position, other whales will try to surround the target prey position. The calculation formula for this process is as follows:

[0271] D=| C×X best (t)-X(t)|;

[0272] X(t+1)=X best (t)-A×D;

[0273] It gradually decreases from 2 to 0 during the iteration process. In the above formula, t represents the current number of iterations, and X best (t) represents the position of the best candidate solution before the generation, X(t+1) and X(t) represent the solutions of the generation and the next generation respectively. A and C are coefficient vectors that need to be calculated separately. The calculation formula is as follows:

[0274] A=2a1·r1;

[0275] C = 2·r2;

[0276] Among them, r1 and r2 represent two random numbers between 0 and 1. a1 is a number that decreases linearly from 2 to 0 with iterations and is expressed as:

[0277]

[0278] Where G represents the total number of iterations.

[0279] 3) Bubble Net Attack Phase

[0280] When the algorithm detects a promising solution, it enters the Surround Prey phase. In this phase, other solutions (whales) begin to focus on this potential best solution, performing a more purposeful search. This is similar to how a whale discovers a school of fish and begins to swim around them, preparing to prey. The position updates of the solutions become more precise and focused in this phase, and they will adjust their direction and position based on the location of the current best solution. The purpose of this phase is to narrow the search area and focus on finding the optimal solution in the most promising area.

[0281] In order to simulate the spiral process of the humpback whale, that is, the foaming net attack process, we established a spiral equation between the whale and the prey to simulate the spiral movement of the humpback whale. The formula and schematic diagram are as follows:

[0282] D′=|X*(t)-X(t)|;

[0283] X(t+1)=D′·e bl ·cos(2πl)+X * (t);

[0284] Among them, is a constant, usually set to 1. l is a random number between [-1,1].

[0285] In this embodiment, the whale optimization algorithm (WOA) is used to simulate the swimming and social behavior of whales to optimize and adjust the support vector machine model parameters to achieve a higher classification prediction accuracy. The algorithm flow is as follows: Fig.12 shown.

[0286] Specifically, the steps include:

[0287] S31: Set algorithm parameters and randomly generate initial individuals;

[0288] Among them, the optimization algorithm first generates a population and sets the search range. The C search range is: [0.01, 20], the step size is 0.01, and the gamma search range is [0.001, 5], the step size is 0.001, which corresponds to the batch generation of candidate solutions in the optimization problem.

[0289] S32: Calculate individual fitness and record the best individual.

[0290] S33: Determine whether the algorithm has reached the maximum number of iterations. If so, output the optimal solution. If not, the following steps are included:

[0291] The generation strategy controls the random number p;

[0292] Update the step control a, and update the step control parameter a and parameter C according to the formula;

[0293] Determine whether the strategy control random number p is less than 0.5; when the strategy control random number p is less than 0.5, the spiral mechanism updates the individual position; when the strategy control random number p is not less than 0.5, determine whether |A| is less than 1; when |A| is less than 1, the search mechanism updates the individual position; when |A| is not less than 1, the encirclement mechanism updates the individual position;

[0294] Determine again whether the algorithm has reached the maximum number of iterations. If not, perform the above steps again.

[0295] Specifically, calculate the fitness of all randomly generated individuals, record the parameters and fitness of the best individual, verify the candidate solutions in the corresponding optimization problem, and record the best solution.

[0296] The search process and fitness relationship of parameters C and gamma are as follows: Fig.13 As shown. Next, randomly select a whale, which is the solution, and then update the step control parameters a and C according to the formula. The formulas for A and C are:

[0297] A = 2ar1-a;

[0298] C = 2·r2;

[0299]

[0300] Where r1 and r2 are random numbers between 0 and 1, i is the current iteration number, and T is the maximum number of iterations.

[0301] During the evolution of the WOA algorithm, the number of random control patterns generated, P, can be used to control whether the algorithm moves toward local development or global exploration. The range of P is 0 to 1. When p ≥ 0.5, the spiral process of the humpback whale is simulated, that is, the foaming net attack process. We establish a spiral equation between the whale and the prey to simulate the spiral movement of the humpback whale and develop the current optimal solution individual. This corresponds to improving the accuracy of the optimal solution in the optimization problem. The formula is as follows:

[0302] D′=|X best (t)-X(t)|

[0303] X(t+1)=D′·e bl ·cos(2πl)+X best (t)

[0304] Among them, b is a constant, generally set to 1. l is a random number between [-1, 1]. That is, the whale group follows the current best fitness whale to hunt with a bubble net according to the above movement formula (roughly a spiral curve). That is, the initial group searches according to the C and gamma solution values ​​of the best whale. When p<0.5, the algorithm enters the global exploration mode. When |A|>1, random search behavior is adopted. The search mechanism updates the individual position in order to jump out of the local optimal solution and move away from the current individual. Its mathematical expression is:

[0305] D=|C×X rand (t)-X(t)|;

[0306] X(t+1)=X rand (t)-A×D;

[0307] Among them, X rand (t) is a randomly selected whale (solution). It can be seen that the formula for random search is very similar to the formula for surrounding prey. The only difference is that the direction is adjusted from the current optimal candidate solution to a random whale. That is, here, the current C and gamma solution values ​​are changed according to the randomly selected solution. This step can expand the search range and jump out of the local optimal solution.

[0308] When |A|<1, the strategy of surrounding prey is adopted. This means that the current whale is about to approach the current optimal individual, which corresponds to finding the optimal solution position in the optimization problem. Its mathematical expression is:

[0309] D=| C×X best (t)-X(t)|;

[0310] X(t+1)=X best (t) = A × D;

[0311] The purpose of this step is to move the population towards the local optimal solution and get closer to the potential global optimal solution.

[0312] The relationship between each iteration and accuracy is as follows: Fig.14 The classification report after the whale optimization algorithm optimizes the SVM parameters is shown in Fig.15 shown.

[0313] A fault warning method for cable joint temperature fault warning, based on a cable joint temperature fault warning model, comprises the following steps:

[0314] Acquire measured data of cable environment; wherein the measured data of cable environment includes measured cable load, ambient temperature and cable outer sheath temperature;

[0315] According to the constructed cable joint temperature fault warning model and the measured data of the cable environment, the cable joint temperature warning data is obtained.

[0316] Specifically, after the cable joint temperature fault warning model obtains the measured data of the cable environment, it will output the cable joint core temperature accordingly. The measured data of the cable environment includes the cable load, the ambient temperature and the cable outer sheath temperature.

[0317] Optionally, after obtaining the measured cable environment data, the cable joint temperature fault warning model needs to preprocess the measured cable environment data, such as converting the cable load, ambient temperature and cable outer sheath temperature into two-dimensional environmental parameters through principal component analysis, and / or normalization in the above steps.

[0318] A computer device comprises a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the fault warning model construction method or the fault warning method.

[0319] A computer program product comprises a computer program, which implements the steps of the fault warning model construction method or the fault warning method when executed by a processor.

[0320] Embodiment 2

[0321] Although the whale optimization algorithm has many advantages such as simple principle, easy implementation, and few algorithm parameters, it still faces some disadvantages such as low population diversity, easy to fall into local optimal solutions, and slow convergence in the later iterations.

[0322] Therefore, this embodiment improves the whale optimization algorithm based on the first embodiment. The specific flow chart is as follows: Fig.16 As shown below:

[0323] The steps include:

[0324] S31: Set algorithm parameters and randomly generate multiple initial individuals of the population;

[0325] Among them, the optimization algorithm first generates a population and sets the search range. The C search range is: [0.01, 20], the step size is 0.01, and the gamma search range is [0.001, 5], the step size is 0.001, which corresponds to the batch generation of candidate solutions in the optimization problem.

[0326] S32: Calculate the fitness of each individual in the population and record the best individual in the best population.

[0327] S33: Determine whether the algorithm has reached the maximum number of iterations. If so, output the optimal solution. If not, perform the following steps:

[0328] Adjust the search step size;

[0329] The generation strategy controls the random number p;

[0330] Update the step size control parameter a, and update the step size control parameter a and parameter C according to the formula; the formulas of A and C are:

[0331] A = 2ar1-a;

[0332] C = 2·r2;

[0333]

[0334] Where r1 and r2 are random numbers between 0 and 1, i is the current iteration number, and T is the maximum number of iterations.

[0335] Determine whether the strategy control random number p is less than 0.5; when the strategy control random number p is less than 0.5, the spiral mechanism updates the individual position; when the strategy control random number p is not less than 0.5, determine whether |A| is less than 1; when |A| is less than 1, the search mechanism updates the individual position; when |A| is not less than 1, the encirclement mechanism updates the individual position;

[0336] Share the best individual population information;

[0337] Determine again whether the algorithm has reached the maximum number of iterations. If not, perform the above steps again.

[0338] This application adds a multi-population information exchange mechanism and an adaptive step size on the basis of the whale optimization algorithm. Specifically, by creating four populations, information is exchanged between populations after each iteration, and the best fitness whales of other populations are added to the current population for information sharing, thereby increasing the breadth of the search solution and helping the algorithm to escape from the local optimum.

[0339] And after the improvement, the search step size can be adaptively adjusted in the process of searching for solutions. In the present invention, the initial step size of the algorithm is 2, and in each iteration, the step size is reduced by 0.01. In the early stage of iteration, the search step size is large, which makes the algorithm more inclined to the global search strategy, which helps to increase the search breadth and improve the search rate. As the number of iterations increases, the search step size is gradually shortened, making the algorithm more inclined to the local search strategy, which helps to improve the convergence speed and accuracy.

[0340] Comparison of the convergence speed of the whale optimization algorithm and the improved whale optimization algorithm Fig.17 As shown in the figure, in the first 25 iterations, the convergence stability of the improved multi-population algorithm is significantly higher than that of the unimproved one, and the accuracy rate quickly increases to about 0.95. In the middle of the iteration, when the number of iterations is between 25 and 100, the improved algorithm has only a fluctuation of about 0.01, while the unimproved one has a fluctuation of 0.04. In the late convergence stage, that is, after the number of iterations reaches 100, the accuracy of the original whale optimization algorithm still has slight fluctuations, while the improved whale optimization algorithm has basically stabilized. The improved classification accuracy report is as follows: Fig.13 As shown in the table below, the accuracy rate after improvement has also increased by 0.01.

[0341] It is to be understood that the present invention is described by some embodiments, and it is known to those skilled in the art that various changes or equivalent substitutions may be made to these features and embodiments without departing from the spirit and scope of the present invention. In addition, under the teachings of the present invention, these features and embodiments may be modified to adapt to specific circumstances and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the scope of protection of the present invention.

[0342] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A method for constructing a fault warning model, characterized in that: Used for cable joint temperature fault early warning, including the following steps: Obtaining a number of cable environmental data and cable joint core temperatures corresponding to the cable environmental data; Reduce the dimension of cable environmental data to obtain two-dimensional environmental parameters; Optimize the C parameter and gamma parameter of the support vector machine through the whale optimization algorithm; According to the two-dimensional environmental parameters, the support vector machine is trained to obtain the cable joint temperature fault warning model.

2. The fault warning model construction method according to claim 1, characterized in that: The cable environmental data includes cable load, ambient temperature and cable outer sheath temperature.

3. The fault warning model construction method according to claim 2, characterized in that: The cable joint core temperature corresponding to the cable environment data is the cable joint core simulation temperature. The acquisition of the cable joint core simulation temperature includes the following steps: Acquire cable model data, and construct a cable connector core model according to the cable model data; According to the cable environment data and the cable joint core model, the cable joint core simulation temperature is obtained; Among them, the cable joint core model is constructed by finite element simulation software; the cable model data includes cable joint structural parameters and cable joint core loss parameters, and the cable joint core loss parameters include conductor core loss, insulation layer dielectric loss, metal shielding layer loss and armor layer loss.

4. The fault warning model construction method according to claim 3 is characterized in that: Before obtaining the cable joint wire core simulation temperature according to the cable environment data and the cable joint wire core model, the method further includes: preprocessing the cable environment data; Wherein, the preprocessing of the cable environment data comprises the following steps: Impute missing values ​​in cable environment data; Detect and handle outliers in cable environmental data; Normalized cable environment data; The missing value is estimated by the following formula: Where x(t) is the estimated missing value, x(t1) and x(t2) are two known data points adjacent to the missing value, and t1 and t2 are the times corresponding to the data points; The detection and processing of outliers in cable environment data comprises the following steps: Calculate the first quartile and the third quartile, and calculate the interquartile range based on the first quartile and the third quartile; According to the interquartile range, define the outlier range and obtain the outlier value; Replace the resulting outliers with the median.

5. The fault warning model construction method according to claim 2, characterized in that: The method of reducing the dimension of the cable environment data to obtain two-dimensional environment parameters includes the following steps: The cable load, ambient temperature and cable outer sheath temperature are reduced in dimension by principal component analysis, so that the cable load, ambient temperature and cable outer sheath temperature are converted into two-dimensional environmental parameters, wherein the two-dimensional environmental parameters include two types of data.

6. The fault warning model construction method according to claim 3, characterized in that: The method of optimizing the C parameter and gamma parameter of the support vector machine by the whale optimization algorithm includes the following steps: Set algorithm parameters and randomly generate initial individuals; Calculate individual fitness and record the best individual; Determine whether the algorithm has reached the maximum number of iterations. If so, output the optimal solution. If the maximum number of iterations is not reached, the following steps are included: The generation strategy controls the random number p; Update the step length control parameter a, and update the step length control parameter a and parameter C according to the formula; Determine whether the random number p is less than 0.5; when the random number p is less than 0.5, the spiral mechanism updates the individual position; when the random number p is not less than 0.5, determine whether |A| is less than 1; when |A| is less than 1, the search mechanism updates the individual position; when |A| is not less than 1, the encirclement mechanism updates the individual position; Determine again whether the algorithm has reached the maximum number of iterations. If not, perform the above steps again.

7. The fault warning model construction method according to claim 3, characterized in that: The method of optimizing the C parameter and gamma parameter of the support vector machine by the whale optimization algorithm includes the following steps: Set algorithm parameters and randomly generate multiple initial individuals of the population; Calculate the fitness of each individual in the population and record the best individual in the best population; Determine whether the algorithm has reached the maximum number of iterations. If so, output the optimal solution. If not, perform the following steps: Adjust the search step size; The generation strategy controls the random number p; Update the step length control parameter a, and update the step length control parameter a and parameter C according to the formula; Determine whether the random number p is less than 0.5; when the random number p is less than 0.5, the spiral mechanism updates the individual position; when the random number p is not less than 0.5, determine whether |A| is less than 1; when |A| is less than 1, the search mechanism updates the individual position; when |A| is not less than 1, the encirclement mechanism updates the individual position; Share the best individual population information; Determine again whether the algorithm has reached the maximum number of iterations. If not, perform the above steps again.

8. A fault warning method, characterized in that: Used for cable joint temperature fault early warning, including the following steps: Obtain measured data of cable environment; According to the cable joint temperature fault warning model constructed according to any one of claims 1 to 7 and the measured parameters of the cable environment, the cable joint temperature warning data is obtained.

9. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the method according to any one of claims 1 to 7 or the steps of the method according to claim 8.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 7 or the steps of the method according to claim 8 are implemented.

Images