Building intelligent electrical fire monitoring system and method

Abstract

This application relates to the field of intelligent fire monitoring, and provides a building intelligent electrical fire monitoring system and method. It employs deep learning-based artificial intelligence data processing technology to monitor for electrical fires by real-time analysis of temperature values ​​and leakage currents of distribution boxes on different floors of the target monitoring area. Upon detection of an electrical fire, it promptly sends alarm information to the fire control center. This achieves automated and intelligent electrical fire monitoring, improving its accuracy and timeliness.

Description

Technical Field

[0001] This application relates to the field of intelligent fire monitoring, and more specifically, to a building intelligent electrical fire monitoring system and method. Background Technology

[0002] In today's rapidly developing society, electrical equipment is increasingly used in various fields, bringing great convenience to people's production and lives. However, the hidden dangers of electrical fires are also ever-present, becoming a safety issue that cannot be ignored.

[0003] Traditional electrical fire monitoring systems have numerous drawbacks in practical applications. Firstly, they often rely on manual inspections or periodic checks, which not only requires significant manpower but also struggles to achieve real-time, continuous monitoring of critical electrical equipment like distribution boxes due to limited human energy and time. In practice, manual inspections can lead to oversights and misjudgments, failing to promptly identify potential electrical fire hazards. Secondly, existing data processing methods are ineffective at handling large volumes of high-dimensional electrical monitoring data. With the increasing number of electrical devices, the scale of monitoring data is constantly expanding. Traditional data processing methods are struggling to cope with such massive amounts of data, and the problem of untimely information processing is becoming increasingly prominent. This results in the inability to quickly and accurately obtain crucial information in the early stages of an electrical fire, leading to a delayed response and posing a significant threat to people's lives and property.

[0004] Therefore, an optimized intelligent electrical fire monitoring solution for buildings is needed. Summary of the Invention

[0005] This application addresses the shortcomings of existing technologies by providing a building intelligent electrical fire monitoring system and method.

[0006] According to one aspect of this application, a building intelligent electrical fire monitoring system is provided, comprising:

[0007] The electrical fire monitoring data acquisition module is used to acquire the leakage current and temperature values ​​of the distribution boxes at multiple predetermined time points on different floors within the target monitoring area.

[0008] The electrical fire monitoring related data regularization module is used to arrange the leakage current value and temperature value of the distribution box at multiple predetermined time points on different floors according to the time dimension and the sample dimension to obtain the time sequence input tensor of electrical fire monitoring parameters.

[0009] An electrical fire monitoring related data feature encoding module is used to perform feature encoding on the time-series input tensor of the electrical fire monitoring parameters to obtain a time-series feature map of the dynamic changes of the electrical fire monitoring parameters.

[0010] An electrical fire monitoring related data feature enhancement module is used to enhance the features of the time-series feature map of the dynamic change of the electrical fire monitoring parameters to obtain an electrical fire monitoring enhancement feature map;

[0011] The alarm result generation module is used to obtain alarm results based on the electrical fire monitoring emphasis feature map.

[0012] According to another aspect of this application, a building intelligent electrical fire monitoring method is provided, comprising:

[0013] Obtain leakage current and temperature values ​​of distribution boxes on different floors within the target monitoring area at multiple predetermined time points within a predetermined time period;

[0014] The leakage current and temperature values ​​of the distribution boxes on different floors at multiple predetermined time points are arranged according to the time dimension and the sample dimension to obtain the time sequence input tensor of electrical fire monitoring parameters.

[0015] The time-series input tensor of the electrical fire monitoring parameters is feature-encoded to obtain a time-series feature map of the dynamic changes of the electrical fire monitoring parameters;

[0016] The time-series feature map of the dynamic changes of the electrical fire monitoring parameters is highlighted to obtain the highlighted feature map of electrical fire monitoring;

[0017] Based on the electrical fire monitoring feature map, alarm results are obtained.

[0018] This application, by adopting the above technical solution, has significant technical effects:

[0019] The building intelligent electrical fire monitoring system and method provided in this application employs deep learning-based artificial intelligence data processing technology. It monitors for electrical fires by real-time analysis of temperature values ​​and leakage currents of distribution boxes on different floors of the target monitoring area. Upon detection of an electrical fire, it promptly sends alarm information to the fire control center. This achieves automated and intelligent electrical fire monitoring, improving its accuracy and timeliness. Attached Figure Description

[0020] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0021] Figure 1This is a flowchart of a building intelligent electrical fire monitoring method according to an embodiment of this application.

[0022] Figure 2 This is a schematic diagram of data flow in a building intelligent electrical fire monitoring method according to an embodiment of this application.

[0023] Figure 3 This is a flowchart illustrating the process of emphasizing the time-series feature map of dynamic changes in electrical fire monitoring parameters to obtain an emphasized feature map of electrical fire monitoring in the building intelligent electrical fire monitoring method according to an embodiment of this application.

[0024] Figure 4 This is a flowchart illustrating how an alarm result is obtained based on the electrical fire monitoring feature map in a building intelligent electrical fire monitoring method according to an embodiment of this application.

[0025] Figure 5 This is a system block diagram of a building intelligent electrical fire monitoring system according to an embodiment of this application. Detailed Implementation

[0026] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0027] As mentioned in the background section, traditional electrical fire monitoring systems often rely on manual inspections or periodic checks, making it difficult to achieve real-time and continuous monitoring of critical electrical equipment such as distribution boxes. Furthermore, previous data processing methods cannot effectively handle large volumes of high-dimensional electrical monitoring data, leading to untimely information processing and consequently, a lack of rapid and timely response to electrical fires. Therefore, an optimized intelligent electrical fire monitoring solution for buildings is needed.

[0028] In recent years, deep learning and neural networks have been widely applied in fields such as computer vision, natural language processing, and text signal processing. Furthermore, deep learning and neural networks have demonstrated near-human or even surpassed human-level performance in areas such as image classification, object detection, semantic segmentation, and text translation. The development of deep learning and neural networks has provided new solutions and approaches for intelligent electrical fire monitoring in buildings.

[0029] Figure 1 This is a flowchart of a building intelligent electrical fire monitoring method according to an embodiment of this application. Figure 2 This is a schematic diagram of the data flow in a building intelligent electrical fire monitoring method according to an embodiment of this application. Figure 1 and Figure 2As shown, the building intelligent electrical fire monitoring method according to an embodiment of this application includes: S110, acquiring leakage current values ​​and temperature values ​​of distribution boxes at multiple predetermined time points on different floors within a target monitoring area within a predetermined time period; S120, arranging the leakage current values ​​and temperature values ​​of the distribution boxes at multiple predetermined time points on different floors according to the time dimension and the sample dimension to obtain a time-series input tensor of electrical fire monitoring parameters; S130, performing feature encoding on the time-series input tensor of electrical fire monitoring parameters to obtain a time-series feature map of dynamic changes in electrical fire monitoring parameters; S140, emphasizing features on the time-series feature map of dynamic changes in electrical fire monitoring parameters to obtain an emphasized feature map of electrical fire monitoring; S150, obtaining an alarm result based on the emphasized feature map of electrical fire monitoring.

[0030] Specifically, in the technical solution of this application, it is first necessary to obtain the leakage current and temperature values ​​of distribution boxes on different floors within the target monitoring area at multiple predetermined time points within a predetermined time period. It should be understood that the operating status of electrical equipment is dynamically changing. By acquiring data at multiple time points, the operating status of the distribution boxes can be monitored in real time, and trend analysis can be performed. This trend analysis helps identify potential electrical fire risks. Specifically, electrical fires are often accompanied by abnormal changes in electrical parameters, such as a sudden increase in leakage current or a sharp rise in temperature. By continuously monitoring data at multiple time points, these abnormal behaviors can be more easily captured, and alarm signals can be issued in a timely manner. Buildings typically contain multiple floors, and the configuration of electrical equipment, power load, and operating status of each floor may differ. By acquiring data from different floors, comprehensive monitoring of the entire building's electrical system can be achieved, ensuring that no potential fire risk points are missed. In summary, acquiring the leakage current and temperature values ​​of distribution boxes on different floors within the target monitoring area at multiple predetermined time points within a predetermined time period enables comprehensive, accurate, and timely electrical fire monitoring, which is beneficial for achieving real-time and rapid response to electrical fires. In particular, in one possible implementation of this application, leakage current and temperature data of the distribution boxes at multiple time points can be collected by leakage current sensors and temperature sensors installed in the distribution boxes on each floor.

[0031] Considering that the operating status and electrical parameters (such as leakage current and temperature) of electrical equipment change over time, in order to capture early signs before an electrical fire occurs, the leakage current and temperature values ​​of distribution boxes at multiple time points can be arranged according to the time dimension. It is easy to understand that arranging data according to the time dimension allows for a clear understanding of the changing trends of electrical parameters over time, thus helping to identify potential electrical fire risks. Furthermore, considering that different floors in a building have different electrical equipment configurations and power loads, in order to achieve independent monitoring and analysis of the electrical equipment on each floor, and thus accurately assess the fire risk of each floor, the collected monitoring data needs to be arranged according to the floor sample dimension. That is, in the technical solution of this application, the leakage current and temperature values ​​of distribution boxes on different floors at multiple predetermined time points are arranged according to the time dimension and the sample dimension to obtain the electrical fire monitoring parameter time-series input tensor. In the spatial dimension of the arranged electrical fire monitoring parameter time-series input tensor, each row represents the monitoring data of a certain floor at a certain time point, each column represents a specific monitoring index value, and the channel dimension of the tensor represents the time dimension.

[0032] The time-series input tensor of electrical fire monitoring parameters contains a large amount of time-series data on electrical fire monitoring parameters (such as leakage current, temperature, etc.). Although this data contains rich information, the raw data is often difficult to use directly for analyzing and judging the risk of electrical fires. In order to extract useful feature information from this raw data, such as abnormal fluctuations and periodic changes of electrical parameters, and thus analyze the risk of electrical fires, the time-series input tensor of electrical fire monitoring parameters needs to be passed through an electrical fire monitoring parameter time-series feature encoder to obtain a time-series feature map of dynamic changes in electrical fire monitoring parameters. The occurrence of electrical fires is often closely related to the dynamic changes of electrical parameters. Through the encoding processing of the electrical fire monitoring parameter time-series feature encoder, the dynamic changes of electrical parameters can be effectively captured, thereby improving the accuracy of the analysis of the risk of electrical fires. In this application, the electrical fire monitoring parameter time-series feature encoder is a convolutional neural network model using a three-dimensional convolutional kernel. The electrical fire monitoring parameter time-series input tensor is a three-dimensional tensor containing three dimensions: time, monitoring indicators, and floor samples. Traditional two-dimensional convolutional neural networks cannot directly process this three-dimensional data, while three-dimensional convolutional kernels can perform convolution operations on these three dimensions simultaneously. By using the temporal feature encoder for electrical fire monitoring parameters, the spatiotemporal variation characteristics of electrical parameters can be captured, thereby more effectively extracting useful feature information from the temporal input tensor of electrical fire monitoring parameters.

[0033] In an embodiment of this application, one possible implementation of obtaining a dynamic change time-series feature map of electrical fire monitoring parameters by passing the time-series input tensor of electrical fire monitoring parameters through a time-series feature encoder of electrical fire monitoring parameters can be as follows: During the forward propagation of each layer of the time-series feature encoder of electrical fire monitoring parameters, the input data is processed as follows: The input data is convolved using the convolution units of each layer of the time-series feature encoder based on a three-dimensional convolution kernel to obtain a convolutional feature map; the convolutional feature map is pooled using the pooling units of each layer of the time-series feature encoder based on a local feature matrix to obtain a pooled feature map; and the feature values ​​at each position in the pooled feature map are nonlinearly activated using the activation units of each layer of the time-series feature encoder to obtain an activation feature map; wherein, the output of the last layer of the time-series feature encoder of electrical fire monitoring parameters is the dynamic change time-series feature map of electrical fire monitoring parameters.

[0034] Figure 3 This is a flowchart illustrating the process of emphasizing the dynamic change time-series feature map of electrical fire monitoring parameters to obtain an emphasized feature map in a building intelligent electrical fire monitoring method according to an embodiment of this application. For example... Figure 3 As shown, the step of emphasizing the time-series feature map of the dynamic changes of the electrical fire monitoring parameters to obtain an electrical fire monitoring emphasized feature map includes: S141, passing the time-series feature map of the dynamic changes of the electrical fire monitoring parameters through an electrical monitoring time-series feature emphasizer based on a channel attention mechanism to obtain an electrical fire monitoring parameter dynamic change time-series feature emphasized map; S142, passing the time-series feature map of the dynamic changes of the electrical fire monitoring parameters through a floor electrical monitoring feature emphasizer based on a spatial attention mechanism to obtain an electrical fire monitoring parameter dynamic change time-series floor emphasized feature map; S143, fusing the electrical fire monitoring parameter dynamic change time-series feature emphasized map and the electrical fire monitoring parameter dynamic change time-series floor emphasized feature map to obtain the electrical fire monitoring emphasized feature map.

[0035] The time-series feature map of dynamic changes in electrical fire monitoring parameters, obtained through 3D convolutional kernel encoding, has extracted temporal and spatial feature information. However, the importance of feature information from different channels (i.e., different time points) for the final assessment of electrical fire risk may vary. To focus more on the temporal features that have a decisive impact on electrical fire prediction during data processing and reduce interference from redundant information, the time-series feature map of dynamic changes in electrical fire monitoring parameters needs to be processed by an electrical monitoring time-series feature emphasizer based on a channel attention mechanism to obtain an emphasized time-series feature map of dynamic changes in electrical fire monitoring parameters. In this application, the electrical monitoring time-series feature emphasizer based on a channel attention mechanism is a convolutional neural network model using a channel attention mechanism. The channel attention mechanism assigns different attention weights to different feature channels, thereby focusing more on those channels that are more critical to assessing electrical fire risk (i.e., electrical parameter monitoring features at important time points) during data processing, while ignoring relatively unimportant channels. This improves the accuracy and real-time performance of electrical fire monitoring, providing strong support for subsequent fire prevention and emergency rescue work.

[0036] In an embodiment of this application, one possible implementation of obtaining an emphasized time-series feature map of the dynamic changes of electrical fire monitoring parameters by passing the electrical monitoring time-series feature emphasizer based on a channel attention mechanism can be as follows: During the forward propagation of the layer of the electrical monitoring time-series feature emphasizer based on a channel attention mechanism, each layer performs the following operations on the input data: convolution processing on the input data based on a two-dimensional convolution kernel to generate a convolutional feature map; pooling processing on the convolutional feature map to generate a pooled feature map; activation processing on the pooled feature map to generate an activated feature map; calculating the quotient of the sum of the eigenvalues ​​of the feature matrices corresponding to each channel in the activated feature map and the eigenvalues ​​of the feature matrices corresponding to all channels as the weighting coefficient of the feature matrices corresponding to each channel; weighting the feature matrices of each channel using the weighting coefficients of each channel in the activated feature map to generate a channel attention feature map; wherein, the input of the first layer of the electrical monitoring time-series feature emphasizer is the electrical fire monitoring parameter dynamic change time-series feature map, and the output of the last layer of the electrical monitoring time-series feature emphasizer is the electrical fire monitoring parameter dynamic change time-series feature map.

[0037] Considering the significant differences in electrical fire risk across different floors due to variations in usage, equipment configuration, and personnel density, and the varying importance of different monitoring indicators for electrical fire prediction (some indicators more sensitively reflect electrical fire risk, while others are relatively less important), this application aims to selectively emphasize features for different floors and monitoring indicators. This optimizes feature representation and improves the predictive ability for electrical fire risks. The dynamic change time-series feature map of electrical fire monitoring parameters needs to be processed by a floor electrical monitoring feature emphasizer based on a spatial attention mechanism to obtain a dynamic change time-series floor emphasized feature map of electrical fire monitoring parameters. In this application, the floor electrical monitoring feature emphasizer based on a spatial attention mechanism is a convolutional neural network model using this mechanism. The spatial attention mechanism can assign different attention weights along the spatial dimension of the feature map. Specifically, it dynamically adjusts the weights at various locations on the feature map based on the importance of the floor samples and different monitoring indicators. This allows the encoder to focus more on floors and monitoring indicators that are more critical for electrical fire prediction during feature map processing, while ignoring relatively less important parts. By accurately capturing key feature changes in electrical fire monitoring parameters, it is beneficial to identify fire risks and issue alarm signals in a timely manner.

[0038] Monitoring and analyzing electrical fires requires comprehensive consideration of information across multiple dimensions, including dynamic changes over time and differences between floors in space. The time-series feature map of dynamic changes in electrical fire monitoring parameters captures the patterns of these changes over time and emphasizes key time points. The floor-level feature map of dynamic changes in electrical fire monitoring parameters, in addition to expressing the time-series changes in electrical parameters, also highlights the differences between different floors. To more comprehensively reflect the potential risks of electrical fires, the technical solution of this application integrates the time-series feature map of dynamic changes in electrical fire monitoring parameters and the floor-level feature map to obtain an electrical fire monitoring emphasis feature map. By integrating these two feature maps, their respective shortcomings can be compensated for, thereby improving the accuracy of electrical fire risk prediction. Furthermore, the integrated feature map makes the data more robust in the face of complex and ever-changing electrical fire monitoring environments. Even if one feature map deviates due to certain reasons (such as data noise, sensor failure, etc.), other feature maps can still provide useful information, thus ensuring the reliability of the overall prediction results.

[0039] In an embodiment of this application, one possible implementation for fusing the time-series feature emphasis map of the dynamic changes in electrical fire monitoring parameters and the floor-level feature emphasis map of the dynamic changes in electrical fire monitoring parameters to obtain the electrical fire monitoring emphasis feature map can be as follows: The time-series feature emphasis map of the dynamic changes in electrical fire monitoring parameters and the floor-level feature emphasis map of the dynamic changes in electrical fire monitoring parameters are fused using the following fusion formula to obtain the electrical fire monitoring emphasis feature map; wherein, the fusion formula is: ;in, This is an emphasis feature map for the electrical fire monitoring system. This is an emphasis diagram of the time-series characteristics of the dynamic changes in the electrical fire monitoring parameters. This is a time-series floor feature map illustrating the dynamic changes in the electrical fire monitoring parameters. This represents the sum of elements at corresponding positions in the time-series feature emphasis map of the dynamic changes of the electrical fire monitoring parameters and the floor-level feature emphasis map of the time-series changes of the electrical fire monitoring parameters. and The weighted parameter is used to balance the dynamic change time-series feature emphasis map of electrical fire monitoring parameters and the dynamic change time-series floor feature emphasis map of electrical fire monitoring parameters in the electrical fire monitoring emphasis feature map.

[0040] Figure 4 This is a flowchart illustrating how an alarm result is obtained based on the emphasized feature map of electrical fire monitoring in a building intelligent electrical fire monitoring method according to an embodiment of this application. For example... Figure 4 As shown, obtaining the alarm result based on the electrical fire monitoring emphasis feature map includes: S151, performing mean pooling on each feature matrix along the channel dimension of the electrical fire monitoring emphasis feature map to obtain the electrical fire monitoring emphasis feature vector; S152, performing local perturbation adjustment based on feature offset compensation on the electrical fire monitoring emphasis feature vector to obtain an optimized electrical fire monitoring emphasis feature vector; S153, passing the optimized electrical fire monitoring emphasis feature vector through a classifier-based alarm result generator to obtain a classification result, the classification result being used to indicate whether to send alarm information to the fire control center.

[0041] Electrical fire monitoring emphasis feature maps typically contain a large amount of feature information, but this may include redundant information or unnecessary details that are not crucial for electrical fire detection. Therefore, in the technical solution of this application, the electrical fire monitoring emphasis feature map is subjected to mean pooling along each feature matrix of the channel dimension to obtain the electrical fire monitoring emphasis feature vector. Pooling operations (such as mean pooling) effectively reduce the size of the feature map by calculating the average value of each feature matrix along the channel dimension, achieving data dimensionality reduction and compression. This not only reduces the computational load of subsequent processing but also improves the operational efficiency of electrical fire monitoring.

[0042] Specifically, considering the errors introduced by sensors when measuring leakage current and temperature, these errors directly affect the accuracy of monitoring data. Furthermore, environmental factors such as temperature, humidity, and electromagnetic interference can all influence sensor readings, leading to noise in the collected electrical fire monitoring data. This results in local feature perturbations in the final electrical fire monitoring emphasis feature vector obtained by encoding the electrical monitoring parameters. To more accurately identify electrical fire risks and reduce false alarms or missed alarms caused by input perturbations (false alarms lead to unnecessary alarms and resource waste), and to improve the reliability of electrical fire monitoring, this application's technical solution performs local perturbation adjustment based on feature offset compensation on the electrical fire monitoring emphasis feature vector to obtain an optimized electrical fire monitoring emphasis feature vector.

[0043] Specifically, in the embodiments of this application, the local perturbation adjustment based on feature offset compensation is performed on the electrical fire monitoring emphasis feature vector to obtain an optimized electrical fire monitoring emphasis feature vector, including: determining the first classification weight matrix and the second classification weight matrix of the classifier before and after each iteration update; performing matrix multiplication of the first classification weight matrix and the second classification weight matrix with the electrical fire monitoring emphasis feature vector to obtain the first electrical fire monitoring modulation feature vector and the second electrical fire monitoring modulation feature vector; and calculating the positional difference between the first electrical fire monitoring modulation feature vector and the second electrical fire monitoring modulation feature vector to obtain the electrical fire monitoring shift information representation. Vector; calculate the F-norm of the position-mean vector between the first electrical fire monitoring modulation feature vector and the second electrical fire monitoring modulation feature vector as a shift compensation scaling factor; linearly scale the electrical fire monitoring shift information representation vector with the shift compensation scaling factor to obtain a scaled electrical fire monitoring shift information representation vector, and input the scaled electrical fire monitoring shift information representation vector into a Sigmoid activation function to obtain an electrical fire monitoring backpropagation shift compensation representation vector; calculate the position-mean multiplication of the electrical fire monitoring backpropagation shift compensation representation vector and the electrical fire monitoring emphasis feature vector to obtain the optimized electrical fire monitoring emphasis feature vector.

[0044] In the embodiments of this application, another way to express the optimized electrical fire monitoring emphasis feature vector by performing local perturbation adjustment based on feature offset compensation on the electrical fire monitoring emphasis feature vector can be: processing the electrical fire monitoring emphasis feature vector with the following formula to obtain the optimized electrical fire monitoring emphasis feature vector; wherein, the formula is: ; ;in, This is the first classification weight matrix. This is the second classification weight matrix. The electrical fire monitoring emphasizes the feature vector. Indicates matrix multiplication. This indicates adding based on position points. The F-norm of the eigenvectors is represented by... This represents the shift compensation scaling factor. This indicates dot product by position. This indicates subtraction based on position. This represents the activation function. The optimized electrical fire monitoring emphasizes the feature vector.

[0045] More specifically, in this embodiment of the application, the electrical fire monitoring shift information representation vector is linearly scaled using the shift compensation scaling factor to obtain a scaled electrical fire monitoring shift information representation vector, and the scaled electrical fire monitoring shift information representation vector is input into a Sigmoid activation function to obtain an electrical fire monitoring backpropagation shift compensation representation vector. This includes: creating a Spring Boot project; creating a server class in the Spring Boot project, the server class being used to perform vector linear scaling and vector input activation function processing; creating a controller class in the Spring Boot project, the controller class being used to accept requests and call the server class; packaging the Spring Boot project into a JAR file; and deploying the JAR file to a server.

[0046] In this embodiment of the application, the electrical fire monitoring shift information representation vector is linearly scaled using the shift compensation scaling factor to obtain a scaled electrical fire monitoring shift information representation vector, and the scaled electrical fire monitoring shift information representation vector is input into a Sigmoid activation function to obtain a backward propagation shift compensation representation vector for electrical fire monitoring. A partial implementation method for this is as follows:

[0047] / / Services

[0048] import org.springframework.stereotype.Service;

[0049] @Service

[0050] public class FireMonitoringService {

[0051] / **

[0052] * The displacement information representation vector for electrical fire monitoring is linearly scaled and activated using the Sigmoid activation function.

[0053] *

[0054] * @param inputVector The vector representing the input electrical fire monitoring shift information.

[0055] * @param scaleFactor Shift compensation scaling factor

[0056] * @return Backpropagation shift compensation representation vector for electrical fire monitoring

[0057] /

[0058] public double[] scaleAndApplySigmoid(double[] inputVector,double scaleFactor) {

[0059] double[] scaledVector = new double[inputVector.length];

[0060] for (int i = 0; i < inputVector.length; i++) {

[0061] scaledVector[i] = inputVector[i] * scaleFactor;

[0062] }

[0063] double[] sigmoidVector = applySigmoid(scaledVector);

[0064] return sigmoidVector;

[0065] }

[0066] private double[] applySigmoid(double[] input) {

[0067] double[] output = new double[input.length];

[0068] for (int i = 0; i < input.length; i++) {

[0069] output[i] = 1 / (1 + Math.exp(-input[i]));

[0070] }

[0071] return output;

[0072] }

[0073] }

[0074] / / Controller class

[0075] import org.springframework.beans.factory.annotation.Autowired;

[0076] import org.springframework.web.bind.annotation.*;

[0077] @RestController

[0078] @RequestMapping(" / api / fire-monitoring")

[0079] public class FireMonitoringController {

[0080] @Autowired

[0081] private FireMonitoringService fireMonitoringService;

[0082] @PostMapping(" / scale-and-sigmoid")

[0083] public double[] scaleAndApplySigmoid(@RequestBodyFireMonitoringVectorDTO input, @RequestParam double scaleFactor) {

[0084] return fireMonitoringService.scaleAndApplySigmoid(input.getVector(), scaleFactor);

[0085] }

[0086] }

[0087] It is understandable that the advantages of using the Spring framework to deploy and implement "linearly scaling the electrical fire monitoring shift information representation vector with the shift compensation scaling factor to obtain the scaled electrical fire monitoring shift information representation vector, and inputting the scaled electrical fire monitoring shift information representation vector into the Sigmoid activation function to obtain the electrical fire monitoring backpropagation shift compensation representation vector" specifically include: a. Spring has a broad ecosystem, including Spring Boot, Spring Data, etc., which can quickly integrate other functions, such as database operations, message queues, etc.; b. Spring supports multiple configuration methods (such as Java configuration, XML configuration, etc.), and the appropriate method can be selected according to the team's needs; c. The Spring framework is optimized to handle high-concurrency requests, which is suitable for electrical fire monitoring solutions that require real-time processing; d. Spring's dependency injection feature reduces the coupling between components, making it easier to perform unit testing and replace implementations; e. Spring's design allows the system to be expanded as requirements change, and new functions or services can be easily integrated.

[0088] To enhance the robustness of the classifier to local perturbations in the input feature vector, a shift compensation based on backpropagation representation is applied to the electrical fire monitoring emphasis feature vector. The key to this shift compensation lies in capturing information about the changes in the classification weight matrix during training and using this information to generate an electrical fire monitoring shift information representation vector. This is then used to generate a backpropagation shift compensation representation vector through scaling and a nonlinear transformation (Sigmoid function). In this way, the local shift of the electrical fire monitoring emphasis feature vector is described by using the distribution differences of the classifier across different classification scenarios during training. This not only quantifies the displacement of the feature vector caused by changes in the classification weight matrix but also generates a compensation vector based on the displacement information. This compensation vector adjusts the weights of each element in the feature vector according to the degree of perturbation, thereby reducing the impact of perturbations on the classification results. This significantly improves the classifier's robustness to local perturbations in the input data, enhances the classifier's stability and generalization ability, and reduces the impact of local perturbations on the classification results, enabling the classifier to maintain high classification accuracy even when facing noisy or variable data.

[0089] Finally, the optimized electrical fire monitoring emphasis feature vector is passed through a classifier-based alarm result generator to obtain a classification result, which indicates whether to send an alarm message to the fire control center. It should be understood that in order to convert the optimized electrical fire monitoring emphasis feature vector, which contains complex electrical monitoring feature information, into a more intuitive result, the optimized electrical fire monitoring emphasis feature vector needs to be passed through a classifier-based alarm result generator. A classifier is a machine learning algorithm whose main function is to map input data to predefined category labels. The classification result directly determines whether an alarm message needs to be sent to the fire control center. When the classifier determines that there is an electrical fire risk, it will automatically trigger the alarm mechanism and send an alarm signal to the fire control center so that timely countermeasures can be taken. In this way, automatic alarm for electrical fires is achieved, which can further improve the overall performance of building electrical fire monitoring, making it more reliable and efficient.

[0090] In an embodiment of this application, the optimized electrical fire monitoring emphasis feature vector is processed by a classifier-based alarm result generator to obtain a classification result. One possible implementation for this classification result to indicate whether an alarm message should be sent to the fire control center is as follows: the optimized electrical fire monitoring emphasis feature vector is fully encoded using the fully connected layer of the classifier to obtain a fully connected encoded feature vector for electrical fire monitoring; the fully connected encoded feature vector is input into the Softmax classification function of the classifier to obtain probability values ​​for each classification label, where the classification labels include those indicating that an alarm message should be sent to the fire control center at the current time and those indicating that an alarm message should not be sent to the fire control center at the current time; the classification label corresponding to the largest probability value is determined as the classification result.

[0091] In summary, the building intelligent electrical fire monitoring method based on the embodiments of this application has been clarified. It employs deep learning-based artificial intelligence data processing technology to monitor the presence of electrical fires by real-time analysis of the temperature values ​​and leakage currents of distribution boxes on different floors of the target monitoring area. In response to the presence of an electrical fire, it promptly sends alarm information to the fire control center. This achieves automated and intelligent electrical fire monitoring, which helps improve the accuracy and timeliness of electrical fire monitoring.

[0092] Figure 5 This is a system block diagram of a building intelligent electrical fire monitoring system according to an embodiment of this application. Figure 5 As shown, the building intelligent electrical fire monitoring system 100 according to an embodiment of this application includes: an electrical fire monitoring related data acquisition module 110, used to acquire leakage current values ​​and temperature values ​​of distribution boxes at multiple predetermined time points on different floors within a target monitoring area within a predetermined time period; an electrical fire monitoring related data normalization module 120, used to arrange the leakage current values ​​and temperature values ​​of the distribution boxes at multiple predetermined time points on different floors according to the time dimension and the sample dimension to obtain a time-series input tensor of electrical fire monitoring parameters; an electrical fire monitoring related data feature encoding module 130, used to perform feature encoding on the time-series input tensor of electrical fire monitoring parameters to obtain a time-series feature map of dynamic changes in electrical fire monitoring parameters; an electrical fire monitoring related data feature emphasis module 140, used to emphasize the features of the time-series feature map of dynamic changes in electrical fire monitoring parameters to obtain an emphasized feature map of electrical fire monitoring; and an alarm result generation module 150, used to obtain an alarm result based on the emphasized feature map of electrical fire monitoring.

[0093] Here, those skilled in the art will understand that the specific functions and operations of each unit and module in the above-mentioned building intelligent electrical fire monitoring system 100 have been referenced above. Figures 1 to 4 The description of the building intelligent electrical fire monitoring method is detailed here, and therefore, its repeated description will be omitted.

[0094] In summary, the building intelligent electrical fire monitoring system 100 based on the embodiments of this application is explained. It employs deep learning-based artificial intelligence data processing technology to monitor for electrical fires by real-time analysis of the temperature values ​​and leakage currents of distribution boxes on different floors of the target monitoring area. In response to the presence of an electrical fire, it promptly sends alarm information to the fire control center. This achieves automated and intelligent electrical fire monitoring, improving the accuracy and timeliness of electrical fire monitoring.

[0095] As described above, the building intelligent electrical fire monitoring system 100 according to the embodiments of this application can be implemented in various wireless terminals, such as servers for building intelligent electrical fire monitoring. In one example, the building intelligent electrical fire monitoring system 100 according to the embodiments of this application can be integrated into a wireless terminal as a software module and / or hardware module. For example, the building intelligent electrical fire monitoring system 100 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the building intelligent electrical fire monitoring system 100 can also be one of many hardware modules of the wireless terminal.

[0096] Alternatively, in another example, the building intelligent electrical fire monitoring system 100 and the wireless terminal can also be separate devices, and the building intelligent electrical fire monitoring system 100 can connect to the wireless terminal via wired and / or wireless networks and transmit interactive information in accordance with an agreed data format.

Claims

1. A method for monitoring electrical fires in buildings, characterized in that, include: Obtain leakage current and temperature values ​​of distribution boxes on different floors within the target monitoring area at multiple predetermined time points within a predetermined time period; The leakage current and temperature values ​​of the distribution boxes on different floors at multiple predetermined time points are arranged according to the time dimension and the sample dimension to obtain the time sequence input tensor of electrical fire monitoring parameters. The time-series input tensor of the electrical fire monitoring parameters is feature-encoded to obtain a time-series feature map of the dynamic changes of the electrical fire monitoring parameters; The time-series feature map of the dynamic changes of the electrical fire monitoring parameters is highlighted to obtain the highlighted feature map of electrical fire monitoring; Based on the aforementioned electrical fire monitoring feature map, alarm results are obtained; Among them, based on the electrical fire monitoring emphasis feature map, alarm results are obtained, including: The electrical fire monitoring emphasis feature map is subjected to mean pooling along each feature matrix of the channel dimension to obtain the electrical fire monitoring emphasis feature vector; The electrical fire monitoring emphasis feature vector is adjusted by local perturbation based on feature offset compensation to obtain an optimized electrical fire monitoring emphasis feature vector; The optimized electrical fire monitoring emphasis feature vector is passed through a classifier-based alarm result generator to obtain a classification result, which is used to indicate whether to send an alarm message to the fire control center. The optimized electrical fire monitoring emphasis feature vector is obtained by performing local perturbation adjustment based on feature offset compensation on the emphasized feature vector, including: Determine the first and second classification weight matrices of the classifier before and after each iteration update; The first classification weight matrix and the second classification weight matrix are respectively multiplied by the electrical fire monitoring emphasis feature vector to obtain the first electrical fire monitoring modulation feature vector and the second electrical fire monitoring modulation feature vector; Calculate the positional difference between the first electrical fire monitoring modulation feature vector and the second electrical fire monitoring modulation feature vector to obtain the electrical fire monitoring shift information representation vector; The F-norm of the position mean vector between the first electrical fire monitoring modulation feature vector and the second electrical fire monitoring modulation feature vector is calculated as the shift compensation scaling factor. The electrical fire monitoring shift information representation vector is linearly scaled using the shift compensation scaling factor to obtain a scaled electrical fire monitoring shift information representation vector, and the scaled electrical fire monitoring shift information representation vector is input into a Sigmoid activation function to obtain an electrical fire monitoring backward propagation shift compensation representation vector. The optimized electrical fire monitoring emphasis feature vector is obtained by multiplying the backpropagation shift compensation representation vector of the electrical fire monitoring with the emphasis feature vector of the electrical fire monitoring by position.

2. The building intelligent electrical fire monitoring method according to claim 1, characterized in that, Encoding the time-series input tensor of the electrical fire monitoring parameters to obtain a dynamic change time-series feature map of the electrical fire monitoring parameters includes: passing the time-series input tensor of the electrical fire monitoring parameters through an electrical fire monitoring parameter time-series feature encoder to obtain a dynamic change time-series feature map of the electrical fire monitoring parameters.

3. The building intelligent electrical fire monitoring method according to claim 2, characterized in that, The time-series feature map of the dynamic changes of the electrical fire monitoring parameters is used to highlight features to obtain an electrical fire monitoring highlighted feature map, including: The time-series feature map of the dynamic change of the electrical fire monitoring parameters is passed through an electrical monitoring time-series feature accenter based on a channel attention mechanism to obtain an accent map of the dynamic change of the electrical fire monitoring parameters. The time-series feature map of the dynamic changes of the electrical fire monitoring parameters is used to obtain the floor emphasis feature map of the dynamic changes of the electrical fire monitoring parameters by passing the floor electrical monitoring feature emphasis tool based on the spatial attention mechanism; The electrical fire monitoring parameter dynamic change time-series feature emphasis map and the electrical fire monitoring parameter dynamic change time-series floor emphasis feature map are combined to obtain the electrical fire monitoring emphasis feature map.

4. The building intelligent electrical fire monitoring method according to claim 3, characterized in that, The electrical fire monitoring parameter timing feature encoder is a convolutional neural network model using a three-dimensional convolutional kernel; the electrical monitoring timing feature accenter based on the channel attention mechanism is a convolutional neural network model using the channel attention mechanism; and the floor electrical monitoring feature accenter based on the spatial attention mechanism is a convolutional neural network model based on the spatial attention mechanism.

5. The building intelligent electrical fire monitoring method according to claim 4, characterized in that, The electrical fire monitoring shift information representation vector is linearly scaled using the shift compensation scaling factor to obtain a scaled electrical fire monitoring shift information representation vector. This scaled vector is then input into a Sigmoid activation function to obtain an electrical fire monitoring backpropagation shift compensation representation vector, including: Create a Spring Boot project; Create a server class in the Spring Boot project. This server class is used to perform vector linear scaling and vector input activation function processing. Create a controller class in the Spring Boot project. This controller class will accept requests and invoke the server class. Package the Spring Boot project into a JAR file and deploy the JAR file to the server.

6. A building intelligent electrical fire monitoring system, characterized in that, include: The electrical fire monitoring data acquisition module is used to acquire the leakage current and temperature values ​​of the distribution boxes at multiple predetermined time points on different floors within the target monitoring area. The electrical fire monitoring related data regularization module is used to arrange the leakage current value and temperature value of the distribution box at multiple predetermined time points on different floors according to the time dimension and the sample dimension to obtain the time sequence input tensor of electrical fire monitoring parameters. An electrical fire monitoring related data feature encoding module is used to perform feature encoding on the time-series input tensor of the electrical fire monitoring parameters to obtain a time-series feature map of the dynamic changes of the electrical fire monitoring parameters. An electrical fire monitoring related data feature enhancement module is used to enhance the features of the time-series feature map of the dynamic change of the electrical fire monitoring parameters to obtain an electrical fire monitoring enhancement feature map; The alarm result generation module is used to obtain alarm results based on the electrical fire monitoring emphasis feature map; The alarm result generation module includes: The feature map pooling unit is used to perform mean pooling on each feature matrix along the channel dimension of the electrical fire monitoring emphasis feature map to obtain the electrical fire monitoring emphasis feature vector; The feature vector optimization unit is used to perform local perturbation adjustment on the electrical fire monitoring emphasis feature vector based on feature offset compensation to obtain an optimized electrical fire monitoring emphasis feature vector; The feature vector parsing unit is used to pass the optimized electrical fire monitoring emphasis feature vector through a classifier-based alarm result generator to obtain a classification result, which is used to indicate whether to send an alarm message to the fire control center. The optimized electrical fire monitoring emphasis feature vector is obtained by performing local perturbation adjustment based on feature offset compensation on the emphasized feature vector, including: Determine the first and second classification weight matrices of the classifier before and after each iteration update; The first classification weight matrix and the second classification weight matrix are respectively multiplied by the electrical fire monitoring emphasis feature vector to obtain the first electrical fire monitoring modulation feature vector and the second electrical fire monitoring modulation feature vector; Calculate the positional difference between the first electrical fire monitoring modulation feature vector and the second electrical fire monitoring modulation feature vector to obtain the electrical fire monitoring shift information representation vector; The F-norm of the position mean vector between the first electrical fire monitoring modulation feature vector and the second electrical fire monitoring modulation feature vector is calculated as the shift compensation scaling factor. The electrical fire monitoring shift information representation vector is linearly scaled using the shift compensation scaling factor to obtain a scaled electrical fire monitoring shift information representation vector, and the scaled electrical fire monitoring shift information representation vector is input into a Sigmoid activation function to obtain an electrical fire monitoring backward propagation shift compensation representation vector. The optimized electrical fire monitoring emphasis feature vector is obtained by multiplying the backpropagation shift compensation representation vector of the electrical fire monitoring with the emphasis feature vector of the electrical fire monitoring by position.

7. The building intelligent electrical fire monitoring system according to claim 6, characterized in that, The electrical fire monitoring related data feature encoding module is used to: pass the electrical fire monitoring parameter time series input tensor through the electrical fire monitoring parameter time series feature encoder to obtain the dynamic change time series feature map of the electrical fire monitoring parameters.

Images