Environment perception based cable shaft fire safety monitoring method and fire extinguishing system

Abstract

The application discloses a cable shaft fire safety monitoring method and fire extinguishing system based on environment perception, and the method comprises the following steps: building a multi-dimensional environment perception monitoring network of the cable shaft, collecting multi-source environment data and equipment operation data of cable operation temperature, environment smoke concentration, air temperature and humidity, combustible gas concentration and airflow velocity in the shaft based on the monitoring network, and obtaining a standardized monitoring data set after pre-processing the collected data. A cable shaft fire grading identification model is constructed to output real-time fire risk grade and fault positioning information; a grading linkage disposal strategy is triggered based on the real-time fire risk grade, and warning information is generated. The application realizes risk grading identification and accurate fault positioning, greatly reduces the false positive and false negative rates, constructs a whole-chain fire prevention and control system, significantly improves the accuracy and timeliness of cable shaft fire warning and disposal, and ensures the safe and stable operation of power equipment.

Description

Technical Field

[0001] This invention relates to the field of fire protection engineering technology, specifically to a fire safety monitoring method and fire extinguishing system for cable shafts based on environmental perception. Background Technology

[0002] Cable shafts are the core vertical channels for cable laying and centralized wiring in high-rise buildings, rail transit, industrial and mining enterprises, and power hub projects, undertaking the critical functions of power transmission and signal communication. Because cable shafts are long, narrow, and enclosed vertical spaces, densely packed with various power and control cables, and characterized by poor ventilation, difficulty in heat dissipation due to cable stacking, and continuous high-load operation, they are highly susceptible to overheating, smoldering, and even open-flame fires caused by cable insulation aging, overload overheating, short circuits, and poor contact. Furthermore, cable shafts exhibit a significant "chimney effect," meaning that in the event of a fire, the flames can spread rapidly upwards and downwards along the vertical channel, accompanied by the diffusion of toxic and harmful gases. This can cause large-scale power outages and equipment damage, and in severe cases, lead to major fire safety accidents, threatening building operation safety and the lives and property of personnel. Therefore, cable shafts represent a key and challenging scenario for power fire safety management.

[0003] Currently, fire monitoring methods for cable shafts remain relatively traditional, with the industry commonly employing single smoke sensors, fixed-point temperature sensors, and conventional infrared thermography equipment. These monitoring technologies are mostly single-point, discrete monitoring modes, inherently flawed due to their limited monitoring dimensions, limited coverage, and insufficient data acquisition accuracy. They can only passively detect open flames or noticeable smoke within the shaft, failing to achieve comprehensive and continuous perception of the entire cable's temperature and multi-dimensional environmental parameters within the shaft. Existing monitoring systems rely solely on single parameter thresholds to determine fire status, neglecting complex coupled factors such as airflow, temperature and humidity changes, combustible gas accumulation, and vertical fire spread within the cable shaft. This makes it difficult to capture early, subtle risk characteristics such as initial cable overheating, localized micro-hot spots, and smoldering hazards, resulting in low accuracy in early fire hazard identification, high false alarm and missed alarm rates, and an inability to predict and warn of fire hazards in advance. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method and system for monitoring fire safety in cable shafts based on environmental perception, which solves the problems of low accuracy in early fire hazard identification, high false alarm and missed alarm rates, and poor accuracy in classifying and determining fire risk levels.

[0005] The technical solution of this invention is: a method for monitoring fire safety in cable shafts based on environmental perception, comprising: S1. Establish a multi-dimensional environmental sensing and monitoring network for cable shafts. Based on the monitoring network, collect multi-source environmental data and equipment operation data such as cable operating temperature, ambient smoke concentration, air temperature and humidity, combustible gas concentration, and airflow velocity in the shaft. After preprocessing the collected data, a standardized monitoring dataset is obtained. S2: Construct a fire classification and identification model for cable shafts, input standardized monitoring datasets into the fire classification and identification model, and output real-time fire risk level and fault location information; The fire classification and identification model includes a multi-parameter coupled feature extraction module, a temporal phase difference feature extraction module, a spatiotemporal graph neural network module, a re-ignition risk comprehensive assessment module, and an operating condition fusion judgment module. The working condition fusion and determination module is used to fuse and determine the outputs of the multi-parameter coupled feature extraction module, the temporal phase difference feature extraction module, the spatiotemporal graph neural network module, and the re-ignition risk comprehensive assessment module. S3: Trigger a graded linkage response strategy based on the real-time fire risk level and generate early warning information.

[0006] In S1, the monitoring network adopts a three-dimensional adaptive monitoring topology structure combining a distributed optical fiber baseline with self-organizing floating nodes to carry out three-dimensional monitoring; distributed optical fibers are laid along the entire length of the cable as the monitoring baseline, and self-organizing floating sensing nodes are deployed at different heights and floors in the cable shaft.

[0007] In S2, the multi-parameter coupling features extracted by the multi-parameter coupling feature extraction module include: temperature-combustible gas concentration rise rate ratio, smoke concentration-airflow velocity phase difference, and dynamic change rate of CO / CO2 ratio; The method for calculating the ratio of temperature to combustible gas concentration rise rate is as follows: it is obtained by calculating the ratio of the time-series change rate of cable operating temperature to the time-series change rate of combustible gas concentration. The method for calculating the smoke concentration-airflow velocity phase difference characteristic is as follows: it is obtained by calculating the absolute difference between the phase of smoke concentration change and the phase of airflow velocity change. The method for calculating the dynamic change rate of the CO / CO2 ratio is as follows: it is obtained by calculating the derivative of the ratio of carbon monoxide concentration to carbon dioxide concentration with respect to time.

[0008] In S2, the time-series phase difference feature extraction module is used to calculate the sliding window phase difference of each monitoring data from the standardized monitoring dataset.

[0009] In S2, the spatiotemporal graph neural network module includes: The monitoring nodes distributed vertically in the shaft are used as graph nodes, and the physical connection relationships between monitoring nodes in adjacent layers are used as graph edges to construct a spatiotemporal graph of the shaft. Each monitoring node corresponds to the multi-dimensional environmental data of a monitoring layer in the shaft. The graph edges represent the spatial adjacency relationship between adjacent monitoring nodes. Each row in the monitoring node feature matrix corresponds to the feature vector of a graph node. The adjacency matrix represents the spatial connection relationship between graph nodes. Temporal features and spatial features are extracted using a temporal feature extraction module and a spatial feature extraction module, respectively. The temporal feature extraction module is used to capture the temporal evolution trend of features of each graph node based on a gating mechanism in the time dimension. The spatial feature extraction module is used to capture the spread pattern of fire in the vertical direction of the shaft through graph convolution operations in the spatial dimension. After fusing temporal and spatial characteristics, the predicted values ​​for fire risk and fire spread direction are output.

[0010] In S2, the comprehensive assessment module for reignition risk includes: The following three reignition risk indicators were calculated: the second-order change rate of the CO / CO2 ratio, the slope of the temperature rise in local hotspots, and the residual concentration of air ions. The characteristic of the second-order rate of change of the CO / CO2 ratio is obtained by calculating the derivative of the dynamic rate of change of the CO / CO2 ratio with respect to time. The slope characteristic of the temperature rise of local hotspots is obtained by calculating the ratio of the temperature difference of local hotspots at two different times after the fire extinguishing operation to the time difference. Air ion concentration residual amount: After the fire extinguishing operation is completed, monitor the residual ion concentration in the air inside the shaft. The three reignition risk indicators are weighted and summed with their corresponding weight coefficients to obtain a comprehensive reignition risk score. When the comprehensive reignition risk score exceeds the preset reignition risk score threshold, it is determined to be a reignition risk condition.

[0011] The working condition fusion judgment module is used to classify the vertical shaft fire status into four fire risk levels: abnormal cable overheating, initial smoldering fire, open flame fire, and risk of reignition.

[0012] In S2, the fault location information is determined by the fire risk prediction values ​​of each graph node output by the spatiotemporal graph neural network module. The fire risk prediction values ​​of each graph node are compared, and the graph node with the largest fire risk prediction value is taken as the fire source location node. The sequence of adjacent nodes pointed to by the fire spread direction prediction vector from the fire source location node is taken as the fire spread path. The output fault location information includes the vertical shaft level location of the fire source and the spread direction.

[0013] In S3, the hierarchical linkage response strategy includes: When the cable is in an abnormal overheating condition, a cooling warning is triggered, and the cable load is adjusted or ventilation is activated for cooling. In the initial stage of a smoldering fire, initiate localized targeted fire extinguishing operations and deploy extinguishing agents into the fault location area; In the event of an open flame fire, initiate full-area firefighting operations. When there is a risk of reignition, initiate continuous suppression and extinguishing operations, continuously delivering extinguishing agents to the area at risk of reignition until the second change rate of the CO / CO2 ratio is less than or equal to zero, the slope of the temperature rise of local hot spots is less than or equal to zero, and the residual air ion concentration is below the ion concentration threshold.

[0014] An environment-aware fire suppression system for cable shafts, the system comprising: Adaptive fire suppression control module and fire suppression execution component, The adaptive inactivation control module is used to adaptively control the spray flow rate, spray angle, spray duration, and coverage of the fire extinguishing execution components based on the determined fire risk level and the corresponding fire extinguishing level, for zoned fire extinguishing of cable shaft fires; specifically: The adaptive fire suppression control module receives real-time information on fire risk level and fault location. The built-in hierarchical strategy database is retrieved, and based on the current fire risk level, a unique match is made with the preset corresponding fire extinguishing level to establish a one-to-one correspondence between fire risk level and fire extinguishing level. Based on the matched fire extinguishing level, the corresponding standard control parameter group is adaptively called. The control parameter group includes preset spray flow threshold, spray angle adjustment range, single and continuous spray duration, and spatial coverage parameters. Based on the fire location zone information, only the fire extinguishing execution components corresponding to the fire zone of the cable shaft are activated, while the fire extinguishing devices in non-fire zones are locked into a standby state. The fire suppression execution component includes fire suppression nozzles arranged in zones. Control commands are issued to the activated zone fire suppression execution components to adaptively adjust the real-time spray flow, adjust the spray angle to the fire area, and set the spray duration to match the fire scale, so as to complete the full coverage fire suppression operation of the fire area according to the preset coverage range.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention constructs a three-dimensional sensing network combining distributed optical fibers and self-organizing floating nodes, capable of continuously collecting temperature data along the entire cable line and multi-source environmental data from vertical shafts. Standardized data is obtained through preprocessing, effectively improving monitoring coverage and data reliability. Multi-parameter coupling and temporal phase difference features are extracted and combined with an ST-GNN spatiotemporal graph neural network to accurately capture the vertical spread pattern and dynamic evolution trend of fires, enabling early fire prediction. A reignition risk assessment model is constructed, filling the gap in existing technologies for identifying reignition hazards, effectively avoiding secondary disasters, achieving risk classification identification and precise fault location, and significantly reducing false alarms and missed alarms. A full-chain fire prevention and control system is constructed, significantly improving the accuracy and timeliness of early warning and response to cable shaft fires, ensuring the safe and stable operation of power equipment, and demonstrating significant engineering application value. Attached Figure Description

[0016] Figure 1 This is a flowchart of the monitoring method of the present invention; Figure 2 This is a four-condition clustering distribution diagram of the multi-parameter coupling characteristics of the present invention; Figure 3 This is a comparative heat map for predicting the spread of fires in cable shafts, as presented in this invention. Figure 4 This is a comparison chart of fire risk prediction values ​​for key nodes in this invention. Detailed Implementation

[0017] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0018] Reference Figure 1 As shown, the environmental perception-based fire safety monitoring method for cable shafts includes: S1. Establish a multi-dimensional environmental sensing and monitoring network for cable shafts. Based on the monitoring network, collect multi-source environmental data and equipment operation data such as cable operating temperature, ambient smoke concentration, air temperature and humidity, combustible gas concentration, and airflow velocity in the shaft. After preprocessing the collected data, a standardized monitoring dataset is obtained. S2: Construct a fire classification and identification model for cable shafts, input standardized monitoring datasets into the fire classification and identification model, and output real-time fire risk level and fault location information; The fire classification and identification model includes a multi-parameter coupled feature extraction module, a temporal phase difference feature extraction module, a spatiotemporal graph neural network module, a re-ignition risk comprehensive assessment module, and an operating condition fusion judgment module. The standardized monitoring dataset is input into a multi-parameter coupled feature extraction module, a temporal phase difference feature extraction module, a spatiotemporal graph neural network module, and a resurgence risk comprehensive assessment module. The working condition fusion and determination module is used to fuse and determine the outputs of the multi-parameter coupled feature extraction module, the temporal phase difference feature extraction module, the spatiotemporal graph neural network module, and the re-ignition risk comprehensive assessment module. S3: Trigger a graded linkage response strategy based on the real-time fire risk level and generate early warning information.

[0019] Specifically, S1. Construction and data acquisition of a multi-dimensional environmental perception and monitoring network; In the fire safety monitoring system for cable shafts, a multi-dimensional environmental perception and monitoring network covering the entire shaft space is constructed. This monitoring network adopts a three-dimensional adaptive monitoring topology combining distributed fiber optic baselines and self-organizing floating nodes to achieve comprehensive, high-precision, and real-time acquisition of environmental parameters within the shaft.

[0020] The distributed fiber optic baseline deployment method involves laying distributed fiber optic temperature sensors along the entire length of the cable. These sensors are laid along the cable surface or cable tray, enabling continuous sensing of temperature changes at various points along the cable. The distributed fiber optics utilize backscattering Raman scattering or Brillouin scattering principles for temperature measurement, achieving continuous temperature monitoring with a spatial resolution better than 1 meter and a temperature resolution better than 0.1 degrees Celsius along the fiber's length. As a monitoring baseline, the distributed fiber optics provide continuous temperature monitoring coverage throughout the shaft, filling monitoring blind spots that traditional point-type temperature sensors cannot cover.

[0021] The deployment method for self-organizing floating nodes is as follows: multiple self-organizing floating sensing nodes are deployed at different heights and between different floors in the cable shaft. These self-organizing floating sensing nodes use wireless self-organizing network technology to form a mesh communication topology, enabling data relay transmission between adjacent nodes and ensuring communication coverage throughout the entire shaft space. Each self-organizing floating sensing node is equipped with multiple high-precision sensor modules, including smoke concentration sensors, combustible gas concentration sensors, air temperature and humidity sensors, gas flow rate sensors, and air ion concentration sensors. The deployment density of the self-organizing floating sensing nodes is adaptively adjusted according to the shaft height, number of floors, and cable laying density. Typically, they are densely deployed at each floor level, corner locations, and areas with dense cables to ensure the integrity and reliability of the monitoring data.

[0022] During data acquisition, the distributed fiber optic demodulator demodulates the fiber optic sensor signals at a sampling frequency of at least 10 Hz to acquire the temperature distribution curve along the entire length of the cable in real time. The ad hoc network floating nodes acquire the output signals of each sensor at a sampling frequency of at least 5 Hz and transmit the acquired data to the aggregation node via a wireless ad hoc network. The aggregation node performs unified aggregation and preliminary processing of the data from the distributed fiber optic network and ad hoc network nodes.

[0023] The system performs preprocessing on the real-time collected multi-source raw data, removing interference noise and abnormal data, filling in missing data, and performing time-series alignment and format normalization on the multi-source data to generate a standardized and structured cable shaft monitoring dataset.

[0024] S2. Construction and operation of the fire classification identification model; By inputting standardized monitoring datasets into a pre-built cable shaft fire classification and identification model, intelligent identification of the fire status and risk level determination of the shaft can be achieved.

[0025] The fire classification identification model includes: a multi-parameter coupled feature extraction module, a temporal phase difference feature extraction module, a spatiotemporal graph neural network module, a comprehensive assessment module for reignition risk, and a condition fusion judgment module.

[0026] Construction of a multi-parameter coupled feature extraction module: Based on experimental data and historical monitoring data of cable shaft fires, multi-source time-series data such as temperature, smoke concentration, combustible gas concentration, airflow velocity, CO concentration, and CO2 concentration are collected under different combustion conditions (abnormal cable overheating, initial smoldering fire, open flame fire, and risk of reignition). The statistical distribution of three coupled features—temperature-combustible gas concentration rise rate ratio, smoke concentration-airflow velocity phase difference, and dynamic change rate of CO / CO2 ratio—is calculated for each condition. The feature threshold ranges corresponding to each condition are calibrated (e.g., the first preset range of 0.5-2.0 corresponds to the thermal decomposition condition, and the second preset range of 2.0-5.0 corresponds to the smoldering condition), forming a coupled feature discrimination rule base.

[0027] Construction of the spatiotemporal graph neural network module: Based on the multi-layer monitoring node layout of cable shafts, a spatiotemporal graph data structure (node ​​feature matrix, adjacency matrix) is constructed for the shafts. Cable shaft fire simulation data (such as multi-layer temperature field evolution data generated by FDS fire dynamics simulation software) and historical fire event data are used as the training set. The future fire risk level label (level 0-3) and spread direction label of each node are used as supervision signals to train the graph convolutional layer weights and gated recurrent unit parameters of the ST-GNN spatiotemporal model. The Adam optimizer is used for training, and the loss function is a weighted sum of the cross-entropy loss of the fire risk prediction value and the cosine distance loss of the spread direction prediction vector.

[0028] Construction of the comprehensive assessment module for reignition risk: Based on historical monitoring data after fire extinguishing operations, the statistical distribution of three indicators—the second change rate of the CO / CO2 ratio, the slope of the temperature rise in local hotspots, and the residual amount of air ion concentration—is collected under different reignition / non-reignition scenarios. The weight coefficients of the three indicators (0.40 for the second change rate of the CO / CO2 ratio, 0.35 for the slope of the temperature rise, and 0.25 for the residual amount of ion concentration) and the reignition risk scoring threshold are determined by logistic regression.

[0029] The construction of the working condition fusion judgment module involves fusing the coupling feature discrimination results output by the multi-parameter coupling feature extraction module, the sliding window phase difference of each monitoring parameter calculated from the dataset by the temporal phase difference feature extraction module, the node risk prediction value and spread direction prediction value output by the spatiotemporal graph neural network module, and the reignition risk score output by the reignition risk comprehensive assessment module. The fusion strategy employs a hierarchical judgment logic: first, it determines whether the fire is in the post-extinguishing stage; if so, the reignition risk is assessed first; otherwise, the coupling feature discrimination results and ST-GNN risk prediction values ​​are combined to distinguish four working conditions.

[0030] Extraction of multi-parameter coupled features is the foundation of fire classification identification, referring to... Figure 2 The diagram illustrates the four-condition clustering distribution of multi-parameter coupling features. First, three core multi-parameter coupling features are extracted from the standardized monitoring dataset: the temperature-combustible gas concentration rise rate ratio, the smoke concentration-airflow velocity phase difference, and the dynamic change rate of the CO / CO2 ratio. The temperature-combustible gas concentration rise rate ratio is obtained by calculating the ratio of the time-series change rate of cable operating temperature to the time-series change rate of combustible gas concentration; this feature characterizes the dynamic relationship between temperature change and combustible gas release. The smoke concentration-airflow velocity phase difference is obtained by calculating the absolute difference between the phase change of smoke concentration and the phase change of airflow velocity; this feature characterizes the temporal correlation between smoke diffusion behavior and airflow movement. The dynamic change rate of the CO / CO2 ratio is obtained by calculating the derivative of the carbon monoxide concentration to carbon dioxide concentration ratio with respect to time; this feature characterizes the changing trend of combustion completeness.

[0031] Temporal phase difference feature extraction focuses on the phase relationship of monitoring data over time. A sliding window sampling method is applied to each monitoring data point, and the phase angle of the data change trend within adjacent time windows is calculated. The temporal phase difference feature is obtained by calculating the difference in these phase angles. These temporal phase difference features can capture the sequence and correlation strength of changes in different parameters during fire development.

[0032] The construction and operation of the spatiotemporal graph neural network is the core technical feature of this method. N monitoring nodes distributed vertically along the shaft are treated as graph nodes, and the physical connections between adjacent monitoring nodes are treated as graph edges, thus constructing a spatiotemporal graph of the shaft. Each monitoring node corresponds to multi-dimensional environmental data of a monitoring layer in the shaft. Each row in the monitoring node feature matrix corresponds to a feature vector of a graph node, and the adjacency matrix represents the spatial connections between graph nodes. In the temporal dimension, a gating mechanism is used to capture the temporal evolution trend of each graph node's features, and in the spatial dimension, graph convolution operations are used to capture the fire spread pattern along the vertical direction of the shaft. Through spatiotemporal feature fusion, the fire classification and identification model simultaneously outputs the predicted fire risk value and fire spread direction for each node at future moments.

[0033] A monitoring node is a logical abstraction of a self-organizing network floating sensing node in the spatiotemporal map construction. The relationship between the monitoring node and the self-organizing network floating sensing node is as follows: the self-organizing network floating sensing node is a physical entity (a wireless communication node equipped with multiple sensors), and the monitoring node is the graph node corresponding to these physical nodes when constructing the spatiotemporal map of the shaft. A monitoring node corresponds to one or a group of self-organizing network floating sensing nodes at a certain height / floor location in the shaft, and the multidimensional environmental data it collects constitutes the feature vector of the monitoring node.

[0034] The reignition risk comprehensive assessment module is used to determine whether there is a risk of reignition after firefighting operations. This model is based on a comprehensive assessment of three core indicators: the second-order change rate of the CO / CO2 ratio, the slope of the temperature recovery of local hotspots, and the residual concentration of air ions. A weighted comprehensive calculation of these three indicators yields a reignition risk score; when the score exceeds a preset threshold, the condition is identified as a reignition risk situation.

[0035] The fire condition differentiation and output are achieved through model fusion of multi-dimensional features. The fire classification and identification model comprehensively analyzes multi-parameter coupling features, temporal phase difference features, spatiotemporal graph neural network output features, and reignition risk assessment results to classify shaft fire states into four typical conditions: cable overheating anomaly, initial smoldering fire, open flame fire, and reignition risk. The fire classification and identification model outputs the current fire risk level in real time, which is divided into four levels, each corresponding to one of the four conditions. Simultaneously, the fire classification and identification model outputs fault location information, including the shaft level location of the fire source and the possible direction and path of fire spread.

[0036] The fire classification identification model's condition fusion and judgment module receives the outputs of four modules and distinguishes between the four conditions through hierarchical judgment logic. The specific fusion analysis mechanism is as follows: The standardized monitoring dataset is the unified data source output in step S1. All four modules use this dataset as input and extract features from different dimensions: the multi-parameter coupled feature extraction module calculates the ratio of temperature to combustible gas concentration rise rate, the phase difference between smoke concentration and airflow velocity, and the dynamic change rate of the CO / CO2 ratio from the dataset; the temporal phase difference feature extraction module calculates the sliding window phase difference of each monitoring data from the dataset; the spatiotemporal graph neural network module constructs a node feature matrix using multi-dimensional data from each monitoring node in the dataset for spatiotemporal inference; and the re-ignition risk comprehensive assessment module extracts the second-order change rate of the CO / CO2 ratio, the temperature recovery slope, and the residual ion concentration from the dataset in the post-fire extinguishing stage.

[0037] The correspondence between the four operating conditions and their characteristic combinations is as follows: Condition 1: Cable overheating anomaly – Triggering conditions: The ratio of temperature to combustible gas concentration increase rate is within the first preset range (0.5-2.0), indicating that temperature and combustible gas concentration are rising synchronously; the node risk prediction value output by ST-GNN is below 0.3 (low risk); the dynamic change rate of the CO / CO2 ratio tends to zero or is negative. Overall judgment: The cable insulation layer has begun to decompose due to heat but has not yet formed an independent combustion source.

[0038] Condition 2: Initial Smoldering Fire – Triggering Conditions: The ratio of temperature to combustible gas concentration increase rate is within the second preset range (2.0-5.0), indicating that the combustible gas concentration increases significantly faster than the temperature; the dynamic change rate of the CO / CO2 ratio is positive and continuously increasing, indicating an intensified degree of incomplete combustion; the node risk prediction value output by ST-GNN is between 0.3 and 0.7 (medium risk); the temporal phase difference characteristics show a significant phase shift between changes in smoke concentration and changes in airflow velocity. Overall Judgment: A pyrolysis reaction occurs inside the insulation layer, combustible gas is rapidly released, and smoldering transitions to open flame.

[0039] Condition 3: Open Flame Fire – Triggering Conditions: The dynamic change rate of the CO / CO2 ratio continuously increases and the absolute value exceeds the preset open flame threshold; the node risk prediction value output by ST-GNN exceeds 0.7 (high risk), and the spread direction prediction vector points to adjacent graph nodes (the fire is spreading vertically); multi-parameter coupled features show that temperature, smoke, and combustible gas concentrations rise rapidly and synchronously. Comprehensive Judgment: Open flame combustion has formed, and the fire is spreading vertically along the shaft.

[0040] Condition 4: Reignition Risk – Triggering Conditions: The fire is in the post-extinguishing stage; the comprehensive score for reignition risk exceeds the preset threshold (60 points), specifically: the secondary change rate of the CO / CO2 ratio is greater than zero (weight 0.40, value 1.0); the slope of the temperature rise in local hotspots is positive (weight 0.35, value 1.0); and the residual concentration of air ions exceeds the threshold (weight 0.25, value 1.0). The weighted sum of these three indicators exceeds 60 points. Overall Judgment: The combustion reaction has not completely ceased after the fire extinguishing operation, posing a secondary fire hazard.

[0041] S3. Triggering and execution of tiered and coordinated response strategies; Based on the real-time fire risk level, the system triggers corresponding graded linkage response strategies and generates corresponding early warning information.

[0042] When an abnormal cable overheating condition is detected, the system generates a cooling warning, prompting maintenance personnel to take intervention measures, such as adjusting cable load distribution or activating the shaft ventilation and cooling system. When an initial smoldering fire condition is detected, the system initiates localized targeted fire suppression operations, precisely delivering extinguishing agents to the fault location area. When an open flame fire condition is detected, the system initiates full-area fire suppression operations, increasing the amount and coverage of extinguishing agents. When a reignition risk condition is detected, the system initiates continuous fire suppression operations, continuously delivering extinguishing agents to the reignition risk area until all risk indicators fall below the safe threshold.

[0043] The detailed technical implementation of multi-parameter coupled feature extraction in this invention is as follows: The temperature-combustible gas concentration rise rate ratio is used to characterize the dynamic coupling relationship between temperature change and combustible gas release during the thermal decomposition of cable insulation.

[0044] The calculation method is as follows: Within a preset sliding time window, the time-series data of cable operating temperature and combustible gas concentration are sampled synchronously, and the ratio of the time-series change rate of temperature to the time-series change rate of combustible gas concentration is calculated. The time-series change rate of temperature is defined as the difference between the current temperature value and the previous temperature value divided by the sampling time interval, and the time-series change rate of combustible gas concentration is defined as the difference between the current combustible gas concentration value and the previous concentration value divided by the sampling time interval. A logarithmic transformation is used in the ratio calculation to enhance the comparability and stability of the features.

[0045] The judgment rule adopts a two-level threshold judgment mechanism. The first preset range corresponds to the condition where the temperature and combustible gas concentration rise simultaneously. When the calculated ratio is within the first preset range, it indicates that the cable insulation layer begins to decompose upon heating, and combustible gas is released simultaneously with the temperature rise, pointing to the condition of thermal decomposition of the cable insulation layer. The second preset range corresponds to the condition where the combustible gas concentration rises significantly faster than the temperature rise. When the calculated ratio is within the second preset range, it indicates that a pyrolysis reaction occurs inside the insulation layer, and combustible gas is released rapidly, but the heat has not yet been fully conducted to the cable surface, pointing to the condition of initial smoldering fire. The first preset range is 0.5-2.0; the second preset range is 2.0-5.0.

[0046] The smoke concentration-airflow velocity phase difference feature is used to characterize the temporal correlation between smoke diffusion behavior and vertical shaft airflow movement, and then to infer the relative relationship between the fire source location and the airflow direction.

[0047] The calculation method is as follows: Hilbert transform is applied to the time-series data of smoke concentration and airflow velocity to extract their respective instantaneous phase information. The instantaneous phase of smoke concentration change is obtained by calculating the phase angle of the analytic signal of the smoke concentration signal, and the instantaneous phase of airflow velocity change is obtained by calculating the phase angle of the analytic signal of the airflow velocity signal. Then, the absolute difference between the two instantaneous phases is calculated as the characteristic value of the smoke concentration-airflow velocity phase difference.

[0048] The judgment rule sets a preset phase difference threshold. When the calculated phase difference exceeds the preset phase difference threshold, it indicates that the smoke diffusion behavior is driven by airflow, resulting in a leading or lagging effect. When the smoke concentration change phase leads the airflow velocity change phase, it indicates that the fire source is located upstream of the airflow, and the smoke diffuses towards the monitoring point under the push of the airflow; when the smoke concentration change phase lags behind the airflow velocity change phase, it indicates that the fire source is located downstream of the airflow, and the airflow reaches the monitoring point first, while the smoke diffuses there later. The sign and magnitude of the phase difference can be used to infer the directional relationship between the fire source and the monitoring node; the preset phase difference threshold is obtained based on experimental calibration, and in this embodiment, it is set to 0.5.

[0049] The dynamic change rate of the CO / CO2 ratio is used to characterize the changing trend of the degree of combustion completeness, and thus determine whether the fire is in the transitional stage of converting into open flame.

[0050] The calculation method is as follows: continuously monitor the ratio of carbon monoxide concentration to carbon dioxide concentration in the vertical shaft, and calculate the derivative of this ratio with respect to time to obtain the dynamic rate of change of the CO / CO2 ratio. The dynamic rate of change is calculated using the moving difference method, calculating the rate of change of the ratio within a preset time window.

[0051] The judgment rule is based on the sign and trend of the dynamic change rate. When the dynamic change rate is positive and continuously increasing, it indicates that the proportion of carbon monoxide to carbon dioxide is constantly increasing, the degree of incomplete combustion is intensifying, the combustion reaction is in progress but has not yet reached a complete combustion state, pointing to the transitional condition from smoldering to open flame. When the dynamic change rate is negative or tends to zero, it indicates that the degree of complete combustion is improving or has tended to stabilize, which can be used as one of the criteria for effective fire control.

[0052] The detailed technical implementation of the Spatiotemporal Graph Neural Network (ST-GNN) is as follows: The vertical shaft spatiotemporal graph is the basic data structure for processing vertical shaft fire monitoring data by spatiotemporal graph neural networks. It uses a graph structure to represent the spatial adjacency relationship in the vertical direction of the shaft and the multidimensional temporal characteristics of each monitoring node.

[0053] A graph node is defined as N monitoring nodes distributed vertically along a cable shaft. The monitoring nodes correspond to the floor or height range divisions of the cable shaft, with each node representing a monitoring unit at a specific vertical location within the shaft. The number of monitoring nodes N is determined based on the shaft height and monitoring resolution requirements, typically equal to the number of floors in the shaft multiplied by the node density coefficient for each floor.

[0054] The definition of a graph edge is: the physical connection between monitoring nodes on adjacent layers is used as the graph edge. Graph edges represent the spatial adjacency between monitoring layers on adjacent shafts. Fire mainly spreads vertically within the shaft, and the probability of fire spread between adjacent layers is the highest. Therefore, spatial connection edges are established between adjacent nodes.

[0055] The node feature matrix is ​​constructed as follows: For each graph node, its corresponding multi-dimensional environmental monitoring data is used to form the node's feature vector. The dimension of the feature vector is equal to the number of monitoring parameters, including cable temperature, ambient smoke concentration, combustible gas concentration, air temperature and humidity, airflow velocity, and multi-parameter coupling features extracted in the previous embodiment, all collected at the node location. The dimension of the node feature matrix is ​​N multiplied by F, where N is the number of nodes and F is the feature dimension of each node.

[0056] The adjacency matrix is ​​constructed as follows: It represents the spatial connectivity between nodes in the graph. The element in the i-th row and j-th column of the adjacency matrix represents the connection weight between node i and node j. For adjacent vertical shaft monitoring layer nodes, the connection weight is 1; for non-adjacent nodes, the connection weight is 0 or weighted according to physical distance. The adjacency matrix needs to be normalized before graph convolution operations using a symmetric normalized Laplacian operator.

[0057] The spatial feature extraction module uses graph convolutional layers to spatially aggregate node features at each time step, capturing the spread pattern of fire in the vertical direction of the shaft.

[0058] The computation process of graph convolution is as follows: For the input node feature matrix X and normalized adjacency matrix A, the graph convolutional layer first performs weighted aggregation on the node feature matrix. The aggregation method uses the normalized adjacency matrix to multiply the node feature matrix, so that the features of each node are updated by the weighted sum of its own features and the features of its neighboring nodes. The aggregated feature matrix undergoes a linear transformation through the spatial convolution weight matrix W, and then a nonlinear mapping is performed through an activation function to obtain the spatial feature representation. The activation function uses a modified linear unit or an exponential linear unit to enhance the nonlinear expressive power of the model.

[0059] To enhance the extraction of spatial features, spatiotemporal graph neural networks typically stack multiple graph convolutional layers. Multi-layer graph convolutions expand the receptive field of nodes, enabling each node to aggregate information from more distant neighbors. The first graph convolutional layer aggregates information from adjacent nodes, the second layer aggregates information from second-order neighbors, and so on, achieving multi-hop spatial information aggregation through layer stacking.

[0060] The temporal feature extraction module performs temporal modeling of the feature time sequence of each node through a gated loop unit, capturing the dependencies and evolution trends of node features in the time dimension.

[0061] The internal structure of the gated loop unit includes two gating mechanisms: an update gate and a reset gate. The update gate controls the proportion of hidden state information from the previous time step that needs to be passed to the current time step, while the reset gate controls the proportion of hidden state information from the previous time step that needs to be forgotten. Through these two gating mechanisms, the gated loop unit can selectively retain and update temporal information, effectively capturing long-distance temporal dependencies.

[0062] The calculation process for temporal feature extraction is as follows: For each node, the current input features are concatenated with the hidden state of the previous time step, and a linear transformation is performed using the weight matrix. The values ​​of the update gate, reset gate, and candidate hidden state are calculated respectively. Then, the hidden state of the current time step is updated according to the update gate and reset gate, which is used as the temporal feature output of that node. The gated recurrent unit independently models the time of each node in the graph, thus enabling the simultaneous capture of the temporal evolution trend of all nodes.

[0063] The spatiotemporal fusion output module integrates spatial and temporal features and outputs fire risk prediction values ​​and fire spread direction prediction values ​​based on the fused features.

[0064] The spatiotemporal fusion strategy employs a weighted fusion mechanism, linearly combining spatial and temporal features according to preset spatiotemporal fusion weight coefficients. These weight coefficients are optimized based on the contribution of spatial and temporal features to the fire prediction task. The fused features simultaneously include information on the spatial spread of fire in the vertical direction of the shaft and its evolutionary trend over time.

[0065] The output layer design includes two parallel output branches. The first branch predicts the future fire risk value of each node, mapping the fused features to scalar risk values ​​through a fully connected layer. The risk value range is normalized to the interval between zero and one. The second branch predicts the fire spread direction vector, mapping the fused features to an N-dimensional vector through a fully connected layer. Each dimension of the vector represents the probability of spread to the corresponding neighboring node.

[0066] The detailed technical implementation of the comprehensive assessment of reignition risk and fault location is as follows: The reignition risk comprehensive assessment module is used to determine whether there is a risk of reignition in the shaft after the fire extinguishing operation is completed. It is based on a comprehensive assessment of three independent reignition risk indicators.

[0067] The calculation method for the second-order rate of change of the CO / CO2 ratio is as follows: calculate the derivative of the dynamic rate of change of the CO / CO2 ratio with respect to time, that is, calculate the rate of change of the dynamic rate of change of the CO / CO2 ratio. When the second-order rate of change of the CO / CO2 ratio is greater than zero, it indicates that the first-order rate of change of the CO / CO2 ratio is accelerating, the proportion of carbon monoxide to carbon dioxide is still increasing, the chemical process of the combustion reaction has not been completely terminated, and there is a risk of reignition. When the second-order rate of change of the CO / CO2 ratio is less than or equal to zero, it indicates that the combustion reaction has tended to terminate or has been completely terminated.

[0068] The calculation method for the local hotspot temperature recovery slope index is as follows: After the fire extinguishing operation is completed, monitor the temperature change at the local hotspot location on the cable, and calculate the ratio of the temperature difference between two different moments to the time difference as the temperature recovery slope. When the temperature recovery slope is positive, it indicates that the local hotspot temperature has risen instead of falling after the fire extinguishing operation, and the residual heat inside the cable insulation layer is being conducted to the surface, posing a risk of reignition. When the temperature recovery slope is less than or equal to zero, it indicates that the local hotspot temperature has decreased or remained stable.

[0069] The method for determining the residual air ion concentration is as follows: After the fire extinguishing operation is completed, monitor the residual ion concentration in the air inside the shaft. When the residual air ion concentration exceeds a preset ion concentration threshold, it indicates that a large amount of ionized active substances still exist in the combustion products, the chemical chain of the combustion reaction has not been completely interrupted, and there is a risk of reignition. When the residual air ion concentration is below the ion concentration threshold, it indicates that the chemically active substances have basically dissipated; the ion concentration threshold is set at 1.5 × 10⁻⁶. 4 pcs / cm³

[0070] The comprehensive scoring uses a weighted summation mechanism, summing the three reignition risk indicators with their corresponding weighting coefficients to obtain the comprehensive reignition risk score. Each indicator takes its first value when the corresponding reignition condition is met, and its second value when it is not. The first and second values ​​are set according to the contribution of each indicator to the reignition risk assessment, typically with the first value set to 1.0 and the second value to 0.0. The weighting coefficients are optimized based on historical fire data and expert experience. Specifically, the weighting coefficient for the quadratic change rate of the CO / CO2 ratio is 0.40, the weighting coefficient for the slope of the temperature rise in local hotspots is 0.35, and the weighting coefficient for residual air ion concentration is 0.25, with the sum of these three coefficients being 1.0. When the comprehensive reignition risk score exceeds a preset reignition risk score threshold, the condition is considered a reignition risk condition; the reignition risk score threshold is set at 60.

[0071] An environment-aware fire suppression system for cable shafts includes an adaptive fire suppression control module and fire suppression execution components. The adaptive fire suppression control module, as the system's control and decision-making unit, is responsible for receiving fire risk level and fault location information. Based on the hierarchical strategy database, it matches the corresponding fire suppression level and adaptively adjusts the operating parameters of the fire suppression execution components. The adaptive fire suppression control module integrates a hierarchical strategy database, a standard control parameter library, and an adaptive control algorithm. The hierarchical strategy database stores the correspondence between fire risk levels and fire suppression levels, as well as the standard control parameter sets for each fire suppression level. The standard control parameter library stores the spray flow threshold, spray angle adjustment range, spray duration parameters, and spatial coverage parameters corresponding to different fire suppression levels. The adaptive control algorithm calculates the optimal control parameters in real time based on the current fire situation and sends control commands to lower-level fire suppression execution components.

[0072] The fire suppression execution component, as the system's actuator, includes zoned fire suppression nozzles, solenoid valve assemblies, piping networks, and fire extinguishing agent storage devices. It receives control commands from the adaptive fire suppression control module, drives the solenoid valves to control the on / off state and flow rate of the fire extinguishing agent, and drives the nozzles to adjust the spray angle and coverage area, achieving precise zoned fire suppression for cable shaft fires.

[0073] Through the above description of specific embodiments, the present invention achieves the following technical effects: A three-dimensional adaptive monitoring topology combining distributed fiber optic baselines and self-organizing floating nodes is adopted. It continuously monitors temperature changes along the entire length of the cable while covering environmental parameters at various spatial locations in the shaft, eliminating blind spots of traditional point monitoring and achieving comprehensive coverage of shaft fire monitoring.

[0074] The three multi-parameter coupled features of temperature-combustible gas concentration rise rate ratio, smoke concentration-airflow velocity phase difference, and CO / CO2 ratio dynamic change rate can effectively distinguish different combustion conditions such as thermal decomposition of cable insulation, initial smoldering, and open flame combustion, thus improving the accuracy of fire type identification.

[0075] By simultaneously capturing the vertical spread pattern of fire in space and its evolution trend in time through a spatiotemporal graph neural network, the constructed vertical shaft spatiotemporal graph structure retains the spatial adjacency relationship in the vertical direction of the shaft, and the gated loop unit captures the time dependence of fire development, realizing the early prediction of fire spread direction and risk escalation.

[0076] By comprehensively evaluating three independent indicators—the rate of change of the CO / CO2 ratio, the slope of the temperature rise in local hotspots, and the residual concentration of air ions—it is possible to accurately determine whether there is a risk of reignition after the firefighting operation is completed, providing a scientific basis for determining the termination of continuous firefighting operations.

[0077] Through the hierarchical strategy matching and zoned precise control mechanism of the adaptive fire suppression control module, the precise correspondence between fire suppression response and fire risk level is achieved. Fire suppression resources are only deployed to the fire-affected area, and the dynamic fine-tuning mechanism ensures real-time matching between fire suppression intensity and fire scale.

[0078] Reference Figure 3 As shown, the fire risk value is visually displayed from two dimensions: spatial (shaft level, y-axis) and temporal (time step, x-axis) (the redder the color, the higher the risk): The left image (real fire spread) shows the actual process of a fire spreading gradually upwards and downwards from the fire source layer (approximately the 5th floor). The risk distribution is smooth and conforms to physical laws, serving as a benchmark for evaluating the effectiveness of the prediction.

[0079] The middle image (spatial feature prediction only): relying solely on spatial convolution modeling, the temporal rhythm of fire spread is significantly delayed, and the risk distribution shows a large number of fragmented noise points, which deviates significantly from the actual situation.

[0080] The right figure (ST-GMN spatiotemporal model prediction) closely resembles the actual spread process in terms of spatial distribution and temporal rhythm, with less noise and the ability to accurately capture the upward and downward spread trends of fire.

[0081] The ST-GMN spatiotemporal model, by integrating spatiotemporal features, is significantly superior to methods that only use spatial features, and more closely approximates the actual fire spread patterns.

[0082] Reference Figure 4 The chart shows a time-series comparison of fire risk at key nodes. These three line graphs focus on three core locations: below the fire source (node ​​3), the fire source layer (node ​​5), and above the fire source (node ​​7), comparing the prediction accuracy of different methods. Actual value (solid black line): Represents the curve of the actual fire risk at each node changing over time, and is the standard for assessment.

[0083] Spatial characteristics only (gray dashed line): There is a significant lag at all nodes, especially at and above the ignition source layer, where the risk rises much slower than the true value and fluctuates dramatically.

[0084] ST-GMN spatiotemporal model (red dashed line): At the fire source layer (node ​​5): it almost completely overlaps with the actual value, and can accurately capture the process of risk rising rapidly and approaching saturation.

[0085] Below (node ​​3) and above (node ​​7) the fire source: not only does it match the upward trend of the actual value, but it can also provide early warning of risks about 2 steps in advance (marked by the red dotted line in the figure), achieving more timely prediction.

[0086] The ST-GMN spatiotemporal model is far superior to methods based solely on spatial features in terms of both temporal accuracy and early warning capability, and can more reliably provide early warnings of fire risks at critical nodes.

[0087] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A method for monitoring fire safety in cable shafts based on environmental perception, characterized in that, include: S1. Establish a multi-dimensional environmental sensing and monitoring network for cable shafts. Based on the monitoring network, collect multi-source environmental data and equipment operation data such as cable operating temperature, ambient smoke concentration, air temperature and humidity, combustible gas concentration, and airflow velocity in the shaft. After preprocessing the collected data, a standardized monitoring dataset is obtained. S2: Construct a fire classification and identification model for cable shafts, input standardized monitoring datasets into the fire classification and identification model, and output real-time fire risk level and fault location information; The fire classification and identification model includes a multi-parameter coupled feature extraction module, a temporal phase difference feature extraction module, a spatiotemporal graph neural network module, a re-ignition risk comprehensive assessment module, and an operating condition fusion judgment module. The working condition fusion and determination module is used to fuse and determine the outputs of the multi-parameter coupled feature extraction module, the temporal phase difference feature extraction module, the spatiotemporal graph neural network module, and the re-ignition risk comprehensive assessment module. S3: Trigger a graded linkage response strategy based on the real-time fire risk level and generate early warning information.

2. The method for monitoring fire safety in cable shafts based on environmental perception according to claim 1, characterized in that: In S1, the monitoring network adopts a three-dimensional adaptive monitoring topology structure combining a distributed optical fiber baseline with self-organizing floating nodes to carry out three-dimensional monitoring; distributed optical fibers are laid along the entire length of the cable as the monitoring baseline, and self-organizing floating sensing nodes are deployed at different heights and floors in the cable shaft.

3. The method for monitoring fire safety in cable shafts based on environmental perception according to claim 1, characterized in that, In S2, the multi-parameter coupling features extracted by the multi-parameter coupling feature extraction module include: temperature-combustible gas concentration rise rate ratio, smoke concentration-airflow velocity phase difference, and dynamic change rate of CO / CO2 ratio; The method for calculating the ratio of temperature to combustible gas concentration rise rate is as follows: it is obtained by calculating the ratio of the time-series change rate of cable operating temperature to the time-series change rate of combustible gas concentration. The method for calculating the smoke concentration-airflow velocity phase difference characteristic is as follows: it is obtained by calculating the absolute difference between the phase of smoke concentration change and the phase of airflow velocity change. The method for calculating the dynamic change rate of the CO / CO2 ratio is as follows: it is obtained by calculating the derivative of the ratio of carbon monoxide concentration to carbon dioxide concentration with respect to time.

4. The method for monitoring fire safety in cable shafts based on environmental perception according to claim 1, characterized in that, In S2, the time-series phase difference feature extraction module is used to calculate the sliding window phase difference of each monitoring data from the standardized monitoring dataset.

5. The method for monitoring fire safety in cable shafts based on environmental perception according to claim 1, characterized in that, In S2, the spatiotemporal graph neural network module includes: The monitoring nodes distributed vertically in the shaft are used as graph nodes, and the physical connection relationships between monitoring nodes in adjacent layers are used as graph edges to construct a spatiotemporal graph of the shaft. Each monitoring node corresponds to the multi-dimensional environmental data of a monitoring layer in the shaft. The graph edges represent the spatial adjacency relationship between adjacent monitoring nodes. Each row in the monitoring node feature matrix corresponds to the feature vector of a graph node. The adjacency matrix represents the spatial connection relationship between graph nodes. Temporal features and spatial features are extracted using a temporal feature extraction module and a spatial feature extraction module, respectively. The temporal feature extraction module is used to capture the temporal evolution trend of features of each graph node based on a gating mechanism in the time dimension. The spatial feature extraction module is used to capture the spread pattern of fire in the vertical direction of the shaft through graph convolution operations in the spatial dimension. After fusing temporal and spatial characteristics, the predicted values ​​for fire risk and fire spread direction are output.

6. The method for monitoring fire safety in cable shafts based on environmental perception according to claim 1, characterized in that, In S2, the comprehensive assessment module for reignition risk includes: The following three reignition risk indicators were calculated: the second-order change rate of the CO / CO2 ratio, the slope of the temperature rise in local hotspots, and the residual concentration of air ions. The characteristic of the second-order rate of change of the CO / CO2 ratio is obtained by calculating the derivative of the dynamic rate of change of the CO / CO2 ratio with respect to time. The slope characteristic of the temperature rise of local hotspots is obtained by calculating the ratio of the temperature difference of local hotspots at two different times after the fire extinguishing operation to the time difference. Air ion concentration residual amount: After the fire extinguishing operation is completed, monitor the residual ion concentration in the air inside the shaft. The three reignition risk indicators are weighted and summed with their corresponding weight coefficients to obtain a comprehensive reignition risk score. When the comprehensive reignition risk score exceeds the preset reignition risk score threshold, it is determined to be a reignition risk condition.

7. The method for monitoring fire safety in cable shafts based on environmental perception according to claim 1, characterized in that: The working condition fusion judgment module is used to classify the vertical shaft fire status into four fire risk levels: abnormal cable overheating, initial smoldering fire, open flame fire, and risk of reignition.

8. The method for monitoring fire safety in cable shafts based on environmental perception according to claim 1, characterized in that: In S2, the fault location information is determined by the fire risk prediction values ​​of each graph node output by the spatiotemporal graph neural network module. The fire risk prediction values ​​of each graph node are compared, and the graph node with the largest fire risk prediction value is taken as the fire source location node. The sequence of adjacent nodes pointed to by the fire spread direction prediction vector from the fire source location node is taken as the fire spread path. The output fault location information includes the vertical shaft level location of the fire source and the spread direction.

9. The method for monitoring fire safety in cable shafts based on environmental perception according to claim 1, characterized in that, In S3, the tiered and coordinated response strategy includes: When the cable is in an abnormal overheating condition, a cooling warning is triggered, and the cable load is adjusted or ventilation is activated for cooling. In the initial stage of a smoldering fire, initiate localized targeted fire extinguishing operations and deploy extinguishing agents into the fault location area; In the event of an open flame fire, initiate full-area firefighting operations. When there is a risk of reignition, initiate continuous suppression and extinguishing operations, continuously delivering extinguishing agents to the area at risk of reignition until the second change rate of the CO / CO2 ratio is less than or equal to zero, the slope of the temperature rise of local hot spots is less than or equal to zero, and the residual air ion concentration is below the ion concentration threshold.

10. An environmentally-aware cable shaft fire extinguishing system, applied to the environmentally-aware cable shaft fire safety monitoring method according to any one of claims 1 to 7, characterized in that, The system includes: an adaptive fire suppression control module and a fire suppression execution component. The adaptive inactivation control module is used to adaptively control the spray flow rate, spray angle, spray duration, and coverage of the fire extinguishing execution components based on the determined fire risk level and the corresponding fire extinguishing level, for zoned fire extinguishing of cable shaft fires; specifically: The adaptive fire suppression control module receives real-time information on fire risk level and fault location. The built-in hierarchical strategy database is retrieved, and based on the current fire risk level, a unique match is made with the preset corresponding fire extinguishing level to establish a one-to-one correspondence between fire risk level and fire extinguishing level. Based on the matched fire extinguishing level, the corresponding standard control parameter group is adaptively called. The control parameter group includes preset spray flow threshold, spray angle adjustment range, single and continuous spray duration, and spatial coverage parameters. Based on the fire location zone information, only the fire extinguishing execution components corresponding to the fire zone of the cable shaft are activated, while the fire extinguishing devices in non-fire zones are locked into a standby state. The fire suppression execution component includes fire suppression nozzles arranged in zones. Control commands are issued to the activated zone fire suppression execution components to adaptively adjust the real-time spray flow, adjust the spray angle to the fire area, and set the spray duration to match the fire scale, so as to complete the full coverage fire suppression operation of the fire area according to the preset coverage range.

Images