Cable fire prediction method based on concentration data of cable thermal degradation characteristic gases

Abstract

The present invention relates to the technical field of cable fire prediction. Disclosed is a cable fire prediction method based on concentration data of cable thermal degradation characteristic gases. The method comprises: establishing a data fusion model, and processing first data; performing data fusion and error analysis; and by means of establishing a fire prediction model, generating fire occurrence probability information. By means of establishing a data fusion model, using concentration data of gases released in a very early stage of cable thermal degradation as features for data fusion, and establishing a fire prediction model for prediction, the method of the present invention can provide more accurate and real-time cable fire prediction, and realize very early warning to reduce the risk of cable fire, thereby ensuring the stable operation of a power system.

Description

A Cable Fire Prediction Method Based on Cable Thermal Deterioration Characteristic Gas Concentration Data Technical Field

[0001] This invention relates to the field of cable fire prediction technology, specifically a cable fire prediction method based on cable thermal degradation characteristic gas concentration data. Background Technology

[0002] As a key medium for power systems and information transmission, cables are crucial to the infrastructure of modern society. Their reliability directly affects the stability of energy supply and the continuity of information transmission. However, cables may experience thermal degradation during long-term operation. This is caused by factors such as overload, poor contact, excessively high ambient temperature, or material aging, which leads to an increase in the internal temperature of the cable and affects its insulation performance. Thermal degradation not only shortens the service life of the cable but may also cause cable fires. Cable fires can lead to power outages and cause significant economic losses. More seriously, they may trigger secondary disasters, such as the release of toxic fumes and the spread of fire, posing a serious threat to personnel safety and the environment.

[0003] To prevent cable thermal degradation and fire accidents, existing prediction algorithms mainly focus on temperature monitoring, load analysis, and material property evaluation. While these algorithms can assess the operating status of cables to some extent, they often have certain limitations. For example, some algorithms may rely on a large amount of historical data and are not sensitive enough to real-time changes. Therefore, developing a new real-time prediction algorithm is of great significance for improving the safety and reliability of cable operation. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by this invention is that existing cable fire prediction methods are not responsive, rely on a large amount of historical data, and cannot assess changes in cable operating status in real time. The invention also addresses how to improve the real-time performance and accuracy of cable fire prediction.

[0006] To address the aforementioned technical problems, this invention provides the following technical solution: a cable fire prediction method based on cable thermal degradation characteristic gas concentration data, comprising establishing a data fusion model, processing the first data, performing data fusion and error analysis, and generating fire occurrence probability information by establishing a fire prediction model.

[0007] As a preferred embodiment of the cable fire prediction method based on cable thermal degradation characteristic gas concentration data described in this invention, the data fusion model includes feature extraction, preprocessing, preliminary fusion, and error analysis.

[0008] As a preferred embodiment of the cable fire prediction method based on cable thermal degradation characteristic gas concentration data according to the present invention, the processing of the first data includes feature extraction and preprocessing of node data of different cable thermal degradation characteristic gas sensor clusters.

[0009] As a preferred embodiment of the cable fire prediction method based on cable thermal degradation characteristic gas concentration data according to the present invention, the data fusion includes preliminary fusion of preprocessed data at cluster head nodes based on a data fusion model.

[0010] As a preferred embodiment of the cable fire prediction method based on cable thermal degradation characteristic gas concentration data described in this invention, the error analysis includes classifying sensor data and performing error analysis based on a data fusion model.

[0011] As a preferred embodiment of the cable fire prediction method based on cable thermal degradation characteristic gas concentration data described in this invention, the fire prediction model includes determining multiple probabilities of fire risk, outputting the levels of no-fire probability P1, smoldering probability P2 and open flame probability P3 as large, medium and small, and classifying the fire probability P level into four levels: general, relatively large, serious and extremely serious.

[0012] As a preferred embodiment of the cable fire prediction method based on cable thermal degradation characteristic gas concentration data described in this invention, the method for generating fire probability information includes establishing a fire probability discrimination rule base based on the fused data. There are 27 possible fire probability levels (P), including 5 moderate, 6 relatively high, 9 serious, and 7 extremely serious. When at least two of the fire probability levels (P1, P2, and P3) are low, and the fire probability level (P3) is not medium or high, the fire probability level (P) is output as moderate. When at least two of the fire probability levels (P1, P2, and P3) are high, and the fire probability level (P2) is not low, or when the fire probability level (P1) is low, the fire probability level (P2) is high, and the fire probability level (P3) is medium, the fire probability level (P) is output as extremely serious.

[0013] Another objective of this invention is to provide a cable fire prediction system based on the concentration data of characteristic gases in cable thermal degradation. This system can solve the problem of insensitive response in current cable fire prediction by establishing a data fusion model to extract features and preprocess the first data.

[0014] As a preferred embodiment of the cable fire prediction system based on cable thermal degradation characteristic gas concentration data according to the present invention, it includes: a data processing module, an error analysis module, and a fire probability prediction module; the data processing module is used to establish a data fusion model, perform feature extraction and data preprocessing on the first data; the error analysis module is used for data fusion and error analysis, classifying and analyzing the processed data; the fire probability prediction module is used to establish a fire prediction model, generate fire occurrence probability information based on the fused data, output no-fire probability, smoldering probability, and open flame probability, and finally evaluate the fire probability, outputting fire level probability information of general, relatively large, serious, and particularly serious.

[0015] A computer device includes a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement a cable fire prediction method based on cable thermal degradation characteristic gas concentration data.

[0016] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a cable fire prediction method based on cable thermal degradation characteristic gas concentration data.

[0017] The beneficial effects of this invention are as follows: The cable fire prediction method based on characteristic gas concentration data of cable thermal degradation provided by this invention establishes a data fusion model and processes the first data. It collects characteristic gas concentration data of cable thermal degradation in real time, enabling comprehensive integration of data from multiple sensors. This not only ensures the diversity and wide range of data sources but also effectively improves the overall quality of the data. Feature extraction and preprocessing during the processing remove obvious outliers, ensuring higher accuracy and reliability of the data used in subsequent analysis. Furthermore, data integration reveals the interrelationships between different gas concentrations, providing more comprehensive information to support fire early warning and laying a solid foundation for subsequent fire risk prediction. This improves the system's response speed and accuracy to potential fires. In the classification and error analysis of the processed data, the cluster head node based on the data fusion model is used to classify the received data. By effectively classifying sensor data and further eliminating erroneous and unreliable data, and by applying logical algorithms to distinguish and integrate data from different sensors, more accurate statistical results can be obtained. In particular, the error analysis stage can identify which sensor data behaves abnormally under specific environments or conditions, and then correct the impact of these data, thereby achieving refined data analysis. This makes the data on which the fire prediction model is based more reliable and accurate when inputting data, effectively reducing the possible risk of erroneous judgments. By constructing a fire prediction model and generating fire occurrence probability information based on the fused data, a comprehensive assessment of different fire risk levels can be achieved. It not only provides multi-dimensional fire risk assessment, but also simplifies complex data relationships into clear output levels by establishing a fire occurrence probability discrimination rule base. This invention achieves better results in terms of efficiency, accuracy, and reliability. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 is an overall flowchart of a cable fire prediction method based on cable thermal degradation characteristic gas concentration data provided in the first embodiment of the present invention.

[0020] Figure 2 shows a data fusion model optimized by K-Means clustering algorithm for a cable fire prediction method based on cable thermal degradation characteristic gas concentration data provided in the first embodiment of the present invention.

[0021] Figure 3 is a structural diagram of a fire prediction model based on a fuzzy neural network, which is a cable fire prediction method based on the concentration data of characteristic gases of cable thermal degradation provided in the first embodiment of the present invention.

[0022] Figure 4 is a schematic diagram of a cable fire prediction system based on cable thermal degradation characteristic gas concentration data provided in the third embodiment of the present invention. Detailed Implementation

[0023] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0024] Example 1, referring to Figures 1-3, is an embodiment of the present invention, providing a cable fire prediction method based on cable thermal degradation characteristic gas concentration data, including:

[0025] S1: Establish a data fusion model to process the first data.

[0026] Furthermore, a data fusion model includes feature extraction, preprocessing, preliminary fusion, and error analysis.

[0027] It should be noted that processing the first data includes feature extraction and preprocessing of the gas sensor cluster node data with different cable thermal degradation characteristics.

[0028] It should also be noted that a data fusion model can be established using various methods, such as K-Means clustering algorithm or outlier removal method based on IQR.

[0029] As a preferred embodiment, a data fusion model is established based on the K-Means clustering algorithm, and its principle and steps are as follows:

[0030] The data fusion model optimized based on the K-Means clustering algorithm is shown in Figure 2. First, the first data of the characteristic gas sensors of cable thermal degradation are collected in a centralized manner. The first data includes, but is not limited to, CO concentration data, CO2 concentration data, CH4 concentration data, H2 concentration data and smoke concentration data, etc., and feature extraction is performed. Then, simple and effective low-level preprocessing is carried out on the raw data collected by each sensor node to remove obvious outliers and ensure the accuracy of the data.

[0031] The preprocessed data will be sent to the cluster head node for initial fusion. In the initial fusion stage, the data will be filtered again to further remove any possible abnormal data. The processed sensor data will be sorted according to their numerical values, and the maximum, median and minimum values ​​will be determined and defined as cluster centers.

[0032] Next, all sensor data are categorized into the nearest cluster center according to the minimum distance principle, and the values ​​of each cluster center are recalculated. In this process, the three cluster centers identified will be used for error analysis, specifically to calculate the distances of the largest and smallest cluster centers to the middle cluster center.

[0033] Finally, based on the preset allowable error range, the distance from each cluster center to the intermediate cluster center is compared with the error range. If the distance from a cluster center to the intermediate cluster center exceeds the error range, the relevant data group will be deleted; otherwise, it will be retained and input into the fire prediction model. The data fusion model optimized based on the K-Means clustering algorithm ensures the reliability of the data, thereby providing a more accurate basis for subsequent fire prediction.

[0034] It should also be noted that, in another feasible embodiment, a data fusion model is established using an IQR-based outlier removal method, the principle and steps of which are as follows:

[0035] First, data from gas sensors identifying cable thermal degradation characteristics are collected, including but not limited to CO, CO2, CH4, H2, and smoke concentration data. Next, relevant features are extracted from the collected sensor data for further processing. Based on this, preliminary low-level preprocessing is performed on the collected data to remove obvious outliers. Then, the first quartile (Q1) and third quartile (Q3) of the dataset are calculated, and the interquartile range (IQR), i.e., Q3 minus Q1, is calculated. Based on the calculated IQR, the criterion for identifying outliers is a value lower than Q1 minus 1.5. Data with an outlier of 1.5 times the IQR or higher than Q3 will be removed to ensure the accuracy and authenticity of the remaining data. For the outlier-free data after this step, further data fusion is performed at the cluster head node. During the data fusion process, error analysis is performed on the fused data to ensure that the data error is within the allowable range, thereby enabling more accurate subsequent prediction analysis. The outlier removal and data fusion model based on the IQR method makes the final data more reliable and accurate through strict outlier identification and screening, effectively improving the accuracy and real-time performance of fire prediction.

[0036] S2: Perform data fusion and error analysis.

[0037] Furthermore, data fusion includes a data fusion model that performs preliminary fusion of preprocessed data at the cluster head node, further removing some abnormal data.

[0038] It should be noted that the error analysis includes classifying the sensor data received by the cluster head node based on a data fusion model and performing error analysis based on the error range.

[0039] It should also be noted that, based on a preferred data fusion model, this embodiment adopts a data fusion model optimized by the K-Means clustering algorithm. After the central node receives the data sent by each cluster member node, it assigns a reasonable weight value to the data of each cluster member node based on the degree of support of each cluster member node for the collected concentration data, considering that there may be some biased data in the data that the cluster member nodes cannot eliminate. High weight is assigned to data with high reliability, and low weight is assigned to data with low reliability, so as to achieve the best fusion effect.

[0040] S3: By establishing a fire prediction model, fire occurrence probability information is generated.

[0041] Furthermore, a fire prediction model includes determining multiple probabilities of fire risk and outputting the levels of no-fire probability P1, smoldering probability P2, and open flame probability P3 as high, medium, and low, and classifying the fire probability P into four levels: general, relatively high, serious, and extremely serious.

[0042] It should be noted that generating fire probability information involves establishing a fire probability discrimination rule base based on the fused data. There are 27 possible fire probability levels (P), including 5 moderate, 6 relatively high, 9 serious, and 7 extremely serious. When at least two of the fire probability levels (P1, P2, and P3) are low, and the fire probability level (P3) is neither medium nor high, the output fire probability level (P) is moderate. When at least two of the fire probability levels (P1, P2, and P3) are high, and the fire probability level (P2) is not low, or when the fire probability level (P1) is low, the fire probability level (P2) is high, and the fire probability level (P3) is medium, the output fire probability level (P) is extremely serious.

[0043] It should also be noted that there are various methods for establishing a fire prediction model, such as those based on fuzzy neural networks or decision trees.

[0044] As a preferred embodiment, a fire prediction model based on a fuzzy neural network is established in this embodiment, and its principle and steps are as follows:

[0045] Figure 3 shows the structure of a fire prediction model based on a fuzzy neural network. It combines fuzzy logic reasoning with a neural network, including an input layer, hidden layers, an output layer, a defined fuzzy set, and a fuzzy rule base. The fuzzy rule base is a rule base for determining the probability of fire occurrence. The neural network algorithm simulates the human brain's neural network. Humans quickly detect fires through visual thinking and can learn to adapt to various environmental changes, demonstrating strong fault tolerance. The fuzzy logic algorithm, designed to mimic human logical thinking, has strong recognition capabilities for nonlinear structural problems like fire prediction. Through the adaptive learning ability of the neural network's real-time feedback, combined with fuzzy rules, a fuzzy fire occurrence probability is output. Finally, after defuzzification, a more accurate prediction of the cable fire level is made.

[0046] Based on the data processed by the data fusion model, the number of nodes in the input layer of the neural network is determined to be 5, and the input variables are CO concentration, CO2 concentration, CH4 concentration, H2 concentration, and smoke concentration. At the same time, the number of nodes in the output layer of the neural network is determined to be 3, and the output variables are open flame probability P1, smoldering probability P2, and no-flame probability P3.

[0047] The number of hidden layer nodes is calculated as follows:

[0048] Where, n H Let n be the number of hidden layer nodes, n1 be the number of input nodes, n2 be the number of output nodes, and n be an integer between 1 and 10. Therefore, the number of hidden layer nodes is determined to be 8.

[0049] After determining the number of nodes in each layer of the neural network, the neural network model needs to be trained. Using the error gradient descent principle of neural networks, the training process involves continuous iteration to obtain the optimal weights and thresholds, thus refining the neural network structure. Specific steps include...

[0050] A1, Initialization: Select the maximum acceptable error Emax during neural network training, and set the weight coefficients w ij ,ν jk and threshold θ j r t Initialize to a small random value.

[0051] A2 is the input training sample (A1, Y1), where A1 is the input value and Y1 is the expected output value.

[0052] A3, calculate the hidden layer output, represented as:

[0053] Among them, b j Let f be the output of the j-th node in the hidden layer, f be the activation function used to introduce nonlinearity, and s be the output of the j-th node in the hidden layer. jLet w be the input signal of the j-th node in the hidden layer, m be the total number of input layer nodes, and w be the input signal of the j-th node in the hidden layer. ij Let x be the connection weight from the i-th node in the input layer to the j-th node in the hidden layer. i Let θ be the input value of the i-th node in the input layer. j is the threshold for the j-th node in the hidden layer, and p is the total number of nodes in the hidden layer.

[0054] A4, calculate the output of the output layer, represented as:

[0055] in, To output the predicted value, l k ν is the input signal of the k-th node in the output layer. jk Let r be the connection weight from the j-th node in the hidden layer to the k-th node in the output layer. k is the threshold for the k-th node of the output layer, and n is the total number of nodes in the output layer.

[0056] A5, calculate the correction error of each unit in the output layer, expressed as:

[0057] Where, d k Let be the correction error of the k-th node in the output layer. Let y be the predicted value of the k-th node in the output layer. k This is the expected output value.

[0058] A6, calculate the correction error of each element in the hidden layer, expressed as:

[0059] Among them, e j This represents the correction error of the j-th node in the hidden layer.

[0060] A7, calculate the new connection weights between the hidden layer and the output layer, represented as: v jk (L+1)=v jk (L)+αd k b j r k (L+1)=r k (L)+αd k

[0061] Among them, v jk (L) represents the connection weight from the j-th node in the hidden layer to the k-th node in the output layer, v jk (L+1) represents the updated connection weight from the j-th node in the hidden layer to the k-th node in the output layer, r k (L) represents the threshold of the k-th node in the current output layer, r k(L+1) is the threshold of the k-th node of the updated output layer, and α is the learning rate between the hidden layer and the output layer.

[0062] A8, calculate the new connection weights between the input layer and the hidden layer, represented as: w ij (L+1)=w ij (L)+βe j x i θ j (L+1)=θ j (L)+βe j

[0063] Among them, w ij (L) represents the connection weight from the i-th node in the current input layer to the j-th node in the hidden layer, w ij (L+1) represents the updated connection weight from the i-th node in the input layer to the j-th node in the hidden layer, θ j (L) is the threshold of the j-th node in the current hidden layer, θ j (L+1) is the threshold of the j-th node of the updated hidden layer, and β is the learning rate between the input layer and the hidden layer.

[0064] A9. Repeat steps A3 to A8 for the training samples from group 2 to N to complete the first training.

[0065] A10, begin the second training, repeating the process of the first training. Training ends when the error E is less than the target value, or when the number of training iterations reaches the maximum value.

[0066] After the neural network model is constructed, it is combined with the fuzzy logic algorithm. The input and output variables of the fuzzy logic algorithm are fuzzified. The input of the fuzzy logic algorithm is the output of the neural network, and the output is the probability of fire occurrence. The fuzzification levels of the probability of no fire P1, the probability of smoldering P2, and the probability of open flame P3 output by the neural network are defined as three levels: large (L), medium (M), and small (S). The output probability of fire P is defined as four levels of fuzzification: general (S), relatively large (M), serious (L), and extremely serious (B).

[0067] Define a fuzzy discrimination rule base, which is the fire occurrence probability discrimination rule base. Fuzzy discrimination rules are usually expressed in the form of "IF (a set of conditions) THEN (derive a set of results)". For example, when the no-fire probability level is S, the smoldering probability level is S, and the open flame probability level is S, the output fire occurrence probability is S. There are 27 possible results. See Table 1 for details.

[0068] Table 1 Fuzzy discrimination rule base

[0069] Finally, the centroid method is used to defuzzify the output fire occurrence probability to obtain an accurate fire detection result, expressed as:

[0070] Where u is the output of the defuzzification process, i.e., the centroid (center of gravity) of the fuzzy set, μ(u a ) is the membership function, representing the a-th element u. a In a fuzzy set, the membership degree is A, where A is the total number of elements in the fuzzy set.

[0071] It should also be noted that, in another feasible embodiment, a fire prediction model is established based on a decision tree, and its principle and steps are as follows:

[0072] First, the data preparation phase requires collecting various fire-related feature data, including but not limited to CO concentration, CO2 concentration, CH4 concentration, H2 concentration, and smoke concentration, processed by the data fusion model. Simultaneously, corresponding fire occurrence data needs to be obtained as label data for subsequent model training. By calculating the information gain or Gini coefficient of the features, features highly correlated with fire risk can be identified. For example, if CO concentration is found to have a significantly higher predictive effect on fire occurrence than other features, this feature should be preferentially selected as the root node of the decision tree. During the decision tree construction, the system will recursively partition the selected features from top to bottom, forming a series of decision nodes and subsets until the node purity condition is met, such as reaching a specific sample purity threshold. Once the threshold or information gain is reduced to a certain level, the decision tree will eventually generate multiple leaf nodes through this layer-by-layer analysis. Each leaf node corresponds to a set of fire probability prediction results. These results can be verified through cross-validation during the model evaluation phase. Evaluation metrics include classification accuracy and confusion matrix. To ensure the effectiveness and generalization ability of the model, it is important to implement pruning operations in the later stages of model building to remove unnecessary branches and reduce the risk of overfitting. After the model training is completed, the decision tree can process specific input data and make real-time fire probability predictions based on decision rules for newly collected gas concentration data. Finally, it outputs the fire occurrence probability in the form of fuzzy levels to help decision-makers assess and respond to potential fire risks.

[0073] Example 2 is an embodiment of the present invention, which provides a cable fire prediction method based on the concentration data of characteristic gases of cable thermal degradation. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiment.

[0074] In this embodiment, a data fusion model based on the K-Means clustering algorithm was established, mainly used to process the data collected by the gas sensor for cable thermal degradation characteristics, so as to improve the accuracy and reliability of fire prediction. First, the experimental group selected multiple advanced gas sensors, specifically including CO, CO2, CH4, H2 and smoke concentration sensors, and constructed a multi-sensor network for collecting fire-related data. These sensors were distributed in a real climate environment to obtain real-time data. The experiment lasted for 24 hours.

[0075] After data collection, feature extraction is performed first, and a basic dataset is constructed based on the gas concentration values ​​collected by the sensors. Then, low-level preprocessing is performed on the raw data. Based on the concentration value of each gas, obvious outliers are removed to improve the accuracy and reliability of the data. During this process, preliminary statistical analysis is performed on all gas concentration data to ensure that the removed outliers meet the standard deviation range.

[0076] The preprocessed data is initially fused by the cluster head node. At this stage, the K-Means algorithm is applied to further cluster the data, identify central trends, and assign reasonable weight values ​​to each detected cluster. By assigning higher weights to data with high reliability, it is ensured that the final retained data can accurately reflect the real environmental conditions. After the fusion is completed, the central node analyzes the data of each cluster and performs detailed error analysis on the clustering results to optimize the final output dataset and ensure that it is suitable for subsequent fire risk prediction.

[0077] Next, a fire prediction model based on a fuzzy neural network was constructed using the obtained fused data. This model combines fuzzy logic reasoning with a neural network. The input layer has five nodes, representing the concentration variables of CO, CO2, CH4, H2, and smoke, respectively. The number of hidden layer nodes was calculated using an empirical formula and determined to be eight to enhance the network's learning ability and fault tolerance. By setting connection weights and thresholds, the neural network can continuously adjust while learning from new data, ensuring that the output layer reflects the accurate values ​​of the probability of open flame, smoldering, and no-flame.

[0078] Finally, in the output stage of the model, fuzzy logic algorithms are used to fuzzify the output of the neural network. By defining fuzzification levels and corresponding fuzzy discrimination rule bases, the effective output of the fire occurrence probability is finally achieved. The centroid method is used for defuzzification processing, thereby obtaining more accurate fire risk discrimination results, thus providing an important basis for fire prevention and response.

[0079] Refer to Table 2 for comparative analysis of the experimental data.

[0080] Table 2 Experimental Data Recording Table

[0081] Analysis of the experimental data in Table 2 reveals that the data fusion model optimized using the K-Means clustering algorithm in this embodiment demonstrates advantages over traditional fire early warning systems. Table 2 shows the gas concentration data collected by different sensors within the same time period and their corresponding risk levels. First, the gas concentrations recorded by sensors A, B, and E are all within a relatively safe range, with a low risk level, indicating that the fire risk in these areas is relatively small. However, the data from sensor C shows a medium risk level, especially with a certain increase in CO, CO2, and smoke concentrations, suggesting a gradual increase in fire hazards in this area. Furthermore, the data from sensors D and F show high and particularly serious risk levels, with CO concentration being particularly prominent, reflecting an increased fire risk at this point that requires attention.

[0082] This data analysis clearly demonstrates that implementing a data fusion model based on the K-Means clustering algorithm ensures reliable data output, a feature unavailable in traditional fire early warning systems. In particular, the innovative methods of outlier removal and data weighting effectively improve data accuracy and usability. Especially in multi-sensor systems, the diversity and complexity of data often present challenges to traditional methods, such as poor optimization and adaptability. The K-Means clustering-optimized data fusion model of this invention, through precise data fusion and weight allocation, fully leverages the advantages of data acquired from different sensors.

[0083] Finally, a fire prediction model based on fuzzy neural networks was constructed, making the determination of fire probability more flexible and adaptable. Through the reasoning mechanism of fuzzy logic, the network can still achieve efficient self-learning ability even when facing uncertain and complex environmental changes, providing a more accurate basis for the prevention and control of cable fires. Experimental results show that the combination of the improved data fusion model and prediction algorithm significantly improves the accuracy and timeliness of cable fire prediction. This innovative solution effectively overcomes the shortcomings of existing technologies and has broad application prospects and market value.

[0084] Example 3, referring to Figure 4, is an embodiment of the present invention, providing a cable fire prediction system based on cable thermal degradation characteristic gas concentration data, including a data processing module, an error analysis module, and a fire probability prediction module.

[0085] The data processing module is used to establish a data fusion model, extract features and preprocess the first data; the error analysis module is used for data fusion and error analysis, classifying and analyzing the processed data; the fire probability prediction module is used to establish a fire prediction model, generate fire occurrence probability information based on the fused data, output no-fire probability, smoldering probability and open flame probability, and finally evaluate the fire probability, outputting fire level probability information of general, large, serious and extremely serious fire.

[0086] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0087] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0088] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0089] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc. It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0090] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A cable fire prediction method based on cable thermal degradation characteristic gas concentration data, characterized by, include: Establish a data fusion model to process the first set of data; Perform data fusion and error analysis; By establishing a fire prediction model, fire occurrence probability information is generated.

2. The cable fire prediction method based on cable thermal degradation characteristic gas concentration data according to claim 1, characterized by: The data fusion model includes feature extraction, preprocessing, preliminary fusion, and error analysis.

3. The cable fire prediction method based on cable thermal degradation signature gas concentration data as recited in claim 2, wherein: The processing of the first data includes feature extraction and preprocessing of the node data of gas sensor clusters with different cable thermal degradation characteristics.

4. The cable fire prediction method based on cable thermal degradation characteristic gas concentration data according to claim 3, characterized by: The data fusion includes preliminary fusion of preprocessed data at the cluster head node based on a data fusion model.

5. The cable fire prediction method based on cable thermal degradation signature gas concentration data as recited in claim 4, wherein: The error analysis includes classifying sensor data and performing error analysis based on a data fusion model.

6. The cable fire prediction method based on cable thermal degradation signature gas concentration data as recited in claim 5, wherein: The fire prediction model includes determining various probabilities of fire risk, outputting the levels of no-fire probability P1, smoldering probability P2, and open flame probability P3 as large, medium, and small, and classifying the fire probability P into four levels: general, relatively large, serious, and extremely serious.

7. The cable fire prediction method based on cable thermal degradation signature gas concentration data as recited in claim 6, wherein: The generation of fire occurrence probability information includes establishing a fire occurrence probability discrimination rule base based on the fused data. There are 27 possible fire probability levels (P), including 5 general levels, 6 relatively high levels, 9 major levels, and 7 extremely major levels. When at least two of the three probabilities of no fire (P1), smoldering (P2), and open flame (P3) are of low level, and the open flame probability (P3) is neither medium nor large level, the output fire probability level (P) is moderate. When at least two of the following probabilities—no fire probability P1, smoldering probability P2, and open flame probability P3—are large, and the open flame probability P2 is not small, or when the no fire probability P1 is small, the smoldering probability P2 is large, and the open flame probability P3 is medium, the output fire probability P level is extremely serious.

8. A system employing the cable fire prediction method based on the concentration data of the gas generated by thermal degradation of the cable according to any one of claims 1 to 7, characterized by: Includes a data processing module, an error analysis module, and a fire probability prediction module; The data processing module is used to establish a data fusion model and perform feature extraction and data preprocessing on the first data. The error analysis module is used for data fusion and error analysis, and classifies and analyzes the processed data for errors. The fire probability prediction module is used to establish a fire prediction model, generate fire occurrence probability information based on the fused data, output no-fire probability, smoldering probability and open flame probability, and finally evaluate the fire probability, outputting fire level probability information of general, large, serious and extremely serious. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. When the processor executes the computer program, it implements the steps of the cable fire prediction method based on cable thermal degradation characteristic gas concentration data as described in any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the cable fire prediction method based on cable thermal degradation characteristic gas concentration data as described in any one of claims 1 to 7.

Images