Fire early warning system based on internet of things and edge computing

Abstract

This invention discloses a fire early warning system based on the Internet of Things (IoT) and edge computing, relating to the field of fire early warning. It primarily addresses the shortcomings of sensors in the early stages of a fire, where crystal oscillator frequency drift and local clock rate changes can negatively impact early warning effectiveness due to temperature rise. This invention utilizes IoT coverage to construct distributed edge computing nodes, using sensors at various locations as edge nodes for data acquisition and preprocessing. It identifies and eliminates clock interference by using timestamp micro-jitter, verifies early warning accuracy and avoids false alarms through sudden anomaly detection, and improves system efficiency after prevention and control through recovery phase detection. Ultimately, it achieves early warning, precise prevention and control, and source tracing assistance for fires, solving problems such as delayed fire warnings, high false alarm rates, weak interference resistance, and slow system recovery after prevention and control, thereby improving the intelligence and precision of fire early warning and prevention.

Description

Technical Field

[0001] This invention relates to the field of fire early warning technology, specifically a fire early warning system based on the Internet of Things and edge computing. Background Technology

[0002] With the rapid development of IoT and edge computing technologies, fire early warning systems are gradually upgrading from traditional single-point monitoring to distributed and intelligent systems. However, the industry still faces many technical bottlenecks. Conventional fire early warning systems rely on a single sensor to collect data and adopt a centralized computing model, which has problems such as high data transmission latency and uneven load on edge nodes, making it difficult to achieve early warning of fires.

[0003] Meanwhile, in the early stages of a fire, when the temperature rises, the sensor is prone to crystal frequency drift and local clock rate changes, which can cause timestamp jitter. This can lead to inaccurate alignment of data collected by different edge nodes, resulting in false alarms or missed alarms. Furthermore, the fire early warning decision-making lacks an effective secondary verification mechanism. Relying solely on a single data threshold makes it susceptible to environmental interference, leading to a high false alarm rate. After fire prevention and control are completed, the system lacks a targeted recovery and compensation mechanism, making it difficult to quickly restore the efficiency of data collection and analysis at edge nodes, which affects the continuity and reliability of subsequent monitoring.

[0004] To address the aforementioned technical shortcomings, a fire early warning system based on the Internet of Things and edge computing is proposed. A dedicated identification and processing unit is designed to address the key interference factor of timestamp micro-jitter, eliminating the impact of clock deviation on early warning decisions. Furthermore, after the early warning decision is made, a secondary verification method based on edge node load changes is provided to facilitate the distinction between valid early warnings and false alarms. Summary of the Invention

[0005] The purpose of this invention is to solve the problems mentioned above by proposing a fire early warning system based on the Internet of Things and edge computing.

[0006] The objective of this invention can be achieved through the following technical solution: a fire early warning system based on the Internet of Things and edge computing, including a fire early warning platform, wherein the fire early warning platform has communication connections with: The timestamp micro-jitter recognition unit performs sensor timestamp micro-jitter recognition and detection in the fire monitoring area, and performs data acquisition compensation based on the jitter recognition and detection. The sudden anomaly detection unit analyzes the sensor distribution network based on the current early warning decision to determine the accuracy of the early warning decision; The recovery phase testing unit conducts recovery phase testing after fire prevention and control are completed.

[0007] Furthermore, the timestamp micro-jitter recognition and detection process in the timestamp micro-jitter recognition unit is as follows: Sensors are distributed according to the fire monitoring area, and each sensor is set as an edge computing node through Internet of Things coverage. Sensor data is collected from each location in the area and the collected data is processed to provide regional fire early warning. The determined environmental parameters are compared with the current environmental parameters using similar data. If the corresponding numerical deviation under the comparison is lower than the set numerical deviation threshold, the current stage is marked as a steady-state safe stage; otherwise, if the corresponding numerical deviation under the comparison is not lower than the set numerical deviation threshold, the current stage is marked as a steady-state risk stage. Based on the steady-state security stage, the difference distribution between the timestamp of the same sensor and the receiving time of the edge gateway is extracted when each sensor collects data, and the normal deviation range of the timestamp of each sensor is determined based on the difference distribution.

[0008] Furthermore, micro-jitter identification and detection are performed during the steady-state risk phase. The drift of the crystal oscillator frequency of each sensor under the temperature rise environment and the range of change of the local clock rate of the current sensor are obtained within the steady-state risk phase. If the drift of the crystal oscillator frequency of each sensor under the temperature rise environment exceeds the drift threshold, or the range of change of the local clock rate of the current sensor exceeds the range of rate change threshold, then the current sensor is marked as a timestamp jitter object. If the drift of the crystal oscillator frequency of each sensor under the temperature rise environment exceeds the drift threshold, or the range of change of the local clock rate of the current sensor exceeds the range of rate change threshold, then the current sensor is marked as a timestamp non-jitter object.

[0009] Furthermore, type matching and identification are performed on the entire sensor distribution network to obtain two adjacent timestamp jitter objects, and the time misalignment deviation range of the two timestamp jitter objects under the corresponding temperature difference distribution is recorded. At the same time, the temperature gradient of the adjacent edge computing nodes within the time misalignment deviation range is extracted and marked as jitter temperature gradient. Get two adjacent timestamp non-jitter objects and record the time misalignment deviation range of the two timestamp jitter objects under the corresponding temperature difference distribution. At the same time, extract the temperature gradient of the adjacent edge computing nodes within the time misalignment deviation range and mark it as the steady-state temperature gradient. The jitter temperature gradient and steady-state temperature gradient are merged to extract the temperature gradient deviation. At the same time, the spatial distance between the timestamp jitter object and the timestamp non-jitter object is identified, and the temperature deviation value under the corresponding spatial distance is determined based on the temperature gradient deviation. If the current spatial distance is lower than the set distance threshold and the temperature deviation exceeds the deviation threshold, a timestamp anomaly signal is generated and sent to the fire early warning platform. If the current spatial distance is not lower than the set distance threshold or the temperature deviation does not exceed the deviation threshold, a timestamp offset signal is generated and sent to the fire early warning platform.

[0010] Furthermore, after receiving the timestamp anomaly signal and timestamp offset signal, the fire early warning platform, based on the data collected at each time point during the continuous monitoring phase, and based on the data sequence, derives the mapping relationship between data collection and delay at each stage when the corresponding timestamp is fluctuating. Based on the continuous update of the mapping relationship, it obtains the delay trend under the abnormal state and the delay trend under the offset state. Combined with the data collection data analysis of multiple types of sensors, it makes alarm decisions for edge computing nodes and performs sensor compensation simultaneously.

[0011] Furthermore, the sensor distribution network analysis process in the sudden anomaly detection unit is as follows: Based on the early warning decision, the area where the edge computing node of the early warning is located is obtained and marked as the early warning area. Conversely, the remaining areas within the fire monitoring area are marked as non-early warning areas. Based on the location of the early warning area, the boundary points and non-boundary points within the non-early warning area are selected, that is, the points close to the early warning area and the points in the opposite direction. The real-time CPU utilization, chip temperature, and frequency reduction status of edge nodes are determined as node parameters. The deviation of node parameters before and after the warning decision time of edge computing nodes in the warning area is obtained, and the fluctuation trend of node parameter deviation is obtained as the time increases. If the fluctuation range of the node parameter deviation trend before and after the warning decision time does not exceed the set fluctuation range threshold, a warning signal maintenance instruction is generated and sent to the fire early warning platform. If the load of the corresponding edge computing node does not increase at the current warning decision time, a warning signal review instruction is generated and sent to the fire early warning platform.

[0012] Furthermore, the fire early warning platform performs load analysis on all edge computing nodes within the corresponding warning area. If the area where the load of the edge node is increasing does not continue to increase, the area where the load of the edge computing node is increasing is set as a fire warning point and fire warning point check and control is performed. If the area where the load of the edge node is increasing continues to increase, the area where the load of the edge computing node is increasing is set as a fire source point and fire prevention and control are performed. If the fluctuation span of the node parameter deviation trend before and after the warning decision exceeds the set fluctuation span threshold range, a warning decision confirmation signal is generated and sent to the fire early warning platform, and fire prevention and control are executed.

[0013] Furthermore, the recovery phase detection process in the recovery phase detection unit is as follows: After the control measures are completed, smoke concentration is used as the data type for edge computing nodes to collect smoke concentration data in the warning area. The warning area is divided into several sub-areas based on the data collection range of the edge computing nodes. Continuous data collection is carried out on the smoke concentration before, during, and after the control measures are implemented in the sub-areas to construct a smoke concentration curve. The smoke concentration values ​​corresponding to the control time points are marked and set as action inflection points. Trend change inflection points are extracted based on the trend of the smoke concentration curve.

[0014] Furthermore, after an action inflection point is generated, the interval duration of the trend reversal inflection point is obtained, and the curve descent rate after the trend reversal inflection point is also obtained. If the interval duration of the trend reversal inflection point is lower than the interval duration threshold, or the curve descent rate after the trend reversal inflection point exceeds the curve descent rate threshold, then the current action inflection point is marked as a valid inflection point and set as the prevention and control starting point. If the interval duration of the trend reversal inflection point is not lower than the interval duration threshold, and the curve descent rate after the trend reversal inflection point does not exceed the curve descent rate threshold, then the current action inflection point is marked as an invalid inflection point, and an invalid prevention and control signal is generated and sent to the fire early warning platform.

[0015] Furthermore, after determining the starting point of prevention and control, the smoke concentration curve is identified by nodes. If the current node of the smoke concentration curve shows an inflection point of rising concentration, a false extinguishing signal is generated and sent to the fire early warning platform. The fire early warning system collects smoke concentrations before and after the starting point of prevention and control synchronously and establishes a continuous data sequence to facilitate source tracing during secondary prevention and control. If the current node of the smoke concentration curve does not show an inflection point of rising concentration and the smoke concentration is lower than the set fire red line value, a fire cancellation signal is generated and sent to the fire early warning platform.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This method accurately identifies micro-jitter in sensor timestamps under the initial temperature rise conditions of a fire, eliminating the impact of clock interference on fire early warning and improving the accuracy of early warning decisions. Specifically, by detecting and analyzing the micro-jitter in sensor timestamps, it distinguishes between jittered and non-jittered objects, and combines temperature gradient deviation and spatial distance to determine the substantial impact of jitter interference. It generates timestamp anomaly and offset signals, providing accurate interference feedback to the early warning platform, avoiding data alignment deviations caused by timestamp jitter, and thus reducing false alarms and missed alarms. At the same time, by determining the normal deviation range of sensor timestamps, it provides a benchmark for data synchronization of distributed edge computing nodes, improving the synchronization and reliability of data acquisition, and providing accurate data support for early fire warning. By analyzing the parameter deviations and fluctuation trends of edge nodes in the warning and non-warning areas, the accuracy of warning decisions can be accurately judged, and effective warnings, warnings requiring review, and invalid warnings can be distinguished. This avoids false alarms caused by environmental interference and sensor deviations, reducing unnecessary waste of prevention and control resources. At the same time, by judging whether the area of ​​increased load on edge nodes continues to expand, the fire source and warning point can be accurately located, providing clear targets for fire prevention and control and improving prevention and control efficiency. In addition, this unit uses the real-time operating parameters of edge nodes as the basis for judgment, eliminating the need for additional monitoring equipment, reducing the hardware cost of the system, and achieving synergy between warning verification and node load monitoring, thereby improving the system's integration level. This system enables rapid recovery after fire prevention and control, improves the continuity and reliability of subsequent monitoring, and provides traceability support for secondary prevention and control. By constructing smoke concentration curves, it identifies effective prevention and control inflection points, accurately judges the fire extinguishing status, and generates timely fire clearance signals or false extinguishing signals to avoid untimely secondary prevention and control or waste of resources due to misjudgment. By restoring and compensating edge nodes, it rapidly improves the data collection and analysis efficiency of nodes, ensuring that the system can quickly return to normal monitoring status after prevention and control, and guaranteeing the continuity of monitoring. In addition, by establishing a continuous data sequence of smoke concentration before and after prevention and control, it provides traceability data for secondary prevention and control, facilitating the analysis of prevention and control effects, optimization of prevention and control strategies, and improvement of the system's intelligence level and subsequent prevention and control capabilities. Attached Figure Description

[0017] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0018] Figure 1 This is a system principle block diagram of the present invention; Figure 2 This is a flowchart of the method for the timestamp micro-jitter identification unit in this invention. Detailed Implementation

[0019] To enable those skilled in the art to better understand the present invention, 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. 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.

[0020] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0021] Please see Figures 1-2 As shown, the fire early warning system based on the Internet of Things and edge computing includes a fire early warning platform, which is connected to a timestamp micro-jitter recognition unit, a sudden anomaly detection unit, and a recovery phase detection unit. The fire early warning platform generates a timestamp micro-jitter identification signal and sends it to the timestamp micro-jitter identification unit. After receiving the timestamp micro-jitter identification signal, the timestamp micro-jitter identification unit performs sensor timestamp micro-jitter identification and detection on the fire monitoring area. Sensors are distributed according to the fire monitoring area, and each sensor is set as an edge computing node through IoT coverage. Sensor data is collected from each location in the area, and the collected data is processed to provide regional fire early warning. The distributed edge computing nodes rely on Network Time Protocol (NTP) or IEEE1588 to align the timestamps of the sensor data. The environmental parameters of steady-state time status, such as temperature peak or temperature rise span peak, are determined according to the sensor equipment specifications. The determined environmental parameters are compared with the current environmental parameters using similar data. If the corresponding numerical deviation under the comparison is lower than the set numerical deviation threshold, the current stage is marked as a steady-state safe stage; otherwise, if the corresponding numerical deviation under the comparison is not lower than the set numerical deviation threshold, the current stage is marked as a steady-state risk stage. Based on the steady-state security stage, extract the difference distribution between the timestamp of the same sensor and the receiving time of the edge gateway when each sensor collects data, and determine the normal deviation range of the timestamp of each sensor based on the difference distribution; Micro-jitter detection is performed during the steady-state risk phase. The drift of the crystal oscillator frequency of each sensor under the temperature rise environment and the range of change of the local clock rate of the current sensor are obtained within the steady-state risk phase. If the drift of the crystal oscillator frequency of each sensor under the temperature rise environment exceeds the drift threshold, or the range of change of the local clock rate of the current sensor exceeds the range of change of the rate threshold, the current sensor is marked as a timestamp jitter object. If the drift of the crystal oscillator frequency of each sensor under the temperature rise environment exceeds the drift threshold, or the range of change of the local clock rate of the current sensor exceeds the range of change of the rate threshold, the current sensor is marked as a timestamp non-jitter object. Type matching and identification are performed on the entire sensor distribution network to obtain two adjacent timestamp jitter objects, and the time misalignment deviation range of the two timestamp jitter objects under the corresponding temperature difference distribution is recorded. At the same time, the temperature gradient of the adjacent edge computing nodes within the time misalignment deviation range is extracted and marked as jitter temperature gradient. Get two adjacent timestamp non-jitter objects and record the time misalignment deviation range of the two timestamp jitter objects under the corresponding temperature difference distribution. At the same time, extract the temperature gradient of the adjacent edge computing nodes within the time misalignment deviation range and mark it as the steady-state temperature gradient. The jitter temperature gradient and steady-state temperature gradient are merged to extract the temperature gradient deviation. At the same time, the spatial distance between the timestamp jitter object and the timestamp non-jitter object is identified, and the temperature deviation value under the corresponding spatial distance is determined based on the temperature gradient deviation. If the current spatial distance is lower than the set distance threshold and the temperature deviation exceeds the deviation threshold, it is inferred that the current timestamp jitter interference exists and has an impact, and a timestamp anomaly signal is generated and sent to the fire early warning platform. If the current spatial distance is not lower than the set distance threshold or the temperature deviation value does not exceed the deviation threshold, it is inferred that the current timestamp jitter interference has not had an impact, and a timestamp offset signal is generated and sent to the fire early warning platform. After receiving timestamp anomaly and timestamp offset signals, the fire early warning platform uses the collected data at each time point during the continuous monitoring phase to derive the mapping relationship between data collection and delay at each stage when the corresponding timestamp fluctuates. For example, when the unit time is 1 minute, the mapping relationship between data collection and delay is 2 time points, and the collected data at the current edge gateway reception time is the data collected two minutes ago. Based on the continuous update of the mapping relationship, the delay trend under abnormal conditions and the delay trend under offset conditions are obtained. Combining the analysis of collected data from multiple types of sensors (such as temperature sensors and smoke sensors), the edge computing node makes alarm decisions and performs sensor compensation simultaneously. That is, based on the mapping relationship, the platform makes early warning decisions based on the current collected data. For example, if the collected data is data from two minutes ago, and the value of the corresponding data's distance from the red line value is compared with the value of the red line value over two minutes, and the value of the value's movement speed exceeds the recorded rate of change of the collected data, then an early warning is issued directly. The abnormal state and the offset state correspond to a deviation in the time range of delay within the mapping relationship, that is, the number of abnormal times in the abnormal state is much higher than the number of abnormal times in the offset state. In actual early warning decisions, a decision reference standard can be used.

[0022] What needs to be explained is: Numerical deviation threshold: used to determine whether the deviation between the current environmental parameters and the steady-state environmental parameters exceeds the safe range, and to distinguish between the steady-state safe stage and the steady-state risk stage.

[0023] Origin: Based on the normal operating environment parameter range specified in the sensor equipment specification manual, combined with historical environmental data of the fire monitoring area (such as historical temperature peaks and temperature rise range peaks), the influence of environmental fluctuations on the sensor is comprehensively considered to determine a reasonable deviation range and avoid misjudging slight environmental fluctuations as steady-state risks.

[0024] Acquisition method: By collecting historical environmental data of the fire monitoring area for the past 1-3 years, statistical analysis methods (such as mean and standard deviation calculation) are used to remove abnormal data and determine the normal fluctuation range of environmental parameters. The upper limit of this range is set as the numerical deviation threshold. At the same time, the threshold is corrected in combination with the parameter tolerances provided by the sensor equipment manufacturer to ensure the rationality and adaptability of the threshold.

[0025] Crystal frequency drift threshold: Used to determine whether the frequency drift of the sensor crystal oscillator exceeds the normal range, serving as one of the bases for marking timestamp jitter objects.

[0026] Origin: Based on the technical parameters of the sensor, such as the crystal oscillator material and operating temperature range, and combined with the crystal oscillator frequency drift pattern under the initial temperature rise environment of a fire, the maximum allowable drift of the crystal oscillator frequency is determined. Exceeding this range will lead to excessive timestamp deviation and affect data alignment.

[0027] Acquisition method: The temperature rise environment in the early stage of a fire is simulated in the laboratory (gradually rising from room temperature to the critical temperature in the early stage of a fire). Crystal oscillator frequency data of the sensor at different temperatures are collected, and the correspondence between frequency drift and temperature is recorded. The maximum drift that can ensure the accuracy of timestamp alignment is set as the threshold. At the same time, the drift standards of similar sensors in the industry are referenced for optimization and adjustment.

[0028] Clock rate variation threshold: Used to determine whether the local clock rate variation of the sensor exceeds the normal range, serving as one of the bases for marking timestamp jitter objects.

[0029] Origin: Based on the timestamp alignment accuracy requirements of the NTP or IEEE1588 protocol, and combined with the stability parameters of the sensor's local clock, the maximum allowable range of clock rate variation is determined. Exceeding this range will cause timestamp misalignment and affect the synchronization of data acquisition.

[0030] Acquisition method: By building a distributed edge computing node test environment, timestamp alignment tests are conducted using NTP or IEEE1588 protocols. Timestamp alignment errors under different clock rate change spans are recorded. The maximum rate change span of the alignment error within the allowable range is set as a threshold, and dynamic corrections are made in combination with network environment fluctuations in actual monitoring scenarios.

[0031] Spatial distance threshold: used to determine whether the spatial distance between the timestamp jittering object and the non-jittering object is too close, and to judge the impact of jittering interference in combination with the temperature deviation value.

[0032] Origin: Based on factors such as the monitoring range of sensors and the speed of fire spread, the effective monitoring distance between adjacent sensors is determined. When the spatial distance between the shaking object and the non-shaking object is lower than this threshold, the shaking interference may affect the monitoring data of adjacent nodes, thereby affecting the early warning decision.

[0033] Acquisition method: Combine the sensor distribution density and sensor monitoring radius in the fire monitoring area to calculate the average distance between adjacent sensors. Set half of the average distance as the initial threshold. By simulating fire scenarios, test the impact of jitter interference on monitoring data under different spatial distances, and optimize and adjust the threshold to ensure that the actual impact of jitter interference can be accurately judged.

[0034] Temperature deviation threshold: used to determine whether the deviation between the jitter temperature gradient and the steady-state temperature gradient exceeds the normal range, and to determine the impact of jitter interference in combination with spatial distance.

[0035] Origin: Based on the normal temperature gradient distribution in the fire monitoring area (under no fire and no interference), and combined with the changing pattern of the temperature gradient in the early stage of a fire, the normal deviation range of the temperature gradient is determined. If it exceeds this range, it indicates that the jitter interference has a substantial impact on temperature monitoring.

[0036] Acquisition method: Collect temperature gradient data in the fire monitoring area under no-fire and no-interference conditions, calculate the mean and standard deviation of temperature gradients in different areas, set twice the standard deviation as the temperature deviation threshold, and combine it with temperature gradient change data in the early stage of fire to make corrections to ensure that normal temperature fluctuations and temperature deviations caused by shaking interference can be distinguished.

[0037] Because the current fire early warning system may issue early warnings due to timestamp anomalies, the current warnings require secondary verification to avoid false alarms. After the fire early warning platform makes a warning decision, it generates a sudden anomaly detection signal and sends it to the sudden anomaly detection unit. After receiving the sudden anomaly detection signal, the sudden anomaly detection unit performs sensor distribution network analysis based on the current warning decision to determine the accuracy of the warning decision. The area where the edge computing node that issued the early warning is located is determined based on the early warning decision and marked as the early warning area; otherwise, the remaining areas within the fire monitoring area are marked as non-early warning areas. Based on the location of the warning area, select boundary points and non-boundary points within the non-warning area, that is, points close to the warning area and points in the opposite direction; The real-time CPU utilization, chip temperature, and frequency reduction status of edge nodes are determined as node parameters. The deviation of node parameters before and after the warning decision time of edge computing nodes in the warning area is obtained, and the fluctuation trend of node parameter deviation is obtained as the time increases. If the fluctuation range of the node parameter deviation trend before and after the warning decision time does not exceed the set fluctuation range threshold, it is inferred that the node parameter deviation before and after the corresponding time is low, that is, the load intensity of the edge computing node is close. If the load of the edge computing node corresponding to the current warning decision time continues to increase, a warning signal maintenance instruction is generated and sent to the fire early warning platform. If the load of the edge computing node corresponding to the current warning decision time does not increase, a warning signal review instruction is generated and sent to the fire early warning platform. The fire early warning platform performs load analysis on all edge computing nodes within the corresponding warning area. If the area where the load of the edge computing node is increasing does not continue to increase, the area where the load of the edge computing node is increasing will be set as a fire warning point and the fire warning point will be checked and controlled. If the area where the load of the edge computing node is increasing continues to increase, the area where the load of the edge computing node is increasing will be set as a fire source point and fire prevention and control will be carried out. If the fluctuation range of the node parameter deviation trend before and after the early warning decision exceeds the set fluctuation range threshold, it is inferred that the node parameter deviation at the corresponding time point is high, that is, there is a deviation in the load intensity of the edge computing node. An early warning decision determination signal is generated and sent to the fire early warning platform, and fire prevention and control are executed. It should be explained that the floating span threshold is used to determine whether the floating trend of the edge node parameter deviation before and after the early warning decision exceeds the normal range, and serves as a basis for judging the accuracy of the early warning decision.

[0038] Origin: Based on the normal operating load parameters of edge nodes (CPU utilization, chip temperature, frequency reduction status), combined with the load change pattern of edge nodes during fire early warning, the maximum allowable fluctuation range of parameter deviation is determined. If it exceeds this range, it indicates that there is a significant deviation in the node load intensity, and the early warning decision may be abnormal.

[0039] Acquisition method: Collect parameter data of edge nodes under normal working conditions (no early warning, no fire) and record the normal fluctuation range of parameters; at the same time, simulate fire early warning scenarios, collect node parameter data before and after the early warning time, calculate the fluctuation range of parameter deviation, and set the upper limit of the normal fluctuation range as the fluctuation range threshold; combine with the node hardware performance to make optimization adjustments.

[0040] After fire prevention and control is completed, the fire early warning platform generates a recovery phase detection signal and sends it to the recovery phase detection unit. After receiving the recovery phase detection signal, the recovery phase detection unit performs recovery phase detection after the fire prevention and control is completed, thereby improving the efficiency of data collection and analysis of edge computing nodes. After the control measures are completed, smoke concentration is used as the data type for edge computing nodes to collect smoke concentration data in the warning area. The warning area is divided into several sub-areas based on the data collection range of the edge computing nodes. Continuous data collection is carried out on the smoke concentration before, during, and after the control measures are implemented in the sub-areas to construct a smoke concentration curve. The smoke concentration values ​​corresponding to the control time points are marked and set as action inflection points. And based on the trend of the smoke concentration curve, the inflection point of trend change is extracted; After an action inflection point occurs, the interval duration of the trend change inflection point and the rate of decline of the curve after the trend change inflection point are obtained. If the interval duration of the trend change inflection point is lower than the interval duration threshold, or the rate of decline of the curve after the trend change inflection point exceeds the rate of decline of the curve threshold, the current action inflection point is marked as a valid inflection point and set as the prevention and control starting point. If the interval duration of the trend change inflection point is not lower than the interval duration threshold, and the rate of decline of the curve after the trend change inflection point does not exceed the rate of decline of the curve threshold, the current action inflection point is marked as an invalid inflection point, an invalid prevention and control signal is generated and sent to the fire early warning platform. After determining the starting point of prevention and control, the nodes of the smoke concentration curve are identified. When the current node of the smoke concentration curve shows an inflection point of concentration increase, a false extinguishing signal is generated and sent to the fire early warning platform. The fire early warning system collects smoke concentrations before and after the starting point of prevention and control synchronously and establishes a continuous data sequence to facilitate source tracing during secondary prevention and control. If the current smoke concentration curve node does not show an inflection point of concentration increase, and the smoke concentration is lower than the set fire red line value, then a fire cancellation signal will be generated and sent to the fire early warning platform.

[0041] What needs to be explained is: Interval duration threshold: Used to determine whether the interval between trend reversal inflection point and action inflection point is reasonable, and serves as one of the bases for marking valid inflection points.

[0042] Origin: Based on the effective time of fire prevention and control measures (such as the response time of fire extinguishing equipment and the decay time of smoke diffusion), combined with the changing pattern of smoke concentration curve, the minimum reasonable interval between the trend change inflection point and the action inflection point is determined. If the interval is less than this, it indicates that the prevention and control measures are effective quickly, and the action inflection point is an effective inflection point.

[0043] Acquisition method: By simulating different types of fire prevention and control scenarios, the change data of smoke concentration curve after the implementation of prevention and control measures are recorded. The interval between the inflection point of action and the inflection point of trend change is statistically analyzed. The minimum interval time under most effective prevention and control scenarios is set as the threshold and then corrected in combination with the performance parameters of actual prevention and control equipment.

[0044] Curve descent rate threshold: Used to determine whether the descent rate of the smoke concentration curve meets the requirements after the inflection point of trend change occurs, and serves as one of the bases for marking valid inflection points.

[0045] Origin: Based on the standard requirements for fire prevention and control (such as the requirement that smoke concentration must be reduced to a safe range within a specified time), and combined with the decay law of smoke concentration, the minimum rate of decrease of the smoke concentration curve is determined. Exceeding this rate indicates that the prevention and control measures are effective, and the inflection point of the action is the effective inflection point.

[0046] Acquisition method: Refer to relevant industry standards for fire prevention and control to determine the required safe rate of decrease of smoke concentration. By simulating prevention and control scenarios, collect data on the rate of decrease of smoke concentration curves and set the minimum rate of decrease that meets the standard requirements as the threshold. Adjustments are made dynamically based on the actual size of the monitoring area and ventilation conditions.

[0047] Fire red line value: used to determine whether the smoke concentration has dropped to a safe range, serving as the basis for generating a fire clearance signal.

[0048] Origin: Based on relevant national and industry standards for fire prevention and control (such as the "Code for Design of Automatic Fire Alarm Systems"), and combined with the degree of harm of smoke concentration to human body and environment, a safe threshold for smoke concentration is determined, namely the fire red line value. If it is lower than this value, it means that the fire has been effectively controlled and the warning can be lifted.

[0049] Acquisition method: Directly refer to the smoke concentration safety threshold specified in relevant national and industry standards; combine the usage scenario of the fire monitoring area (such as densely populated areas, flammable and explosive areas) to fine-tune the threshold to ensure its applicability.

[0050] In summary, constructing a closed-loop management system throughout the entire process enables fire early warning, interference identification, secondary verification, precise prevention and control, recovery compensation, and source tracing assistance to be completed in a closed loop. With each unit working collaboratively, the system's integration and intelligence levels are improved, providing comprehensive and end-to-end technical support for fire prevention and control, and promoting the upgrade of fire early warning systems towards distributed, intelligent, and precise systems.

[0051] These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A fire early warning system based on the Internet of Things and edge computing, characterized in that, This includes a fire early warning platform, whose communication connections include: The timestamp micro-jitter recognition unit performs sensor timestamp micro-jitter recognition and detection in the fire monitoring area, and performs data acquisition compensation based on the jitter recognition and detection. The sudden anomaly detection unit analyzes the sensor distribution network based on the current early warning decision to determine the accuracy of the early warning decision; The recovery phase testing unit conducts recovery phase testing after fire prevention and control are completed.

2. The fire early warning system based on the Internet of Things and edge computing according to claim 1, characterized in that, The timestamp micro-jitter detection process in the timestamp micro-jitter detection unit is as follows: Sensors are distributed according to the fire monitoring area, and each sensor is set as an edge computing node through Internet of Things coverage. Sensor data is collected from each location in the area and the collected data is processed to provide regional fire early warning. The determined environmental parameters are compared with the current environmental parameters using similar data. If the corresponding numerical deviation under the comparison is lower than the set numerical deviation threshold, the current stage is marked as a steady-state safe stage; otherwise, if the corresponding numerical deviation under the comparison is not lower than the set numerical deviation threshold, the current stage is marked as a steady-state risk stage. Based on the steady-state security stage, the difference distribution between the timestamp of the same sensor and the receiving time of the edge gateway is extracted when each sensor collects data, and the normal deviation range of the timestamp of each sensor is determined based on the difference distribution.

3. The fire early warning system based on the Internet of Things and edge computing according to claim 2, characterized in that, Micro-jitter identification and detection are performed during the steady-state risk stage. The drift of the crystal oscillator frequency of each sensor under the temperature rise environment and the range of change of the local clock rate of the current sensor are obtained within the steady-state risk stage. If the drift of the crystal oscillator frequency of each sensor under the temperature rise environment exceeds the drift threshold, or the range of change of the local clock rate of the current sensor exceeds the range of change threshold, then the current sensor is marked as a timestamp jitter object. If the frequency drift of the crystal oscillator of each sensor exceeds the drift threshold under the temperature rise environment during the steady-state risk stage, or if the current local clock rate change span of the sensor exceeds the rate change span threshold, then the current sensor will be marked as a timestamp non-jitter object.

4. The fire early warning system based on the Internet of Things and edge computing according to claim 3, characterized in that, Type matching and identification are performed on the entire sensor distribution network to obtain two adjacent timestamp jitter objects, and the time misalignment deviation range of the two timestamp jitter objects under the corresponding temperature difference distribution is recorded. At the same time, the temperature gradient of the adjacent edge computing nodes within the time misalignment deviation range is extracted and marked as jitter temperature gradient. Get two adjacent timestamp non-jitter objects and record the time misalignment deviation range of the two timestamp jitter objects under the corresponding temperature difference distribution. At the same time, extract the temperature gradient of the adjacent edge computing nodes within the time misalignment deviation range and mark it as the steady-state temperature gradient. The jitter temperature gradient and steady-state temperature gradient are merged to extract the temperature gradient deviation. At the same time, the spatial distance between the timestamp jitter object and the timestamp non-jitter object is identified, and the temperature deviation value under the corresponding spatial distance is determined based on the temperature gradient deviation. If the current spatial distance is lower than the set distance threshold and the temperature deviation exceeds the deviation threshold, a timestamp anomaly signal is generated and sent to the fire early warning platform. If the current spatial distance is not lower than the set distance threshold or the temperature deviation value does not exceed the deviation threshold, a timestamp offset signal is generated and sent to the fire early warning platform.

5. The fire early warning system based on the Internet of Things and edge computing according to claim 4, characterized in that, After receiving timestamp anomaly and timestamp offset signals, the fire early warning platform uses the collected data at each time point during the continuous monitoring phase and the data sequence to derive the mapping relationship between data collection and delay at each stage when the corresponding timestamp is fluctuating. Based on the continuous updating of the mapping relationship, it obtains the delay trend under abnormal conditions and the delay trend under offset conditions. Combined with the data collection data from multiple types of sensors, it makes alarm decisions for edge computing nodes and performs sensor compensation simultaneously.

6. The fire early warning system based on the Internet of Things and edge computing according to claim 1, characterized in that, The sensor distribution network analysis process in the sudden anomaly detection unit is as follows: Based on the early warning decision, the area where the edge computing node of the early warning is located is obtained and marked as the early warning area. Conversely, the remaining areas within the fire monitoring area are marked as non-early warning areas. Based on the location of the early warning area, the boundary points and non-boundary points within the non-early warning area are selected, that is, the points close to the early warning area and the points in the opposite direction. The real-time CPU utilization, chip temperature, and frequency reduction status of edge nodes are determined as node parameters. The deviation of node parameters before and after the warning decision time of edge computing nodes in the warning area is obtained, and the fluctuation trend of node parameter deviation is obtained as the time increases. If the fluctuation range of the node parameter deviation trend before and after the warning decision time does not exceed the set fluctuation range threshold, a warning signal maintenance instruction is generated and sent to the fire early warning platform. If the load of the corresponding edge computing node does not increase at the current warning decision time, a warning signal review instruction is generated and sent to the fire early warning platform.

7. The fire early warning system based on the Internet of Things and edge computing according to claim 6, characterized in that, The fire early warning platform performs load analysis on all edge computing nodes within the corresponding warning area. If the area where the load of an edge node is increasing does not continue to increase, the area where the load of the edge computing node is increasing is set as a fire warning point and fire warning point check and control are performed. If the area where the load of an edge node is increasing continues to increase, the area where the load of the edge computing node is increasing is set as a fire source point and fire prevention and control are performed. If the fluctuation span of the node parameter deviation trend before and after the warning decision exceeds the set fluctuation span threshold range, a warning decision confirmation signal is generated and sent to the fire early warning platform, and fire prevention and control are executed.

8. The fire early warning system based on the Internet of Things and edge computing according to claim 1, characterized in that, The recovery phase detection process in the recovery phase detection unit is as follows: After the control measures are completed, smoke concentration is used as the data type for edge computing nodes to collect smoke concentration data in the warning area. The warning area is divided into several sub-areas based on the data collection range of the edge computing nodes. Continuous data collection is carried out on the smoke concentration before, during, and after the control measures are implemented in the sub-areas to construct a smoke concentration curve. The smoke concentration values ​​corresponding to the control time points are marked and set as action inflection points. Trend change inflection points are extracted based on the trend of the smoke concentration curve.

9. The fire early warning system based on the Internet of Things and edge computing according to claim 8, characterized in that, After the action inflection point is generated, the interval time of the trend change inflection point is obtained, and the curve decline rate after the trend change inflection point is generated. If the interval time of the trend change inflection point is lower than the interval time threshold, or the curve decline rate after the trend change inflection point is generated exceeds the curve decline rate threshold, then the current action inflection point is marked as a valid inflection point and set as the prevention and control starting point. If the interval between the trend reversal inflection point and the curve decline rate after the trend reversal inflection point is not lower than the interval duration threshold, and the curve decline rate after the trend reversal inflection point is not higher than the curve decline rate threshold, then the current action inflection point will be marked as an invalid inflection point and an invalid prevention and control signal will be generated and sent to the fire early warning platform.

10. The fire early warning system based on the Internet of Things and edge computing according to claim 9, characterized in that, After determining the starting point of prevention and control, the smoke concentration curve is identified by nodes. If the current node of the smoke concentration curve shows an inflection point of rising concentration, a false extinguishing signal is generated and sent to the fire early warning platform. The fire early warning system collects smoke concentrations before and after the starting point of prevention and control synchronously and establishes a continuous data sequence to facilitate source tracing during secondary prevention and control. If the current node of the smoke concentration curve does not show an inflection point of rising concentration and the smoke concentration is lower than the set fire red line value, a fire cancellation signal is generated and sent to the fire early warning platform.

Images