Intelligent electrical fire risk assessment method and device based on data analysis and cloud platform

Abstract

The invention relates to the technical field of fire risk assessment, in particular to an intelligent electrical fire risk assessment method and device based on data analysis and a cloud platform. The method comprises the following steps: acquiring operation data of electrical equipment; uploading the electrical equipment operation data to a cloud platform in real time for data integration, and generating an electrical equipment operation monitoring data set; performing abnormal electrical equipment analysis on the electrical equipment operation monitoring data set to obtain abnormal electrical equipment identification data; performing IP position extraction on the electrical equipment operation monitoring data set based on the abnormal electrical equipment identification data to obtain abnormal electrical equipment IP address data; and address range camera calling is carried out through the abnormal electrical equipment IP address data to obtain abnormal electrical equipment range camera calling data. According to the invention, through real-time data integration, anomaly detection, fire spreading prediction and intelligent decision support, the accuracy and comprehensiveness of fire risk assessment are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of fire risk assessment, and particularly to an intelligent electrical fire risk assessment method, device and cloud platform based on data analysis. Background Art

[0002] In the early stage, electrical fires were mainly prevented through regular inspections and on-site manual detections, and the risk assessment relied on the experience of firefighters and the maintenance status of infrastructure. However, this method has great limitations, such as the inability to monitor the dynamic state of electrical equipment in real time and the inability to comprehensively evaluate the risks of electrical equipment under different environmental conditions. With the progress of sensor technology and the Internet of Things, the real-time monitoring of electrical equipment has become possible. The wide application of sensors such as temperature, humidity, and current provides basic data support for the fire warning system. However, the analysis of single-sensor data has problems of insufficient information and high false alarm rates. Entering the era of big data and artificial intelligence, intelligent electrical fire risk assessment methods have gradually incorporated machine learning and deep learning technologies. Through the mining and analysis of massive sensor data, algorithms can predict the fault trends of electrical equipment, identify potential fire risks, and provide intelligent decision-making support for electrical fire prevention and control. However, currently, traditional methods often cannot accurately identify the spread trend and specific impact area after a fire occurs, and lack the accurate identification of key factors (such as the distribution and aggregation of combustibles) during the fire spread process, thus resulting in low accuracy and comprehensiveness of fire risk assessment. Summary of the Invention

[0003] Based on this, it is necessary to provide an intelligent electrical fire risk assessment method, device and cloud platform based on data analysis to solve at least one of the above technical problems.

[0004] To achieve the above object, an intelligent electrical fire risk assessment method based on data analysis, the method includes the following steps: Step S1: Obtain the operation data of electrical equipment; upload the operation data of electrical equipment to the cloud platform in real time for data integration to generate an electrical equipment operation monitoring data set; perform analysis on abnormal electrical equipment for the electrical equipment operation monitoring data set to obtain abnormal electrical equipment identification data; Step S2: Extract the IP location for the electrical equipment operation monitoring data set based on the abnormal electrical equipment identification data to obtain abnormal electrical equipment IP address data; retrieve cameras within the address range through the abnormal electrical equipment IP address data to obtain abnormal electrical equipment range camera retrieval data; perform analysis on the fire spread trend during the fire for the abnormal electrical equipment IP address data based on the abnormal electrical equipment range camera retrieval data to generate fire spread trend data; adjust the camera view for the abnormal electrical equipment range camera retrieval data based on the fire spread trend data to generate in-fire camera view adjustment data; Step S3: Locate combustibles based on the data of the camera view adjustment during the disaster, and generate data on the spatial position information of the combustibles within a range; extract the aggregation characteristics of the combustibles from the data of the camera view adjustment during the disaster through the data on the spatial position information of the combustibles within a range, and obtain the data on the spatial aggregation characteristics of the combustibles; use the data on the spatial aggregation characteristics of the combustibles to predict the risk of fire spread for the data of the camera view adjustment during the disaster, and generate the data on the prediction of fire spread. Step S4: Mark the fire risk level according to the data on the prediction of fire spread, and generate the data on the marked fire risk level; construct a fire decision based on the data on the marked fire risk level, and generate a fire risk decision report to execute the intelligent electrical fire risk control operation.

[0005] The present invention realizes the centralized storage and efficient processing of data by obtaining the operation data of electrical equipment in real time and uploading it to the cloud platform. Through data integration, the status of the equipment can be comprehensively monitored, abnormalities can be detected in a timely manner, and the limitations of traditional manual inspections are avoided. The analysis of abnormal electrical equipment can quickly identify potentially faulty equipment, give early warnings, and reduce the probability of accidents. By extracting the IP address to locate the position of the abnormal electrical equipment, the accuracy of monitoring is improved, ensuring that measures can be taken in a timely manner. By retrieving the relevant camera data, the situation at the fire scene can be obtained in real time, ensuring an accurate assessment of the fire spread trend. The analysis of the fire spread trend helps to predict the speed and scope of the fire spread, adjust the camera view, make the fire monitoring more comprehensive, and improve the disaster response ability. The location of combustibles provides an important influencing factor in the process of fire spread, helping to better understand the development of the fire. By extracting the spatial aggregation characteristics of combustibles, the key factors that exacerbate the fire spread in the fire can be identified, providing data support for the prediction of the fire spread. The prediction of the risk of fire spread can identify the severity of the fire in advance, provide more accurate prevention and control measures, and help reduce fire losses. The marking of the fire risk level quantifies the fire risk in a data-driven manner, improves the scientificity and accuracy of risk assessment, and the generated fire risk decision report provides data-based intelligent support for decision-makers, reduces the deviation of human judgment, and improves the timeliness and effectiveness of response. It provides a basis for the intelligent electrical fire risk control operation, optimizes the post-disaster emergency response and control strategies, and helps to extinguish the fire in a timely and effective manner. Therefore, the present invention improves the accuracy and comprehensiveness of fire risk assessment through real-time data integration, anomaly detection, fire spread prediction, and intelligent decision support.

[0006] Preferably, step S1 includes the following steps: Step S11: Obtain the operation data of electrical equipment through multi-source sensors; Step S12: Perform data preprocessing on the operation data of electrical equipment to generate standard operation data of electrical equipment, where the data preprocessing includes data cleaning, data denoising, missing value filling, and data standardization; Step S13: Real-time upload the standard electrical equipment operation data to the cloud platform for equipment identification, generating electrical equipment identification data; based on the electrical equipment identification data, perform data integration on the standard electrical equipment operation data to generate an electrical equipment operation monitoring data set; Step S14: Analyze the abnormal electrical equipment in the electrical equipment operation monitoring data set to obtain abnormal electrical equipment identification data.

[0007] The present invention can comprehensively monitor various performance indicators of electrical equipment by using multi-source sensors to real-time obtain the operation data of electrical equipment (such as temperature, current, power, vibration, etc.). This data acquisition method can reflect the operation status of the equipment in real time, ensuring the comprehensiveness and accuracy of monitoring, and providing a rich information source for subsequent analysis and decision-making. Data preprocessing of the electrical equipment operation data, including data cleaning, denoising, missing value filling, and standardization, can ensure data quality and remove irrelevant noise. Data cleaning and denoising help improve the accuracy of data and avoid deviations in analysis results caused by data problems. Standardization processing can eliminate data scale differences caused by factors such as different equipment and different environments, facilitating subsequent comparison and analysis, and enhancing the usability and reliability of data. Real-time upload the standard electrical equipment operation data to the cloud platform for equipment identification. This process can assign a unique identification code to each electrical equipment, realizing accurate tracking and management of the equipment. Integrate the operation data through the equipment identification data to generate an electrical equipment operation monitoring data set, forming a complete equipment operation data file. Data integration makes the equipment monitoring information more systematic, contributing to the long-term tracking of equipment performance and historical data analysis. Analyze the abnormalities in the electrical equipment operation monitoring data set, which can help detect abnormal situations in the equipment operation in a timely manner, such as problems like overloading operation, too high temperature, and abnormal vibration of electrical equipment. Through abnormal analysis, potential failure risks can be quickly identified, enabling early maintenance or repair measures to be taken, reducing equipment failures and downtime, and extending the service life of the equipment.

[0008] Preferably, step S14 includes the following steps: Step S141: Extract the operation time-domain characteristics of the electrical equipment operation monitoring data set to obtain equipment operation time-domain characteristic data; perform fast Fourier transform on the equipment operation time-domain characteristic data to generate equipment operation frequency-domain characteristic data; Step S142: Analyze the trend change of the equipment operation frequency-domain characteristic data to generate an equipment operation frequency-domain trend change curve; based on a preset time range, segment the equipment operation frequency-domain trend change curve to generate an equipment operation short-term trend change curve and an equipment operation long-term trend change curve; Step S143: Perform zero-value mutation anomaly monitoring on the short-term trend change curve of the device operation to generate the first abnormal state data of the abnormal device; perform curve change rate anomaly monitoring on the long-term trend change curve of the device operation to generate the second abnormal state data of the abnormal device. Step S144: Screen the abnormal electrical equipment from the electrical equipment operation monitoring dataset through the first abnormal state data of the abnormal device and the second abnormal state data of the abnormal device to obtain the abnormal electrical equipment operation data; perform device identification on the abnormal electrical equipment operation data to obtain the abnormal electrical equipment identification data.

[0009] Through the extraction of the operation time-domain characteristics of the electrical equipment operation monitoring dataset, the present invention can effectively capture the dynamic behavior of the device, such as conventional operation parameters such as current and voltage fluctuations. Transforming the time-domain data into frequency-domain characteristics through the fast Fourier transform (FFT) can reveal the frequency components in the device operation process and help identify whether the device is in a normal working state. The frequency-domain characteristic data can provide more detailed information, especially the frequency anomaly performance when the device has potential faults. Performing trend change analysis on the device operation frequency-domain characteristic data can help capture the long-term trend changes of the frequency-domain characteristics in the device operation process. For example, the fluctuations or gradual changes in frequency indicate changes in the device load, environment, or other working conditions. By segmenting the frequency-domain trend change curve into a short-term trend change curve and a long-term trend change curve, the short-term and long-term operation trends of the device can be analyzed more carefully, and the short-term fluctuations and long-term operation trends can be analyzed separately. Performing zero-value mutation anomaly monitoring on the short-term trend change curve helps detect whether the device has sudden state changes at certain time points, such as sudden increases in current and device failures, and quickly identify potential short-term faults. Performing curve change rate anomaly monitoring on the long-term trend change curve can capture the changes in the device operation status over a long time, especially those gradually accumulating problems, such as device aging and performance degradation. Through these two monitoring methods, comprehensive anomaly detection of electrical equipment from short-term to long-term can be achieved. Combining the abnormal state data obtained from short-term and long-term trend analysis for screening abnormal electrical equipment helps identify which devices have potential faults or instability risks. Through device identification, abnormal devices can be accurately distinguished from normal devices, ensuring that problem devices receive timely attention and maintenance. Through accurate device identification, managers can quickly locate problem devices during device repair, replacement, or monitoring, improving the fault response efficiency.

[0010] Preferably, step S2 includes the following steps: Step S21: Based on the abnormal electrical equipment identification data, perform real geographical mapping of abnormal electrical equipment on the electrical equipment operation monitoring data set to generate real geographical mapping data of abnormal electrical equipment; extract the IP location from the real geographical mapping data of abnormal electrical equipment to obtain the IP address data of abnormal electrical equipment; Step S22: Use the IP address data of abnormal electrical equipment to retrieve cameras within the address range from the cloud platform to obtain camera retrieval data within the range of abnormal electrical equipment; collect regional smoke sensor data for the IP address data of abnormal electrical equipment based on the camera retrieval data within the range of abnormal electrical equipment to obtain regional smoke sensing data of abnormal electrical equipment; Step S23: Based on the regional smoke sensing data of abnormal electrical equipment, perform layer separation during the disaster on the camera retrieval data within the range of abnormal electrical equipment to generate smoke layer data and fire alarm data during the disaster; Step S24: Analyze the spread trend of the fire alarm data through the smoke layer data during the disaster to generate fire spread trend data; adjust the camera perspective of the camera retrieval data within the range of abnormal electrical equipment based on the fire spread trend data to generate camera perspective adjustment data during the disaster.

[0011] Through the realistic geographical mapping of the identification data of abnormal electrical equipment, the present invention can accurately determine the location of the equipment and combine it with geographical coordinates, thereby realizing the precise positioning of abnormal electrical equipment. The generated realistic geographical mapping data of abnormal electrical equipment provides accurate data support for subsequent on-site monitoring, investigation, and response. Through IP address extraction, a specific network address is further provided for equipment positioning, which helps the cloud platform to remotely monitor and manage the equipment. By using the IP address data of abnormal electrical equipment to retrieve cameras within the address range of the cloud platform, real-time video monitoring images around the equipment can be obtained, further evaluating the safety status around the equipment. In addition, the collection of regional smoke sensor data helps to monitor the air quality around the equipment in real time, especially the change in smoke concentration, and can quickly detect potential fire risks caused by fires or equipment abnormalities. Based on the regional smoke sensing data of abnormal electrical equipment, layer separation is carried out during a disaster, and the smoke layer and fire alarm data are extracted and analyzed, which helps to distinguish the scope and intensity of the fire. The smoke layer data during a disaster can show the spread range of the smoke, helping to determine the source and spreading area of the fire; while the fire alarm data reflects the urgency and warning level of the fire, providing a basis for further disaster emergency response. By analyzing the spreading trend of the smoke layer data and fire alarm data during a disaster, the spreading direction and speed of the fire can be predicted, and countermeasures can be taken in advance to reduce casualties and property losses. Based on the spreading trend of the fire, the camera perspective adjustment can be optimized in real time according to the expansion of the fire, ensuring that the camera can monitor the fire area comprehensively. This provides precise perspective support for fire extinguishing work and helps the command center to timely dispatch resources for emergency handling.

[0012] Preferably, step S23 includes the following steps: Step S231: Screen the frame images of the retrieved data of the cameras within the range of abnormal electrical equipment to obtain the frame images of abnormal electrical equipment during a disaster; convert the frame images of abnormal electrical equipment during a disaster into grayscale images to generate the grayscale images of abnormal electrical equipment during a disaster; Step S232: Detect the smoke area in the grayscale image of abnormal electrical equipment during a disaster through the regional smoke sensing data of abnormal electrical equipment to generate the detected data of the smoke area of abnormal electrical equipment during a disaster; perform the first layer separation on the grayscale image of abnormal electrical equipment during a disaster according to the detected data of the smoke area of abnormal electrical equipment during a disaster, thereby obtaining the smoke layer data during a disaster; Step S233: Use the smoke layer data during a disaster to perform the second layer separation on the frame images of abnormal electrical equipment during a disaster to generate the fire layer data during a disaster; perform precise fire alarm analysis on the fire layer data during a disaster to generate the fire alarm data.

[0013] By screening the frame images retrieved from the camera, the present invention can extract key frame images from the video stream, which helps reduce the computational amount and improve the processing efficiency. Only focusing on the key frames containing potential dangers ensures that the image data is accurate, real-time, and of high priority. Converting the screened in-disaster frame images to grayscale simplifies the complexity of the image data, reduces the interference of color information, and facilitates subsequent image processing and feature extraction. This process ensures the efficiency of image processing and enhances the ability to extract important details (such as smoke, fire, etc.). Using the data of the smoke sensor and combining image processing techniques to detect the smoke area in the grayscale image can accurately identify the areas with smoke in the image. This step helps improve the initial recognition accuracy of disasters, reduce false alarms, and precisely locate the disaster source. Based on the smoke area detection data, the first layer separation of the grayscale image is performed to extract the smoke layer data, which is convenient for subsequent analysis and judgment of the smoke spread range. Through effective layer separation technology, the smoke in the fire scene can be clearly distinguished from other environmental factors, improving the accuracy of disaster location. Through the in-disaster smoke layer data, the second layer separation of the frame image is performed to generate the fire layer data. The separation of the fire layer makes the visual information of smoke and flame further clearly separated, thus providing an accurate image basis for the assessment and spread prediction of the fire. This step enables the image processing system to accurately extract the flame and heat source information related to the fire and exclude interference factors. Conducting precise fire situation analysis on the generated fire layer data helps timely and accurately assess the severity and development trend of the fire. This process can quickly generate fire situation data to assist on-site decision-makers in formulating emergency response measures and optimizing resource allocation and emergency response.

[0014] Preferably, the precise fire situation analysis of the in-disaster fire layer data includes: Performing infrared image conversion on the in-disaster fire layer data to generate an in-disaster fire infrared image; extracting the pixel pigment of the in-disaster fire infrared image to obtain the pixel pigment data of the in-disaster fire image; identifying the flame pixel points based on the preset pixel pigment range for the pixel pigment data of the in-disaster fire image to generate the flame pixel point identification data; Using the flame pixel point identification data to segment the flame area of the in-disaster fire infrared image to generate the in-disaster fire core area image; calculating the temperature gradient of the regional pixel points of the in-disaster fire core area image to obtain the pixel point temperature gradient data; Performing fire intensity analysis on the in-disaster fire core area image according to the pixel point temperature gradient data to generate the fire situation data.

[0015] By converting the in-disaster fire layer data into an infrared image, the present invention can accurately capture the thermal radiation information of the flame source, which is crucial for fire monitoring and analysis. The infrared image can effectively penetrate smoke and other obstacles, providing a clear location of the fire source and the scope of the fire, avoiding the limitations of traditional visible light images under conditions such as smoke and haze. After extracting the pixel pigment data in the image, the color change and intensity of the flame can be accurately distinguished, further enhancing the real-time monitoring ability of the fire scene. The extraction of pigment data can better reflect the changes and temperature gradients in the flame area, assisting in judging the severity of the fire. Based on the preset pixel pigment range, the pixel points of the flame can be effectively identified. This process ensures that only the areas related to the flame are focused on, eliminating other interfering factors in the image, such as smoke and hot air currents, thereby improving the accuracy of fire detection. By applying the flame pixel point identification data to the in-disaster fire infrared image, the core flame area is segmented. In this way, the spatial range of the fire source can be accurately delimited, helping the on-site emergency response team to clarify the location of the fire source and formulate more precise fire extinguishing measures. Calculating the temperature gradient of the pixel points in the core area of the fire can reveal the thermal distribution of the flame, helping to judge the intensity and spread trend of the flame. The temperature gradient data can provide the location of the hottest area of the flame and provide the necessary basic information for the analysis of the fire intensity. This process is crucial for understanding the propagation speed and expansion direction of the fire, especially in large-scale fires. Through further analysis of the pixel temperature gradient data, the intensity, burning rate, and potential hazards of the fire can be accurately evaluated. The fire intensity directly affects the speed and scope of the fire spread, providing the core decision-making basis for fire emergency response. By comprehensively analyzing the temperature gradient data and the fire intensity, fire alarm data is generated to reflect the current state of the fire in real time. These data can be provided to the command center to assist personnel in making decisions on formulating key response measures such as fire extinguishing strategies or evacuation routes.

[0016] Preferably, step S24 includes the following steps: Step S241: Calculate the smoke concentration of the fire alarm data through the in-disaster smoke layer data to obtain the in-disaster fire smoke concentration data; based on the in-disaster fire smoke concentration data, screen the adjacent timestamp layers of the in-disaster smoke layer to obtain the first in-disaster smoke layer and the second in-disaster smoke layer; Step S242: Use the in-disaster fire smoke concentration data to confirm the extreme concentration center points of the first in-disaster smoke layer and the second in-disaster smoke layer to obtain the first in-disaster smoke concentration extreme center point and the second in-disaster smoke concentration extreme center point; analyze the moving trend of the center points of the first in-disaster smoke concentration extreme center point and the second in-disaster smoke concentration extreme center point to generate center point moving trend data; Step S243: Confirm the smoke spreading direction for the first in-disaster smoke layer and the second in-disaster smoke layer according to the central point movement trend data to generate the fire spreading trend data; adjust the camera view of the abnormal electrical equipment range camera data based on the fire spreading trend data to generate the in-disaster camera view adjustment data.

[0017] The present invention can accurately monitor the smoke concentration distribution at the fire scene by calculating the smoke concentration of the in-disaster smoke layer data. This data helps to judge the severity and spread range of the fire. By measuring the smoke concentration in real time, it can provide first-hand disaster information after the fire occurs, helping decision-makers to make more timely responses. Based on the smoke concentration data, screening the layers at different timestamps can track the smoke spread and changes during the fire process. By comparing the smoke layers of the previous and subsequent timestamps, the spread speed and direction of the fire can be clearly understood, providing a basis for fire spread prediction and fire emergency handling. By confirming the extreme concentration central point of the in-disaster smoke layer, the core area of the smoke can be clarified. This data helps to determine the combustion point and the most dangerous area of the fire, thus providing key fire source information for the emergency department and helping to mobilize necessary resources for fire extinguishing or evacuation. Analyzing the movement trend of the concentration extreme central point can reveal the direction and speed of the fire spread and predict the next development of the fire situation. This analysis is crucial for formulating the emergency response plan, especially when the fire spreads rapidly and unpredictably. Through the movement trend data, the fire extinguishing strategy can be adjusted in time to reduce the losses caused by the fire. Based on the central point movement trend analysis, the fire spreading direction obtained can help emergency management personnel better understand the fire development trend. Through the analysis of the spreading direction, the most extended area of the fire can be determined, and personnel evacuation or centralized deployment of fire-fighting resources can be carried out in time. Based on the fire spreading trend data, adjusting the camera view helps to ensure that the monitoring equipment always covers the core area of the fire. This measure can continuously obtain the latest information of the fire scene and provide accurate fire situation dynamics for the command center. After the view adjustment, the camera can more effectively monitor the changes of the fire, support emergency personnel to obtain the intelligence of the fire scene in real time, and optimize the emergency response efficiency.

[0018] Preferably, step S3 includes the following steps: Step S31: Identify the combustibles within the adjustment range for the in-disaster camera view adjustment data to generate the range combustible identification data; perform spatial positioning on the range combustible identification data to generate the range combustible spatial position information data; Step S32: Calibrate the dangerous area for the in-disaster camera view adjustment data through the range combustible spatial position information data to generate the in-disaster combustible area calibration data; extract the combustible aggregation characteristics from the in-disaster combustible area calibration data to obtain the combustible spatial aggregation characteristic data; Step S33: Divide the combustible spatial aggregation feature data into datasets to generate a model training set and a model test set; train the model using the model training set through a convolutional neural network algorithm to generate a preliminary fire risk prediction model; use the model test set to optimize and iterate the preliminary fire risk prediction model to generate a fire risk prediction model; Step S34: Import the in-disaster camera perspective adjustment data into the fire risk prediction model to predict the fire spread risk and generate fire spread prediction data.

[0019] The present invention can effectively monitor potential fire sources at the fire scene by adjusting the in-disaster camera perspective and identifying combustibles. The identified combustible data provides crucial information for the occurrence and spread of fires. Timely identification of on-site combustibles, especially in the initial stage of a fire, can help the fire department take effective measures to prevent the spread of the fire. Locating the combustibles in space and generating combustible spatial location information can accurately mark the positions of combustibles in the disaster area. Through accurate spatial positioning, decision-makers can more efficiently dispatch fire-fighting resources, especially in complex fire scenes, which helps to better implement fire-fighting and evacuation strategies. Based on the combustible spatial location information, by calibrating the dangerous areas through the in-disaster camera perspective adjustment data, high-risk areas for fire occurrence and spread can be identified, which helps to determine which areas are most vulnerable to fire threats, thus providing accurate fire-fighting key areas for firefighters. Area calibration can also guide the planning of emergency evacuation routes to ensure the safety of personnel. By extracting the spatial aggregation characteristics of combustibles, the potential danger of fire occurrence can be evaluated. For example, the aggregation of multiple combustibles leads to an accelerated fire spread speed, and extracting these characteristics can provide a strong basis for fire risk assessment, helping to take preventive measures in advance. By dividing the combustible aggregation feature data into datasets to generate a training set and a test set, and using a convolutional neural network (CNN) to train the training set, a preliminary fire risk prediction model is generated. The CNN algorithm can effectively extract spatial and temporal features from a large amount of data, has strong pattern recognition ability, and is suitable for dealing with complex fire risk prediction problems. Using the test set to optimize and iterate the preliminary model gradually improves the prediction accuracy of the model. Through this optimization process, the finally generated fire risk prediction model can more accurately predict the fire spread trend and potential risks, which provides a scientific basis for disaster prevention and emergency response and helps the disaster prevention and mitigation work to be more targeted. Import the in-disaster camera perspective adjustment data into the fire risk prediction model to predict the fire spread risk. By predicting the fire spread risk, the expansion range and speed of the fire can be understood in advance, thus helping the relevant departments to take emergency measures in a timely manner.

[0020] In this specification, an intelligent electrical fire risk assessment device based on data analysis is provided, which is used to execute the above-mentioned intelligent electrical fire risk assessment method based on data analysis. The intelligent electrical fire risk assessment device based on data analysis includes: An abnormal device identification module, which is used to obtain the operation data of electrical equipment; upload the operation data of electrical equipment to the cloud platform in real time for data integration to generate an operation monitoring data set of electrical equipment; analyze the abnormal electrical equipment in the operation monitoring data set of electrical equipment to obtain abnormal electrical equipment identification data; A fire situation analysis module, which is used to extract the IP location from the operation monitoring data set of electrical equipment based on the abnormal electrical equipment identification data to obtain the IP address data of abnormal electrical equipment; retrieve the cameras within the address range through the IP address data of abnormal electrical equipment to obtain the retrieved data of cameras within the range of abnormal electrical equipment; analyze the fire spread trend during the disaster for the IP address data of abnormal electrical equipment based on the retrieved data of cameras within the range of abnormal electrical equipment to generate fire spread trend data; adjust the camera view for the retrieved data of cameras within the range of abnormal electrical equipment based on the fire spread trend data to generate the adjusted camera view data during the disaster; A fire spread prediction module, which is used to locate the combustibles for the adjusted camera view data during the disaster to generate the spatial location information data of combustibles within the range; extract the aggregation characteristics of combustibles from the adjusted camera view data during the disaster through the spatial location information data of combustibles within the range to obtain the spatial aggregation characteristics data of combustibles; predict the fire growth risk for the adjusted camera view data during the disaster using the spatial aggregation characteristics data of combustibles to generate the fire growth prediction data; A risk assessment module, which is used to label the fire risk level according to the fire growth prediction data to generate the labeled data of fire risk level; construct a fire decision based on the labeled data of fire risk level to generate a fire risk decision report for performing intelligent electrical fire risk control operations.

[0021] A cloud platform, which is communicatively connected to multiple intelligent electrical fire risk assessment devices based on data analysis and is used to execute the above-mentioned intelligent electrical fire risk assessment method based on data analysis.

[0022] The beneficial effects of the present invention are as follows: By adjusting the perspective of the in-disaster camera and identifying combustibles, it is possible to effectively monitor potential fire sources at the fire scene. The identified combustible data provides crucial information for the occurrence and spread of fires. Timely identification of on-site combustibles, especially in the initial stage of a fire, can help the fire department take effective measures to prevent the spread of the fire. Spatial positioning of combustibles and generation of combustible spatial location information can accurately mark the positions of combustibles in the disaster area. Through accurate spatial positioning, decision-makers can more efficiently dispatch fire-fighting resources, especially in complex fire scenes, which helps to better implement fire-fighting and evacuation strategies. Based on the combustible spatial location information, by adjusting the data of the in-disaster camera perspective for dangerous area calibration, high-risk areas for fire occurrence and spread can be identified, which helps to determine which areas are most vulnerable to fire threats, thus providing accurate fire-fighting key areas for firefighters. Area calibration can also guide the planning of emergency evacuation routes to ensure the safety of personnel. By extracting the spatial aggregation characteristics of combustibles, the potential danger of fire occurrence can be evaluated. For example, the aggregation of multiple combustibles leads to an accelerated fire spread speed, and extracting these characteristics can provide a strong basis for fire risk assessment and help to take preventive measures in advance. By dividing the combustible aggregation characteristic data into a training set and a test set, and using a convolutional neural network (CNN) to train the training set, a fire risk prediction pre-model is generated. The CNN algorithm can effectively extract spatial and temporal characteristics from a large amount of data, has strong pattern recognition ability, and is suitable for dealing with complex fire risk prediction problems. Using the test set to optimize and iterate the pre-model, gradually improving the prediction accuracy of the model. Through this optimization process, the finally generated fire risk prediction model can more accurately predict the spread trend and potential risks of fires, which provides a scientific basis for disaster prevention and emergency response and helps disaster prevention and mitigation work to be more targeted. Importing the in-disaster camera perspective adjustment data into the fire risk prediction model for fire growth risk prediction. By predicting the fire growth risk, the expansion range and speed of the fire can be understood in advance, thus helping relevant departments to take emergency measures in a timely manner. Therefore, the present invention improves the accuracy and comprehensiveness of fire risk assessment through real-time data integration, anomaly detection, fire spread prediction, and intelligent decision support. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a schematic diagram of the step flow of an intelligent electrical fire risk assessment method based on data analysis; Figure 2 is Figure 1 a detailed implementation step flow diagram of step S2 in Figure 3 is Figure 1 a detailed implementation step flow diagram of step S3 in The realization, functional features and advantages of the present invention will be further described with reference to the embodiments and the accompanying drawings. Detailed implementation manners

[0024] The technical method of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0025] In addition, the accompanying drawings are only schematic diagrams of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, and thus their repeated description will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0026] It should be understood that although terms such as "first" and "second" may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly, the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the associated listed items.

[0027] To achieve the above object, please refer to Figures 1 to 3 , an intelligent electrical fire risk assessment method based on data analysis, the method comprising the following steps: Step S1: Obtain the operation data of electrical equipment; upload the operation data of electrical equipment to the cloud platform in real time for data integration to generate an operation monitoring data set of electrical equipment; analyze the abnormal electrical equipment from the operation monitoring data set of electrical equipment to obtain abnormal electrical equipment identification data; Step S2: Extract the IP location from the electrical equipment operation monitoring data set based on the abnormal electrical equipment identification data to obtain the abnormal electrical equipment IP address data; retrieve the cameras within the address range through the abnormal electrical equipment IP address data to obtain the camera retrieval data within the abnormal electrical equipment range; analyze the fire spread trend of the abnormal electrical equipment IP address data based on the camera retrieval data within the abnormal electrical equipment range to generate the fire spread trend data; adjust the camera view of the camera retrieval data within the abnormal electrical equipment range based on the fire spread trend data to generate the camera view adjustment data during the disaster. Step S3: Locate the combustibles in the camera view adjustment data during the disaster to generate the spatial location information data of the combustibles within the range; extract the aggregation characteristics of the combustibles from the camera view adjustment data during the disaster through the spatial location information data of the combustibles within the range to obtain the spatial aggregation characteristics data of the combustibles; predict the fire growth risk of the camera view adjustment data during the disaster using the spatial aggregation characteristics data of the combustibles to generate the fire growth prediction data. Step S4: Mark the fire risk level according to the fire growth prediction data to generate the fire risk level marking data; construct a fire decision based on the fire risk level marking data to generate a fire risk decision report for performing the intelligent electrical fire risk control operation.

[0028] The present invention realizes the centralized storage and efficient processing of data by obtaining the operation data of electrical equipment in real time and uploading it to the cloud platform. Through data integration, the equipment status can be comprehensively monitored, anomalies can be detected in a timely manner, and the limitations of traditional manual inspections are avoided. The analysis of abnormal electrical equipment can quickly identify potentially faulty equipment, issue early warnings, and reduce the probability of accidents. By extracting the IP address to locate the position of abnormal electrical equipment, the accuracy of monitoring is improved, ensuring that measures can be taken in a timely manner. By retrieving relevant camera data, the situation at the fire scene can be obtained in real time, ensuring an accurate assessment of the fire spread trend. The analysis of the fire spread trend helps to predict the speed and scope of fire spread, adjust the camera angle, make the fire monitoring more comprehensive, and improve the disaster response ability. The location of combustibles provides an important influencing factor in the process of fire spread, helping to better understand the development of the fire. By extracting the spatial aggregation characteristics of combustibles, the key factors that exacerbate the fire spread in a fire can be identified, providing data support for fire spread prediction. The prediction of fire growth risk can identify the severity of a fire in advance, provide more accurate prevention and control measures, and help reduce fire losses. The annotation of fire risk levels quantifies fire risks in a data-driven manner, improves the scientificity and accuracy of risk assessment, and the generated fire risk decision-making report provides data-based intelligent support for decision-makers, reduces human judgment biases, and improves the timeliness and effectiveness of responses. It provides a basis for intelligent electrical fire risk control operations, optimizes post-disaster emergency response and control strategies, and helps to extinguish fires in a timely and effective manner. Therefore, the present invention improves the accuracy and comprehensiveness of fire risk assessment through real-time data integration, anomaly detection, fire spread prediction, and intelligent decision support.

[0029] In an embodiment of the present invention, referring to Figure 1 as described, it is a schematic diagram of the step flow of a method for intelligent electrical fire risk assessment based on data analysis according to the present invention. In this example, the method for intelligent electrical fire risk assessment based on data analysis includes the following steps: Step S1: Obtain the operation data of electrical equipment; upload the operation data of electrical equipment to the cloud platform in real time for data integration to generate an operation monitoring dataset of electrical equipment; perform analysis of abnormal electrical equipment on the operation monitoring dataset of electrical equipment to obtain abnormal electrical equipment identification data; In the embodiments of the present invention, the operating state data of the equipment is collected in real time through sensors installed on electrical equipment (such as current sensors, voltage sensors, temperature sensors, vibration sensors, etc.). These sensors can detect the power load, operating temperature, voltage fluctuation, vibration condition, etc. of the equipment. According to the requirements of equipment operation, an appropriate collection frequency (such as per second, per minute, etc.) is set to ensure the real-time nature of the data. The collected operation data is uploaded to the cloud platform in real time through Internet of Things (IoT) devices or gateways using wireless networks (such as Wi-Fi, 4G, 5G, etc.). The role of the cloud platform is to centrally store, process, and analyze the data. Standard protocols such as MQTT, HTTP, CoAP, etc. can be used for data transmission to ensure the stability and efficiency of data upload. The cloud platform integrates the operation data from different devices to establish a dataset for monitoring the operation of electrical equipment. This dataset includes various state indicators of all devices, facilitating subsequent analysis. Through the monitoring system of the cloud platform, the operation states of each electrical equipment are displayed in real time to ensure that the operation conditions of the equipment can be timely feedback. Machine learning or statistical analysis methods are used to analyze the dataset for monitoring the operation of electrical equipment. Common anomaly detection methods include threshold-based detection, deviation detection based on historical data trends, and anomaly pattern recognition based on deep learning. If certain indicators (such as voltage, current, temperature, etc.) exceed the set safety thresholds, the system will mark them as anomalies. By analyzing the trends of the equipment's historical data, it is judged whether the current operation data deviates from the normal trend. For example, support vector machines (SVM) or neural networks (such as LSTM) are used to identify potential anomaly patterns. The analysis results will output the identification data of the abnormal electrical equipment, such as equipment ID, anomaly type (such as overload, over-temperature, vibration anomaly, etc.), timestamp of the anomaly occurrence, anomaly severity, etc. The system generates alarm information based on the anomaly identification data to notify maintenance personnel or management personnel to ensure timely handling of equipment failures and prevent larger-scale equipment damage or shutdown.

[0030] Step S2: Extract the IP location from the dataset for monitoring the operation of electrical equipment based on the identification data of the abnormal electrical equipment to obtain the IP address data of the abnormal electrical equipment; retrieve the cameras within the address range through the IP address data of the abnormal electrical equipment to obtain the camera retrieval data within the range of the abnormal electrical equipment; analyze the fire spread trend in the disaster for the IP address data of the abnormal electrical equipment based on the camera retrieval data within the range of the abnormal electrical equipment to generate the fire spread trend data; adjust the camera view for the camera retrieval data within the range of the abnormal electrical equipment based on the fire spread trend data to generate the camera view adjustment data during the disaster; In the embodiments of the present invention, by searching for the IP address of the abnormal electrical device in the network according to the identification data of the device, this process usually needs to rely on the network management system of the device to retrieve the corresponding IP address in the data set or database through the device ID or serial number. Record the extracted IP address as "abnormal electrical device IP address data", which is the basic data for subsequent operations. Based on the IP address data of the location where the electrical device is located, determine the geographical location of the device. If the device is connected to a certain camera or within the monitoring range of a camera, the system will automatically retrieve the corresponding camera video data. For example, through data matching between the IP address and the camera management system, identify the monitoring area where the device is located, and use the positioning system or internal map of the camera to determine which cameras can cover the device. The system organizes the obtained camera monitoring data (such as video streams, pictures, etc.) into "abnormal electrical device range camera retrieval data". Combining the on-site pictures in the camera data and the abnormal information of the device (such as high temperature, electrical fire, etc.), analyze the fire spread trend through the fire spread prediction model. This process requires integrating fire simulation algorithms, such as a fire spread model based on a physical model (such as a heat transfer model, a smoke diffusion model, etc.) or a data-driven machine learning model (such as a fire prediction based on time series). Information such as smoke and fire extracted from the camera image or video, combined with environmental parameters (such as wind speed, temperature, humidity, etc.), is input into the fire spread model. The model processes the real-time camera data and environmental data to analyze the direction, speed, dangerous areas, etc. of the fire spread, and generates "fire spread trend data". According to the fire spread trend data, adjust the viewing angle of the camera to ensure that the monitoring range covers the area where the fire spreads. The viewing angle adjustment of the camera can be completed through an automatic control system or manual intervention to ensure that the camera always focuses on the high-risk areas of the fire. According to the fire spread trend data (such as the spread direction and speed of the fire), calculate and adjust the viewing angle of the camera so that it can cover more areas affected by the fire. If the system detects that the fire spread speed is relatively fast or there are abnormal changes, it can trigger the automatic adjustment of the camera viewing angle. At the same time, the system can also provide the viewing angle adjustment suggestions to security or emergency response personnel for manual intervention. Finally, generate "in-disaster camera viewing angle adjustment data", including the new viewing angle of the camera, the adjusted monitoring area, the adjustment time point, etc. This data can be fed back to the monitoring platform to assist in further emergency response decisions.

[0031] Step S3: Locate combustibles for the in-disaster camera viewing angle adjustment data to generate range combustible spatial location information data; extract the combustible aggregation characteristics from the in-disaster camera viewing angle adjustment data through the range combustible spatial location information data to obtain combustible spatial aggregation characteristic data; use the combustible spatial aggregation characteristic data to predict the fire growth risk for the in-disaster camera viewing angle adjustment data to generate fire growth prediction data; In the embodiments of the present invention, by adjusting data (video or image data) based on the perspective of cameras during a disaster, image processing technologies (such as computer vision and deep learning) are used to detect the positions of combustibles. Common techniques include: using deep learning models (such as YOLO, Faster R-CNN, RetinaNet, etc.) for combustible detection to identify combustible objects (such as paper, wood, plastic, etc.) in the camera view. Using semantic segmentation techniques (such as U-Net, DeepLab, etc.) to perform pixel-level annotation on camera images to accurately separate the combustible areas. Through these methods, the system can determine the positions of combustibles and mark them as "spatial position information data of the combustible range", including the coordinates of the combustibles in the camera view, the object types, and the descriptions of their surrounding environments. Based on the positioning of the combustibles, through spatial aggregation analysis, the aggregation characteristics of the combustibles in space are extracted. This step can help identify which areas have a higher fire risk during a fire. Using clustering algorithms (such as K-means, DBSCAN, etc.) to cluster the position data of the combustibles to identify the distribution patterns of the combustibles and determine which areas have a higher combustible density. Based on the clustering results, calculate the distribution density of the combustibles in space. For example, the number of combustibles per unit area in each monitoring area can be calculated to generate "spatial aggregation characteristic data of the combustibles". According to the combustible density data, generate a heat map to indicate which areas have a higher risk of combustible aggregation. Using the spatial aggregation characteristic data of the combustibles, combined with a fire propagation model to predict the growth risk of the fire. These models are usually built based on factors such as heat transfer, airflow models, and flame spread speed. The prediction of the fire growth is achieved through the following steps: Combine environmental data (such as temperature, humidity, wind speed, etc.) and analyze their impacts on the fire growth. When the wind speed is high, the fire spreads faster; when the humidity is high, the risk of fire spread is lower. According to the types, sizes, and distribution densities of the combustibles, calculate the fire growth potential in different areas. For example, substances with a faster burning speed such as wood and plastic will accelerate the spread of the fire. Combine the aforementioned environmental factors and combustible characteristics, and use machine learning algorithms (such as regression analysis, decision tree, support vector machine, etc.) to predict the fire growth. The model will output the fire growth risk of each area, including information such as the fire spread speed and the severity of the fire, to generate "fire growth prediction data". Combine the prediction results with the camera images and display the fire growth path and risk areas through a graphical interface to help decision-makers quickly assess the fire risk.

[0032] Step S4: Perform fire risk level annotation according to the fire growth prediction data to generate fire risk level annotation data; construct a fire decision based on the fire risk level annotation data to generate a fire risk decision report to execute the intelligent electrical fire risk control operation.

[0033] In the embodiments of the present invention, different risk levels are set according to information such as the fire spread speed, fire severity, and fire spread direction provided in the fire growth prediction data. Common risk levels may include: Low risk: The fire spread speed is slow, and the combustible density is low, and the fire does not pose an immediate threat to the equipment. Medium risk: The fire spread speed is moderate, and there is a certain possibility of fire spread, and vigilance needs to be maintained. High risk: The fire spread speed is fast, and the combustibles are dense, and the risk of fire spread is large, and immediate measures need to be taken. Determine the fire risk level according to the following factors: The faster the fire spread speed, the higher the fire risk level. The more concentrated the combustibles, the greater the possibility of fire spread, and the higher the risk level. The influence of environmental factors such as high temperature, wind speed, and humidity on fire growth. Use a risk scoring algorithm based on prediction data (such as the weighted average method, the analytic hierarchy process, etc.) to calibrate the fire risk of each area and generate "fire risk level annotation data", including: risk level annotation (low / medium / high) and information such as the fire spread speed and severity of the relevant area. According to the fire risk level annotation data, construct a fire emergency decision-making model, comprehensively consider the severity, influence range of the fire and the feasibility of prevention and control measures, and propose corresponding control plans. The decision-making model can combine methods such as expert systems, fuzzy logic, and decision trees to automatically generate decisions according to the risk level. If the fire is at a low risk level, the system can recommend conventional monitoring and evacuation plans to ensure safety prevention. If the fire is at a medium risk level, the system can recommend strengthening monitoring within the area, dispatching fire-fighting equipment, and preparing fire-fighting equipment in advance. For high-risk areas, the system can recommend measures such as emergency evacuation, activation of automatic fire extinguishing systems, evacuation of personnel, and rapid response of the fire brigade.

[0034] Preferably, step S1 includes the following steps: Step S11: Obtain electrical equipment operation data through multi-source sensors; Step S12: Perform data preprocessing on the electrical equipment operation data to generate standard electrical equipment operation data, where data preprocessing includes data cleaning, data denoising, missing value filling, and data standardization; Step S13: Real-time upload the standard electrical equipment operation data to the cloud platform for equipment identification to generate electrical equipment identification data; perform data integration on the standard electrical equipment operation data based on the electrical equipment identification data to generate an electrical equipment operation monitoring dataset; Step S14: Analyze the abnormal electrical equipment in the electrical equipment operation monitoring dataset to obtain abnormal electrical equipment identification data.

[0035] In the embodiments of the present invention, various sensors (such as temperature sensors, vibration sensors, current / voltage sensors, etc.) installed on electrical equipment are used to collect the operation data of the equipment in different states. The sensors should cover all important indicators of the equipment operation to ensure the comprehensiveness of the data. High-precision measurement technologies (such as fiber optic sensors, MEMS sensors, etc.) should be used for the sensors to ensure the reliability of the data, and a real-time data acquisition system should be adopted to ensure the timeliness of the data. Invalid data (such as abnormal and incorrect readings) and duplicate data should be removed to ensure the data quality. Filtering algorithms (such as Kalman filtering, wavelet denoising, etc.) are used to remove the noise data and improve the accuracy of the data. For the missing values during the acquisition process, interpolation methods (such as linear interpolation or machine learning-based methods) are used to fill in the missing values. The sensor data from different sources are unified and standardized to ensure the comparability of various indicators. Data processing libraries (such as Pandas, Numpy in Python) and machine learning preprocessing tools (such as scikit-learn) can be used to complete the data standardization and cleaning operations. The standardized electrical equipment data are uploaded to the cloud platform in real time through Internet of Things (IoT) devices or communication networks (such as 5G, Wi-Fi, etc.). The cloud platform uses device identification algorithms (such as device ID, device type, etc.) to identify the devices. In the cloud platform, the system classifies and marks the device data according to the unique identifier of the device (such as device serial number, model, etc.) to generate electrical equipment identification data. Lightweight protocols such as MQTT, CoAP, etc. can be used to implement the data upload and transmission, and the cloud platform can be based on big data analysis and device management platforms (such as AWS IoT, Azure IoT Hub, etc.) for device identification and data storage. Based on the electrical equipment identification data, the standard electrical equipment operation data from different devices and different sensors are uniformly integrated to generate a complete electrical equipment operation monitoring data set. This process needs to ensure the timeliness and consistency of the data for subsequent analysis. During the data integration process, the data from different sources need to be aligned (such as aligned by timestamp) and merged into a unified data structure (such as table, database, etc.). ETL tools (such as Apache Nifi) or data integration platforms (such as Talend) can be used for data extraction, transformation, loading, and relational or non-relational databases (such as MySQL, MongoDB) are used for storage. Machine learning or statistical analysis methods are used to perform anomaly detection on the electrical equipment operation monitoring data set. For example, based on historical data, an anomaly detection model (such as K-Means based on clustering, Isolation Forest, neural network, etc.) is constructed to identify the devices with abnormal operations. The anomaly device identification data can include information such as device ID, anomaly type, anomaly occurrence time, etc., which are used for subsequent alarm and maintenance operations.Adopt popular anomaly detection algorithms (such as ARIMA based on time series analysis, LSTM neural network based on models, etc.), and combine with data visualization tools (such as Grafana, PowerBI) to display the anomaly analysis results.

[0036] Preferably, step S14 includes the following steps: Step S141: Extract the operation time domain features from the electrical equipment operation monitoring data set to obtain the equipment operation time domain feature data; perform fast Fourier transform on the equipment operation time domain feature data to generate the equipment operation frequency domain feature data; Step S142: Analyze the trend change of the equipment operation frequency domain feature data to generate the equipment operation frequency domain trend change curve; segment the curve time period of the equipment operation frequency domain trend change curve based on the preset time range to generate the equipment operation short-term trend change curve and the equipment operation long-term trend change curve; Step S143: Monitor the zero-value mutation anomaly of the equipment operation short-term trend change curve to generate the first anomaly state data of the abnormal equipment; monitor the curve change rate anomaly of the equipment operation long-term trend change curve to generate the second anomaly state data of the abnormal equipment; Step S144: Screen the abnormal electrical equipment from the electrical equipment operation monitoring data set through the first anomaly state data of the abnormal equipment and the second anomaly state data of the abnormal equipment to obtain the abnormal electrical equipment operation data; identify the equipment for the abnormal electrical equipment operation data, so as to obtain the abnormal electrical equipment identification data.

[0037] In the embodiments of the present invention, time-domain features of device operation are extracted from the operation monitoring dataset of electrical equipment, such as time series data of current, voltage, power, temperature, etc. Common time-domain features include mean, standard deviation, kurtosis, skewness, maximum value, minimum value, etc. Statistical analysis methods (such as sliding window, clustering analysis, etc.) are used to extract features from the time-domain data to obtain time-domain feature data reflecting the operation state of the device. The fast Fourier transform (FFT) is applied to the time-domain feature data to convert it from the time domain to the frequency domain and extract frequency features. The Fourier transform can help identify the working state of electrical equipment at different frequencies, especially in the case of abnormal fluctuations or noises, generating frequency-domain feature data, including frequency components, amplitudes, phases, etc., to help further analyze the working frequency and fault mode of the device. By performing trend analysis on the frequency-domain feature data of the device (such as using the sliding window method, linear regression or polynomial regression method), frequency-domain trend change curves of device operation are generated, and these curves can reflect the change of the working state of the device at different frequency bands over time. The frequency-domain trend change curves are segmented into time periods according to a preset time range, divided into short-term trend change curves and long-term trend change curves. The short-term trend mainly focuses on the fluctuations of the device in a short period of time, while the long-term trend reflects the long-term change trend of the device. The time window segmentation or sliding window method is adopted to dynamically divide the data according to different device requirements and monitoring objectives. The sudden change to zero value (such as the current and voltage jumping from the normal value to zero or a very small value) in the short-term trend change curve usually indicates that the device has a fault or anomaly. By setting a threshold detection, when the short-term trend data has a sudden change to zero value, it can be marked as the first abnormal state. The sudden change to zero value can be detected by combining control limit methods (such as Shewhart control chart) or rule-based methods. The abnormal change rate in the long-term trend change curve can be identified by calculating the change rate of each time period (such as slope change). If there are severe fluctuations in the change rate during the long-term operation of the device, it is specifically caused by device aging, damage or external interference. Techniques such as the dynamic time warping (DTW) method, acceleration / acceleration ratio analysis, etc. are used to identify the anomalies where the change rate exceeds the set threshold. Using the first abnormal state data (the result of sudden change to zero value detection) and the second abnormal state data (the result of abnormal change rate detection), the operation data of the device is screened to find out the devices with anomalies. By combining the two abnormal state data, the faulty devices can be identified more accurately. Combining information such as device type and operation history records, the faulty devices are screened out. The data of the screened abnormal devices are marked, and detailed abnormal electrical device identification data are generated, including device ID, abnormal type, abnormal occurrence time, abnormal level and other information, so that further fault diagnosis and maintenance can be carried out later.

[0038] As an example of the present invention, refer to Figure 2 shown, in this example, step S2 includes: Step S21: Based on the abnormal electrical equipment identification data, perform real geographical mapping of the electrical equipment operation monitoring data set for abnormal electrical equipment to generate real geographical mapping data of abnormal electrical equipment; extract the IP location from the real geographical mapping data of abnormal electrical equipment to obtain the IP address data of abnormal electrical equipment; Step S22: Use the IP address data of abnormal electrical equipment to retrieve cameras within the address range from the cloud platform to obtain camera retrieval data within the range of abnormal electrical equipment; collect regional smoke sensor data for the IP address data of abnormal electrical equipment based on the camera retrieval data within the range of abnormal electrical equipment to obtain regional smoke sensor data of abnormal electrical equipment; Step S23: Based on the regional smoke sensor data of abnormal electrical equipment, perform disaster - middle layer separation on the camera retrieval data within the range of abnormal electrical equipment to generate disaster - middle smoke layer data and fire alarm data; Step S24: Analyze the spread trend of the fire alarm data through the disaster - middle smoke layer data to generate fire spread trend data; adjust the camera perspective of the camera retrieval data within the range of abnormal electrical equipment based on the fire spread trend data to generate disaster - middle camera perspective adjustment data.

[0039] In the embodiments of the present invention, by performing anomaly analysis on the operation monitoring data of electrical equipment, the identification information of abnormal equipment (such as equipment ID, location, fault type, etc.) is obtained. Based on the physical location of the equipment (such as GPS coordinates, building location, etc.), the equipment is mapped to the actual geographical location. For equipment without GPS positioning, the equipment location can be deduced through the location code or IP address in the building automation system. By combining the equipment identification data with the location data, geographical mapping data of abnormal electrical equipment is generated, and this data can be used for geographical distribution display by visualization tools (such as on a GIS platform). Using the IP address of the electrical equipment, the IP information of the equipment is extracted from the network management system (such as routers, servers) or the device communication protocol (such as Modbus, MQTT). The geographical location data of the equipment is matched with its IP address to generate abnormal electrical equipment IP address data containing the equipment location information and IP address. The IP location is extracted using network protocols (such as ARP, ICMP) or the device management system. The location information and IP address of the equipment are stored in a database (such as MySQL, MongoDB) for extraction and integration. Based on the IP address of the abnormal electrical equipment, a query is made through the cloud platform and the connected camera system to obtain the camera coverage data for the equipment. The association between the equipment and the camera can be achieved through geographical information, IP address location matching, or the monitoring camera location database within the equipment area. The cloud platform obtains the image data around the equipment in real time by retrieving camera video streams, snapshots, etc. According to the area where the abnormal equipment is located, monitoring is carried out through smoke sensors (such as MQ-2, MiCS-5524, etc.). The sensor data can be sent to the cloud platform through protocols such as wireless network, LoRa, ZigBee, etc. The cloud platform determines the area to be collected (i.e., the monitoring range of the abnormal electrical equipment) based on the retrieved data from the camera and obtains data from the smoke sensors. The collected smoke sensor data includes smoke concentration values, temperature, humidity, etc., which reflect the environmental changes around the equipment in real time. The Internet of Things (IoT) device communication protocol (such as MQTT, CoAP) is used for data transmission with the smoke sensors. The sensor data is uploaded to the cloud platform (such as AWS IoT, Google Cloud IoT) for storage and analysis. The image data obtained from the camera is combined with the smoke sensor data for multimodal analysis of disaster events. Using image processing and data fusion techniques, disaster features such as smoke and fire are extracted from the camera images to generate in-disaster smoke layer data and fire alarm data. The smoke layer shows the smoke distribution in the fire area, while the fire alarm data represents information such as the intensity and development direction of the fire. According to the smoke layer and fire alarm data, a fire spread model (such as a wind speed model, heat propagation model, etc.) is used to predict the trend of fire spread.Conduct dynamic analysis of the fire spread trend in combination with environmental data (such as wind speed, air pressure, humidity, etc.), generate data reflecting the fire development direction and speed, which can be updated in real time and used to adjust the monitoring strategy. Use physical modeling methods (such as FDS, CFD simulation) to simulate the fire spread, and combine machine learning methods (such as random forest, SVM) for trend prediction. Based on the fire spread trend data, adjust the camera view angle to ensure that the camera can cover the fire spread area and its surrounding environment. Through an automated camera control system, adjust the camera in real time according to the spread trend to ensure the optimal view angle, generate control instructions, and change parameters such as the rotation angle and focal length of the camera to ensure panoramic monitoring. Use an automated camera control system (such as PTZ cameras, pan-tilt control) for view angle adjustment, and realize the transmission and execution of device control instructions based on the IoT platform.

[0040] Preferably, step S23 includes the following steps: Step S231: Screen the frame images of the abnormal electrical equipment range camera to obtain the in-disaster frame images of the abnormal electrical equipment; convert the in-disaster frame images of the abnormal electrical equipment into grayscale images to generate the in-disaster grayscale images of the abnormal electrical equipment. Step S232: Detect the smoke area in the in-disaster grayscale images of the abnormal electrical equipment through the smoke sensing data of the abnormal electrical equipment area to generate the in-disaster smoke area detection data of the abnormal electrical equipment; perform the first layer separation on the in-disaster grayscale images of the abnormal electrical equipment according to the in-disaster smoke area detection data of the abnormal electrical equipment to obtain the in-disaster smoke layer data. Step S233: Use the in-disaster smoke layer data to perform the second layer separation on the in-disaster frame images of the abnormal electrical equipment to generate the in-disaster fire layer data; perform precise fire alarm analysis on the in-disaster fire layer data to generate the fire alarm data.

[0041] In the embodiments of the present invention, images of the disaster event moment captured by a camera are obtained through a cloud platform. According to the device location and the time point when the fire occurs, in-time images frames are obtained. The image sequence obtained by the camera is screened, and frame images with disaster characteristics (for example, the moment when the fire starts and the smoke spreads) are selected, and irrelevant or poorly defined frames are eliminated. The frame difference method and image edge detection (such as Canny edge detection) are used to screen important image frames. Through time series analysis (such as anomaly detection based on time series), according to information such as the timestamp of the camera data and the device location, images at critical moments during the disaster are selected. The in-disaster frame images of abnormal electrical equipment are converted into grayscale images because grayscale images simplify color information and can more efficiently process contrast changes related to smoke. The smoke sensor data (such as smoke concentration, temperature, humidity, etc.) in the area of the abnormal electrical equipment is used to guide the detection of the smoke area in the image. The sensor data provides actual data on the smoke concentration, while image processing helps to locate the spatial distribution of the smoke. Canny edge detection combined with threshold segmentation is used to detect the area related to smoke in the grayscale image. The smoke characteristics in the image are extracted through texture analysis (such as gray level co-occurrence matrix), and the generated smoke area detection data includes the location where the smoke appears, the concentration range, and the distribution, and these data can be fused with the sensor data to further verify the smoke area in the image. Based on the smoke area detection data, the smoke layer is separated from the grayscale image. The separation process extracts the part related to smoke in the image and eliminates factors such as background information, other irrelevant objects, and light changes. Image segmentation techniques (such as GrabCut, Watershed algorithm) are used to perform image segmentation of the smoke area. Combining the smoke concentration data and the image analysis results, methods such as setting thresholds or weighted averaging are used for area extraction. The separated smoke layer data contains information such as the location, shape, and concentration change of the in-disaster smoke, and can be used for further analysis of the fire spread trend and the device state. Based on the in-disaster smoke layer data, the image is further separated into a fire layer. The goal of this step is to distinguish the fire-related flames, heat sources, etc. from the smoke and extract the fire information. Thermal imaging image processing or color distribution analysis techniques are used to extract the area related to the flame. Flames usually have a higher temperature, so they can be enhanced through thermal imaging data. If there is no thermal imaging image, the color characteristics of the image (such as typical flame colors like red, yellow, etc.) can be used to extract the fire area. The characteristics of the fire area are extracted through multi-channel separation (such as RGB, HSV channel analysis), and the generated fire layer data will contain information such as the spread of the fire, the size of the flame, and the intensity of the fire. The fire layer will help to determine the severity of the fire and the risk area. Based on the fire layer data, an analysis algorithm is used to accurately evaluate the fire spread trend, which includes the spread speed of the fire, the intensity of the flame, the change direction, etc.Use a fire spread simulation model (such as the FDS model based on environmental conditions like wind speed, humidity, etc.) to conduct dynamic simulation of fires. Utilize deep learning (such as CNN) to train the fire layer data, predict the trend of fire spread, and analyze the fire alarm situation in real time, generating fire alarm data including the current state of the fire (fire area, spread speed), the areas affected by the fire, and the risk level. These data can be used in a decision support system to notify relevant personnel to take emergency response measures.

[0042] Preferably, the precise fire alarm analysis of the in-fire fire layer data includes: Convert the in-fire fire layer data into an infrared image to generate an in-fire fire infrared image; extract the pixel point pigments of the in-fire fire infrared image to obtain the in-fire fire image pixel point pigment data; based on a preset pixel point pigment range, identify the flame pixel points from the in-fire fire image pixel point pigment data to generate flame pixel point identification data; Use the flame pixel point identification data to segment the flame area of the in-fire fire infrared image to generate an in-fire fire core area image; calculate the temperature gradient of the regional pixel points of the in-fire fire core area image to obtain the pixel point temperature gradient data; Conduct fire intensity analysis on the in-fire fire core area image according to the pixel point temperature gradient data to generate fire alarm data.

[0043] In the embodiments of the present invention, by using an infrared sensor to obtain the fire situation layer data in a fire, the image data is converted from a visible light image to an infrared image through a conversion algorithm. The infrared image can more clearly show the heat sources and temperature distributions in the fire area, especially the flame and high-temperature areas. An infrared thermal imaging camera is used to obtain the infrared image of the fire area and record the temperature information of the fire area. The fire situation layer data in the fire is converted into a thermal image (infrared image) through data correction (such as the mapping relationship between temperature and color), and this image will show the temperature gradients in different areas. A dedicated infrared image processing tool (such as the FLIR tool) is used for image conversion and temperature calibration. Infrared thermal imaging technology, such as a FLIR thermal imaging camera, is used for real-time image acquisition and analysis. In the infrared image of the fire situation, each pixel point represents a different temperature value. Through pigment extraction technology, the pigment value of each pixel point (usually proportional to the temperature) is extracted from the infrared image and converted into numerical pixel point pigment data. An algorithm based on color space conversion (such as RGB to HSV conversion) is used to extract the pigment value of each pixel from the infrared image. Algorithms such as K-means clustering are used to separate the areas with high temperature values in order to further identify the flame area. According to the preset pixel point pigment range (usually based on the temperature threshold and the color range of the infrared image), the flame pixel points in the image are identified. The pixels in the high-temperature area usually appear as red, orange, or yellow, etc., and the pixel points corresponding to these areas can be identified as flames. By setting a threshold range (for example, pixel points with a temperature higher than 100 °C are considered as the flame area), the flame area is identified based on the pigment data of the pixel points. The qualified pixel points are calibrated as flame pixel points to generate the corresponding flame pixel point identification data. The infrared image of the fire situation is segmented using the flame pixel point identification data. Through region growing algorithms or contour detection techniques, the flame area is segmented from the image, and the core area of the fire situation is extracted. Based on the identification data of the flame pixel points, the flame area is further expanded using the region growing algorithm to extract a more accurate flame core area. The contour of the flame area is detected by methods such as Canny edge detection to segment the core part of the flame. The temperature gradient of each pixel point in the image of the core area of the fire situation is calculated to evaluate the intensity of the fire. Areas with a high temperature gradient usually represent the core area of the flame. The temperature gradient of each pixel point in the image is calculated using the Sobel operator or the gradient operator. The Sobel operator or the Laplacian operator is used to calculate the gradient of the core area of the fire situation to obtain the temperature change rate of each pixel point, reflecting the intensity of the fire. According to the temperature gradient, the spreading speed and intensity of the flame are evaluated. Based on the calculated temperature gradient data, the intensity of the fire is analyzed. The intensity of the fire can be comprehensively evaluated by the magnitude of the temperature gradient, the temperature distribution in the area, and the spreading speed of the flame. The fire is classified by setting an intensity threshold (for example, a gradient value greater than a certain threshold indicates a strong fire) to generate the fire intensity data.Generate fire alarm data based on fire intensity data, spread speed, and affected areas to help predict the scope and risk of fire spread.

[0044] Preferably, step S24 includes the following steps: Step S241: Calculate the smoke concentration of the fire alarm data through the in-disaster smoke layer data to obtain the in-disaster fire smoke concentration data; based on the in-disaster fire smoke concentration data, screen the adjacent timestamp layers of the in-disaster smoke layer to obtain the first in-disaster smoke layer and the second in-disaster smoke layer; Step S242: Use the in-disaster fire smoke concentration data to confirm the extreme concentration center points of the first in-disaster smoke layer and the second in-disaster smoke layer to obtain the first in-disaster smoke concentration extreme center point and the second in-disaster smoke concentration extreme center point; analyze the center point movement trend of the first in-disaster smoke concentration extreme center point and the second in-disaster smoke concentration extreme center point to generate center point movement trend data; Step S243: Confirm the smoke spread direction of the first in-disaster smoke layer and the second in-disaster smoke layer according to the center point movement trend data to generate fire spread trend data; adjust the camera view of the abnormal electrical equipment range camera call data based on the fire spread trend data to generate in-disaster camera view adjustment data.

[0045] In the embodiments of the present invention, by using the in-disaster smoke layer data, the smoke concentration values of each area are estimated by analyzing the smoke concentrations in different areas. The smoke concentration is usually indirectly calculated by sensing the number density of smoke particles or environmental factors such as temperature and humidity. Using pixel density analysis based on image processing, combined with smoke sensor data, the smoke concentration of each pixel point in the layer is extracted. According to the mapping relationship between the color value of each pixel point in the image and the smoke concentration, the smoke concentration of each area is calculated, and the interpolation method is used to fill the blank areas. The in-disaster smoke layer data is sorted according to the time stamp, and two adjacent layers are selected, usually the image data collected twice before and after. These two layers represent the smoke distribution at adjacent time points. According to the collected time stamp data, the adjacent layer pairs (such as the first layer and the second layer) are matched, and the changes in the smoke concentration in these two layers are analyzed. Using the filtered layers as the analysis basis, the change in the smoke concentration between the two time stamp layers is calculated, and the changed areas are extracted. Through time stamp sorting and time window analysis, the layers with adjacent time stamps are screened for further analysis. The differential image method (for example: the difference calculation between adjacent time layers) is used to screen the changes between the layers, and the changed areas of the two-layer data are obtained. By analyzing the smoke concentration data of each layer, the extreme value center points of the smoke concentration are identified, that is, the areas with the highest concentration. The local maximum detection algorithm is used to identify the positions of the pixel points with the highest concentration in the layer, and these positions are determined as the extreme value center points of the smoke concentration. On each layer, the pixel points in the area with the highest concentration are extracted, and their coordinates are calibrated as the extreme value center points. The local extreme value detection algorithm (for example: the local extreme value detection based on the gradient) is used to find the area with the highest concentration. By comparing the extreme value concentration data of the two layers, the extreme value center points of the first in-disaster smoke layer and the second in-disaster smoke layer are obtained respectively. By comparing the extreme value center points in the two layers, the change trend in the time series is calculated, and the moving direction of the smoke center is analyzed. By calculating the position change of the extreme value center point in two adjacent layers, the moving trend of the center point is obtained, which can be achieved by Euclidean distance calculation or trajectory regression analysis. Based on the time stamp comparison, the moving path of the center point at different times is generated, and the trend analysis of the path is carried out to judge its moving direction (for example, whether it spreads to the southeast). According to the moving trend data of the center point, the main direction of the fire spread is confirmed. By analyzing the moving direction of the extreme value center point, the trend and direction of the fire spread are speculated. The vector analysis method is used to analyze the change trajectory of the extreme value center point to confirm the direction of the smoke spread. Based on the analysis results, the trend data of the fire spread is generated, for example, in which direction the fire spreads and the spread speed. According to the fire spread trend data, the viewing angle of the camera is adjusted to ensure real-time monitoring of the key areas of the fire spread. The camera should follow the direction of the fire spread to provide more efficient monitoring.Using the fire spread trend data, adjust the position or angle of the camera to ensure that it covers the path and area of fire spread. Transmit the perspective adjustment data to the camera control system and adjust the direction of the camera through automatic control.

[0046] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes: Step S31: Identify combustibles within the adjustment range for the in-disaster camera perspective adjustment data to generate range combustible identification data; perform spatial positioning on the range combustible identification data to generate range combustible spatial position information data; Step S32: Calibrate the dangerous area for the in-disaster camera perspective adjustment data through the range combustible spatial position information data to generate in-disaster combustible area calibration data; extract the combustible aggregation characteristics from the in-disaster combustible area calibration data to obtain combustible spatial aggregation characteristic data; Step S33: Divide the combustible spatial aggregation characteristic data into a dataset to generate a model training set and a model test set; train the model training set through a convolutional neural network algorithm to generate a preliminary fire risk prediction model; use the model test set to optimize and iterate the preliminary fire risk prediction model to generate a fire risk prediction model; Step S34: Import the in-disaster camera perspective adjustment data into the fire risk prediction model to predict the fire growth risk and generate fire growth prediction data.

[0047] In the embodiments of the present invention, image preprocessing is performed on the data of the camera view adjustment during the disaster, such as denoising, brightness equalization, contrast enhancement, etc., to ensure that the images are clear and easy to identify. Object detection algorithms based on deep learning (such as YOLO, Faster R-CNN, etc.) are used to identify combustibles in the images. The annotated combustible image dataset is used for training to ensure that the network can efficiently identify combustibles in the images. The combustible areas in the images are marked, and combustible identification data is generated, which includes the category and position coordinates of each combustible object. By combining the camera view adjustment data with the position of the camera, three-dimensional space reconstruction is performed using stereo vision technology or depth perception technology to identify the precise position of each combustible object in space. By matching the combustible identification data with the actual space coordinates, the position and distribution of the combustibles in space are obtained, thereby generating combustible spatial position information data. OpenCV is used for image processing, and a deep learning framework (such as TensorFlow or PyTorch) is combined to implement combustible identification. Through the camera view adjustment information, the spatial position of the combustibles is obtained by combining SLAM (Simultaneous Localization and Mapping) or stereo vision technology. Based on the combustible spatial position information data, the dangerous areas are defined as those areas with a relatively high density of combustibles, close to buildings or electrical equipment. Spatial clustering algorithms (such as DBSCAN or K-means clustering) are used to perform spatial clustering analysis on the combustibles to identify areas with a relatively high fire risk. The clustering results divide the combustibles into different dangerous areas according to their spatial distribution. For each clustering area, the fire risk level of the area (such as low risk, medium risk, high risk) is calibrated to generate in-disaster combustible area calibration data. By analyzing the spatial distribution of the combustibles, aggregation characteristics such as aggregation degree, density distribution, and distance distribution are extracted, and these characteristics can help evaluate the concentration of combustibles and the fire risk in a specific area. Spatial statistical methods (such as spatial information entropy, cluster degree analysis) are used to extract combustible spatial aggregation feature data from the combustible area calibration data. The DBSCAN (Density-based Spatial Clustering of Applications with Noise) or K-means clustering algorithm is used to group the combustibles spatially. Spatial statistical methods (such as k-nearest neighbor analysis, entropy value calculation) are used to extract the aggregation characteristics of the combustibles. The combustible spatial aggregation feature data is randomly divided into a training set and a test set. Usually, the training set accounts for 80% and the test set accounts for 20%. The dataset is standardized or normalized to ensure that the data input into the neural network has the same scale. A convolutional neural network (CNN) is designed, which is suitable for feature extraction and classification tasks of spatial data. The network structure can include multiple convolutional layers, pooling layers, fully connected layers, etc. The training set data is used to train the CNN model, the loss function (such as cross-entropy loss) is optimized, and the weights are continuously adjusted. After training, a preliminary fire risk prediction pre-model is generated. The trained model is evaluated using the test set to check indicators such as prediction accuracy and recall rate.According to the evaluation results, adjust the model parameters (such as learning rate, network structure, etc.) and optimize until the model can accurately predict the fire risk. After multiple optimization iterations, generate the final fire risk prediction model. Input the data for adjusting the camera view during the fire (including the distribution of combustibles in the disaster area, spatial aggregation characteristics, fire spread trend, etc.) into the trained fire risk prediction model. The model predicts the growth risk of the fire based on the input data, including the speed, direction, and high-risk areas of fire spread. The model outputs the predicted data of the fire growth, including the area of fire expansion, estimated time, and intensity.

[0048] In this specification, an intelligent electrical fire risk assessment device based on data analysis is provided for performing the above-mentioned intelligent electrical fire risk assessment method based on data analysis. The intelligent electrical fire risk assessment device based on data analysis includes: An abnormal device identification module, which is used to obtain the operation data of electrical equipment; upload the operation data of electrical equipment to the cloud platform in real time for data integration to generate an operation monitoring data set of electrical equipment; perform analysis of abnormal electrical equipment on the operation monitoring data set of electrical equipment to obtain abnormal electrical equipment identification data; A fire situation analysis module, which is used to extract the IP location from the operation monitoring data set of electrical equipment based on the abnormal electrical equipment identification data to obtain the IP address data of abnormal electrical equipment; retrieve the cameras within the address range through the IP address data of abnormal electrical equipment to obtain the retrieved data of cameras within the range of abnormal electrical equipment; perform analysis of the fire spread trend during the fire on the IP address data of abnormal electrical equipment based on the retrieved data of cameras within the range of abnormal electrical equipment to generate fire spread trend data; adjust the camera view based on the fire spread trend data for the retrieved data of cameras within the range of abnormal electrical equipment to generate the data for adjusting the camera view during the fire. A fire spread prediction module, which is used to locate the combustibles for the data of adjusting the camera view during the fire to generate the spatial position information data of combustibles within the range; extract the aggregation characteristics of combustibles from the data of adjusting the camera view during the fire through the spatial position information data of combustibles within the range to obtain the spatial aggregation characteristics data of combustibles; use the spatial aggregation characteristics data of combustibles to predict the fire growth risk for the data of adjusting the camera view during the fire to generate the fire growth prediction data. A risk assessment module, which is used to label the fire risk level according to the fire growth prediction data to generate the fire risk level labeling data; construct a fire decision based on the fire risk level labeling data to generate a fire risk decision report for performing intelligent electrical fire risk control operations.

[0049] A cloud platform, which is communicatively connected to multiple intelligent electrical fire risk assessment devices based on data analysis and is used to perform the intelligent electrical fire risk assessment method as described above.

[0050] The beneficial effects of the present invention are as follows: by adjusting the perspective of the in-disaster camera and identifying combustibles, it is possible to effectively monitor potential fire sources at the fire scene. The identified combustible data provides crucial information for the occurrence and spread of fires. Timely identification of on-site combustibles, especially in the initial stage of a fire, can help the fire department take effective measures to prevent the spread of the fire. Spatially locating the combustibles and generating combustible spatial location information can accurately mark the positions of combustibles in the disaster area. Through accurate spatial positioning, decision-makers can more efficiently dispatch fire-fighting resources, especially in complex fire scenes, which helps to better implement fire-fighting and evacuation strategies. Based on the combustible spatial location information, by adjusting the data of the in-disaster camera perspective for hazard area calibration, high-risk areas for fire occurrence and spread can be identified, which helps to determine which areas are most vulnerable to fire threats, thus providing accurate fire-fighting key areas for firefighters. Area calibration can also guide the planning of emergency evacuation routes to ensure the safety of personnel. By extracting the spatial aggregation characteristics of combustibles, the potential danger of fire occurrence can be evaluated. For example, the aggregation of multiple combustibles leads to an accelerated fire spread speed, and extracting these characteristics can provide a strong basis for fire risk assessment, helping to take preventive measures in advance. By partitioning the dataset of combustible aggregation feature data into a training set and a test set, and using a convolutional neural network (CNN) to train the training set, a fire risk prediction pre-model is generated. The CNN algorithm can effectively extract spatial and temporal features from a large amount of data, has strong pattern recognition ability, and is suitable for dealing with complex fire risk prediction problems. Using the test set to optimize and iterate the pre-model gradually improves the prediction accuracy of the model. Through this optimization process, the finally generated fire risk prediction model can more accurately predict the spread trend and potential risks of fires, which provides a scientific basis for disaster prevention and emergency response, and helps disaster prevention and mitigation work to be more targeted. Importing the in-disaster camera perspective adjustment data into the fire risk prediction model for fire growth risk prediction. By predicting the fire growth risk, the expansion range and speed of the fire can be understood in advance, thus helping relevant departments to take emergency measures in a timely manner. Therefore, the present invention improves the accuracy and comprehensiveness of fire risk assessment through real-time data integration, anomaly detection, fire spread prediction, and intelligent decision support.

[0051] Therefore, in any regard, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to cover all changes falling within the meaning and scope of the equivalent elements of the application documents within the present invention.

[0052] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features invented herein.

Claims

1. An intelligent electrical fire risk assessment method based on data analysis, characterized in that: The following steps are involved: Step S1: Acquire electrical equipment operation data; upload the electrical equipment operation data to the cloud platform in real time for data integration to generate an electrical equipment operation monitoring data set; perform abnormal electrical equipment analysis on the electrical equipment operation monitoring data set to obtain abnormal electrical equipment identification data; Step S2: extracting the IP location of the electrical equipment operation monitoring data set based on the abnormal electrical equipment identification data to obtain the abnormal electrical equipment IP address data; performing address range camera acquisition through the abnormal electrical equipment IP address data to obtain abnormal electrical equipment range camera acquisition data; Based on the data retrieved by the camera in the abnormal electrical equipment range, the fire spread trend analysis of the abnormal electrical equipment IP address data was performed to generate fire spread trend data; Based on the fire spread trend data, the camera angle is adjusted by retrieving the data from the camera in the abnormal electrical equipment range to generate the camera angle adjustment data during the disaster; Step S3: locating combustibles from the camera angle adjustment data during the disaster, generating range combustible spatial position information data; extracting combustible aggregation features from the camera angle adjustment data during the disaster through the range combustible spatial position information data, obtaining combustible spatial aggregation feature data; predicting the fire growth risk from the camera angle adjustment data during the disaster using the combustible spatial aggregation feature data, generating fire growth prediction data; Step S4: marking the fire risk level according to the fire growth prediction data to generate fire risk level marking data; Fire decision-making is constructed based on the fire risk level annotation data, and a fire risk decision report is generated to perform intelligent electrical fire risk control operations.

2. The intelligent electrical fire risk assessment method based on data analysis according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: Acquire electrical equipment operation data through multi-source sensors; Step S12: performing data preprocessing on the electrical equipment operation data to generate standard electrical equipment operation data, wherein the data preprocessing includes data cleaning, data denoising, missing value filling and data standardization; Step S13: uploading the standard electrical equipment operation data to the cloud platform in real time for equipment identification to generate electrical equipment identification data; integrating the standard electrical equipment operation data based on the electrical equipment identification data to generate an electrical equipment operation monitoring data set; Step S14: performing abnormal electrical equipment analysis on the electrical equipment operation monitoring data set to obtain abnormal electrical equipment identification data.

3. The intelligent electrical fire risk assessment method based on data analysis according to claim 2 is characterized in that: Step S14 includes the following steps: Step S141: extracting the operation domain characteristics of the electrical equipment operation monitoring data set to obtain the equipment operation domain characteristic data; performing fast Fourier transform on the equipment operation domain characteristic data to generate equipment operation frequency domain characteristic data; Step S142: performing trend change analysis on the equipment operation frequency domain characteristic data to generate an equipment operation frequency domain trend change curve; dividing the equipment operation frequency domain trend change curve into curve time periods based on a preset time range to generate an equipment operation short-term trend change curve and an equipment operation long-term trend change curve; Step S143: Perform zero-value mutation abnormality monitoring on the short-term trend change curve of the equipment operation to generate first abnormal state data of the abnormal equipment; perform curve change rate abnormality monitoring on the long-term trend change curve of the equipment operation to generate second abnormal state data of the abnormal equipment; Step S144: Screen the electrical equipment operation monitoring data set for abnormal electrical equipment using the abnormal equipment first abnormal state data and the abnormal equipment second abnormal state data to obtain abnormal electrical equipment operation data; perform equipment identification on the abnormal electrical equipment operation data to obtain abnormal electrical equipment identification data.

4. The intelligent electrical fire risk assessment method based on data analysis according to claim 1 is characterized in that: Step S2 includes the following steps: Step S21: Based on the abnormal electrical equipment identification data, the electrical equipment operation monitoring data set is subjected to abnormal electrical equipment real-time geographic mapping to generate abnormal electrical equipment real-time geographic mapping data; the IP location of the abnormal electrical equipment real-time geographic mapping data is extracted to obtain the abnormal electrical equipment IP address data; Step S22: Using the IP address data of the abnormal electrical equipment, the cloud platform performs address range camera acquisition to obtain abnormal electrical equipment range camera acquisition data; based on the abnormal electrical equipment range camera acquisition data, regional smoke sensor data is collected for the abnormal electrical equipment IP address data to obtain abnormal electrical equipment regional smoke sensor data; Step S23: Based on the smoke sensor data of the abnormal electrical equipment area, the camera data of the abnormal electrical equipment area is retrieved to separate the disaster layer, and generate the disaster smoke layer data and the fire alarm data; Step S24: Analyze the fire alarm data for spreading trends through the smoke layer data during the disaster to generate fire spreading trend data; adjust the camera viewing angle of the abnormal electrical equipment range camera based on the fire spreading trend data to generate camera viewing angle adjustment data during the disaster.

5. The intelligent electrical fire risk assessment method based on data analysis according to claim 4 is characterized in that: Step S23 includes the following steps: Step S231: perform frame image screening on the data retrieved by the camera in the range of abnormal electrical equipment to obtain a frame image of the abnormal electrical equipment in the disaster; perform grayscale conversion on the frame image of the abnormal electrical equipment in the disaster to generate a grayscale image of the abnormal electrical equipment in the disaster; Step S232: performing smoke area detection on the grayscale image of abnormal electrical equipment during disaster using smoke sensing data of the abnormal electrical equipment area to generate smoke area detection data of abnormal electrical equipment during disaster; performing first layer separation on the grayscale image of abnormal electrical equipment during disaster based on the smoke area detection data of abnormal electrical equipment during disaster to obtain smoke layer data of disaster; Step S233: Use the disaster smoke layer data to perform a second layer separation on the disaster frame image of abnormal electrical equipment to generate the disaster fire layer data; perform precise fire warning analysis on the disaster fire layer data to generate fire warning data.

6. The intelligent electrical fire risk assessment method based on data analysis according to claim 5 is characterized in that: Accurate fire warning analysis of fire layer data during a disaster includes: Perform infrared image conversion on the fire layer data in the disaster to generate an infrared image of the fire in the disaster; perform pixel pigment extraction on the infrared image of the fire in the disaster to obtain pixel pigment data of the fire in the disaster image; perform flame pixel recognition on the pixel pigment data of the fire in the disaster image based on a preset pixel pigment range to generate flame pixel recognition data; The flame pixel recognition data is used to segment the flame area of ​​the infrared image of the fire in the disaster, and the core area image of the fire in the disaster is generated; the temperature gradient of the regional pixels in the core area image of the fire in the disaster is calculated to obtain the temperature gradient data of the pixel points; The fire intensity of the core area of ​​the fire in the disaster is analyzed based on the pixel temperature gradient data to generate fire warning data.

7. The intelligent electrical fire risk assessment method based on data analysis according to claim 4 is characterized in that: Step S24 includes the following steps: Step S241: Calculate the smoke density of the fire alarm data through the smoke layer data to obtain the smoke density data of the fire in the disaster; filter the smoke layer of the disaster based on the smoke density data of the fire in the disaster to obtain the first smoke layer of the disaster and the second smoke layer of the disaster; Step S242: using the smoke density data of the fire intensity in the disaster, confirming the center point of the extreme concentration of the first disaster smoke layer and the second disaster smoke layer, obtaining the center point of the extreme concentration of the first disaster smoke layer and the center point of the extreme concentration of the second disaster smoke layer; performing center point movement trend analysis on the center point of the extreme concentration of the first disaster smoke layer and the center point of the extreme concentration of the second disaster smoke layer, generating center point movement trend data; Step S243: Confirm the smoke spread direction of the first disaster smoke layer and the second disaster smoke layer according to the center point movement trend data to generate fire spread trend data; adjust the camera perspective of the abnormal electrical equipment range camera based on the fire spread trend data to generate disaster camera perspective adjustment data.

8. The intelligent electrical fire risk assessment method based on data analysis according to claim 1 is characterized in that: Step S3 includes the following steps: Step S31: Identify combustibles within the adjustment range of the camera angle adjustment data during the disaster, and generate combustible identification data within the range; perform spatial positioning on the combustible identification data within the range, and generate spatial position information data of the combustibles within the range; Step S32: calibrate the dangerous area of ​​the camera viewing angle adjustment data in the disaster by using the spatial position information data of the combustibles in the range to generate the calibration data of the combustible area in the disaster; extract the combustible aggregation features from the calibration data of the combustible area in the disaster to obtain the combustible spatial aggregation feature data; Step S33: dividing the combustible spatial aggregation feature data into a data set to generate a model training set and a model test set; training the model training set using a convolutional neural network algorithm to generate a fire risk prediction pre-model; optimizing and iterating the fire risk prediction pre-model using the model test set to generate a fire risk prediction model; Step S34: Import the camera angle adjustment data during the disaster into the fire risk prediction model to perform fire growth risk prediction and generate fire growth prediction data.

9. An intelligent electrical fire risk assessment device based on data analysis, characterized in that: For executing the intelligent electrical fire risk assessment method based on data analysis as claimed in claim 1, the intelligent electrical fire risk assessment device based on data analysis comprises: The abnormal equipment identification module is used to obtain the operation data of electrical equipment; upload the operation data of electrical equipment to the cloud platform in real time for data integration to generate an electrical equipment operation monitoring data set; perform abnormal electrical equipment analysis on the electrical equipment operation monitoring data set to obtain abnormal electrical equipment identification data; The fire situation analysis module is used to extract the IP location of the electrical equipment operation monitoring data set based on the abnormal electrical equipment identification data to obtain the abnormal electrical equipment IP address data; to retrieve the address range camera through the abnormal electrical equipment IP address data to obtain the abnormal electrical equipment range camera retrieval data; to analyze the fire spread trend during the disaster on the abnormal electrical equipment IP address data based on the abnormal electrical equipment range camera retrieval data to generate the fire spread trend data; to adjust the camera angle of view of the abnormal electrical equipment range camera retrieval data based on the fire spread trend data to generate the disaster camera angle adjustment data; The fire prediction module is used to locate combustibles based on the camera angle adjustment data during the disaster, and generate the range combustible spatial position information data; extract the combustible aggregation features from the camera angle adjustment data during the disaster through the range combustible spatial position information data, and obtain the combustible spatial aggregation feature data; use the combustible spatial aggregation feature data to predict the fire growth risk of the camera angle adjustment data during the disaster, and generate the fire growth prediction data; The risk assessment module is used to label the fire risk level according to the fire growth prediction data and generate fire risk level labeling data; to construct fire decisions based on the fire risk level labeling data and generate fire risk decision reports to execute intelligent electrical fire risk control operations.

10. A cloud platform, characterized in that: The cloud platform is communicatively connected to a plurality of intelligent electrical fire risk assessment devices based on data analysis, and is used to execute the intelligent electrical fire risk assessment method based on data analysis as described in claim 1 above.

Images