Fire information transmission method and system based on edge computing

Abstract

The application discloses a fire-fighting information transmission method and system based on edge computing, and relates to the technical field of data transmission; each edge node acquires initial fire-fighting data uploaded by a corresponding target area; data cleaning is performed on the initial fire-fighting data to obtain effective fire-fighting data; abnormality detection is performed on the effective fire-fighting data to obtain a detection result of each type of fire-fighting data; target upload data uploaded to the cloud is determined according to the detection result of each type of fire-fighting data; the total amount of data of all target upload data is counted, and the theoretical upload time length required for the edge node to upload the target upload data to the cloud server is calculated according to the current available upload bandwidth of the edge node; if the theoretical upload time length is greater than a preset time threshold, a data offloading request is sent from the edge computing node to a neighboring edge computing node, and fire-fighting information is uploaded. Thus, in the case of high-bandwidth data, the transmission efficiency of fire-fighting information is still guaranteed.

Description

Technical Field

[0001] This invention belongs to the field of data transmission technology, specifically relating to a fire information transmission method and system based on edge computing. Background Technology

[0002] High-rise buildings and urban complexes are equipped with a large number of fire protection equipment and monitoring terminals. To ensure fire safety, existing systems typically deploy various types of sensors and monitoring equipment on each floor of the building, in equipment rooms and pipe shafts. The monitoring equipment continuously collects data such as fire water source level, water supply network pressure, fire pump operating current and voltage, fault arc high-frequency waveform data, smoke concentration data, ambient temperature data, combustible gas concentration data, and video surveillance data. This data is then aggregated through an IoT gateway and uploaded to a remote cloud server, where it is used for data storage, analysis, processing, and decision-making.

[0003] However, the large number of fire protection equipment in high-rise buildings, especially the widespread deployment of video surveillance equipment and arc fault detectors, has led to an exponential increase in the amount of data that the system needs to process. Existing technologies typically upload all sensor data to the cloud for centralized processing without discrimination. In this mode, massive amounts of high-bandwidth data need to be uploaded via a wide area network, which consumes a great deal of uplink bandwidth resources. When multiple areas in the building report data at the same time, it is very easy to cause network congestion, resulting in increased data transmission time. Summary of the Invention

[0004] The purpose of this invention is to solve the problem of low efficiency in fire information transmission when faced with high bandwidth data, and to propose a fire information transmission method and system based on edge computing.

[0005] In a first aspect of this invention, a fire information transmission method based on edge computing is first proposed, the method comprising: Each edge node acquires the initial fire protection data uploaded to the corresponding target area; the initial fire protection data is the data collected by each fire protection device in the target area during the collection period; The initial fire data is cleaned to obtain effective fire data; Anomaly detection is performed on the valid fire protection data to obtain the detection results for each type of fire protection data; For each type of fire-fighting data type, determine the target data to be uploaded to the cloud based on the test results; The total amount of data uploaded by all targets is counted, and the theoretical upload time required for the edge node to upload all the target data to the cloud server is calculated based on the current available upload bandwidth of the edge node. If the theoretical upload duration exceeds the preset time threshold, the edge computing node sends a data offload request to the neighboring edge computing node to upload fire information.

[0006] Optionally, the initial fire protection data includes: fire water source level data, water supply network pressure data, fire pump operating current and voltage data, fault arc high-frequency waveform data, smoke concentration data, ambient temperature data, combustible gas concentration data, and video surveillance data.

[0007] Optionally, data cleaning of the initial fire data to obtain effective fire data includes: The fire water source level data, water supply network pressure data, fire pump operating current and voltage data, smoke concentration data, ambient temperature data, and combustible gas concentration data are sequentially subjected to amplitude limiting filtering, median filtering, and sliding window mean filtering to denoise them and obtain the corresponding denoised data. For high-frequency waveform data of fault arc, the high-frequency waveform data of fault arc is decomposed into multiple scales by db4 wavelet, and the corresponding denoised data is obtained by noise reduction processing by wavelet modulus maxima theory. Based on smoke concentration data and ambient temperature data, the video surveillance data is used to extract corresponding valid data by video frame extraction. The processing results of all data categories are obtained to obtain valid fire protection data.

[0008] Initial fire data is cleaned through differentiated classification: fire water source level data, water supply network pressure data, fire pump operating current and voltage data, smoke concentration data, ambient temperature data, and combustible gas concentration data are filtered out by a three-level joint filter to remove pulse interference; fault arc waveforms are decomposed with db4 wavelet decomposition to reduce noise while fully preserving fault characteristics; key video frames are extracted based on temperature and smoke data to reduce video redundancy, thereby integrating various processing results to obtain effective fire data. This not only significantly reduces the amount of data to be transmitted and shortens the upload time, solving the problem of inefficient high-bandwidth data transmission, but also eliminates various signal interferences, ensuring the accuracy of fire and equipment fault identification, and providing a data foundation for edge collaborative offloading and transmission.

[0009] Optionally, extracting corresponding valid data from the video surveillance data by extracting video frames based on smoke concentration data and ambient temperature data includes: If the rate of increase of the smoke concentration data within a preset time exceeds a preset limit, then video frames of the smoke concentration data before and after the abnormal point are obtained as the first valid video data. If the rate of change of temperature data within a preset time exceeds a preset limit, then video frames of the temperature data before and after the abnormal point are obtained as the second valid video data. The valid data corresponding to the video surveillance data is obtained by finding the intersection of the first and second valid video data.

[0010] By triggering frame capture through the linkage of smoke and temperature sensor change rates, only key video clips before and after the abnormal moment are retained. Then, irrelevant images are filtered out from the intersection of the two types of abnormal videos, which greatly reduces the amount of redundant video data, reduces the bandwidth consumption of subsequent transmission, and shortens the upload time. At the same time, only effective images related to the fire are retained to avoid invalid videos occupying cloud storage, thus balancing the effectiveness of video evidence and the efficiency of fire information transmission.

[0011] Optionally, the anomaly detection of the valid fire protection data to obtain the detection results for each type of fire protection data includes: The effective fire protection data corresponding to the fire water source level data, water supply network pressure data, fire pump operating current and voltage data, smoke concentration data, ambient temperature data, and combustible gas concentration data are substituted into a dynamic threshold detection model based on historical window statistics to obtain the detection results for each type of fire protection data; the detection results include those that found anomalies and those that did not. For the effective fire protection data of the high-frequency waveform data of the fault arc, the corresponding current change rate, wavelet high-frequency energy and interharmonic factor are extracted; If the rate of change of current, wavelet high-frequency energy, and interharmonic factor are all greater than the corresponding thresholds, the detection result is recorded as an anomaly detected; otherwise, it is recorded as no anomaly detected. For the valid data corresponding to the video surveillance data, the valid data is substituted into the preset fire anomaly video detection model to obtain the detection result; the detection result includes the detection of anomalies and the absence of anomalies.

[0012] Conventional sensor data relies on historical window dynamic threshold judgment to adapt to slowly fluctuating environmental scenarios; fault arc waveform extraction uses multiple features for joint identification to avoid false alarms from a single indicator; video data identifies fire situations through a dedicated fire video model, and multi-dimensional layered detection can accurately distinguish various fire hazards, reduce missed detections and false detections, and quickly mark abnormal data to facilitate subsequent screening of data to be uploaded, reduce bandwidth consumption from useless data transmission, and improve the overall transmission efficiency of high-bandwidth fire information.

[0013] Optionally, the target data to be uploaded to the cloud for each type of fire protection data includes: For any data type among fire water source level data, water supply network pressure data, fire pump operating current and voltage data, smoke concentration data, ambient temperature data, combustible gas concentration data, and fault arc high-frequency waveform data, if the detection result for this data type is no abnormality, then the data characteristics of this data type within the detection period are extracted as the target uploaded data; the data characteristics include the mean, maximum, minimum, and rate of change within the detection period. If the detection result of this data type is that an anomaly is found, then the original data segments of this type of data within each preset time period before and after the anomaly point will be uploaded as the target data. For video surveillance data, if the detection result is no abnormality found, no data will be uploaded; otherwise, the valid data corresponding to the video surveillance data will be compressed and used as the target data for upload.

[0014] By configuring the uploaded content differently according to the hierarchical detection results of various fire protection data, the complete original segment of the abnormal period is uploaded when an anomaly occurs, and the video without an anomaly is directly discarded and not uploaded. The abnormal video is then compressed before transmission, which can significantly reduce the amount of data transmitted under normal conditions, reduce bandwidth consumption, and shorten the cloud upload time. At the same time, the complete original information is preserved in abnormal scenarios, which improves the efficiency of high-bandwidth data transmission while ensuring the ability to trace fire hazards.

[0015] Optionally, if the theoretical upload duration exceeds a preset time threshold, sending a data offload request from the edge computing node to neighboring edge computing nodes includes: The unloading request includes information on the amount of data to be unloaded and the data priority; upon receiving the unloading request, the neighboring edge computing node executes the following response steps: Step 1: Neighboring edge computing nodes obtain their current available upload bandwidth and the current length of the data queue to be uploaded; Step 2: The neighboring edge computing nodes calculate the maximum amount of data they can handle based on the current available upload bandwidth and the remaining time before the preset time threshold expires; Step 3: If the upper limit of the data volume is greater than or equal to the total amount of data to be unloaded, the neighboring edge computing node will receive the upload task of the data to be unloaded in its entirety and return an acknowledgment response to the source edge computing node. Step 4: If the upper limit of the data volume is less than the total amount of data to be unloaded, the adjacent edge computing nodes will accept the upload task of the unloaded data in descending order of data priority until the amount of data that can be accepted reaches the upper limit of the data volume. Step 5: Based on the acceptance results returned by the neighboring edge computing nodes, the source edge computing node removes the accepted data from its upload queue and queues the remaining unaccepted data for upload according to priority. Step 6: If the sum of the available capacity of all edge computing nodes adjacent to the source edge computing node is still less than the total amount of data to be unloaded, the source edge computing node forwards the unloading request to a more distant edge computing node until all data to be uploaded is allocated or the preset maximum forwarding hop count is reached.

[0016] When the upload latency of a single node exceeds the limit, an offload request with data volume and priority is initiated. Neighboring nodes calculate the upper limit of data they can carry based on their own bandwidth and remaining transmission time. If the capacity is sufficient, the entire data is shared. If the capacity is insufficient, critical data is prioritized according to the fire situation. The source node adjusts the upload queue synchronously. When the computing power and bandwidth of surrounding nodes are insufficient, the data is forwarded to the remote edge node level by level, and the maximum number of hops is limited. This enables multi-edge node collaborative traffic distribution, effectively distributing the bandwidth pressure of a single node, shortening the overall upload time, and prioritizing the timely uploading of high-risk fire data to the cloud. It also takes into account the real-time transmission and fire response priority, thereby solving the problems of inefficient, congested, and delayed high-bandwidth fire data transmission.

[0017] In a second aspect of this invention, a fire information transmission system based on edge computing is proposed, comprising: The acquisition module is used for each edge node to acquire the initial fire protection data uploaded to the corresponding target area; the initial fire protection data is the data collected by each fire protection device in the target area during the acquisition cycle; The data cleaning module is used to clean the initial fire data to obtain effective fire data; An anomaly detection module is used to perform anomaly detection on the valid fire protection data to obtain the detection results for each type of fire protection data; The target upload data determination module is used to determine the target upload data to the cloud based on the detection results of each type of fire protection data. The theoretical upload time determination module is used to count the total amount of all target upload data and calculate the theoretical upload time required for the edge node to upload all the target upload data to the cloud server based on the current available upload bandwidth of the edge node. The data offloading module is used to send a data offloading request from the edge computing node to the neighboring edge computing node to upload fire information if the theoretical upload duration exceeds a preset time threshold.

[0018] The beneficial effects of this invention are: By implementing differentiated anomaly detection and uploading strategies for different types of fire data at the edge, massive amounts of raw data are classified and processed at the source. Normal data is uploaded only as a statistical feature summary, while abnormal data is uploaded as raw fragments. This significantly reduces the amount of data uploaded under normal circumstances, effectively alleviating uplink bandwidth pressure. When the theoretical upload time of a certain edge node exceeds a preset time threshold, an offload request is sent to neighboring edge nodes, dynamically transferring some upload tasks to neighboring edge nodes with lower loads. This enables coordinated scheduling of edge layer bandwidth resources when a sudden fire causes a large amount of abnormal data to be reported centrally, avoiding queuing and congestion of critical data due to insufficient bandwidth at a single point, and improving the efficiency of fire information transmission. Attached Figure Description

[0019] The invention will now be further described with reference to the accompanying drawings.

[0020] Figure 1 A flowchart illustrating a fire information transmission method based on edge computing provided in an embodiment of the present invention; Figure 2 This is a flowchart illustrating how a computing node sends a data offload request, as provided in an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0022] It should be noted that all formula calculations in the scheme are purely numerical calculations.

[0023] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] This invention provides a fire information transmission method based on edge computing. See also... Figure 1 The method includes the following steps: S101, Each edge node acquires the initial fire protection data uploaded to the corresponding target area; the initial fire protection data is the data collected by each fire protection device in the target area during the collection period; S102, Data cleaning is performed on the initial fire protection data to obtain effective fire protection data; S103, perform anomaly detection on valid fire protection data to obtain the detection results for each type of fire protection data; S104, for each type of fire protection data, determine the target data to be uploaded to the cloud based on the test results; S105, count the total amount of data uploaded by all targets, and calculate the theoretical upload time required for the edge node to upload all the target data to the cloud server based on the current available upload bandwidth of the edge node; S106, if the theoretical upload time exceeds the preset time threshold, the edge computing node sends a data offload request to the neighboring edge computing node to upload fire information.

[0025] In one implementation, the initial fire protection data includes: fire water source level data, water supply network pressure data, fire pump operating current and voltage data, fault arc high-frequency waveform data, smoke concentration data, ambient temperature data, combustible gas concentration data, and video surveillance data.

[0026] In one implementation, the high-rise building is divided into multiple target areas according to floors or fire compartments. Each target area is equipped with an edge computing node. The edge computing node is an industrial computer based on the ARM Cortex-A series architecture, with a quad-core processor and a computing power of 6.0 TOPS or more. A single node can support the connection of no less than 200 fire-fighting devices. Each fire-fighting device is connected to the corresponding edge computing node through heterogeneous communication networks such as RS-485 bus, CAN bus, LoRaWAN, NB-IoT or 4G / 5G.

[0027] In one implementation, the fire-fighting equipment includes a level sensor, a pressure transmitter, a current transformer, a voltage sensor, a fault arc detector, a smoke detector, a temperature sensor, a combustible gas detector, and video surveillance equipment; all of the fire-fighting equipment are common and well-known devices on the market.

[0028] In one implementation, each fire-fighting device continuously collects data according to a preset collection cycle. The collection cycle is divided into two modes: periodic collection and event-triggered collection. Periodic collection reports equipment status data at preset time intervals. Event-triggered collection means that when the collected fire-fighting data exceeds a preset threshold or limit, the sensor immediately generates an alarm data packet and uploads it. Here, the collection cycle, preset threshold, and preset limit are all determined by technical personnel.

[0029] In one implementation, each fire-fighting device uploads the initial fire-fighting data it collects to the edge computing node of the corresponding target area through the aforementioned heterogeneous communication network. The edge computing node performs protocol parsing and unified format conversion on the data of different communication protocols, converting Modbus RTU, LoRaWAN binary protocol, etc. into MQTT protocol format, and storing the converted data in the local cache.

[0030] In one implementation, the theoretical upload time required for the edge node to upload all the target data to the cloud server is calculated based on the current available upload bandwidth of the edge node. Specifically, the total amount of target data to be uploaded is determined, multiplied by 8, and then divided by the current available upload bandwidth to obtain the theoretical upload time. If the theoretical upload time is less than or equal to a preset time threshold, the upload is performed directly.

[0031] In one embodiment, data cleaning of initial fire data to obtain effective fire data includes: For the fire water source level data, water supply network pressure data, fire pump operating current and voltage data, smoke concentration data, ambient temperature data, and combustible gas concentration data, amplitude limiting filtering, median filtering, and sliding window mean filtering are performed sequentially to denoise them and obtain the corresponding denoised data; For high-frequency waveform data of fault arc, the high-frequency waveform data of fault arc is decomposed into multiple scales by db4 wavelet, and the corresponding denoised data is obtained by noise reduction processing by wavelet modulus maxima theory. Based on smoke concentration data and ambient temperature data, video frames are extracted from video surveillance data to obtain corresponding valid data. The processing results of all data categories are obtained to obtain valid fire protection data.

[0032] In one implementation, amplitude limiting filtering, median filtering, and sliding window mean filtering are all existing technologies and will not be described in detail here. For high-frequency sampling data generated by fault arc detectors, the data form is completely different from low-frequency slowly varying data. The signal contains rich time-domain abrupt change features and frequency-domain harmonic components. Simply using the above filtering methods will lead to the loss of a large amount of feature information. Therefore, the edge computing node adopts a noise reduction scheme based on db4 wavelet transform for processing.

[0033] In one implementation, video surveillance data constitutes the largest proportion of the eight types of fire-fighting data. Full acquisition and processing of all video streams would severely consume the storage and computing resources of edge computing nodes. Therefore, instead of using fixed-frequency frame extraction or full storage, selective extraction is performed based on real-time fluctuations in smoke concentration and ambient temperature data as driving signals.

[0034] In one embodiment, the effective data obtained by extracting video frames from video surveillance data based on smoke concentration data and ambient temperature data includes: If the rate of increase of the smoke concentration data within a preset time exceeds a preset limit, then video frames of the smoke concentration data before and after the abnormal point are obtained as the first valid video data. If the rate of change of temperature data within a preset time exceeds a preset limit, then video frames of the temperature data before and after the abnormal point are obtained as the second valid video data. The valid data corresponding to the video surveillance data is obtained by finding the intersection of the first and second valid video data.

[0035] In one implementation, the edge computing node allocates a circular video buffer to each video surveillance device. This buffer continuously stores the monitoring video stream data of the most recent 5 minutes according to the first-in-first-out principle, ensuring that when a video extraction request is triggered at any time, historical video frames before and after the anomaly can be retrieved from the buffer. Simultaneously, the edge computing node continuously receives smoke concentration data from smoke detectors and ambient temperature data from temperature sensors, and performs short-term fluctuation monitoring on both types of data. Short-term fluctuation monitoring and the denoising process in step S102 are executed in parallel: denoising optimizes the quality of the collected data, while short-term fluctuation monitoring performs dynamic analysis of the real-time data stream. Both use data that has undergone preliminary verification through amplitude limiting filtering as input. Specifically, amplitude limiting filtering, as the first stage of data cleaning, discards data exceeding the physically reasonable range, while data within the reasonable range is simultaneously output to two parallel branches—one branch enters median filtering and mean filtering for deep denoising (for subsequent anomaly detection), and the other branch enters short-term fluctuation monitoring.

[0036] In one implementation, the edge computing node performs short-time rate of change calculation on the smoke concentration data stream after amplitude limiting and filtering verification. Specifically, the edge computing node maintains a sliding time window of length T1 seconds, with the window tracing back T1 seconds from the current moment. The edge computing node calculates the difference between the smoke concentration value at the current moment and the smoke concentration value at the beginning of the window, dividing by T1 to obtain the rate of increase. The edge computing node compares the calculated rate of increase with a preset rate of increase limit. If the rate of increase is greater than the preset rate of increase limit, it indicates that the smoke particle concentration in the environment is rising rapidly, potentially indicating a combustion or smoldering event. At this point, the edge computing node immediately records the current moment as the smoke concentration anomaly point. Subsequently, using the smoke concentration anomaly point as the center, the edge computing node extracts video frames from the loop video buffer of the video surveillance equipment in the corresponding target area for each T2 seconds (preset seconds) before and after it. All extracted video frames are used as the first valid video data, and their timestamps are recorded. Both T1 and T2 seconds are determined by technical personnel.

[0037] In one implementation, the edge computing node monitors short-term changes in the ambient temperature data stream after amplitude limiting and filtering. The monitoring logic includes two modes: rate-of-change triggering and absolute value triggering. Specifically, the rate-of-change triggering method involves the edge computing node maintaining a sliding time window of length T2 seconds. The difference between the current temperature value and the temperature value at the beginning of the window is calculated and divided by T3 to obtain the rate of change. This rate of change is compared with a preset rate-of-change limit. If the rate of change is greater than the preset limit, it indicates that the ambient temperature is rapidly and abnormally rising, potentially indicating open flame combustion. The absolute value triggering method involves the current temperature value being greater than a preset absolute temperature threshold, indicating that the ambient temperature has reached a dangerous range. When either of these conditions is met, the edge computing node immediately records the current moment as the temperature anomaly point, extracts video frames from the loop video buffer of the corresponding target area's video surveillance equipment for each preceding and following T2 seconds (preset seconds), uses all extracted video frames as the second valid video data, and records their timestamps.

[0038] In one implementation, when only the smoke condition is triggered and the temperature condition is not triggered (or vice versa), since one of the first and second video intervals is missing, the intersection is empty, and no valid video data is generated. Through this intersection operation mechanism, a single sensor false alarm will not drive invalid acquisition and invalid processing of video data. Only when the smoke concentration anomaly and the temperature anomaly are close in time, that is, when both sensors simultaneously point to an anomaly in the same time period, will the video data be extracted and used for subsequent analysis.

[0039] In one embodiment, the detection results for each type of fire data type obtained by anomaly detection of valid fire data include: The effective fire protection data, including fire water source level data, water supply network pressure data, fire pump operating current and voltage data, smoke concentration data, ambient temperature data, and combustible gas concentration data, are substituted into a dynamic threshold detection model based on historical window statistics to obtain the detection results for each type of fire protection data; the detection results include those that found anomalies and those that did not. For effective fire protection data based on high-frequency waveform data of fault arcs, extract the corresponding current change rate, wavelet high-frequency energy, and interharmonic factor; If the rate of change of current, wavelet high-frequency energy, and interharmonic factor are all greater than the corresponding thresholds, the detection result is recorded as an anomaly detected; otherwise, it is recorded as no anomaly detected. For the valid data corresponding to the video surveillance data, the valid data is substituted into the preset fire anomaly video detection model to obtain the detection result; the detection result includes the detection of anomalies and the absence of anomalies.

[0040] In one implementation, a sliding window containing the most recent N historical valid data points is maintained for each type of data. The mean μ and standard deviation σ of the data within the window are calculated, where N ranges from 10 to 100. When a newly collected data value x exceeds the range of μ ± k·σ, the data is determined to be abnormal. The degree of abnormality is calculated based on the deviation (|x-μ| / σ), and the larger the deviation, the higher the degree of abnormality. k is a preset coefficient with a value range of 2.5 to 3.5.

[0041] In one implementation, the thresholds corresponding to the current change rate, wavelet high-frequency energy, and interharmonic factor are all determined in advance by technicians; the preset fire anomaly video detection model is a lightweight model commonly used in the market and deployed on local edge nodes, such as the MobileNet series.

[0042] In one embodiment, the target data to be uploaded to the cloud for each type of fire-fighting data includes: For any data type among fire water source level data, water supply network pressure data, fire pump operating current and voltage data, smoke concentration data, ambient temperature data, combustible gas concentration data, and fault arc high-frequency waveform data, if the detection result for this data type is no abnormality, then the data characteristics of this data type within the detection period are extracted as the target uploaded data; the data characteristics include the mean, maximum, minimum, and rate of change within the detection period. If the detection result of this data type is that an anomaly is found, then the original data segments of this type of data within each preset time period before and after the anomaly point will be uploaded as the target data. For video surveillance data, if the detection result is no abnormality found, no data will be uploaded; otherwise, the valid data corresponding to the video surveillance data will be compressed and used as the target data for upload.

[0043] In one implementation, the method for determining the time of anomalies is the same for all types of data, referencing smoke concentration data and temperature data to determine the time of anomalies.

[0044] In one embodiment, see Figure 2 A flowchart for a computing node sending a data offload request is provided, including: S1061, The neighboring edge computing node obtains its own current available upload bandwidth and the current length of the data queue to be uploaded; S1062, The adjacent edge computing node calculates the maximum amount of data it can handle based on the current available upload bandwidth and the remaining time before the preset time threshold expires; S1063, If the upper limit of the data volume is greater than or equal to the total amount of data to be unloaded, the neighboring edge computing node will receive the upload task of the data to be unloaded in full and return an acknowledgment response to the source edge computing node. S1064, If the upper limit of the data volume is less than the total amount of data to be unloaded, the adjacent edge computing nodes will accept the upload task of the unloaded data in descending order of data priority until the amount of data that can be accepted reaches the upper limit of the data volume. S1065, the source edge computing node removes the data that has been accepted from its upload queue based on the acceptance results returned by the neighboring edge computing nodes, and queues the remaining unaccepted data for upload in this node according to priority. S1066 If the sum of the available capacity of all edge computing nodes adjacent to the source edge computing node is still less than the total amount of data to be unloaded, the source edge computing node will forward the unloading request to the edge computing node further away until all the data to be uploaded is allocated or the preset maximum forwarding hop count is reached.

[0045] In one implementation, the unloading request includes unloading data volume information and data priority; the unloading request includes unloading data volume information, data priority information, and time delay constraint information; the unloading data volume information includes the total amount of data to be unloaded and the data volume corresponding to each data type; the data priority information is divided into the following order from high to low according to the urgency of the abnormal data types: electrical fire-related data, including high-frequency waveform data of fault arcs and their corresponding video monitoring data, is of the first priority; abnormal smoke concentration data and abnormal temperature data are of the second priority; abnormal water level data and abnormal pressure data are of the third priority; normal operation status data is of the lowest priority; the time delay constraint information is the latest reporting time limit for each type of data; the latest reporting time limit is set according to the urgency of the fire data, with the shortest latest reporting time limit for the first priority data.

[0046] In one implementation, the upper limit of the amount of data that can be handled is calculated by multiplying the current available upload bandwidth by the remaining time before the preset time threshold expires, dividing by 8, and then subtracting the length of the current data queue to be uploaded; the preset time threshold and the preset maximum number of forwarding hops are determined by technical personnel.

[0047] In one implementation, in traditional fire information transmission schemes, each edge node independently undertakes the data upload task for its own area. When a node faces insufficient uplink bandwidth due to a large amount of abnormal data generated by a sudden fire, its surrounding nodes may be in a low-load idle state. However, this bandwidth resource cannot be shared and utilized, resulting in an uneven bandwidth utilization rate of the overall edge layer. By establishing a collaborative offloading mechanism among edge nodes, when the theoretical upload time of the source node exceeds a preset time threshold, it actively requests assistance from neighboring nodes to transfer part of the data upload task to neighboring nodes with lower loads. This allows the previously idle neighboring bandwidth resources to be effectively utilized, achieving real-time dynamic load balancing of the overall bandwidth resources of the edge layer.

[0048] The foregoing has described one embodiment of the present invention in detail, but this content is merely a preferred embodiment and should not be considered as limiting the scope of the present invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the scope of the claims of this invention.

Claims

1. A fire information transmission method based on edge computing, characterized in that, The method includes: Each edge node acquires the initial fire protection data uploaded to the corresponding target area; the initial fire protection data is the data collected by each fire protection device in the target area during the collection period; The initial fire data is cleaned to obtain effective fire data; Anomaly detection is performed on the valid fire protection data to obtain the detection results for each type of fire protection data; For each type of fire-fighting data type, determine the target data to be uploaded to the cloud based on the test results; The total amount of data uploaded by all targets is counted, and the theoretical upload time required for the edge node to upload all the target data to the cloud server is calculated based on the current available upload bandwidth of the edge node. If the theoretical upload duration exceeds the preset time threshold, the edge computing node sends a data offload request to the neighboring edge computing node to upload fire information.

2. The fire information transmission method based on edge computing according to claim 1, characterized in that, The initial fire protection data includes: fire water source level data, water supply network pressure data, fire pump operating current and voltage data, fault arc high-frequency waveform data, smoke concentration data, ambient temperature data, combustible gas concentration data, and video surveillance data.

3. The fire information transmission method based on edge computing according to claim 2, characterized in that, The initial fire data was cleaned to obtain valid fire data, including: The fire water source level data, water supply network pressure data, fire pump operating current and voltage data, smoke concentration data, ambient temperature data, and combustible gas concentration data are sequentially subjected to amplitude limiting filtering, median filtering, and sliding window mean filtering to denoise them and obtain the corresponding denoised data. For high-frequency waveform data of fault arc, the high-frequency waveform data of fault arc is decomposed into multiple scales by db4 wavelet, and the corresponding denoised data is obtained by noise reduction processing by wavelet modulus maxima theory. Based on smoke concentration data and ambient temperature data, the video surveillance data is used to extract corresponding valid data by video frame extraction. The processing results of all data categories are obtained to obtain valid fire protection data.

4. The fire information transmission method based on edge computing according to claim 3, characterized in that, Based on smoke concentration data and ambient temperature data, the video surveillance data was used to extract corresponding valid data by video frame extraction, including: If the rate of increase of the smoke concentration data within a preset time exceeds a preset limit, then video frames of the smoke concentration data before and after the abnormal point are obtained as the first valid video data. If the rate of change of temperature data within a preset time exceeds a preset limit, then video frames of the temperature data before and after the abnormal point are obtained as the second valid video data. The valid data corresponding to the video surveillance data is obtained by finding the intersection of the first and second valid video data.

5. The fire information transmission method based on edge computing according to claim 2, characterized in that, The anomaly detection results for each type of fire data obtained by performing anomaly detection on the valid fire data include: The effective fire protection data corresponding to the fire water source level data, water supply network pressure data, fire pump operating current and voltage data, smoke concentration data, ambient temperature data, and combustible gas concentration data are substituted into a dynamic threshold detection model based on historical window statistics to obtain the detection results for each type of fire protection data; the detection results include those that found anomalies and those that did not. For the effective fire protection data of the high-frequency waveform data of the fault arc, the corresponding current change rate, wavelet high-frequency energy and interharmonic factor are extracted; If the rate of change of current, wavelet high-frequency energy, and interharmonic factor are all greater than the corresponding thresholds, the detection result is recorded as an anomaly detected; otherwise, it is recorded as no anomaly detected. For the valid data corresponding to the video surveillance data, the valid data is substituted into the preset fire anomaly video detection model to obtain the detection result; the detection result includes the detection of anomalies and the absence of anomalies.

6. The fire information transmission method based on edge computing according to claim 5, characterized in that, For each type of fire safety data, the target data to be uploaded to the cloud includes: For any data type among fire water source level data, water supply network pressure data, fire pump operating current and voltage data, smoke concentration data, ambient temperature data, combustible gas concentration data, and fault arc high-frequency waveform data, if the detection result for this data type is no abnormality, then the data characteristics of this data type within the detection period are extracted as the target uploaded data; the data characteristics include the mean, maximum, minimum, and rate of change within the detection period. If the detection result of this data type is that an anomaly is found, then the original data segments of this type of data within each preset time period before and after the anomaly point will be uploaded as the target data. For video surveillance data, if the detection result is no abnormality found, no data will be uploaded; otherwise, the valid data corresponding to the video surveillance data will be compressed and used as the target data for upload.

7. The fire information transmission method based on edge computing according to claim 2, characterized in that, If the theoretical upload duration exceeds a preset time threshold, then sending a data offload request from the edge computing node to neighboring edge computing nodes includes: The unloading request includes information on the amount of data to be unloaded and the data priority; upon receiving the unloading request, the neighboring edge computing node executes the following response steps: Step 1: Neighboring edge computing nodes obtain their current available upload bandwidth and the current length of the data queue to be uploaded; Step 2: The neighboring edge computing nodes calculate the maximum amount of data they can handle based on the current available upload bandwidth and the remaining time before the preset time threshold expires; Step 3: If the upper limit of the data volume is greater than or equal to the total amount of data to be unloaded, the neighboring edge computing node will receive the upload task of the data to be unloaded in its entirety and return an acknowledgment response to the source edge computing node. Step 4: If the upper limit of the data volume is less than the total amount of data to be unloaded, the adjacent edge computing nodes will accept the upload task of the unloaded data in descending order of data priority until the amount of data that can be accepted reaches the upper limit of the data volume. Step 5: Based on the acceptance results returned by the neighboring edge computing nodes, the source edge computing node removes the accepted data from its upload queue and queues the remaining unaccepted data for upload according to priority. Step 6: If the sum of the available capacity of all edge computing nodes adjacent to the source edge computing node is still less than the total amount of data to be unloaded, the source edge computing node forwards the unloading request to a more distant edge computing node until all data to be uploaded is allocated or the preset maximum forwarding hop count is reached.

8. A fire information transmission system based on edge computing, characterized in that, The system includes: The acquisition module is used for each edge node to acquire the initial fire protection data uploaded to the corresponding target area; the initial fire protection data is the data collected by each fire protection device in the target area during the acquisition cycle; The data cleaning module is used to clean the initial fire data to obtain effective fire data; An anomaly detection module is used to perform anomaly detection on the valid fire protection data to obtain the detection results for each type of fire protection data; The target upload data determination module is used to determine the target upload data to the cloud based on the detection results of each type of fire protection data. The theoretical upload time determination module is used to count the total amount of all target upload data and calculate the theoretical upload time required for the edge node to upload all the target upload data to the cloud server based on the current available upload bandwidth of the edge node. The data offloading module is used to send a data offloading request from the edge computing node to the neighboring edge computing node to upload fire information if the theoretical upload duration exceeds a preset time threshold.

Images