Substation early fire early warning method based on multi-source data fusion

Abstract

The invention provides a substation fire early warning method based on multi-source data fusion, and relates to various machine learning technologies. The method mainly comprises the following steps: arranging an integrated terminal comprising a temperature sensor, a humidity sensor, a smoke sensor and a gas sensor at narrow positions such as a transformer substation ground cable groove and a switch cabinet; preprocessing the collected data by using a Grubbs criterion, a forward interpolation fitting method and a PCA principal component analysis method to ensure the reliability of the data; a fuzzy neural network based on multi-source data is constructed, membership functions are applied to all input variables, and membership degrees of the input variables are calculated; and fusing and inputting the multi-source data into the constructed fuzzy neural network, and performing graded early warning by generating a corresponding state probability. The method provided by the invention has relatively high reliability and accuracy, realizes real-time detection in the smoldering state at the early stage of the fire, reduces the occurrence of electrical fire accidents, and ensures the safe and stable operation of the transformer substation.

Description

Technical Field

[0001] The present invention belongs to the technical field of machine learning, and in particular, relates to an early fire warning method for a substation based on multi-source data fusion. Background Art

[0002] Substations are a vital component of the power system. Their functional areas are clearly divided, and similar equipment is usually arranged in a centralized manner. Although this layout is helpful for management and maintenance, it also has certain risks. In particular, in the event of equipment failure, a fire may occur. If the fire is not promptly and effectively controlled in the early stage, the fire may spread rapidly, further threatening surrounding equipment, and eventually leading to chain failures, resulting in immeasurable economic losses and safety hazards.

[0003] Therefore, for the fire protection work of substations, it is crucial to quickly and accurately identify the fire in the early stage of the fire and take effective fire-fighting measures in time. Early fire control can not only prevent the spread of the fire, but also minimize equipment losses and economic damages, and ensure the safety and stable operation of the power system. Establishing an efficient fire monitoring and emergency response system is one of the key measures to ensure the safety of substations and reduce potential risks, which is also of great significance to the reliability of the entire power system.

[0004] In the prior art, for example, the publication number CN114743335A, the invention name “Cable trench fire warning method based on fuzzy neural network” discloses a fire warning method based on fuzzy neural network, which improves the ability to distinguish fire conditions in complex environments by constructing a fuzzy neural network; for example, the publication number CN115641688A, the invention name “Fire very early detection warning alarm system and method” discloses an intelligent fire warning model, which realizes multi-modal real-time monitoring by linking different devices to a variety of visible light sensors; it can effectively avoid false fire alarms and greatly improve the accuracy of fire detector warning alarms.

[0005] Given that traditional fire warnings are prone to false alarms and rely on manual verification, which easily misses the best time to deal with them, and that fire hazards in small and confined spaces (such as switch cabinets and cable trenches) are not discovered and located in a timely manner, it is urgent to propose some relevant methods to detect and warn of the occurrence of substation fire accidents at an extremely early stage.

[0006] In view of this, the present invention is proposed. Summary of the invention

[0007] The technical problem to be solved by the present invention is to overcome the deficiencies of the prior art and provide a substation early fire warning method based on multi-source data fusion, which solves the problems raised in the above background technology.

[0008] In order to solve the above technical problems, the basic concept of the technical solution adopted by the present invention is:

[0009] A substation early fire warning method based on multi-source data fusion includes the following steps:

[0010] Step 1: Arrange integrated terminals including temperature sensors, humidity sensors, smoke sensors, and gas sensors in narrow locations such as substation cable trenches and switch cabinets;

[0011] Step 2: Use the Grubbs criterion, forward interpolation fitting method and PCA principal component analysis method to preprocess the collected data to ensure the reliability of the data; among them, use the Grubbs criterion to eliminate the collected abnormal data. The Grubbs method assumes that the measurement results obey the normal distribution and determines the selection of suspicious data based on the order statistics. Repeat the test n times and the measured results are x1, x2…, xi,…xn, which obey the normal distribution.

[0012] In order to check whether there is suspicious data in xi (i = 1, 2, ..., n), xi can be rearranged in order from small to large, and the result is: x(1) ≤ x(2) ≤ ... ≤ x(n). According to the principle of order statistics, the standardized order statistic g is given:

[0013] When the minimum value is suspicious x(1), then

[0014] When the maximum value x(n) is suspected, then

[0015] Where: —arithmetic mean of the measured values; S—standard deviation of the measured values.

[0016] Sample standard deviation Population standard deviation

[0017] The forward interpolation method uses the Lagrange interpolation method and the Lagrange interpolation polynomial for fitting. This method constructs an interpolation polynomial containing all known data points to fit and estimate the target function. The interpolation polynomial formula is:

[0018]

[0019] where x i ,y i These are known data points, including temperature, humidity, smoke, pyrolytic particles, and other data.

[0020] Finally, the PCA principal component analysis method is used to perform dimensionality reduction analysis on the data. The principal component analysis first calculates the covariance matrix of the data matrix, and then obtains the eigenvalues ​​and eigenvectors of the covariance matrix, and selects the matrix composed of the eigenvectors corresponding to the k features with the largest eigenvalues. In this way, the data matrix can be converted to a new space to achieve dimensionality reduction of data features. There are two methods to solve the covariance matrix. One is the eigenvalue decomposition matrix. For the matrix H, there is a set of eigenvectors R. By orthogonalizing and normalizing this set of vectors, a set of orthogonal unit vectors can be obtained. Finally, the new data is input into the fuzzy neural network;

[0021] Step 3: Construct a fuzzy neural network based on multi-source data, apply membership functions to all input variables, and calculate their membership; wherein, construct a fuzzy neural network based on multi-source data, collect and preprocess data from different sources, and convert them into a format that can be used for model input. Next, select the Gaussian membership function to represent the uncertainty of each input variable. The function is used to map the input variable to a membership between 0 and 1 to reflect its degree of belonging relative to the fuzzy set. Then, apply the membership function to each input variable, calculate its membership, and form a membership vector as the input layer data of the fuzzy neural network. After that, the input membership is processed and learned through the hierarchical structure of the fuzzy neural network, including the fuzzification layer, the inference layer, and the output layer, and the output result is gradually derived. During the model training process, the parameters of the network need to be adjusted according to the objective function to achieve a nonlinear mapping relationship between input and output, thereby improving the accuracy and robustness of the system when processing complex multi-source data. Data input into the fuzzy neural network needs to go through a total of five layers of network structure, starting with the input layer. The number of nodes in this layer is the characteristic dimension of the input data, that is, when the characteristic dimension of the data is n, the number of nodes in the input layer is n.

[0022] Then the data is passed from the input layer to the membership function calculation layer. In this layer, the membership function is needed to calculate the membership of each node. Each node represents a membership function. The number of nodes in this layer is the number of fuzzy conditions that the input variables may constitute. The membership function generally uses the Gaussian function, and the formula is as follows:

[0023]

[0024] where u ij is the output of the node, x i is the input data, c ij is the center point of the membership function, b ij is the width vector of the membership function, and m is the number of fuzzy classifications of the input. The data is then passed to the rule generation layer. The number of nodes in this layer is the number of fuzzy rules. The output of this layer is the applicability of each rule, which is generally calculated using the following formula:

[0025] α j =Πu ij

[0026] Next, the node outputs of the rule generation layer are normalized in the normalization layer. The nodes in this layer are the same as those in the rule generation layer, and are normalized using the following formula:

[0027]

[0028] Finally, the data is weighted in the output layer to obtain the final output. The calculation formula is:

[0029]

[0030] where ω j is the connection weight of the output layer. The above formula is the weight set when the dimension of the output variable of the network is 1. When the dimension of the output variable increases, the weight will be adjusted accordingly.

[0031] Step 4: Fusion of multi-source data is input into the constructed fuzzy neural network, and graded warning is performed by generating corresponding state probabilities; wherein, the processed data is input into the constructed fuzzy neural network, making full use of the information provided by each data source to achieve a comprehensive assessment of the target state. Use the Gaussian membership function to fuzzify each input variable, convert the numerical data into a fuzzy set, calculate its membership degree to reflect the uncertainty and ambiguity of the input data, forward propagate the fuzzified input data through the fuzzy neural network, combine fuzzy rules for reasoning, and calculate the final output result layer by layer. The output of the neural network reflects the current state of the system, and evaluates the health status of the system by generating corresponding state probabilities. According to the size of each state probability, the current state of the system is divided into the following categories: Figure 2-4 There are three levels: no fire, smoldering, and open flame. Smoldering is subdivided into ultra-early smoldering and early smoldering. The output probabilities of the above stages are used to ultimately determine the type of substation fire, thereby achieving system status monitoring and graded early warning.

[0032] Step 5: Combine the hazard level and infrared camera data for comprehensive detection and identification to ensure the safety and stable operation of the substation; among them, the infrared camera collects the thermal imaging data of the equipment operation in real time. These infrared images can directly reflect the temperature distribution of the equipment, which is convenient for quickly identifying the overheating phenomenon and other potential abnormalities of the equipment. By processing and analyzing the infrared images, extracting characteristic information such as temperature hot spots, the health status of the equipment is visually evaluated. Combining the hazard level and infrared camera data, we jointly identify the operating status of the equipment in the substation. When the sensor data indicates that there is a potential risk in the system, the infrared camera data is analyzed to further confirm whether the specific equipment has abnormal conditions such as overheating or failure, so as to improve the accuracy of risk detection. This comprehensive detection method can not only generate the hazard level according to the environmental parameters, but also confirm the risk point by combining the visual information of the infrared image, realize double verification, and provide a more reliable basis for judging the system status. Once a high hazard level is detected and accompanied by an abnormal infrared image, the system will issue a warning signal and prompt the specific risk location and cause, so that the management personnel can take corresponding treatment measures in time to effectively avoid safety hazards.

[0033] After adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art. Of course, any product implementing the present invention does not necessarily need to achieve all the advantages described below at the same time:

[0034] 1. The present invention improves the ability of model feature recognition and better detects in low-light environments. The model method performs well in both detection speed and accuracy, can effectively identify and locate abnormal power equipment in substations, realize real-time detection of equipment, and provide reliable technical support for maintenance and fault diagnosis of power systems.

[0035] 2. Through comprehensive analysis and fusion of multi-source data, the present invention enables the fuzzy neural network to more accurately judge the degree of danger of the current environment, and output the corresponding state probability for graded warning. Finally, it combines with infrared thermal imaging sensors for comprehensive judgment, realizes intelligent monitoring and dynamic warning of substations, and effectively improves the safety and stability of the system.

[0036] The specific implementation modes of the present invention are further described in detail below in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The drawings described below are only some embodiments. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0038] In the figure:

[0039] Figure 1It is a flow chart of the present invention;

[0040] Figure 2 The probability results for each level - no fire state;

[0041] Figure 3 The probability results of each level - open flame status;

[0042] Figure 4 The probability results for each level - smoldering state.

[0043] It should be noted that these drawings and textual descriptions are not intended to limit the conceptual scope of the present invention in any way, but are intended to illustrate the concept of the present invention for those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0044] The present invention will now be described in further detail with reference to the accompanying drawings.

[0045] See also Figure 1-4 As shown, in this embodiment, a substation early fire warning method based on multi-source data fusion is provided, including the following steps:

[0046] Step 1: Arrange integrated terminals including temperature sensors, humidity sensors, smoke sensors, and gas sensors in narrow locations such as cable trenches and switch cabinets in substations; specifically, it is necessary to determine the locations suitable for installing sensors in cable trenches and switch cabinets to ensure that the sensors can accurately collect environmental data, especially to achieve effective monitoring in narrow spaces. Next, select sensors suitable for high electromagnetic interference environments. These sensors must have the ability to work stably for a long time in strong electromagnetic interference and narrow areas to ensure the accuracy and reliability of the measured data. Then, combine the temperature, humidity, smoke, and gas sensors into a compact terminal to reduce complex wiring requirements, reduce space occupancy, and improve installation convenience. Then, carry out wiring and connection, fix the cables of the integrated terminal in appropriate positions, ensure that the wiring does not affect the normal operation of the switch cabinet and other equipment, and ensure the neatness and safety of equipment management. During the entire installation process, we need to fully consider the safety protection measures in narrow spaces to ensure that the sensors can operate stably, while preventing potential safety hazards to equipment or personnel, and ensure that the system can achieve long-term and reliable safety monitoring in the substation;

[0047] Step 2: Use the Grubbs criterion, forward interpolation fitting method and PCA principal component analysis method to preprocess the collected data to ensure the reliability of the data; the specific steps of preprocessing the collected data using the Grubbs criterion include:

[0048] The Grubbs criterion is used to eliminate the abnormal data collected. The Grubbs method assumes that the measurement results obey the normal distribution and determines the selection of suspicious data based on the order statistics. Repeat the test n times and the measured results are x1, x2…, xi,…xn, which obey the normal distribution.

[0049] To check whether there is suspicious data in xi (i = 1, 2, ..., n), xi can be rearranged in order from small to large, and we can get: x(1) ≤ x(2) ≤ ... ≤ x(n). According to the principle of order statistics, the standardized order statistic g is given:

[0050] When the minimum value is suspicious x(1), then

[0051] When the maximum value x(n) is suspected, then

[0052] Where: —arithmetic mean of the measured values; S—standard deviation of the measured values.

[0053] Sample standard deviation Population standard deviation

[0054] The forward interpolation fitting method is the Lagrange interpolation method, and the specific interpolation formula is:

[0055]

[0056] where x i ,y i are known data points, including temperature, humidity, smoke, and pyrolytic particle data;

[0057] The PCA principal component analysis method is used to perform dimensionality reduction analysis on the data. The specific process is to first calculate the covariance matrix of the data matrix, then obtain the eigenvalues ​​and eigenvectors of the covariance matrix, and select the matrix composed of the eigenvectors corresponding to the k features with the largest eigenvalues. The data matrix can be converted to a new space to achieve dimensionality reduction of data features.

[0058] Step 3: Construct a fuzzy neural network based on multi-source data, apply membership functions to all input variables, and calculate their memberships; after constructing the fuzzy neural network based on multi-source data, the different data collected and preprocessed will be converted into a format that can be used for model input. The membership function is a Gaussian function, which can represent the uncertainty of each input variable. The Gaussian function is used to map the input variable to a membership between 0 and 1 to reflect its degree of belonging to the fuzzy set. The membership function is applied to each input variable, and its membership is calculated to form a membership vector as the input layer data of the fuzzy neural network; then, through the hierarchical structure of the fuzzy neural network, including the fuzzification layer, the inference layer, and the output layer, the input membership is processed and learned, and the output results are gradually derived. During the model training process, the parameters of the network need to be adjusted according to the objective function to achieve a nonlinear mapping relationship between input and output, thereby improving the accuracy and robustness of the system when processing complex multi-source data;

[0059] The data input into the fuzzy neural network needs to go through a total of five layers of network structure. The first is the input layer. The number of nodes in the input layer is the characteristic dimension of the input data. That is, when the characteristic dimension of the data is n, the number of nodes in the input layer is n. Then the data is transferred from the input layer to the membership function calculation layer. The membership function is used in the membership function calculation layer to calculate the membership of each node. Each node represents a membership function. The number of nodes in the membership function calculation layer is the number of fuzzy conditions composed of the input variables; Gaussian function, the formula is as follows:

[0060]

[0061] where u ij is the output of the node, x i is the input data, c ij is the center point of the membership function, b ij is the width vector of the membership function, and m is the number of fuzzy classifications of the input; the data is passed to the rule generation layer, the number of nodes in the rule generation layer is the number of fuzzy rules, and the output of the rule generation layer is the applicability of each rule. The specific calculation formula is as follows:

[0062] α j =Πu ij

[0063] Next, the node output of the rule generation layer is normalized in the normalization layer. The nodes of the normalization layer are the same as those of the rule generation layer. The specific calculation formula for normalization is as follows:

[0064]

[0065] Finally, the data is weighted in the output layer to obtain the final output. The specific calculation formula is as follows:

[0066]

[0067] where ω j is the connection weight of the output layer. The above formula is the weight set when the dimension of the output variable of the network is 1. When the dimension of the output variable increases, the weight will be adjusted accordingly.

[0068] Step 4: Input the multi-source data into the constructed fuzzy neural network, and generate the corresponding state probabilities for graded warning; input the processed data into the constructed fuzzy neural network, use the information provided by each data source to achieve a comprehensive evaluation of the target state, use the Gaussian membership function to fuzzify each input variable, convert the numerical data into a fuzzy set, and calculate the membership degree to reflect the uncertainty and ambiguity of the input data; the fuzzified input data is forward propagated through the fuzzy neural network, combined with fuzzy rules for reasoning, and the final output result is calculated layer by layer; the output of the neural network reflects the current state of the system, and evaluates the health state of the system by generating corresponding state probabilities; according to the size of each state probability, the current state of the system is divided into three levels: no fire, smoldering, and open flame, among which smoldering is subdivided into ultra-early smoldering and early smoldering states. The output probabilities of the above stages are used to finally determine the type of substation fire, and the system status monitoring and graded warning are achieved;

[0069] Step 5: Combine the hazard level and infrared camera data for comprehensive detection and identification to ensure safe and stable operation in the substation. To ensure safe and stable operation in the substation, the hazard level and infrared camera data are combined for comprehensive detection and identification. The infrared camera is used to monitor the operating status of the substation in real time. Through infrared image data, equipment overheating and line abnormalities are quickly detected and responded to. The infrared camera data is analyzed through image processing technology to extract temperature hot spots and abnormal image feature information, and combined with sensor data and fuzzy neural network results, they are used together for comprehensive evaluation of system status to achieve substation fire warning with multi-source data fusion.

[0070] Through comprehensive detection and identification of hazard levels and infrared camera data, we can dynamically monitor the operating status of the substation, promptly detect and deal with potential faults and safety issues, and ensure the safe and stable operation of the entire substation in a complex environment.

[0071] The present invention proposes a substation early fire warning method based on multi-source data fusion to reduce the occurrence of substation fire accidents. This method enables the fuzzy neural network to more accurately judge the danger level of the current environment through comprehensive analysis and fusion of multi-source data, and output the corresponding state probability for graded warning, and finally combines with infrared thermal imaging sensors for comprehensive judgment. Intelligent monitoring and dynamic warning of substations are realized, effectively improving the safety and stability of the system.

[0072] The present invention is not limited to the above-mentioned embodiments. Anyone should be aware that any structural changes made under the enlightenment of the present invention, and any technical solutions that are the same or similar to the present invention, fall within the protection scope of the present invention. The technology, shape, and structural parts not described in detail in the present invention are all well-known technologies.

Claims

1. A substation early fire warning method based on multi-source data fusion, characterized in that: The steps include: Step 1: Arrange integrated terminals including temperature sensors, humidity sensors, smoke sensors, and gas sensors in narrow locations such as substation cable trenches and switch cabinets; Step 2: Use Grubbs criterion, forward interpolation fitting method and PCA principal component analysis method to preprocess the collected data to ensure the reliability of the data; Step 3: Construct a fuzzy neural network based on multi-source data, apply membership functions to all input variables, and calculate their membership degrees; Step 4: Fusion of multi-source data is input into the constructed fuzzy neural network, and hierarchical warning is performed by generating corresponding state probabilities; Step 5: Combine the danger level and infrared camera data to conduct comprehensive detection and identification to ensure safe and stable operation within the substation.

2. According to claim 1, a substation early fire warning method based on multi-source data fusion is characterized in that: In step 2, the specific steps of preprocessing the collected data using the Grubbs criterion include: The Grubbs criterion is used to eliminate the abnormal data collected. The Grubbs method assumes that the measurement results obey the normal distribution and determines the selection of suspicious data based on the order statistics. Repeat the test n times and the measured results are x1, x2…, xi,…xn, which obey the normal distribution. To check whether there is suspicious data in xi (i = 1, 2, ..., n), xi can be rearranged in order from small to large, and we can get: x(1) ≤ x(2) ≤ ... ≤ x(n). According to the principle of order statistics, the standardized order statistic g is given: When the minimum value is suspicious x(1), then When the maximum value x(n) is suspected, then Where: —arithmetic mean of the measured values; S—standard deviation of the measured values. Sample standard deviation Population standard deviation 3. According to claim 1, a substation early fire warning method based on multi-source data fusion is characterized in that: In step 2, the forward interpolation fitting method is the Lagrange interpolation method, and the specific interpolation formula is: where x i ,y i These are known data points, including temperature, humidity, smoke, and pyrolytic particle data.

4. The substation early fire warning method based on multi-source data fusion according to claim 1 is characterized in that: In step two, the PCA principal component analysis method is used to perform dimensionality reduction analysis on the data. The specific process is to first calculate the covariance matrix of the data matrix, then obtain the eigenvalues ​​and eigenvectors of the covariance matrix, and select the matrix composed of the eigenvectors corresponding to the k features with the largest eigenvalues. In this way, the data matrix can be converted to a new space to achieve dimensionality reduction of data features.

5. The substation early fire warning method based on multi-source data fusion according to claim 1 is characterized in that: In step three, after constructing the fuzzy neural network based on multi-source data, the collected and pre-processed different data are converted into a format that can be used for model input.

6. The substation early fire warning method based on multi-source data fusion according to claim 1 is characterized in that: In step three, the membership function is a Gaussian function, which can represent the uncertainty of each input variable. The Gaussian function is used to map the input variable to a membership between 0 and 1 to reflect its degree of belonging to the fuzzy set. The membership function is applied to each input variable, and its membership is calculated to form a membership vector as the input layer data of the fuzzy neural network. After that, the input membership is processed and learned through the hierarchical structure of the fuzzy neural network, including the fuzzification layer, the inference layer and the output layer, and the output result is gradually derived. During the model training process, the parameters of the network need to be adjusted according to the objective function to achieve a nonlinear mapping relationship between input and output, thereby improving the accuracy and robustness of the system when processing complex multi-source data.

7. A substation early fire warning method based on multi-source data fusion according to claim 6, characterized in that: Data input into the fuzzy neural network needs to go through a total of five layers of network structure, starting with the input layer. The number of nodes in the input layer is the characteristic dimension of the input data, that is, when the characteristic dimension of the data is n, the number of nodes in the input layer is n. Then the data is passed from the input layer to the membership function calculation layer, where the membership function is used to calculate the membership of each node. Each node represents a membership function, and the number of nodes in the membership function calculation layer is the number of fuzzy conditions composed of the input variables.

8. The substation early fire warning method based on multi-source data fusion according to claim 7 is characterized in that: Gaussian function, the formula is as follows: where u ij is the output of the node, x i is the input data, c ij is the center point of the membership function, b ij is the width vector of the membership function, and m is the number of fuzzy classifications of the input; the data is passed to the rule generation layer, the number of nodes in the rule generation layer is the number of fuzzy rules, and the output of the rule generation layer is the applicability of each rule. The specific calculation formula is as follows: α j ＝∏u ij Next, the node output of the rule generation layer is normalized in the normalization layer. The nodes of the normalization layer are the same as those of the rule generation layer. The specific calculation formula for normalization is as follows: Finally, the data is weighted in the output layer to obtain the final output. The specific calculation formula is as follows: where ω j is the connection weight of the output layer. The above formula is based on the weight set when the dimension of the output variable of the network is 1. When the dimension of the output variable increases, the weight will be adjusted accordingly.

9. The substation early fire warning method based on multi-source data fusion according to claim 1 is characterized in that: In step 4, the processed data is input into the constructed fuzzy neural network, and the information provided by each data source is used to achieve a comprehensive evaluation of the target state. The Gaussian membership function is used to fuzzify each input variable, and the numerical data is converted into a fuzzy set. The membership degree is calculated to reflect the uncertainty and ambiguity of the input data; the fuzzified input data is forward propagated through the fuzzy neural network, combined with fuzzy rules for reasoning, and the final output result is calculated layer by layer; the output of the neural network reflects the current state of the system, and the health state of the system is evaluated by generating corresponding state probabilities; according to the size of each state probability, the current state of the system is divided into three levels: no fire, smoldering, and open flame, among which smoldering is subdivided into ultra-early smoldering and early smoldering states. The output probabilities of the above stages are finally used to determine the type of substation fire, and the system state monitoring and graded early warning are realized.

10. The substation early fire warning method based on multi-source data fusion according to claim 1 is characterized in that: In step five, to ensure the safety and stable operation of the substation, comprehensive detection and identification are achieved by combining the danger level and infrared camera data. The infrared camera is used to monitor the operation status of the substation in real time; Through infrared image data, equipment overheating and line abnormalities can be quickly detected and responded to; The data from the infrared camera is analyzed through image processing technology to extract temperature hot spots and abnormal image feature information. The information is combined with sensor data and the results of the fuzzy neural network and used together for a comprehensive assessment of the system status to achieve substation fire warning based on multi-source data fusion.

Images