Method and system for predicting gas generated by cable thermal degradation
By combining the CEEMDAN and GRU methods, the data noise interference and complexity problems in cable gas monitoring are solved, accurate prediction of cable status and fault warning are achieved, and the safety and reliability of cable operation are improved.
Patent Information
- Application Number
- CN202510627233.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-15
- Publication Date
- 2025-10-17
AI Technical Summary
Existing cable gas monitoring methods are difficult to accurately extract effective features when faced with large data volumes, severe noise interference, and complex gas concentration variations, resulting in poor cable status diagnosis results.
By combining the fully integrated empirical mode decomposition with adaptive noise (CEEMDAN) and the gated recurrent unit (GRU) neural network, accurate prediction of gas concentration can be achieved through data preprocessing, adaptive noise decomposition, gated recurrent unit training and model optimization.
The accuracy and robustness of gas concentration predictions have been improved, enabling more precise prediction of cable status changes, providing reliable support for equipment health monitoring and fault warning.
Smart Images
Figure CN120808967A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of power systems, in particular to a cable thermal degradation gas prediction method and system. BACKGROUND
[0002] As an important component of power systems, cables play a crucial role in energy transmission, communication, industrial automation and other fields in modern society. With the increasing demand for electricity, the operation load of power systems is increasing, and the safety and stability of cables, as one of the important power transmission lines, are of great significance to ensure power supply, reduce power outages and improve power supply reliability. However, the aging and degradation of cables are becoming increasingly serious, especially under harsh working conditions such as high temperature and high current for a long time, the insulation layer and conductor material of the cable will show thermal degradation, which will affect the safe operation of the power system.
[0003] Cable thermal degradation is usually caused by a variety of factors, including high temperature, excessive load, external environment and other factors. As the internal temperature of the cable rises, the insulation material will decompose and release a certain amount of gas. The composition and concentration of these gases can reflect the degree of degradation and possible failure risk inside the cable. In general, the changes in cable thermal degradation product gases have certain regularity, which can provide important basis for the diagnosis and prediction of cable failure.
[0004] However, the diagnosis of cable failure is not simply based on external manifestations such as appearance damage or local short circuit. These traditional fault detection methods often cannot fully reflect the true operating state of the cable. The internal degradation process of the cable is not obvious for a long time, and the early stage of degradation is almost impossible to detect by traditional means, so more accurate technical means are needed to realize real-time monitoring and early warning of the cable state.
[0005] In recent years, the detection technology of cable thermal degradation product gases has attracted more and more attention. Many studies have shown that the composition of gases generated inside the cable can be used as an important indicator for early warning of cable failure. For example, the decomposition process of insulation materials may release gases such as ethylene (C2H4), acetylene (C2H2), methane (CH4), etc., and these gases have different concentration characteristics at different degradation stages and different failure types. Therefore, by monitoring the concentration changes of these gases online, abnormalities can be detected in advance when the cable fails or is about to fail, so that timely handling measures can be taken to prevent the occurrence of failure and reduce power outages and economic losses.
[0006] At present, cable gas monitoring technology mainly relies on online gas analysis equipment, which can monitor the concentration of gas in the cable in real time and transmit data to the monitoring platform. However, traditional gas monitoring methods often face problems such as large data volume, serious noise interference and complex gas concentration changes, which make the application effect of traditional methods in cable state diagnosis not satisfactory. Therefore, how to more accurately extract effective features from gas concentration data, predict the change of gas concentration, and then realize the accurate prediction and fault warning of cable state is a key problem to be solved in the current cable online monitoring field. SUMMARY
[0007] In view of the above problems, the present application is proposed.
[0008] Therefore, the problem to be solved by the present application is how to more accurately extract effective features from gas concentration data, predict the change of gas concentration, and then realize the accurate prediction and fault warning of cable state.
[0009] To solve the above technical problems, the present application provides the following technical scheme: a cable thermal degradation gas prediction method, comprising collecting data, preprocessing data, and identifying abnormal distribution data using box plots and scatter plots; performing adaptive noise complete ensemble empirical mode decomposition on the processed sequence to obtain intrinsic mode function sequences; training and predicting each intrinsic mode function sequence using a gated recurrent unit recurrent neural network; performing adaptive noise complete ensemble empirical mode decomposition inverse process on the predicted intrinsic mode function sequence to obtain a prediction result, calculating the energy of the intrinsic mode function, normalizing the energy of each intrinsic mode function to obtain a weight, and for each intrinsic mode function sequence, the gated recurrent unit outputs a prediction curve for a future period of time, and the total prediction curve is obtained by weighted summation; model evaluation and optimization are performed.
[0010] As a preferred scheme of the cable thermal degradation gas prediction method, the data preprocessing includes obtaining abnormal distribution data of fault gas concentration generated by cable thermal degradation using box plots and scatter plots; the blank items in the abnormal distribution data are filled using linear interpolation method to obtain optimized data; the z-score value of each data point in the optimized data is calculated and obtained using z-score method, and the data whose z-score value does not meet the preset data point threshold is removed as an abnormal value, and finally the gas concentration time sequence is obtained as sample data.
[0011] As a preferred scheme of the cable thermal degradation gas prediction method, the decomposition of the intrinsic mode function sequence includes superimposing one white noise component on the gas concentration time sequence , obtaining new sequences , wherein is the adaptive coefficient of the first iteration, is the time-varying white noise signal of the first iteration, ; performing empirical mode decomposition on the sequences respectively, obtaining the first-order modal component corresponding to the sequences , averaging the to obtain the first-order intrinsic mode function of the completely integrated empirical mode decomposition of the adaptive noise, and the first-order intrinsic mode function is expressed as: , wherein is the first-order intrinsic mode function, is the first-order modal component corresponding to the sequences ; the first residual sequence after the completely integrated empirical mode decomposition of the adaptive noise is calculated and expressed as: , wherein is the first residual sequence; in the second iteration, is the original sequence, and the white noise component is added again on the residual sequence , wherein is the adaptive coefficient of the second iteration, and the empirical mode decomposition is performed to obtain second-order modal components , averaging the to obtain the of the completely integrated empirical mode decomposition of the adaptive noise, and the is expressed as: , wherein is the second-order intrinsic mode function, is the second-order modal component corresponding to the sequences; the steps are repeated to obtain the residual components and after the completely integrated empirical mode decomposition of the adaptive noise, and the is expressed as: , wherein is the residual component after adding the white noise signal, After the empirical mode decomposition of the sequence , , is the first intrinsic mode function, is the first residual component, is the second intrinsic mode function, is the second intrinsic mode function; until the residual signal When the empirical mode decomposition cannot be performed, the adaptive noise complete integrated empirical mode decomposition is terminated.
[0012] As a preferred scheme of the cable thermal degradation gas prediction method, wherein: the gas concentration time sequence includes the gas concentration time sequence After the adaptive noise complete integrated empirical mode decomposition, the intrinsic mode function , , , and the residual , the original sequence is represented as: , wherein, is the intrinsic mode function, is the first intrinsic mode function, is the first residual component.
[0013] As a preferred scheme of the cable thermal degradation gas prediction method, wherein: the training and prediction of the gated recurrent unit recurrent neural network for each intrinsic mode function sequence includes calculating the reset gate and update gate of the current network layer; the state and the input of the current node are obtained The two gate information of the gated recurrent unit, namely the reset gate and the update gate , are represented as: , , wherein, is the reset gate, is the update gate, is the Sigmoid activation function, is the weight matrix of the reset gate, is the weight matrix of the update gate, is the state transmitted from the previous node network, is the input of the current node; the signal screening of the reset gate data is represented as: , wherein, is the information of the hidden layer at the current moment, The weight matrix used to reset the screening process; The network output of the gated recurrent unit is calculated. The network output at the current moment is Expressed as: , in, is the network output at the current moment; each intrinsic mode function sequence is used as a sample input, and after the gated recurrent unit is cyclically trained, the prediction results of each intrinsic mode function sequence are obtained.
[0014] As a preferred solution of the method for predicting gas generation due to thermal degradation of a cable according to the present invention, the predicted intrinsic mode function sequence is subjected to the inverse process of fully integrated empirical mode decomposition of adaptive noise to obtain the prediction result expressed as follows: , in, It is The amplitude of the eigenmode function in the frequency domain, For the The energy of each eigenmode function is normalized to obtain the weight, which is expressed as: in, For the The weight of the eigenmode function; for each eigenmode function sequence, the gated recurrent unit outputs a prediction curve for a period of time in the future , weighted summation, to obtain the total prediction curve Expressed as: , in, is the overall prediction result, It is the forecast curve for a period of time in the future.
[0015] As a preferred solution of the method for predicting gas generation due to thermal degradation of cables described in the present invention, the model evaluation and optimization includes dividing the sample time period, setting the first 80% of the time series as training samples and the last 20% as prediction samples, and using the error indicators root mean square error, mean absolute error, and mean absolute percentage error as a reference for evaluating the carbon price prediction effect of the model; when the model evaluation result is not ideal, the number of layers of the fully integrated empirical mode decomposition of the adaptive noise and the adaptive noise coefficient are adjusted, the structure of the gated recurrent unit is adjusted, the number of units in the hidden layer is increased or decreased, the time step is adjusted, and the learning rate is tried to be adjusted, and the steps are repeated until the expected prediction requirements are met to obtain the optimal model.
[0016] The application further provides a cable thermal degradation gas prediction method system, which can solve the problem of cable thermal degradation gas prediction by constructing a cable thermal degradation gas concentration prediction system.
[0017] To solve the above technical problems, the application provides the following technical scheme: a cable thermal degradation gas prediction system, comprising a data processing module, a decomposition module, a training module, a prediction module and an evaluation and optimization module; the data processing module is used for collecting data, performing data preprocessing, and identifying abnormal distribution data by using a box plot and a scatter plot; the decomposition module is used for performing complete ensemble empirical mode decomposition of adaptive noise on the processed sequence to decompose intrinsic mode function sequences; the training module is used for training and predicting each intrinsic mode function sequence by using a gated recurrent unit recurrent neural network; the prediction module is used for obtaining a prediction result by performing complete ensemble empirical mode decomposition of adaptive noise on the predicted intrinsic mode function sequence, calculating the energy of the intrinsic mode function, normalizing the energy of each intrinsic mode function to obtain a weight, and for each intrinsic mode function sequence, a gated recurrent unit outputs a prediction curve in the future period of time, and the total prediction curve is obtained by weighted summation; and the evaluation and optimization module is used for model evaluation and optimization.
[0018] A computer device comprises a memory and a processor, the memory stores a computer program, and the processor implements the steps of the cable thermal degradation gas prediction method when executing the computer program.
[0019] A computer readable storage medium stores a computer program, and the computer program implements the steps of the cable thermal degradation gas prediction method when executed by a processor.
[0020] The application has the following beneficial effects: the cable thermal degradation gas prediction method provided by the application is different from the traditional time series prediction method, the method combines CEEMDAN and GRU models, fully utilizes the advantages of the two models, and thus realizes high efficiency and high precision of gas concentration prediction. The CEEMDAN algorithm first decomposes the gas concentration time series, converts the complex nonlinear and multi-scale data into multiple intrinsic mode functions, effectively extracts the local features in the data, and captures the time domain characteristics of the gas concentration change. This decomposition method can avoid the errors that may be caused by the traditional method when processing high-frequency noise and non-stationary signals, and better cope with the complexity of the gas concentration sequence. This combination provides a flexible, accurate and efficient prediction framework, which can maintain high accuracy and robustness when processing complex gas concentration time series data.
[0021] The cable thermal degradation gas concentration prediction method based on CEEMDAN and GRU can avoid the single assumption of sequence characteristics in the traditional method, and improve the prediction accuracy. Through this method, not only the change of gas concentration in the operation process of cable thermal degradation can be more accurately predicted, but also the state change of the cable can be more accurately evaluated, providing reliable support for the health monitoring and fault warning of the equipment. BRIEF DESCRIPTION OF DRAWINGS
[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.
[0023] Figure 1 A flow chart of a cable thermal degradation gas prediction method provided for the first embodiment of the present application.
[0024] Figure 2 A structure diagram of the GRU algorithm of the cable thermal degradation gas prediction method provided for the first embodiment of the present application. Figure 3 A structure diagram of a cable thermal degradation gas prediction system provided for the second embodiment of the present application.
[0025] In the figure: 100, data processing module; 200, decomposition module; 300, training module; 400, prediction module; 500, evaluation and optimization module. DETAILED DESCRIPTION
[0026] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings.
[0027] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the connotation of the present application, therefore the present application is not limited by the specific embodiments disclosed below.
[0028] Embodiment 1, refer to Figure 1For the first embodiment of the present application, the embodiment provides a cable thermal degradation gas prediction method, comprising: collecting data, preprocessing data, and identifying abnormal distribution data using box plots and scatter plots; performing adaptive noise complete ensemble empirical mode decomposition on the processed sequence to decompose intrinsic mode function sequences; training and predicting each intrinsic mode function sequence using a gated recurrent unit recurrent neural network, calculating the energy of the intrinsic mode function, normalizing the energy of each intrinsic mode function to obtain weights, and for each intrinsic mode function sequence, the gated recurrent unit outputs a prediction curve for a future period of time, and the weighted sum is obtained. Total prediction curve; perform inverse process on the predicted intrinsic mode function sequence to obtain the prediction result; and perform model evaluation and optimization.
[0029] The core idea of the method is to organically combine the advantages of the improved adaptive noise complete ensemble empirical mode decomposition (CEEMDAN) algorithm and the gated recurrent unit (GRU), and optimize the gas concentration prediction process through the complementarity of the two, thereby improving the accuracy and real-time performance of the prediction. Combined with the signal decomposition characteristics of CEEMDAN and the powerful time series modeling ability of GRU, an efficient prediction framework is constructed.
[0030] Specifically, the CEEMDAN algorithm first effectively decomposes the gas concentration time series, converting complex nonlinear and multi-scale data into multiple intrinsic mode functions (IMF) to extract local features in the data. This decomposition process can better capture the time-domain characteristics of gas concentration changes and avoid errors that may be caused by traditional methods when dealing with high-frequency noise or non-stationary signals. In addition, the adaptability of CEEMDAN allows it to adapt to different types of gas concentration data, effectively separating valuable signal components. GRU further improves the accuracy of gas concentration prediction by capturing long-term dependencies in time series. As a recurrent neural network, GRU can identify complex dynamic changes in time series data and filter input data through a gating mechanism, retaining important information and forgetting unimportant parts. When processing gas concentration time series, GRU can efficiently capture the change rules and accurately predict future concentrations.
[0031] By combining CEEMDAN decomposition with GRU, this method not only improves the understanding of the change law of gas concentration, but also further enhances the prediction ability through the training of GRU. GRU can maintain efficient and accurate prediction ability when processing long time series, ensuring that long-term prediction of gas concentration has high accuracy and robustness, thereby providing strong support for operation state monitoring and fault warning of cables.
[0032] S1, the acquired cable thermal degradation fault gas concentration time series data is preprocessed to obtain sample data.
[0033] The preprocessing is specifically: using the box plot and scatter plot to obtain the abnormal distribution data of the fault gas concentration generated by the cable thermal degradation.
[0034] The blank items in the abnormal distribution data are filled by using linear interpolation method to obtain optimized data.
[0035] The z-score value of each data point in the optimized data is calculated and obtained by using the z-score method, and the data whose z-score value does not meet the preset data point threshold is removed as an abnormal value. Finally, the obtained gas concentration time series is used as sample data.
[0036] S2, the processed sequence is subjected to adaptive noise complete ensemble empirical mode decomposition to decompose the intrinsic mode function sequence.
[0037] In the gas concentration time series , respectively superimposed white noise components , get new sequence , wherein is the adaptive coefficient of the first iteration, is the white noise signal (a ) varying with time. The empirical mode decomposition (EMD) is performed on the sequences respectively to obtain the first modal component corresponding to the sequence , and the average value of is obtained to obtain the first IMF (abbreviation ) of CEEMDAN , and so on , wherein, is the first IMF, is the first order modal component corresponding to the i-th sequence .
[0038] the first residual sequence after CEEMDAN is: , wherein, is the first residual sequence.
[0039] in the second iteration, is the original sequence. In the residual sequence add white noise components again wherein, is the adaptive coefficient of the second iteration, and EMD is performed to obtain second order modal components . Similarly, the average of is obtained is: , wherein, is the second order IMF, is the second order modal component corresponding to the i-th sequence.
[0040] Repeat the above two steps, and the first residual component after CEEMDAN is: , , wherein, is the first residual component added with white noise signal, and the order modal component of the i-th sequence after EMD is , , is the first residual component, is the order IMF. Until the residual signal cannot be subjected to EMD, CEEMDAN is terminated.
[0041] gas concentration time series after CEEMDAN, obtain IMFs , , and residual , then the original sequence can be expressed as: , in, for IMF, For the residual components.
[0042] S3. Perform gated recurrent unit recurrent neural network training and prediction on each intrinsic mode function sequence.
[0043] The GRU training process is as follows: like Figure 2 As shown in Figure 1, it is the structural diagram of the GRU algorithm.
[0044] First, calculate the reset gate and update gate of the current network layer. Combine the state transmitted from the previous layer of the network and the input of the current node Get two gate information of GRU network, namely reset gate and update gate , as shown below: , , in, To reset the gate, To update the gate, is the Sigmoid activation function, To reset the gate weight matrix, is the weight matrix of the update gate, It is the status transmitted from the previous node network. The input of the current node.
[0045] Secondly, perform signal filtering on the reset gate data. If the value of the element in the reset gate is close to 0, the hidden information corresponding to the reset gate should be discarded. The signal filtering of the reset gate is shown in the following formula: , in, is the information of the hidden layer at the current moment, The weight matrix used to reset the filtering process.
[0046] Finally, the network output of GRU is calculated. Update gate Controls the hidden layer at the previous moment The information discarded in the hidden layer and the information that needs to be added to the current hidden layer , its value range is 0~1. The closer the data signal value is to 1, the more data is retained; the closer it is to 0, the more information is discarded. The network output at the current moment is It can be expressed as the following formula: , where, is the network output at the current time.
[0047] denotes the Sigmoid activation function, , denotes the weight matrix of the reset gate and the update gate, and W denotes the weight matrix used in the reset screening process.
[0048] After GRU recurrent training with each IMF sequence as a sample input, the prediction result of each IMF sequence is obtained.
[0049] S4, the predicted intrinsic mode function sequence is subjected to the complete set of adaptive noise empirical mode decomposition inverse process to obtain the prediction result.
[0050] The energy of the intrinsic mode function is calculated, and the energy of each intrinsic mode function is normalized to obtain the weight. For each intrinsic mode function sequence, the gated recurrent unit outputs the prediction curve of the future period of time, and the total prediction curve is obtained by weighted summation.
[0051] First, the weight of each IMF sequence needs to be confirmed. The energy contribution based on IMF is selected as the basis, and the frequency energy is obtained by Fourier transform: , where, is the amplitude of the th IMF in the frequency domain, is the energy of the th IMF.
[0052] Then, the energy of each IMF can be normalized to obtain the weight: , where, is the weight of the th IMF.
[0053] For each IMF sequence, the GRU outputs its prediction curve of the future period of time , and the total prediction curve is obtained by weighted summation: , where, is the total prediction result, is the prediction curve of the future period of time.
[0054] S5, model evaluation and optimization.
[0055] The time period of the sample is divided, the first 80% of the time series is set as the training sample, and the last 20% is set as the prediction sample. The error type indicators root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) are used as reference for evaluating the prediction effect of the carbon price model. Among them, the larger the error type indicator, the worse the prediction effect, and vice versa.
[0056] , , , Among them, represents the sample value, represents the predicted value, and T represents the time length of the sequence.
[0057] When the model evaluation result is not satisfactory, the number of layers and the adaptive noise coefficient of the CEEMDAN decomposition in step one can be adjusted, or the structure of the GRU can be adjusted, the number of units in the hidden layer can be increased or decreased, the time step can be adjusted, and the learning rate can be adjusted. Repeat the steps until the desired prediction requirement is met, and the optimal model is obtained.
[0058] The cable thermal degradation gas concentration prediction method based on CEEMDAN and GRU can avoid the single hypothesis of sequence features in traditional methods and improve the prediction accuracy. Through this method, not only can the change of gas concentration in the cable thermal degradation process be more accurately predicted, but also the state change of the cable can be more accurately evaluated, providing reliable support for the health monitoring and fault warning of the equipment.
[0059] Embodiment 2, refer to Figure 2 , which is different from the previous embodiment, provides a cable thermal degradation gas prediction system, which comprises a data processing module 100, a decomposition module 200, a training module 300, a prediction module 400 and an evaluation optimization module 500.
[0060] The data processing module 100 is used for collecting data and performing data preprocessing, and abnormal distribution data is identified by using box plot and scatter plot.
[0061] The decomposition module 200 is used for adaptive noise complete ensemble empirical mode decomposition of the processed sequence, and intrinsic mode function sequence is decomposed.
[0062] The training module 300 is used for training and prediction of each intrinsic mode function sequence by using the gated recurrent unit recurrent neural network.
[0063] The prediction module 400 is configured to perform a complete ensemble empirical mode decomposition inverse process on the predicted intrinsic mode function sequence to obtain a prediction result, calculate the energy of the intrinsic mode function, normalize the energy of each intrinsic mode function to obtain a weight, and for each intrinsic mode function sequence, a gated recurrent unit outputs a prediction curve for a future period of time, and the prediction curves are summed to obtain a total prediction curve.
[0064] The evaluation and optimization module 500 is configured to perform model evaluation and optimization.
[0065] If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the parts that essentially contribute to the prior art or the parts of the technical solutions can be embodied in the form of software products. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various program code storage media.
[0066] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered a list of executable instructions for implementing logic functions, and can be specifically embodied in any computer-readable medium for use by an instruction execution system, apparatus, or device, such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, apparatus, or device and execute the instructions, or in conjunction with these instruction execution systems, apparatuses, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by an instruction execution system, apparatus, or device, or in conjunction with these instruction execution systems, apparatuses, or devices.
[0067] More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection (electronic) having one or more wires, a portable computer diskette (magnetic), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer readable medium can be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, for example via an optical scanner, then compiled, interpreted or otherwise processed in a suitable manner, if necessary, and then stored in a computer memory.
[0068] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the embodiments described above, various steps or methods can be implemented, for example, in software or firmware that is stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, and in another embodiment, any of the following techniques, which are well known in the art of hardware implementation, can be used: a combination of logic gates in a logic circuit, a combination of processor(s) and memory that forms a special purpose machine and has appropriate software with instructions for performing the desired functions, a combination of application specific integrated circuits (ASICs) and / or other hardware components, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.
[0069] It should be noted that the above-mentioned embodiments illustrate rather than limit the application, since the scope of the application is indicated by the appended claims rather than by the examples that are described above. Although specific terms are employed herein, they are used in a generic and descriptive sense only and not for purposes of limitation, as it will be obvious to those skilled in the art that modifications can be made without departing from the spirit and scope of the application.
Claims
1. A method for predicting gas generation due to thermal degradation of a cable, characterized by: include, Collect data, perform data preprocessing, and use box plots and scatter plots to identify abnormal distribution data; Performing fully integrated empirical mode decomposition with adaptive noise on the processed sequence to decompose the intrinsic mode function sequence; Perform gated recurrent unit recurrent neural network training and prediction on each intrinsic mode function sequence; The predicted intrinsic mode function sequence is subjected to the inverse process of fully integrated empirical mode decomposition with adaptive noise to obtain the prediction result. The energy of the intrinsic mode function is calculated, and the energy of each intrinsic mode function is normalized to obtain the weight. For each intrinsic mode function sequence, the gated recurrent unit outputs the prediction curve for a period of time in the future, and the weighted summation is performed to obtain the total prediction curve. Conduct model evaluation and optimization.
2. The method for predicting gas generation due to thermal degradation of a cable according to claim 1, characterized in that: The data preprocessing includes obtaining abnormal distribution data of fault gas concentration generated by cable thermal degradation using box plots and scatter plots; Use linear interpolation to fill in blank items in abnormal distribution data to obtain optimized data; The z-score method is used to calculate and obtain the z-score value of each data point in the optimized data, and the data whose z-score value does not meet the preset data point threshold are removed as outliers. The final gas concentration time series is used as sample data.
3. The method for predicting gas generation due to thermal degradation of a cable according to claim 2, wherein: The decomposition of the intrinsic mode function sequence includes the gas concentration time series Superimposed on white noise components ,get New sequence ,in, is the adaptive coefficient of the first iteration, The time-varying A white noise signal, ; right Perform empirical mode decomposition on each sequence and get The first-order modal component corresponding to the sequence ,right After averaging, the first-order eigenmode function of the fully integrated empirical mode decomposition of the adaptive noise is expressed as, ， in, is the first-order eigenmode function, For the The first-order modal component corresponding to the sequence ; Calculate the first residual sequence after fully integrated empirical mode decomposition of adaptive noise Expressed as, ， in, is the first residual sequence; In the second iteration, is the original sequence, in the residual sequence Add again white noise components ,in, is the adaptive coefficient of the second iteration, and empirical mode decomposition is performed to obtain Second-order modal components ,right After averaging, the fully integrated empirical mode decomposition of the adaptive noise is obtained. Expressed as, ， in, is the second-order eigenmode function, For the The second-order modal components corresponding to the sequences; Repeat the steps to get the first Residual components and Expressed as, ， ， in, For the After adding white noise signal to the residual component, the After empirical mode decomposition, the sequence The modal components, , For the residual components, for order eigenmode function; Until the residual signal The fully integrated empirical mode decomposition of adaptive noise terminates when the empirical mode decomposition cannot be performed.
4. The method for predicting gas generation due to thermal degradation of a cable according to claim 3, wherein: The gas concentration time series Including gas concentration time series After fully integrated empirical mode decomposition of adaptive noise, we get The eigenmode functions, 、 、…、 and residuals , then the original sequence is expressed as, ， in, for The order eigenmode function, For the residual components.
5. The method for predicting gas generation due to thermal degradation of a cable according to claim 4, characterized in that: The gated recurrent unit recurrent neural network training and prediction for each intrinsic mode function sequence includes calculating the reset gate and update gate of the current network layer; Combined with the status transmitted from the previous node network and the input of the current node Get two gate information of gated recurrent unit, i.e. reset gate and update gate , expressed as, ， ， in, To reset the gate, To update the gate, is the Sigmoid activation function, To reset the gate weight matrix, is the weight matrix of the update gate, It is the status transmitted from the previous node network. is the input of the current node; Perform signal filtering on the reset gate data. The signal filtering of the reset gate is expressed as, ， in, is the information of the hidden layer at the current moment, The weight matrix used to reset the screening process; The network output of the gated recurrent unit is calculated. The network output at the current moment is Expressed as, ， in, is the network output at the current moment; Each intrinsic mode function sequence is used as a sample input, and after the gated recurrent unit is cyclically trained, the prediction results of each intrinsic mode function sequence are obtained.
6. The method for predicting gas generation due to thermal degradation of a cable according to claim 5, characterized in that: The predicted intrinsic mode function sequence is subjected to the inverse process of fully integrated empirical mode decomposition of adaptive noise to obtain the prediction result, which is expressed as: ， in, It is The amplitude of the eigenmode function in the frequency domain, For the The energy of the eigenmode functions; Normalize the energy of each eigenmode function to get the weight, expressed as, ， in, For the The weight of the eigenmode function; For each intrinsic mode function sequence, the gated recurrent unit outputs a prediction curve for a period of time in the future , weighted summation, to obtain the total prediction curve Expressed as, ， in, is the overall prediction result, It is the forecast curve for a period of time in the future.
7. The method for predicting gas generation due to thermal degradation of a cable according to claim 6, characterized in that: The model evaluation and optimization includes dividing the sample time period, setting the first 80% of the time series as training samples and the last 20% as prediction samples, and using error indicators such as root mean square error, mean absolute error, and mean absolute percentage error as a reference for evaluating the carbon price prediction effect of the model; When the model evaluation results are not ideal, adjust the number of layers of the fully integrated empirical mode decomposition of the adaptive noise and the adaptive noise coefficient, adjust the structure of the gated recurrent unit, increase or decrease the number of units in the hidden layer, adjust the time step, try to adjust the learning rate, and repeat the steps until the expected prediction requirements are met to obtain the optimal model.
8. A cable thermal degradation gas generation prediction system, using a cable thermal degradation gas generation prediction method according to any one of claims 1 to 7, characterized in that: It includes a data processing module (100), a decomposition module (200), a training module (300), a prediction module (400) and an evaluation and optimization module (500); The data processing module (100) is used to collect data, perform data preprocessing, and identify abnormal distribution data using box plots and scatter plots; The decomposition module (200) is used to perform fully integrated empirical mode decomposition of adaptive noise on the processed sequence to decompose the intrinsic mode function sequence; The training module (300) is used to train and predict the gated recurrent unit recurrent neural network for each intrinsic mode function sequence; The prediction module (400) is used to perform an inverse process of fully integrated empirical mode decomposition of adaptive noise on the predicted intrinsic mode function sequence to obtain a prediction result, calculate the energy of the intrinsic mode function, normalize the energy of each intrinsic mode function to obtain a weight, and for each intrinsic mode function sequence, the gated recurrent unit outputs a prediction curve for a period of time in the future, and performs weighted summation to obtain a total prediction curve; The evaluation and optimization module (500) is used to perform model evaluation and optimization.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for predicting gas generation due to thermal degradation of a cable according to any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a method for predicting gas generation due to thermal degradation of a cable according to any one of claims 1 to 7 are implemented.