A cable joint temperature prediction method and system

Abstract

The present invention discloses a cable joint temperature prediction method and system, which relates to the field of power cable technology, including constructing and solving a thermal circuit model of a cable joint to be tested, obtaining the temperature distribution of each time step; obtaining the historical operation data of the cable joint to be tested and preprocessing it; constructing and training a temperature prediction model based on the processed historical operation data and temperature distribution to obtain a trained temperature prediction model; collecting the current operation data of the cable joint to be tested and inputting it into the thermal circuit model to calculate the current temperature distribution, inputting the current temperature distribution and the current operation data into the trained temperature prediction model, and predicting the temperature of the cable joint within a future set time period. Combining the physical laws of the thermal circuit model with the data-driven temperature prediction model fully utilizes the accuracy of the physical model and the data-driven advantages of the network model to improve the overall accuracy of temperature prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of power cables, and in particular to a method and system for predicting the temperature of a cable joint. Background Art

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.

[0003] In recent years, with the upgrading and transformation of the power system, the proportion of cables used in power transmission has gradually increased. The power cable joint is the weakest and most prone to failure in the cable line, and its insulation state directly affects the operation safety and service life of the cable line. The failure of cable accessories is much higher than that of the cable body, and the failure of the cable intermediate joint is higher than that of the cable terminal joint. In order to ensure the safety and stability of the power cable operation, it is necessary to minimize or even avoid the paralysis of the power supply system caused by the failure of the power cable intermediate joint, thereby reducing the serious losses caused by the paralysis of the power supply system. Therefore, the analysis and detection of power cable intermediate joint failures is very important.

[0004] The temperature of power cable joints is a key factor in determining whether the cable is operating normally. Real-time monitoring and prediction of cable joint temperature can greatly reduce the failure rate and detect potential faults in a timely manner. With the widespread application of artificial intelligence and deep learning in the field of industrial fault detection, the temperature prediction model of cable joints is becoming more and more widespread.

[0005] However, among the current methods for predicting temperature using neural network models, most existing cable joint temperature prediction models directly use general neural network models, such as convolutional neural networks and long-short time neural networks. However, the structure of general neural network models is relatively complex, and too many training layers make it difficult to improve the prediction accuracy of the model, consuming too many computing resources. At the same time, the existing cable temperature prediction methods do not take into account the physical properties of the cable joint itself. Summary of the invention

[0006] In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides a cable joint temperature prediction method and system, which predicts the cable joint temperature through the 1D-CNN+GRU model, and can learn the complex relationship between the cable temperature and various influencing factors from a large amount of historical data, while taking into account the physical properties of the cable joint itself, thereby achieving more accurate prediction.

[0007] In a first aspect, the present invention provides a cable joint temperature prediction method, comprising:

[0008] Construct and solve the thermal circuit model of the cable joint to be tested to obtain the temperature distribution at each time step;

[0009] Obtain historical operation data of the cable joint to be tested and perform preprocessing;

[0010] Building and training a temperature prediction model based on the processed historical operation data and temperature distribution to obtain a trained temperature prediction model;

[0011] The current operating data of the cable joint to be tested is collected and input into the thermal circuit model to calculate the current temperature distribution. The current temperature distribution and the current operating data are input into the trained temperature prediction model to predict the temperature of the cable joint within the future set time period.

[0012] According to a further technical solution, the historical operation data includes historical temperature data, historical environmental data and historical load data of the monitoring point of the cable joint to be tested.

[0013] In a further technical solution, the heat balance equation in the constructed heat circuit model is:

[0014]

[0015] Where T is the temperature, is the temperature of the previous time step, t is the time, and P is the power loss.

[0016] A further technical solution is to use the Euler method to iteratively solve the heat balance equation. At each time step, the temperature update formula is:

[0017]

[0018] in, is the time step, R is the thermal resistance, is the ambient temperature.

[0019] A further technical solution is that the temperature prediction model is constructed using a one-dimensional convolutional neural network and a gated recurrent unit neural network, including an input layer, a first convolutional layer, a global pooling layer, a second convolutional layer, a global pooling layer, a Dropout layer, a GRU layer and a fully connected layer connected in sequence.

[0020] In a further technical solution, the first convolutional layer and the second convolutional layer have the same structure, and each layer of the convolutional neural network uses convolution kernels of multiple different scales.

[0021] According to a further technical solution, the loss function of the temperature prediction model is a joint loss function of the thermal circuit model error and the temperature prediction model error.

[0022] In a second aspect, the present invention provides a cable joint temperature prediction system, comprising:

[0023] Thermal circuit model building module: build and solve the thermal circuit model of the cable joint to be tested to obtain the temperature distribution at each time step;

[0024] Data acquisition module: obtains historical operation data of the cable joint to be tested and performs preprocessing;

[0025] Model building module: builds and trains a temperature prediction model based on the processed historical operation data and temperature distribution to obtain a trained temperature prediction model;

[0026] Model prediction module: collect the current operating data of the cable joint to be tested and input it into the thermal circuit model to calculate the current temperature distribution. Input the current temperature distribution and the current operating data into the trained temperature prediction model to predict the temperature of the cable joint within the future set time period.

[0027] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in a cable joint temperature prediction method as described in the first aspect.

[0028] In a fourth aspect, the present invention provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in a cable joint temperature prediction method as described in the first aspect are implemented.

[0029] One or more of the above technical solutions have the following beneficial effects:

[0030] The present invention uses the historical operation data of the cable joint to be tested and the temperature distribution calculated by the thermal circuit model to train the temperature prediction model, so that the trained cable temperature prediction model can integrate various complex environmental factors and perform multi-dimensional data training, so as to more accurately predict the temperature of the cable joint, so as to timely detect abnormal heating of the cable, prevent failures caused by overheating of the cable, and extend the service life.

[0031] The present invention combines the temperature prediction model with the thermal circuit model to predict the temperature of the cable joint, combines the physical laws of the thermal circuit model with the data-driven machine learning model, and uses the constraints and guidance provided by the physical model to improve the reliability of the data-driven model, making full use of the accuracy of the physical model and the data-driven advantages of the network model to improve the overall accuracy of temperature prediction. Using the real-time temperature distribution as an input feature, the temperature prediction model can make predictions based on the current actual temperature state, enhancing the robustness of the model so that it can maintain a high prediction accuracy under different operating conditions.

[0032] The present invention adopts the network structure model of 1D-CNN+GRU, extracts features through a multi-layer and multi-scale convolutional network, and introduces a global pooling layer to replace the Flatten layer and the fully connected layer in the traditional CNN network architecture, thereby avoiding the fragmentation of parameter features caused by the Flatten operation and the fully connected layer; at the same time, combined with the time dependence of the gated recurrent unit neural network, the temperature of the cable joint to be tested is quickly predicted, and the prediction accuracy is improved, so that early warning of abnormal temperature can be carried out and timely measures can be taken. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0034] Figure 1 It is a flow chart of a cable joint temperature prediction method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0035] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0036] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0037] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0038] It should be understood that although the following mainly focuses on the application of locomotive-related materials, the embodiments of the present disclosure are not limited to this, and are also applicable to material data extraction for quota analysis in other fields, including material quota analysis of components or equipment in light and heavy industries.

[0039] Embodiment 1

[0040] See attached Figure 1 As shown, this embodiment discloses a cable joint temperature prediction method, which specifically includes the following steps:

[0041] S1: Construct and solve the thermal circuit model of the cable joint to be tested to obtain the temperature distribution at each time step;

[0042] In thermal circuit models, time steps usually refer to the time intervals used in the simulation process. Each time step represents the process of simulation progressing from one time point to the next. For the temperature distribution simulation of cable joints, the time step can be seconds, minutes, hours, or even days, depending on the time scale you want to analyze and the required accuracy.

[0043] S101: Determine the thermal circuit model structure;

[0044] Decomposing the thermal path of the cable joint means decomposing the cable joint into multiple thermal resistance and thermal capacitance components to represent the heat conduction paths of different materials and parts; defining thermal resistance and thermal capacitance, thermal resistance R represents the resistance in the heat conduction process, and thermal capacitance C represents the ability of the material to store heat.

[0045] S102: define thermal circuit parameters and set initialization conditions;

[0046]

[0047] Among them, R is the thermal resistance, L is the length of the heat conduction path, k is the thermal conductivity of the material, and A is the cross-sectional area of ​​the heat conduction path.

[0048]

[0049] Among them, C is the heat capacity, m is the mass of the material, and c is the specific heat capacity of the material.

[0050] Assume that the cable joint to be tested has x nodes, set the initial temperature of each node, and set the ambient temperature , set the power loss P. The initial condition value can be set according to the actual situation, and is not specifically limited in this embodiment.

[0051] S103: According to the principle of conservation of energy, a heat balance equation is established.

[0052] The heat conduction formula is expressed as:

[0053]

[0054] Where Q is the heat flow, For the temperature difference.

[0055] The heat balance equation is:

[0056]

[0057] Where T is the temperature, is the temperature of the previous time step, and t is the time.

[0058] The Euler method is used to iteratively solve the heat balance equation. The temperature update formula is:

[0059]

[0060] in, The ambient temperature is the ambient temperature where the cable connector is located.

[0061] Finally, at each time step, the temperature change of each node is calculated until the specified time range is reached, and the temperature distribution of each time step is obtained. The temperature distribution shows the temperature change trend of the cable joint to be tested over time.

[0062] S2: Obtain historical operation data of the cable joint to be tested and perform preprocessing;

[0063] Historical operation data includes historical temperature data, historical environmental data and historical load data of the cable joint monitoring point to be tested; the cable joint temperature is related to the ambient temperature and load current, and the historical load data includes historical current, voltage and power data, which can be obtained through current sensors, voltage sensors and power metering equipment installed in the cable system. Data from different sources need to be integrated to ensure that they are aligned at the same timestamp.

[0064] The acquired historical operation data is cleaned and smoothed. Data cleaning includes missing value processing and outlier processing. The amount of historical operation data is huge, and the proportion of missing values ​​is very small. Missing values ​​are processed by deleting records with missing values ​​to avoid the supplementary values ​​affecting the accuracy of the data; use the mean and standard deviation to detect outliers and delete abnormal data points. The time series data is smoothed by moving average to reduce the impact of short-term fluctuations. Through these preprocessing steps, the quality and consistency of the data can be ensured, providing a reliable data foundation for the subsequent modeling process, thereby improving the accuracy of cable joint temperature prediction.

[0065] Furthermore, principal component analysis (PCA) is used to reduce feature dimensions and retain the main information of the data, which is beneficial to the subsequent processing of the network model and reduces the complexity of data processing.

[0066] S3: construct and train a temperature prediction model based on the processed historical operation data and temperature distribution to obtain a trained temperature prediction model.

[0067] In this embodiment, the temperature prediction model is constructed using a one-dimensional convolutional neural network and a gated recurrent unit neural network, including an input layer, a first convolutional layer, a global pooling layer, a second convolutional layer, a global pooling layer, a Dropout layer, a GRU layer, and a fully connected layer connected in sequence.

[0068] Specifically, the processed historical operating data and the calculated temperature distribution are input into the constructed temperature prediction model for training; the input data is divided into data in chronological order to ensure that the time spans of the training set, validation set and test set do not overlap, and the training set is used to train the temperature prediction model. The weights and parameters of the model are updated through the back propagation algorithm and gradient descent optimizer to minimize the prediction error.

[0069] In this embodiment, the output of the thermal circuit model is used as one of the input features of the temperature prediction model for model fusion, and the result of the thermal circuit model is corrected and optimized by the temperature prediction model. Furthermore, the thermal circuit model and the temperature prediction model are jointly trained, and the error term of the thermal circuit model is added to the loss function of the temperature prediction model to reduce the deviation between the two.

[0070] The error term of the thermal circuit model is added to the loss function of the temperature prediction model, which specifically includes the following steps:

[0071] 1) Define the joint loss function;

[0072] Calculate the thermal circuit model error, the expression is:

[0073]

[0074] Where N is the number of samples, is the predicted temperature of the thermal circuit model for the ith sample, is the actual observed temperature.

[0075] Calculate the temperature prediction model error, the expression is:

[0076]

[0077] in, is the output temperature of the temperature prediction model for the i-th sample.

[0078] The joint loss function expression is:

[0079]

[0080] in, and is a weight coefficient used to balance the impact of temperature prediction error and thermal circuit model error. By adjusting these two coefficients, the contribution of both in the loss function can be controlled.

[0081] 2) Use the joint loss function to train the temperature prediction model and iteratively adjust the model parameters to minimize the loss function.

[0082] During the training process, the value of the joint loss function should gradually decrease, indicating that the model has made progress in reducing the prediction error and the thermal circuit model error. The performance of the temperature prediction model can be further improved, and the model parameters and structure can be continuously optimized based on the actual temperature feedback.

[0083] Furthermore, the genetic algorithm (GA) is used to jointly optimize the parameters of the thermal circuit model and the temperature prediction model, which can effectively explore the parameter space and find the optimal parameter combination.

[0084] The specific steps of genetic algorithm optimization include:

[0085] 1) Initialize the population and randomly generate a set of initial parameters as the initial population. Each individual (i.e., a parameter set) includes the parameters of the thermal circuit model and the temperature prediction model, and the expression is:

[0086]

[0087] Where M is the population size, It is an individual and contains the parameters of the thermal circuit model and the temperature prediction model, including the boundary condition temperature (ambient temperature) of the thermal circuit model, the number of convolutional layers, convolution kernel size, learning rate, and regularization coefficient of the temperature prediction model.

[0088] 2) Define the evaluation function and calculate the loss function of each individual on the training data, that is, the joint loss function, including the temperature prediction error and the thermal circuit model error, specifically:

[0089] .

[0090] 3) Select the better individual according to the value of the evaluation function, using the roulette wheel selection method, specifically:

[0091]

[0092] in, is the joint loss of individuals, and j is the jth individual.

[0093] 4) Perform a crossover operation on the selected individuals to generate new individuals. The crossover method uses a single-point crossover, specifically:

[0094]

[0095] in, For new individuals, For individual 1, is individual 2, k is the random crossover length, To separate individuals 1 at crossover points according to random crossover lengths, To separate individuals 2 at crossover points according to random crossover lengths.

[0096] 5) Perform mutation operations on individuals, randomly change certain parameters with a certain probability, and increase the diversity of the population. Specifically:

[0097]

[0098] in, is the jth individual, is a small random variation.

[0099] 6) Use selection, crossover and mutation operations to generate a new population, and iterate until the termination condition is met. The termination condition is that the joint loss function converges.

[0100] The parameters of the thermal circuit model and the temperature prediction model are jointly optimized through genetic algorithms to find the optimal parameter combination of the thermal circuit model and the temperature prediction model, so that the joint loss function reaches the minimum value, the comprehensive prediction accuracy of the model is improved, the temperature prediction error and the thermal circuit model error are reduced, and the accuracy and reliability of the cable joint temperature prediction model are improved.

[0101] Specifically, the first convolution layer and the second convolution layer have the same structure, and each convolutional neural network uses three convolution kernels of different scales. In this embodiment, the convolution kernel size with a tolerance of 5 is selected, the scale of the first convolution kernel is 5, the scale of the second convolution kernel is 10, and the scale of the third convolution kernel is 15; CNN uses nonlinear activation functions to solve nonlinear problems in the real world, and selects rectified linear unit (ReLU) as the activation function of the convolutional neural network.

[0102] Choose to build 2 layers of multi-scale convolutional layers; the number of convolutional layers of the one-dimensional multi-scale convolutional neural network model is the key to the model's prediction performance. Each layer of multi-scale convolutional layer will mine different deep features of the input data. At the same time, since the convolutional neural network only mines features of the input data, rather than extracting feature relationships in the data, too many multi-scale convolutional layers are prone to overfitting. Therefore, choose to build 2 layers of multi-scale convolutional layers.

[0103] The input data of the one-dimensional convolution layer is a two-dimensional vector, S is the running data dimension, H is the time series length, and the size of each input sample is (H, S). The convolution kernel Fn of the convolution layer has different scales. The length of Fn is Fn1, Fn2, and Fn3 respectively, and the width is the monitoring data dimension Z of the input convolutional neural network. When the convolution operation of each scale convolution kernel is performed, the convolution step length is 1 and zero padding is used. The feature maps of convolution kernels of different scales are spliced ​​as the input of the next layer. After the convolution operation of the convolution layer, the activation function is used to increase the nonlinearity of the model.

[0104] Specifically, a global pooling layer is introduced to replace the Flatten layer and the fully connected layer in the traditional CNN network architecture, avoiding the fragmentation of parameter features caused by the Flatten operation and the fully connected layer.

[0105] The pooling layer can reduce the dimension of the feature map while maintaining the most important information. It has many types. This embodiment uses the global maximum pooling layer, which is expressed as

[0106]

[0107] in, is the element in the bth feature map of the p+1th layer after pooling, represents the jth pooling region, Represents the elements of the b-th feature map of the p-th layer within the range of the pooling kernel.

[0108] Specifically, the temperature prediction model introduces a random dropout mechanism at the connection between the 1D-CNN part and the GRU part, randomly dropping the weights between neurons, thereby reducing the network's dependence on a certain neuron. This operation can reduce the impact of noise in the input data and randomly reset the neuron weights to 0 in a certain proportion. The expression is:

[0109]

[0110]

[0111] in, represents the probability vector that follows the Bernoulli distribution, The convolutional layer Layer, n is the total number of convolutional layers, represents the output after random discarding mechanism, For the The output of the layer, For the The probability of layer dropping, Indicates the random drop mechanism.

[0112] Specifically, the gated recurrent unit neural network GRU is a variant of the recurrent neural network (RNN). Similar to LATM, it merges the unit state and the output state into a hidden state, relies on the hidden state to transmit information, and integrates the forget gate and the input gate into an update gate. It has the advantages of fewer parameters and faster calculation speed. For sensor time series data, it has the function of capturing the mapping relationship between data.

[0113] The GRU layer includes an update gate and a reset gate. The above two gating vectors determine which information can eventually be output as a gated recurrent unit.

[0114] Specifically, the update gate determines how much old information and new information to continue to pass into the future, which is expressed as follows:

[0115]

[0116] in, To update the gate, is the sigmoid activation function, To update the gate weight, is the hidden state at the previous moment, is the input vector at the current moment.

[0117] The reset gate determines how much past information needs to be forgotten, which is expressed as follows:

[0118]

[0119] in, To reset the gate, To reset the gate weights.

[0120] The current memory content is equivalent to the input information at the current moment, which is expressed as follows:

[0121]

[0122] in, is a candidate hidden state, Hidden state weights.

[0123] Finally, determine the final memory at the current moment, which is expressed as follows:

[0124] .

[0125] In this embodiment, the GRU layer is initialized, specifically by setting the hyperparameters of the GRU neural network model, the dropout coefficient is 0.1, and the optimization method is RMSProp. The deep feature extraction capability of the multi-scale convolutional neural network and the time dependence of the gated recurrent unit neural network are used to predict the temperature of the cable joint to be tested.

[0126] S4: Collect the current operating data of the cable joint to be tested and input it into the thermal circuit model to calculate the current temperature distribution, input the current temperature distribution and the current operating data into the trained temperature prediction model to predict the temperature of the cable joint within a future set time period.

[0127] Collect current operating data, including real-time current value, real-time voltage value, and current ambient temperature, and input the current operating data (current, voltage, load, and environmental data) into the thermal circuit model. The thermal circuit model calculates the temperature distribution of the current cable joint based on the input data, including the temperature at the joint and the temperature gradient along the cable.

[0128] The current temperature distribution and operating data are combined into a feature vector and input into the trained temperature prediction model. The temperature prediction model predicts the temperature of the cable joint in a certain time period in the future (such as the next hour or the next day) based on the input current temperature distribution and operating data. At the same time, the prediction result can be compared with the actual measured temperature to adjust and optimize the temperature prediction model.

[0129] The temperature prediction model can dynamically adjust and update model parameters based on real-time data and prediction errors to improve the adaptability of the model.

[0130] Based on the predicted temperature, it is determined whether the cable joint has the risk of overheating and whether preventive measures are needed. The future temperature of the cable joint can be predicted in real time to prevent and respond to possible overheating in advance and ensure the safe and stable operation of the power system.

[0131] High temperature may cause the joint to overheat, which in turn may lead to serious faults such as insulation damage and joint burning. This embodiment can detect joints with abnormal temperatures in advance by monitoring and predicting the temperature of cable joints, predict possible fault points, prevent fault expansion, and ensure stable operation and power supply reliability of the power system. Through temperature prediction, load distribution and operation mode can also be optimized to avoid long-term operation of equipment at high temperatures, thereby extending the service life of cables and joints.

[0132] Embodiment 2

[0133] This embodiment provides a cable joint temperature prediction system, including:

[0134] Thermal circuit model building module: build and solve the thermal circuit model of the cable joint to be tested to obtain the temperature distribution at each time step;

[0135] Data acquisition module: obtains historical operation data of the cable joint to be tested and performs preprocessing;

[0136] Model building module: builds and trains a temperature prediction model based on the processed historical operation data and temperature distribution to obtain a trained temperature prediction model;

[0137] Model prediction module: collect the current operating data of the cable joint to be tested and input it into the thermal circuit model to calculate the current temperature distribution. Input the current temperature distribution and the current operating data into the trained temperature prediction model to predict the temperature of the cable joint within the future set time period.

[0138] Embodiment 3

[0139] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps in the cable joint temperature prediction method described in the first embodiment are implemented.

[0140] Embodiment 4

[0141] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the cable joint temperature prediction method described in the first embodiment are implemented.

[0142] The steps involved in the apparatus of the above embodiments 3 and 4 correspond to the method embodiment 1, and the specific implementation method can refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0143] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0144] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0145] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A cable joint temperature prediction method, characterized in that: include: Construct and solve the thermal circuit model of the cable joint to be tested to obtain the temperature distribution at each time step; Obtain historical operation data of the cable joint to be tested and perform preprocessing; Building and training a temperature prediction model based on the processed historical operation data and temperature distribution to obtain a trained temperature prediction model; The current operating data of the cable joint to be tested is collected and input into the thermal circuit model to calculate the current temperature distribution, and the current temperature distribution and the current operating data are input into the trained temperature prediction model to predict the temperature of the cable joint within the future set time period; the temperature prediction model is constructed using a one-dimensional convolutional neural network and a gated recurrent unit neural network, including an input layer, a first convolutional layer, a global pooling layer, a second convolutional layer, a global pooling layer, a Dropout layer, a GRU layer and a fully connected layer connected in sequence; the first convolutional layer and the second convolutional layer have the same structure, and each convolutional neural network layer uses a variety of convolution kernels of different scales; the temperature prediction model introduces a random drop mechanism at the connection between the one-dimensional convolutional neural network and the gated recurrent unit neural network, and randomly drops the weights between neurons; The heat balance equation in the constructed heat circuit model is: ； Where C is heat capacity, R is thermal resistance, T is temperature, is the temperature of the previous time step, t is the time, and P is the power loss; The Euler method is used to iteratively solve the heat balance equation. At each time step, the temperature update formula is: ； in, is the time step, R is the thermal resistance, is the ambient temperature.

2. A cable joint temperature prediction method as claimed in claim 1, characterized in that: The historical operation data includes historical temperature data, historical environment data and historical load data of the monitoring point of the cable joint to be tested.

3. A cable joint temperature prediction method as claimed in claim 1, characterized in that: The loss function of the temperature prediction model is a joint loss function of the thermal circuit model error and the temperature prediction model error.

4. A cable joint temperature prediction system, using a cable joint temperature prediction method as claimed in any one of claims 1 to 3, characterized in that: include: Thermal circuit model building module: build and solve the thermal circuit model of the cable joint to be tested to obtain the temperature distribution at each time step; Data acquisition module: obtains historical operation data of the cable joint to be tested and performs preprocessing; Model building module: builds and trains a temperature prediction model based on the processed historical operation data and temperature distribution to obtain a trained temperature prediction model; Model prediction module: collect the current operating data of the cable joint to be tested and input it into the thermal circuit model to calculate the current temperature distribution. Input the current temperature distribution and the current operating data into the trained temperature prediction model to predict the temperature of the cable joint within the future set time period.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in a cable joint temperature prediction method as described in any one of claims 1 to 3 are implemented.

6. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps in the cable joint temperature prediction method according to any one of claims 1 to 3 are implemented.

Images