A control system for tunnel energy saving, fire protection and disaster reduction

Abstract

The present invention discloses a control system for energy-saving, firefighting and disaster reduction in tunnels, which relates to the field of intelligent firefighting technology. The control system comprises the following steps: scanning the electromagnetic environment characteristics in the tunnel to obtain position signal characteristic data; extracting time-frequency energy characteristics based on the position signal characteristic data, and using the Otsu threshold method for anomaly detection to identify abnormal temperature and smoke concentration conditions in the tunnel and generate an abnormal spatial distribution map; calculating the fire extinguishing path weight score using the Dijkstra algorithm, combining the PID control algorithm to determine the action parameters and execution sequence of the fire extinguishing equipment, and generating a fire extinguishing action instruction set; executing the fire extinguishing action instructions and monitoring them in real time, performing real-time monitoring data fusion analysis using the Kalman filter algorithm, generating a fire extinguishing effect score, and dynamically optimizing the fire extinguishing action instruction set. The present invention optimizes the fire extinguishing strategy in real time through Kalman filtering and reinforcement learning, adaptively adjusting the plan according to changes in the fire intensity, and improving fire extinguishing efficiency.

Description

Technical Field

[0001] The present invention relates to the field of intelligent fire protection technology, in particular to a tunnel energy-saving fire protection and disaster reduction control system. Background Art

[0002] Tunnel fire protection and disaster reduction systems are a crucial technical area for tunnel operational safety, encompassing fire monitoring, fire location, and path planning. Currently, most tunnel fire protection systems utilize temperature sensors and smoke detectors for fire monitoring, combined with pre-configured fire extinguishing devices for fire control. Jet fans are also installed in the ventilation system for ventilation, typically with an installed power of 100-200 kW / km for two-lane tunnels and 300-500 kW / km for three-lane tunnels. Furthermore, tunnel ventilation consumes a high amount of energy. Fire location is often determined using signal time difference of arrival (TDOA) algorithms or signal strength location technology. Dynamic control also employs fixed thresholds to trigger fire extinguishing devices, or employs feedback mechanisms to adjust extinguishing strategies. A relatively mature technical framework has been established in the field of tunnel fire protection, providing a fundamental guarantee for tunnel safety.

[0003] Existing technologies also have numerous shortcomings. For example, most fire monitoring methods rely on data from a single sensor, which is severely inadequate in dealing with signal interference in complex electromagnetic environments, leading to detection anomalies. Furthermore, path planning algorithms fail to account for the real-time spread of a fire, and static weight allocations struggle to adapt to real-time changes in the fire situation, potentially impacting the timeliness of firefighting routes. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a tunnel energy-saving fire-fighting and disaster reduction control system to solve the problem of insufficient real-time accurate positioning and dynamic regulation capabilities of fire situations in complex environments.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] The present invention provides a tunnel energy-saving fire-fighting and disaster-reduction control system, which includes a fire foam monitor intelligent fire extinguishing system, including a signal scanning module, an abnormality analysis module, a fire situation mapping module, a path planning module and a dynamic control module;

[0008] Signal scanning module, used to scan the electromagnetic environment characteristics in the tunnel and obtain position signal characteristic data;

[0009] The anomaly analysis module is used to extract time-frequency energy features based on the location signal feature data, and use the Otsu threshold method to perform anomaly detection, identify abnormal temperature and smoke concentration conditions in the tunnel, and generate an abnormal spatial distribution map;

[0010] The fire mapping module is used to identify fire areas based on the abnormal spatial distribution map using a time difference positioning algorithm, and analyze the direction and speed of fire spread in combination with the optical flow method to create a dynamic fire information map;

[0011] The path planning module is used to calculate the weight score of the fire extinguishing path through the Dijkstra algorithm, determine the action parameters and execution sequence of the fire extinguishing equipment in combination with the PID control algorithm, and generate the fire extinguishing action instruction set;

[0012] The dynamic control module executes fire-fighting action instructions and monitors in real time. It uses the Kalman filter algorithm to perform real-time monitoring data fusion analysis, generate fire-fighting effect scores, and dynamically optimize the fire-fighting action instruction set.

[0013] As a preferred solution of the tunnel energy-saving, fire-fighting and disaster-reduction control system of the present invention, the electromagnetic environment characteristics in the tunnel include electromagnetic field intensity characteristics, wireless signal propagation characteristics and electromagnetic noise spectrum;

[0014] The position signal feature data includes time domain feature data, frequency domain feature data, spatial feature data and high-order feature data.

[0015] As a preferred solution of the tunnel energy-saving, fire-fighting and disaster reduction control system of the present invention, wherein: the Otsu threshold method is used for anomaly detection to identify abnormal temperature and smoke concentration in the tunnel and generate an abnormal spatial distribution map. The specific steps are as follows:

[0016] Wavelet packet decomposition and principal component analysis are used to decompose the position signal feature data in the time-frequency domain and output a fused feature data set;

[0017] The Otsu threshold method is used to divide the fused feature dataset into spatial grids, and the anomaly detection results are obtained by comparing them with the data before the fire. The anomaly spatial distribution map with marked abnormal grid points is output.

[0018] As a preferred solution of the tunnel energy-saving, fire-fighting and disaster reduction control system of the present invention, the method comprises the following steps: based on the abnormal spatial distribution map, a time difference positioning algorithm is used to identify the area where the fire exists, and the optical flow method is combined to analyze the direction and speed of the fire spread to establish a dynamic fire information map.

[0019] Using the time difference positioning algorithm, the delay difference of the received signal of each antenna is measured, the coordinates of the fire source are obtained, and the area where the fire is located is output;

[0020] Using optical flow method and turbulence correction, the motion vector of the continuous temperature field is calculated and the direction and speed of fire development are predicted, and the fire spread direction and speed parameters are output;

[0021] Kalman filtering is used to integrate the fire source coordinates, fire spread direction and speed parameters to create a dynamic fire information map.

[0022] As a preferred solution of the tunnel energy-saving fire-fighting and disaster reduction control system of the present invention, wherein: the fire extinguishing path weight score is calculated by the Dijkstra algorithm, the action parameters and execution sequence of the fire extinguishing equipment are determined in combination with the PID control algorithm, and the fire extinguishing action instruction set is generated. The specific steps are as follows:

[0023] Based on the dynamic fire information map, the Dijkstra algorithm is used to calculate the fire extinguishing path weight score, and combined with the real-time fire direction factor, the optimal fire extinguishing path plan is output;

[0024] According to the fire extinguishing path plan, the action parameters and execution sequence of the fire extinguishing equipment are dynamically obtained, and the fire extinguishing action instruction set is output.

[0025] As a preferred solution of the tunnel energy-saving, fire-fighting and disaster reduction control system of the present invention, wherein: the execution of fire-fighting action instructions and real-time monitoring, the real-time monitoring data fusion analysis through the Kalman filter algorithm, the generation of fire-fighting effect scores and the dynamic optimization of the fire-fighting action instruction set are specifically performed as follows:

[0026] Using fuzzy integral fusion algorithm, combined with real-time fire monitoring data and fire extinguishing action instruction set, output comprehensive fire situation assessment report;

[0027] Based on the comprehensive fire situation assessment report, the duel deep Q network is used to calculate the optimal control strategy and output the optimized action instructions;

[0028] Convert optimized action instructions into equipment control signals, use reinforcement learning to collect execution effect data in real time, and generate fire extinguishing effect scores;

[0029] Based on the fire extinguishing effect score, the action parameters are continuously updated through online learning to optimize the fire extinguishing action instruction set.

[0030] As a preferred solution of the tunnel energy-saving, fire-fighting and disaster reduction control system of the present invention, wherein: the wavelet packet decomposition and principal component analysis are used to perform time-frequency domain decomposition on the position signal feature data and output a fusion feature data set. The specific steps are as follows:

[0031] The position signal feature data is decomposed into three layers of wavelet using wavelet packet decomposition, and the dimension is reduced using principal component analysis to output the time-frequency feature matrix after dimension reduction.

[0032] Weighted feature fusion is used to fuse the time-frequency feature matrix and the intensity difference value between nodes according to the optimized weights, and the fused feature dataset is output.

[0033] As a preferred solution of the tunnel energy-saving fire-fighting and disaster reduction control system of the present invention, the optical flow method and turbulence correction are used to calculate the motion vector of the continuous temperature field and predict the direction and speed of fire development, and output the fire spread direction and speed parameters. The specific steps are as follows:

[0034] The optical flow method is combined with turbulence correction to calculate the three-dimensional motion vector and output the fire scene motion vector field;

[0035] The front tracking algorithm is combined with the time series analysis method to deduce the fire spread trend based on the fire motion vector field, and output the fire spread direction and speed parameters.

[0036] As a preferred solution of the tunnel energy-saving fire protection and disaster reduction control system of the present invention, wherein: based on the dynamic fire information graph, the Dijkstra algorithm is used to calculate the fire extinguishing path weight score, and combined with the real-time fire direction factor, the optimal fire extinguishing path solution is output. The specific steps are as follows:

[0037] Based on the dynamic fire information graph, multiple fire extinguishing paths are initialized, the weight of each path is defined according to the dynamic fire information graph, and the initial fire extinguishing path set is output;

[0038] The Dijkstra algorithm is used to calculate the initial fire extinguishing path weight score, identify the optimal fire extinguishing path and execute it, and output the optimal fire extinguishing path plan.

[0039] As a preferred solution of the tunnel energy-saving, fire-fighting and disaster-reduction control system of the present invention, it includes:

[0040] The negative ion smoke suppression air purification system purifies the tunnel air by releasing high-concentration negative ions. It uses the principle of like charges repelling each other to prevent smoke from sinking quickly during a fire, maintaining visibility in the escape route. It also activates the piped smoke exhaust system in a timely manner to keep the tunnel space free of smoke and haze.

[0041] The zero-energy fire protection and antifreeze intelligent control system uses antifreeze and anticorrosion fire extinguishing fluid to replace electric heating. The antifreeze is stored in an epoxy fiberglass cylinder. The water flow sensor triggers the solenoid valve to switch the water supply, achieving zero-energy antifreeze.

[0042] The duct smoke exhaust system uses an axial flow fan to draw the smoke from the top into the exhaust duct and discharge it in the direction of traffic flow to reduce the smoke diffusion. It also cooperates with the negative ion system to purify the residual smoke to a safe concentration.

[0043] The disaster prevention linkage system is used to activate the broadcast and warning display screens inside and outside the tunnel, alerting drivers of a tunnel fire and the location of the fire source, reminding drivers to drive into the nearest emergency parking area, and triggering the simultaneous activation of the fire extinguishing, smoke suppression, and smoke exhaust systems.

[0044] The present invention has the following beneficial effects: By employing the Otsu threshold method and time-frequency energy feature extraction technology, it effectively distinguishes normal environments from abnormal fire signals, reducing the false alarm rate. It also uses a time-difference positioning algorithm and optical flow method to dynamically track the fire source location and spread trend, improving fire location accuracy. It also uses Kalman filtering and reinforcement learning to optimize fire extinguishing strategies in real time, adaptively adjusting the strategy based on changes in fire intensity, and improving fire extinguishing efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0046] Figure 1 Schematic diagram of the control system for energy conservation, fire protection and disaster reduction in tunnels.

[0047] Figure 2 This is a flowchart for multi-module collaborative processing.

[0048] Figure 3 Positioning analysis diagram for the fire mapping module.

[0049] Figure 4 Dynamically optimize the flow chart for fire extinguishing action instructions. DETAILED DESCRIPTION

[0050] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0051] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0052] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0053] Reference Figures 1 to 4 , is an embodiment of the present invention, which provides a tunnel energy-saving, fire-fighting and disaster-reduction control system including:

[0054] The negative ion smoke suppression air purification system purifies the tunnel air by releasing high-concentration negative ions. It uses the principle of like charges repelling each other to prevent smoke from sinking quickly during a fire, maintaining visibility in the escape route. It also activates the piped smoke exhaust system in a timely manner to keep the tunnel space free of smoke and haze.

[0055] The fire foam monitor intelligent fire extinguishing system is used to dynamically adjust the fire extinguishing strategy according to different fire intensity to extinguish the fire quickly and accurately;

[0056] The zero-energy fire protection and antifreeze intelligent control system uses antifreeze and anticorrosion fire extinguishing fluid to replace electric heating. The antifreeze is stored in an epoxy fiberglass cylinder. The water flow sensor triggers the solenoid valve to switch the water supply, achieving zero-energy antifreeze.

[0057] The duct smoke exhaust system uses an axial flow fan to draw the smoke from the top into the exhaust duct and discharge it in the direction of traffic flow to reduce the smoke diffusion. It also cooperates with the negative ion system to purify the residual smoke to a safe concentration.

[0058] The disaster prevention linkage system is used to activate the broadcast and warning display screens inside and outside the tunnel, alerting drivers of a tunnel fire and the location of the fire source, reminding drivers to drive into the nearest emergency parking area, and triggering the simultaneous activation of the fire extinguishing, smoke suppression, and smoke exhaust systems.

[0059] It should be noted that the negative ion smoke suppression air purification system, the zero-energy fire protection and antifreeze intelligent control system, the pipe smoke exhaust system, and the disaster prevention linkage system are all existing technologies and will not be elaborated on here.

[0060] Fire foam monitor intelligent fire extinguishing system includes:

[0061] The signal scanning module scans the electromagnetic environment characteristics in the tunnel and obtains position signal characteristic data.

[0062] Furthermore, a gradient descent method is used to scan the environmental benchmark and output the environmental reflection benchmark.

[0063] Specifically, programmable metasurface antennas are deployed in an equidistant grid, a laser rangefinder is used for position calibration, the programmable metasurface antenna control software is used to load the phase configuration, the scanning area range is set, the scanning step angle is configured, the measurement equipment benchmark is calibrated, the beam pointing is adjusted to transmit the scanning signal and receive the echo signals in all directions, the echo signal intensity and arrival time are recorded and the abnormal reflection feature points are marked, the reflection characteristics under normal environment are used to establish a standard reflection pattern library, an environmental reflection benchmark map is generated, and the environmental reflection benchmark is output.

[0064] The better approach is to directly optimize the antenna phase through the gradient descent method to improve the uniformity of signal coverage; and to use environmental reference scanning to improve the accuracy of capturing environmental features.

[0065] Based on the environmental reflection benchmark, the beam direction is dynamically adjusted, and adaptive frequency hopping is used to suppress interference and output time domain echo signals.

[0066] Specifically, according to the environmental reflection benchmark, the strong reflection area in the echo signal is identified and the coordinates are marked, the optimal beam pointing recommendation scheme under the current environment is generated, and the recommended beam scanning path scheme is output; the beam direction is set according to the recommended beam scanning path scheme, the environmental feedback signal is continuously received, the actual signal strength distribution is detected and compared with the environmental reflection benchmark, and when a sudden strong reflection is detected, the scanning density of the sudden strong reflection area is enhanced, the beam dwell time is adjusted, and the beam scanning trajectory log is output; a frequency hopping strategy is used to establish a list of available channels, and automatically switch to the optimal channel when interference is detected. A pseudo-random sequence is used to control the frequency hopping pattern to maintain time and frequency synchronization with the receiving end, and the interference events and processing measures are recorded; the reflected signal is received by the metasurface antenna array to record the complete waveform including the direct wave and multipath reflection, the received reflected signal is subjected to signal preprocessing and data encapsulation, and the time domain echo signal is output.

[0067] It should be noted that signal preprocessing includes bandpass filtering, automatic gain control and time-domain averaging noise reduction; data encapsulation includes adding timestamps and spatial coordinate tags, calculating signal characteristic parameters and generating data packets in a standard format; the reflected signal is passively received after the transmitted signal is reflected by the environment.

[0068] The better one is to improve the accuracy of scanning path planning through environmental reflection benchmark, adopt frequency hopping strategy to enhance the ability to identify interference sources and improve anti-interference ability; receive reflected signals through metasurface antenna array to improve the signal-to-noise ratio of echo signals and enhance the detection ability of weak reflected signals.

[0069] Density clustering is performed on the time domain echo signal, clutter is eliminated by combining R wall, target features are extracted, and the fire situation fusion feature vector is output.

[0070] Specifically, a 5th-order Butterworth bandpass filter is applied to filter the signal, remove high-frequency noise and low-frequency drift and retain the effective echo signal component, and the maximum amplitude of the signal is calculated by amplitude normalization. The maximum amplitude is used for linear normalization and the dynamic range of the signal is unified; the density clustering algorithm is used to extract the time domain features of the signal to construct a three-dimensional feature space and perform clustering, set the initial neighborhood radius and the minimum number of samples, adjust the parameters until a stable cluster is obtained and optimize the clustering results, merge similar clusters adjacent to the space, eliminate isolated noise points and mark effective signal clusters; random projection dimensionality reduction is used to cluster the clusters. The features are projected into low-dimensional space, the energy distribution of the projected signal is calculated, the R wall is combined to remove clutter, the energy threshold is set, the low-energy projection component is filtered out, the effective signal component is reconstructed, and the signal features after dimensionality reduction are output; the time-frequency features are used to extract the signal features after dimensionality reduction, the signal features after dimensionality reduction are extracted through short-time Fourier transform, and the energy proportion of each frequency band is extracted, the time domain statistical features are calculated, and at the same time, the spatial features are used based on the signal features after dimensionality reduction to extract the spatial distribution parameters and geometric features of the signal cluster, the time-frequency and spatial features are weightedly fused and normalized to generate a fire fusion feature vector.

[0071] It should be noted that the extracted signal time domain features include: peak value, zero crossing rate and energy; the energy threshold can be set to: mean + 3σ; time domain statistical features include skewness and kurtosis; geometric features include area, perimeter and compactness; parameters are adjusted until stable clustering is obtained. The adjustment parameters are the initial neighborhood radius and the minimum number of samples.

[0072] Combined with the feature vector, Kalman filtering is used to correct the error and output the position signal feature data.

[0073] Specifically, calculate the median value of all eigenvectors, adjust all values ​​to between -1 and 1, eliminate data points that exceed the reasonable range, and output the cleaned feature data; set the starting point of the position estimation and configure the allowable error range, send a test signal at a known position, and record the difference between the actual measurement value and the theoretical value; use Kalman filtering to obtain the latest eigenvector, record the precise arrival time and device identification data for fusion processing, dynamically optimize the observation weight according to the filtering residual, and maintain a smooth transition of the output; obtain the three-dimensional coordinates and moving speed of the eigenvector, encapsulate the data in a standard format, add a timestamp and location mark, and output the position signal feature data.

[0074] It should be noted that the static range definition method is used to define reasonable range data points, and the range comes from the physical range of the programmable metasurface antenna.

[0075] The anomaly analysis module extracts time-frequency energy features based on the position signal feature data, and uses the Otsu threshold method to perform anomaly detection, identify temperature anomalies and smoke concentration anomalies in the tunnel, and generate an anomaly spatial distribution map.

[0076] Wavelet packet decomposition and principal component analysis are used to decompose the position signal feature data in the time-frequency domain and output a fused feature data set.

[0077] Furthermore, wavelet packet decomposition is used to perform three-layer wavelet decomposition on the position signal feature data, and principal component analysis is used to reduce the dimension, and the time-frequency feature matrix after dimension reduction is output.

[0078] Specifically, wavelet packet decomposition and energy entropy maximization criteria are adopted, and the number of decomposition layers is defined as three. The first layer decomposes the position signal feature data into low-frequency approximation coefficients and high-frequency detail coefficients. The second layer recursively decomposes the low-frequency approximation coefficients and high-frequency detail coefficients respectively to generate four sub-bands. The third layer further decomposes the four sub-bands of the second layer, outputs 8 final sub-bands and obtains the energy entropy of each sub-band, and outputs a three-layer wavelet packet decomposition coefficient matrix; the three-layer wavelet packet decomposition coefficient matrix is ​​Z-value standardized by column to eliminate the dimension effect, and the covariance matrix is ​​generated through the energy correlation between the 8 final sub-bands. The main directions with the largest energy contribution are found through mathematical transformation, and the principal components covering more than 95% of the original signal energy are retained. At the same time, the noise is eliminated, the covariance matrix is ​​converted to the selected principal component space, and the time-frequency feature matrix after dimensionality reduction is output.

[0079] Weighted feature fusion is used to fuse the time-frequency feature matrix and the intensity difference value between nodes according to the optimized weights, and the fused feature dataset is output.

[0080] Specifically, the time-frequency feature matrix after dimensionality reduction is input, where each row in the matrix structure represents a time point and each column represents a principal component feature. The signal strength of adjacent nodes is subjected to absolute differences. The particle swarm optimization algorithm is used to initialize a set of possible weight combinations and calculate the fusion features. The performance of each weight combination is evaluated through iterative calculation, and the optimal weight combination is recorded after multiple iterations. The time-frequency feature matrix is ​​multiplied by the optimal weight combination to construct a weighted frequency feature matrix. The absolute difference is multiplied by the optimal weight combination to construct an enhanced feature matrix. The weighted frequency feature matrix and the enhanced feature matrix are added correspondingly at the time points, and the fusion feature data set is output.

[0081] It should be noted that the weight range, the sum of the time-frequency feature weight and the spatial feature weight are set to 1, and the sum of 1 is based on normalized mathematical constraints; the optimal weight combination is a weight combination that enables the fusion feature to most accurately represent the dynamic characteristics of the original fire signal through iterative calculation.

[0082] The Otsu threshold method is used to divide the fused feature data into spatial grids, and the anomaly detection results are obtained by comparing them with the data before the fire. The anomaly spatial distribution map with marked abnormal grid points is output.

[0083] Specifically, the number and arrangement of grids are determined according to the actual size of the tunnel, and a unique spatial coordinate identifier is assigned to each grid. Each grid cell independently stores two types of data: real-time monitoring data and benchmark data. The improved Otsu threshold algorithm is used for adaptive anomaly detection. The degree of deviation between the current eigenvalue and the benchmark value is compared with the historical fluctuation range of the grid to evaluate the anomaly probability and analyze the eigenvalue distribution of all grid cells. The eigenvalues ​​of the grid cells are divided into normal or abnormal categories, and the abnormal grid cells are marked. A three-dimensional grid set with abnormal labels is output; adjacent abnormal grids are clustered, and spatially continuous abnormal areas are identified and the geometric center of each abnormal area is located. The abnormal spatial distribution map of the marked abnormal grid points is output.

[0084] It should be noted that the benchmark data includes a standard reflection pattern library, fusion features, and absolute differences.

[0085] The better ones are to enhance the feature extraction capability of low-frequency smoke and high-frequency sudden fire through wavelet packet decomposition and principal component analysis dimensionality reduction, enhance the anti-interference capability through standard reflection pattern comparison, and use Otsu threshold method to improve detection accuracy and reduce errors.

[0086] The fire mapping module uses a time-difference positioning algorithm to identify areas where fires exist based on the abnormal spatial distribution map, and combines it with the optical flow method to analyze the direction and speed of fire spread to create a dynamic fire information map.

[0087] Furthermore, a time difference positioning algorithm is used to measure the time delay difference of the signals received by each antenna, obtain the coordinates of the fire source, and output the area where the fire exists.

[0088] Specifically, at least four ultra-wideband antenna nodes are arranged in the tunnel to receive electromagnetic signals generated by fire, filter the received electromagnetic signals, remove high-frequency interference from vehicle engines, lighting equipment, etc., retain the fire characteristic frequency band, and output the filtered electromagnetic signals received by each antenna; perform cross-correlation calculation on the electromagnetic signals received by every two ultra-wideband antennas to find the time difference of the electromagnetic signal arrival, select an ultra-wideband antenna as a reference node, obtain the time delay difference of other ultra-wideband antennas relative to the reference node, and output a set of time delay difference data, which represents the time difference of the electromagnetic signal arriving at different ultra-wideband antennas; through the delay difference data Based on the known fixed value of electromagnetic wave propagation speed, the time delay difference is converted into distance difference. A spatial coordinate system is established with the position of an ultra-wideband antenna node as the reference, and the coordinates are set to (0,0,0). The distance difference from the fire source to each ultra-wideband antenna forms a set of hyperboloids. The intersection of all the hyperboloids is the fire source position. An initial fire source position is assumed, and the theoretical distance difference between the assumed position and the ultra-wideband antenna is obtained. The error is calculated by comparing it with the actual measured distance difference. The assumed position is adjusted according to the error direction. The above process is repeated until the error is less than the set threshold (such as 0.1 meters). The coordinates of the fire source are obtained and the area where the fire exists is output.

[0089] It should be noted that electromagnetic signals include high temperature radiation signals and smoke scattering signals.

[0090] Using the optical flow method and turbulence correction, the motion vector of the continuous temperature field is calculated and the direction and speed of fire development are predicted, and the fire spread direction and speed parameters are output.

[0091] Furthermore, the optical flow method is combined with turbulence correction to calculate the three-dimensional motion vector and output the fire scene motion vector field.

[0092] Specifically, an infrared thermal imager and an ultrasonic wind speed sensor are used for synchronous data acquisition, and hardware trigger synchronization is used to align the time of temperature data and wind speed data. The infrared thermal imager extracts high-temperature areas through fixed temperature threshold segmentation, and the wind speed data is averaged and filtered through a sliding window; the classical optical flow method is used to analyze the infrared images generated by the infrared thermal imager of adjacent frames, and the pixel displacement is calculated by image gradient. The infrared image is divided into blocks, the movement direction of each block is calculated, and a two-dimensional plane motion vector map is constructed; the ultrasonic wind speed sensor data is combined to supplement the vertical movement information, and the movement intensity of each area is adjusted according to the temperature distribution to construct a preliminary three-dimensional motion field; multiple groups of wind speed sensors are used for cross-validation, the airflow direction is adjusted by ventilation equipment, the angle of the vent guide plate is adjusted, and the fire scene motion vector field is output.

[0093] It should be noted that the fixed temperature threshold may be set to 200°C.

[0094] The better approach is to use hardware-triggered synchronization technology to improve the time alignment accuracy of temperature data and wind speed data; enhance the signal-to-noise ratio of the original data through threshold segmentation and sliding filtering; improve the local accuracy of motion vector detection through block optical flow calculation method; and improve the integrity of vertical motion information through multi-sensor data fusion.

[0095] The front tracking algorithm is combined with the time series analysis method to deduce the fire spread trend based on the fire motion vector field, and output the fire spread direction and speed parameters.

[0096] Specifically, the high-temperature threshold segmentation method is used to scan the infrared image frame by frame, and the adjacent pixel groups with sudden temperature changes are marked. The discontinuous boundaries are connected by the region growing algorithm. The fire front boundary is identified by comparing the temperature difference of adjacent pixels, and a continuous fire front edge line is output; the nearest neighbor displacement matching method is used, and the maximum allowable displacement threshold is set, such as 5 pixels per frame. The fire front edge point of the current frame is matched with the position of the previous frame at the shortest distance, and abnormal matching points with excessive displacement are eliminated; the sliding window trend observation method and the dominant direction voting mechanism are used to count the moving directions of each trajectory point in the window, and the direction with the highest frequency of occurrence is selected as the main spread direction. The average moving distance in the window is calculated as the spread speed, and the fire spread direction (angle value) and speed parameters are output.

[0097] The better ones are to improve the tracking accuracy of the fire front movement trajectory and the accuracy of boundary detection through nearest neighbor displacement matching; improve the real-time judgment ability of the fire development trend through sliding window observation; and enhance the objectivity of the spread direction judgment through the dominant direction voting mechanism.

[0098] The Kalman filter is used to integrate the fire source coordinates with the fire spread direction and speed parameters to create a dynamic fire information map.

[0099] Specifically, it receives the coordinates of the fire source and the fire spread direction and speed parameters, initializes the Kalman filter state vector, process noise covariance matrix and observation noise covariance matrix, and outputs the initialized Kalman filter; integrates the wind speed and direction sensor data in the tunnel, refers to similar scenarios in the historical fire case library, obtains the real-time ventilation system operation parameters, queries similar scenarios in the historical fire case library to match the current fire characteristics, applies the compensation coefficient to adjust the prediction results, and outputs the corrected spread direction and adjusted spread speed; sets the maximum spread speed according to the combustion characteristics of the material, considers the obstacle distribution to correct the prediction path, evaluates the possible impact range at different time points, applies fuzzy logic to deal with uncertainty, outputs the fire coverage prediction map for the next 30 seconds and the hazard level distribution heat map, and establishes a dynamic fire information map.

[0100] It should be noted that the state vector includes the fire source location, fire spread direction and speed parameters; the process noise covariance matrix is ​​constructed based on the fire source location prediction error, fire spread direction estimation deviation and speed parameters; the observation noise covariance matrix is ​​set according to the error characteristics of each measuring device, and is constructed through the observation noise variance of the temperature sensor, the noise variance of the smoke concentration sensor and other observation parameters; the compensation coefficient is composed of environmental component, structural component and material component, which can be set to an environmental component weight of 0.5, a structural component weight of 0.3, and a material component weight of 0.2. The sum of the weights is 1 based on the normalized mathematical constraint.

[0101] The better approach is to improve the sensitivity of fire feature recognition and enhance the ability to suppress interference signals through ultra-wideband signal capture and filtering technology; to improve the reliability of predictions for the next 30 seconds and optimize the rationality of regional division through the combination of rule base and fuzzy logic.

[0102] The path planning module calculates the weight score of the fire extinguishing path through the Dijkstra algorithm, combines it with the PID control algorithm to determine the action parameters and execution sequence of the fire extinguishing equipment, and generates a fire extinguishing action instruction set.

[0103] Based on the dynamic fire information map, the Dijkstra algorithm is used to calculate the fire extinguishing path weight score, and combined with the real-time fire direction factor, the optimal fire extinguishing path plan is output.

[0104] Furthermore, based on the dynamic fire information graph, multiple fire extinguishing paths are initialized, the weight of each path is defined according to the dynamic fire information graph, and the initial fire extinguishing path set is output.

[0105] Specifically, based on a dynamic fire information map, the three-dimensional tunnel space is divided into grid cells. Each grid stores temperature field data, smoke concentration, structural obstacles, and electromagnetic field strength. Starting from the location of firefighting equipment, multiple candidate paths are generated and dynamic weights are assigned to each grid node of each path. The initial fire extinguishing path set is output.

[0106] It should be noted that the dynamic weight formula is:

[0107] W = L + α × t;

[0108] Where W is the comprehensive node weight, W ≥ 0, the smaller the value, the higher the priority, and when W > 8, it is determined to be an impassable path; L represents the path length factor, 0 ≤ L ≤ 1; α represents the threat adjustment coefficient, 0.2 ≤ α ≤ 1, which is PID dynamic control; t is the comprehensive threat value, 0 ≤ t ≤ 10, t < 3 indicates low risk, 3 ≤ t ≤ 7 indicates medium risk, and t > 7 indicates high risk.

[0109] The Dijkstra algorithm is used to calculate the weight score of the fire extinguishing path, identify the optimal fire extinguishing path and execute it, and output the optimal fire extinguishing path plan.

[0110] Specifically, based on the initial fire extinguishing path set and the Dijkstra algorithm, the dynamic weight values ​​of all initial fire extinguishing paths are scanned and normalized to eliminate dimensional differences, and a weight matrix is ​​output; arbitrarily select a node, starting from the current node, select the node with the smallest dynamic weight value among the adjacent nodes, and start the iterative search from the previous node, each time selecting the adjacent node with the lowest weight, and output the local optimal path; scan all node weights of the local optimal path, and mark the inaccessible path; based on the fire spread prediction of the dynamic fire information graph and the action parameters of the fire extinguishing equipment, verify the local optimal path; combine the dynamic weight value and the optimal path to output the optimal fire extinguishing path plan.

[0111] It should be noted that the execution verification includes spatial safety verification, equipment accessibility verification and dynamic weight verification. Spatial safety verification is to eliminate nodes located in the fire prediction coverage area or structural damaged area. Equipment accessibility verification is to retain nodes within the range of fire extinguishing equipment. Dynamic weight verification is to exclude fire extinguishing paths with dynamic weights greater than 8.

[0112] According to the fire extinguishing path plan, the action parameters and execution sequence of the fire extinguishing equipment are dynamically obtained, and the fire extinguishing action instruction set is output.

[0113] Specifically, the straight-line distance from the fire source to the nearest fire foam monitor is used to select sprinklers within the coverage radius; the fire intensity is matched based on the dynamic fire information map, and the output of the fire foam monitor is output; combined with hard-coded logic and priority queues, the fire extinguishing action instruction set is output.

[0114] It should be noted that the fire level is divided into three levels: small, medium and large; the fire level output of the fire foam monitor can be defined as a small fire with a primary output for 30 seconds, a medium fire with a medium output for 60 seconds, and a large fire with a high output for 90 seconds; the hard-coded logic is to generate standardized control instructions based on the device physical interface protocol; the priority queue is defined to block the fire spread path first and then cover the core fire source.

[0115] The better one is to improve the accuracy of fire-fighting path planning and enhance path adaptability through the Dijkstra algorithm; and optimize risk avoidance ability through dynamic weight allocation.

[0116] The dynamic control module executes fire-fighting action instructions and monitors in real time. It uses the Kalman filter algorithm to perform real-time monitoring data fusion analysis, generate fire-fighting effect scores, and dynamically optimize the fire-fighting action instruction set.

[0117] A fuzzy integral fusion algorithm is used to combine real-time fire monitoring data with fire extinguishing action instruction sets to output a comprehensive fire situation assessment report.

[0118] Specifically, the optimal fire extinguishing plan is implemented. Real-time fire monitoring data and fire extinguishing equipment status data are input. The data of each sensor is normalized to the interval [0, 1] according to the maximum and minimum values. For the fire monitoring data of each sensor that are not synchronized in time, the timestamps are aligned using the interpolation method. The standardized data matrix is ​​output, where each row represents a sensor and each column represents the value at a time point. The importance weight of each sensor is defined. If the data of two sensors are abnormal at the same time, the overall weight is increased (for example, by 1.2 times). If only a single sensor is abnormal, the weight remains unchanged. The weight value of each sensor combination is output. All sensor data at the current time point are sorted from small to large in value. The difference between adjacent data is calculated and multiplied by the weight of the corresponding sensor combination. The result of all differences multiplied by the weight is accumulated to obtain the final fusion value. The fire intensity index at the current time point is output and the fire situation is divided. The rate of change of the fusion value is combined to determine whether the fire is spreading rapidly. The coverage rate is compared with the current fire extinguishing equipment status and the expected effect, and a comprehensive fire situation assessment report is output.

[0119] It should be noted that the importance weight can be defined as a temperature weight of 0.7, a smoke weight of 0.6, and a CO2 weight of 0.4; the fire situation classification can be defined as 0-0.3 as green risk, 0.3-0.6 as yellow risk, and above 0.6 as red risk; the fire intensity can be judged as 0.5→0.7→0.9, indicating rapid spread; the coverage rate can be as follows: the sprinkler covers 80% of the fire source area; the real-time fire monitoring data is obtained from the collaborative monitoring of sensors; the fire extinguishing equipment status data is a set of real-time quantitative parameters, including fire extinguishing equipment action parameters, energy and resource status, faults and anomalies.

[0120] Based on the comprehensive fire situation assessment report, the duel deep Q network is used to calculate the optimal control strategy and output optimized action instructions.

[0121] Specifically, the fire temperature is divided into three levels: low, medium, and high, and the smoke concentration is divided into three levels: light, medium, and heavy. At the same time, the status of the fire extinguishing equipment is recorded and a fire status table is output; high temperature priority rules, smoke blocking escape path rules, and resource conservation rules are defined; high temperature priority rules, smoke blocking escape path rules, and resource conservation rules are defined in order of priority. When the conditions are fully matched, the corresponding actions are immediately generated and only the highest priority actions are executed at the same time, and the device action instructions are output; the execution effect is monitored in real time, and the duel deep Q network is used for real-time analysis to check whether the target fire area has cooled down and whether the smoke concentration has decreased. If the expected effect is not achieved, the rule matching is triggered again and the optimized action instructions are output;

[0122] It should be noted that the high-temperature priority rule is defined as follows: when a high level of fire temperature appears in the fire area and the fire-fighting equipment in the area is available, the fire-fighting equipment is started and the output is set to the highest level; the smoke-blocking escape path rule is that when heavy smoke concentration appears in the main channel area and the fan is available, the fan is started and the speed is set to 70%; the resource-saving rule is that when low-level fire temperatures appear in multiple areas and the energy of the fire-fighting equipment is less than 50%, it is only used in the upwind area of ​​the fire source, and the output is adjusted to medium; the fire-fighting equipment is the fire foam cannon; the priority order is high temperature greater than smoke greater than resources.

[0123] Convert optimized action instructions into equipment control signals, use reinforcement learning to collect execution effect data in real time, and generate fire extinguishing effect scores.

[0124] Specifically, the optimization action instruction is received and sent to the equipment controller. When the equipment executes the optimization action instruction, the temperature change of the target area, the smoke concentration change of the target area and the energy consumption of the fire-fighting equipment are read in real time, and a status comparison table is generated for the fire-fighting effect status before and after the execution of the optimization action instruction and the energy consumption status of the fire-fighting equipment; according to the status comparison table, the predefined scoring rules are matched, the effect score is calculated and divided, and the fire-fighting effect score is generated.

[0125] It should be noted that the fire extinguishing effect scoring formula is:

[0126] S=a·△T+b·△C+c·E

[0127] Among them, S represents the fire extinguishing effect score, and a higher score indicates a better fire extinguishing strategy; △T is the temperature difference before and after fire extinguishing in the target area, reflecting the cooling effect; △C represents the change in smoke concentration in the target area, reflecting the smoke removal effect; E is the resource utilization rate of the fire extinguishing equipment, and a higher value indicates better energy consumption control; a represents the fire suppression coefficient, reflecting the direct fire control capability; b is the smoke exhaust efficiency coefficient, reflecting the efficiency of smoke removal after visibility restoration; c represents the energy-saving optimization coefficient, reflecting the resource utilization rate and sustainable fire extinguishing capability; the three coefficients a, b, and c are adjusted according to the actual scenario, a>b>c and a+b+c=1, and can be set to a=0.5, b=0.3, and c=0.2.

[0128] Based on the fire extinguishing effect score, the action parameters are continuously updated through online learning to optimize the fire extinguishing action instruction set.

[0129] Based on the fire extinguishing effect score and combined with the current environmental status (fire temperature, smoke concentration, equipment status), the action space is defined, and the fire extinguishing effect score data is stored using experience playback. The optimal action is selected based on the current status, and continuous optimization action instructions are output. The continuously optimized action instructions are sent to the fire extinguishing equipment, and temperature and smoke data are collected in real time. The new round of fire extinguishing effect scores are calculated, and the changes in the fire extinguishing scores are compared to judge the optimization effect. If the fire extinguishing score increases, the current parameters will be maintained and monitored. If the score decreases, a new round of optimization will be triggered to optimize the fire extinguishing action instruction set.

[0130] It should be noted that the action space includes adjusting the angle of the fire foam monitor, adjusting the energy output of the fire foam monitor and switching the equipment priority to block the fire spread path.

[0131] The better one is to improve the data integration capability through fuzzy integral fusion, enhance the comprehensiveness and accuracy of fire situation assessment; through reinforcement learning and deep Q network, optimize the real-time decision-making ability of fire extinguishing strategy and enhance the adaptive adjustment capability.

[0132] In summary, the present invention employs the Otsu threshold method and time-frequency energy feature extraction technology to effectively distinguish between normal environments and abnormal fire signals, reducing false alarm rates. It also utilizes a time-difference positioning algorithm and optical flow method to dynamically track the fire source location and spread, improving fire location accuracy. Furthermore, it utilizes Kalman filtering and reinforcement learning to optimize fire extinguishing strategies in real time, adaptively adjusting the strategy based on changes in fire intensity to improve fire extinguishing efficiency.

[0133] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A tunnel energy-saving, fire-fighting and disaster-reduction control system, characterized by: include, The fire foam monitor intelligent fire extinguishing system includes a signal scanning module, anomaly analysis module, fire situation mapping module, path planning module, and dynamic control module; Signal scanning module, used to scan the electromagnetic environment characteristics in the tunnel and obtain position signal characteristic data; The anomaly analysis module is used to extract time-frequency energy features based on the location signal feature data, and use the Otsu threshold method to perform anomaly detection, identify abnormal temperature and smoke concentration conditions in the tunnel, and generate an abnormal spatial distribution map; The fire mapping module is used to identify fire areas based on the abnormal spatial distribution map using a time difference positioning algorithm, and analyze the direction and speed of fire spread in combination with the optical flow method to create a dynamic fire information map; The path planning module is used to calculate the weight score of the fire extinguishing path through the Dijkstra algorithm, determine the action parameters and execution sequence of the fire extinguishing equipment in combination with the PID control algorithm, and generate the fire extinguishing action instruction set; The dynamic control module executes fire-fighting action instructions and monitors in real time. It uses the Kalman filter algorithm to perform real-time monitoring data fusion analysis, generate fire-fighting effect scores, and dynamically optimize the fire-fighting action instruction set. The electromagnetic environment characteristics in the tunnel include electromagnetic field strength characteristics, wireless signal propagation characteristics and electromagnetic noise spectrum; The position signal feature data includes time domain feature data, frequency domain feature data, spatial feature data and high-order feature data; The Otsu threshold method is used to detect anomalies, identify abnormal temperature and smoke concentration in the tunnel, and generate an abnormal spatial distribution map. The specific steps are as follows: Wavelet packet decomposition and principal component analysis are used to decompose the position signal feature data in the time-frequency domain and output a fused feature data set; The Otsu threshold method is used to divide the fused feature dataset into spatial grids and compare it with the data before the fire to obtain anomaly detection results. The anomaly spatial distribution map with abnormal grid points marked is output. Based on the abnormal spatial distribution map, the time difference positioning algorithm is used to identify the area where the fire exists, and the optical flow method is combined to analyze the direction and speed of fire spread to establish a dynamic fire information map. The specific steps are as follows: Using the time difference positioning algorithm, the delay difference of the received signal of each antenna is measured, the coordinates of the fire source are obtained, and the area where the fire is located is output; Using optical flow method and turbulence correction, the motion vector of the continuous temperature field is calculated and the direction and speed of fire development are predicted, and the fire spread direction and speed parameters are output; Use Kalman filtering to integrate fire source coordinates, fire spread direction and speed parameters to create a dynamic fire information map; The fire extinguishing path weight score is calculated by the Dijkstra algorithm, and the action parameters and execution sequence of the fire extinguishing equipment are determined by combining the PID control algorithm to generate the fire extinguishing action instruction set. The specific steps are as follows: Based on the dynamic fire information map, the Dijkstra algorithm is used to calculate the fire extinguishing path weight score, and combined with the real-time fire direction factor, the optimal fire extinguishing path plan is output; Dynamically obtain the action parameters and execution sequence of the fire extinguishing equipment according to the fire extinguishing path plan, and output the fire extinguishing action instruction set; The execution of fire extinguishing action instructions and real-time monitoring, the real-time monitoring data fusion analysis through the Kalman filter algorithm, the generation of fire extinguishing effect scores and the dynamic optimization of the fire extinguishing action instruction set are as follows: Using fuzzy integral fusion algorithm, combined with real-time fire monitoring data and fire extinguishing action instruction set, output comprehensive fire situation assessment report; Based on the comprehensive fire situation assessment report, the duel deep Q network is used to calculate the optimal control strategy and output the optimized action instructions; Convert optimized action instructions into equipment control signals, use reinforcement learning to collect execution effect data in real time, and generate fire extinguishing effect scores; Based on the fire extinguishing effect score, the action parameters are continuously updated through online learning to optimize the fire extinguishing action instruction set.

2. The tunnel energy-saving, fire-fighting and disaster-reduction control system according to claim 1, characterized in that: The wavelet packet decomposition and principal component analysis are used to decompose the position signal feature data in the time-frequency domain and output the fusion feature data set. The specific steps are as follows: The position signal feature data is decomposed into three layers of wavelet using wavelet packet decomposition, and the dimension is reduced using principal component analysis to output the time-frequency feature matrix after dimension reduction. Weighted feature fusion is used to fuse the time-frequency feature matrix and the intensity difference value between nodes according to the optimized weights, and the fused feature dataset is output.

3. The tunnel energy-saving, fire-fighting and disaster-reduction control system according to claim 2, characterized in that: The optical flow method and turbulence correction are used to calculate the motion vector of the continuous temperature field and predict the direction and speed of fire development, and output the fire spread direction and speed parameters. The specific steps are as follows: The optical flow method is combined with turbulence correction to calculate the three-dimensional motion vector and output the fire scene motion vector field; The front tracking algorithm is combined with the time series analysis method to deduce the fire spread trend based on the fire motion vector field, and output the fire spread direction and speed parameters.

4. The tunnel energy-saving, fire-fighting and disaster-reduction control system according to claim 3, characterized in that: Based on the dynamic fire information graph, the Dijkstra algorithm is used to calculate the fire extinguishing path weight score, and combined with the real-time fire direction factor, the optimal fire extinguishing path solution is output. The specific steps are as follows: Based on the dynamic fire information graph, multiple fire extinguishing paths are initialized, the weight of each path is defined according to the dynamic fire information graph, and the initial fire extinguishing path set is output; The Dijkstra algorithm is used to calculate the initial fire extinguishing path weight score, identify the optimal fire extinguishing path and execute it, and output the optimal fire extinguishing path plan.

5. The tunnel energy-saving, fire-fighting and disaster-reduction control system according to claim 4, characterized in that: include, The negative ion smoke suppression air purification system purifies the tunnel air by releasing high-concentration negative ions. It uses the principle of like charges repelling each other to prevent smoke from sinking quickly during a fire, maintaining visibility in the escape route. It also activates the piped smoke exhaust system in a timely manner to keep the tunnel space free of smoke and haze. The zero-energy fire protection and antifreeze intelligent control system uses antifreeze and anticorrosion fire extinguishing fluid to replace electric heating. The antifreeze is stored in an epoxy fiberglass cylinder. The water flow sensor triggers the solenoid valve to switch the water supply, achieving zero-energy antifreeze. The duct smoke exhaust system uses an axial flow fan to draw the smoke from the top into the exhaust duct and discharge it in the direction of traffic flow to reduce the smoke diffusion. It also cooperates with the negative ion system to purify the residual smoke to a safe concentration. The disaster prevention linkage system is used to activate the broadcast and warning display screens inside and outside the tunnel, alerting drivers of a tunnel fire and the location of the fire source, reminding drivers to drive into the nearest emergency parking area, and triggering the simultaneous activation of the fire extinguishing, smoke suppression, and smoke exhaust systems.

Images