Liquefied hydrocarbon tank field fire identification system

Abstract

The invention relates to a liquefied hydrocarbon tank field fire identification system, which comprises a video data acquisition module, a detector data acquisition module, an optical fiber sensor data acquisition module, a data processing module and an intelligent judgment module, the video data acquisition module, the detector data acquisition module and the optical fiber sensor data acquisition module transmit respectively acquired data to the data processing module for processing, the data processing module is connected with the intelligent judgment module, and the intelligent judgment module performs fire identification according to the data processed by the data processing module. A multi-sensor information fusion technology and a neural network are combined, multiple factors influencing fire occurrence of the liquefied hydrocarbon tank field are comprehensively analyzed, multi-level parameters of multiple sensors serve as input of the neural network, different levels of fire occurrence serve as output of the neural network, and an optimal model of the neural network is obtained by training the neural network. And the accuracy of identifying and judging the fire in the liquefied hydrocarbon tank field is greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of fire alarm, and in particular to a fire recognition system for a liquefied hydrocarbon tank farm. Background Art

[0002] With the development of the petrochemical industry, the number of storage tanks in the tank farm is increasing, the scale is getting larger, and fire accidents in the tank farm occur frequently. The pressure type liquefied hydrocarbon storage tank is a storage tank used by petrochemical enterprises to store liquefied hydrocarbon hazardous chemical media. During the storage process, once the liquefied hydrocarbon leaks, it is easy to form an explosive vapor cloud, and when encountering a fire source, it will cause large-scale combustion or explosion. Its explosion power is very strong, the damage range is wide, the losses are huge, and the social impact is strong. Moreover, after a fire occurs in a large-scale tank farm, it is relatively difficult to fight the fire, and a chain reaction is likely to occur, resulting in the expansion of the accident; when a fire occurs in a liquefied hydrocarbon tank group, the combustion of hydrocarbons will release strong heat radiation. Under the action of strong heat radiation, the adjacent storage tanks will have their strength reduced, the medium in the tanks will volatilize, the pressure in the tanks will increase, the tank walls will be torn, and the liquid in the tanks will leak, further expanding the fire.

[0003] In the prior art, the conventional fire detection method for a liquefied hydrocarbon tank farm is mostly to use a temperature-sensitive optical fiber sensor to wind around the outer wall of the tank for temperature detection or use a flame detector to detect areas such as valves where leakage is likely to occur for fire detection. For example, a Chinese invention patent application with an application number of 202210406083.9 (publication number CN114821947A) discloses a petrochemical tank farm automatic fire alarm system, which includes an alarm device, a video acquisition module, an outdoor smoke sensor, a wind direction and wind speed acquisition module, and a central control module. The central control module is respectively connected to the alarm device, the video acquisition module, the wind direction and wind speed acquisition module, and the outdoor smoke sensor to determine whether a fire occurs and predict the flame direction; the video acquisition module includes multiple cameras, and each camera is provided with a brightness sensor and a ranging sensor. The above invention focuses on video acquisition of abnormal brightness points. The central control module analyzes the video information of the abnormal brightness points collected, obtains the flashing frequency of the abnormal brightness points, and determines whether it is a fire source according to the flashing frequency. By collecting and analyzing the abnormal brightness, the fire source information can be obtained while the fire source is formed. Another Chinese invention patent application with an application number of CN202210572815.1 (publication number CN114926950A) discloses a tank farm fire automatic alarm device. The tank farm fire automatic alarm device includes: a tank body, which is installed in the factory tank farm, and a plurality of groups of optical fiber detection probes are evenly installed on the outer surface of the tank body. At the same time, adjacent two groups of optical fiber detection probes are connected by optical fibers, and an optical fiber junction box is installed on the outer side of the tank body. The optical fiber junction box is electrically connected to an optical fiber connection box; the optical fiber connection box is electrically connected to an optical fiber signal processor arranged inside the factory area central control room.

[0004] The above-mentioned flame detectors and temperature-sensitive optical fiber sensors each detect only a certain characteristic of a fire for alarm, and cannot comprehensively judge the fire signal, so there is a high false alarm rate. Even when using the logical determination method of multiple sensors, the traditional fire detection algorithm mainly adopts the threshold method, that is, an alarm threshold is set for each individual sensor, and then the fire is judged through "AND" and "OR" logics. However, due to the many uncertainties in the formation and development process of a fire, simple logical judgment alone cannot meet the complex fire scene conditions, which is also one of the reasons for the high false alarm rate of the current composite fire detectors. Secondly, in the existing technology, the occurrence of a fire cannot be determined and warned in a timely manner. Most of them confirm the fire after it has started, and often the fire is already difficult to control. Traditional flame detectors, temperature detectors, image fire detectors, etc. can only detect after there is an open fire, which is not conducive to the timely identification and control of liquefied hydrocarbon fires. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a liquefied hydrocarbon tank farm fire identification system with a high correct rate in view of the current situation of the prior art.

[0006] The technical solution adopted by the present invention to solve the above technical problems is as follows: The liquefied hydrocarbon tank farm fire identification system is characterized in that it includes a camera, a flame detector, a combustible gas detector, a temperature-sensitive optical fiber sensor, a vibration optical fiber sensor, a pressure fiber Bragg grating sensor, a video data acquisition module, a detector data acquisition module, an optical fiber sensor data acquisition module, a data processing module, and an intelligent determination module; the video data acquisition module is connected to the camera and acquires the video image data captured by the camera; the detector data acquisition module is connected to the flame detector and the combustible gas detector and acquires the data of the flame detector and the combustible gas detector; the optical fiber sensor data module acquires the temperature data of the temperature-sensitive optical fiber sensor, the vibration data of the vibration optical fiber sensor, and the pressure data of the pressure fiber Bragg grating sensor. The video data acquisition module, the detector data acquisition module, and the optical fiber sensor data acquisition module are all connected to the data processing module and transmit the data collected by each of them to the data processing module for processing. The data processing module is connected to the intelligent determination module, and the intelligent determination module performs fire identification according to the data processed by the data processing module.

[0007] Furthermore, the temperature-sensitive fiber optic sensors are provided on the outer wall of the liquefied hydrocarbon tank and the valves in the liquefied hydrocarbon tank area; the vibration fiber optic sensors are provided at the pipe joints and other leakage-prone areas such as valves in the liquefied hydrocarbon tank area; the pressure fiber optic grating sensors are pasted on the outer wall of the liquefied hydrocarbon tank in the liquefied hydrocarbon tank area; the flame detectors are provided at the pipe joints and other leakage-prone areas such as valves in the liquefied hydrocarbon tank area; the combustible gas detectors are provided at the top and bottom of the liquefied hydrocarbon tank. Fire early warning can be carried out before the liquefied hydrocarbon leaks but does not catch fire based on the detection data of the combustible gas detector, the temperature-sensitive fiber optic sensor and the vibration fiber optic sensor.

[0008] Furthermore, the fire recognition system for the liquefied hydrocarbon tank area further includes a display output module connected to the data processing module, and the display output module displays the data processed by the data processing module in real time through the client. Displaying the data on the client facilitates the user to view at any time.

[0009] Furthermore, the fire recognition system for the liquefied hydrocarbon tank area further includes an alarm output module connected to the intelligent determination module, and the alarm output module outputs the system determination result through the communication protocol interface and / or the dry contact interface.

[0010] Furthermore, the alarm output module further includes an alarm device, and the alarm device is an audible and visual alarm.

[0011] Furthermore, a camera at a certain position is set to: when the temperature of the temperature-sensitive fiber optic sensor near the camera is too high and triggers a high-temperature alarm, the camera collects video signals of the high-temperature alarm area. The high-temperature alarm at this time may be a fire early warning for the outer perimeter of the tank group or a fire early warning for the tank body; if flame information is collected, a fire alarm for the outer perimeter of the tank group or a fire alarm for the tank body is issued.

[0012] Furthermore, the video data acquisition module performs intelligent analysis on the video images in the camera for parameters such as smoke, flame and brightness. When abnormal flame, smoke and brightness are identified, the abnormal information collected is transmitted to the intelligent determination module. The intelligent determination module establishes a multi-parameter model of the tank body state. In the model, intelligent model analysis is carried out on the tank body temperature, flame, brightness and smoke to obtain the abnormal coefficient of the tank body state. The abnormal coefficient of the tank body state can judge the degree of the fire occurrence and provide help for further rescue.

[0013] Furthermore, temperature-sensitive fiber optic sensors, vibration fiber optic sensors, and combustible gas detectors are installed in the areas of the liquefied hydrocarbon tank that are prone to leakage. When the temperature-sensitive fiber optic sensors, vibration fiber optic sensors, and combustible gas detectors trigger a leakage alarm, the nearby flame detectors are set to: rotate to the leakage alarm area to collect flame signals. The temperature-sensitive fiber optic sensors, vibration fiber optic sensors, and combustible gas detectors here are used to detect the leakage of liquefied hydrocarbons, and the flame detectors are used to detect whether there is an open fire near the leakage location.

[0014] In order to accurately identify the fire that occurs after leakage, the flame detectors are installed at the places where leakage is likely to occur, such as safety valves and tank root valves.

[0015] Furthermore, the intelligent decision-making module has a neural network model. The input values of the input layer of the neural network model are: the detection values of the temperature-sensitive fiber optic sensors, the detection values of the vibration fiber optic sensors, the detection values of the pressure fiber Bragg grating sensors, the detection values of the flame detectors, the alarm values of the combustible gas detectors, and the video image output values; the output values of the output layer are: non-fire, fire warning for the periphery of the tank group, fire alarm for the periphery of the tank group, fire warning for the tank body, fire alarm for the tank body, fire warning for the leakage fire of the tank group, and fire alarm for the leakage fire of the tank group.

[0016] Compared with the prior art, the advantages of the present invention are as follows: By combining the multi-sensor information fusion technology with the intelligent decision-making module, various factors affecting the occurrence of fires in the liquefied hydrocarbon tank area are comprehensively analyzed. Using the multi-level parameters of multiple sensors as the input of the intelligent decision-making module greatly improves the accuracy of fire identification and determination in the liquefied hydrocarbon tank area;

[0017] In the improved solution, the neural network model established by the intelligent decision-making module of the present invention takes the detection values of the temperature-sensitive fiber optic sensors, the detection values of the vibration fiber optic sensors, the detection values of the pressure fiber Bragg grating sensors, the detection values of the flame detectors, the alarm values of the combustible gas detectors, and the video image output values as input values, comprehensively considering the factors affecting fires; taking the different levels of fire occurrence as the output values of the neural network, the output values include 7 types, namely non-fire, fire warning for the periphery of the tank group, fire alarm for the periphery of the tank group, fire warning for the tank body, fire alarm for the tank body, fire warning for the leakage fire of the tank group, and fire alarm for the leakage fire of the tank group. Compared with the determination of the entire tank area by a single type, it improves the timeliness and reliability of fire determination, and reduces the situation of false fire alarms and missed fire alarms;

[0018] In the improved solution, the present invention improves the timeliness of fire prediction. In the past, alarms were often given after the appearance of a flame or the fire burned to a specific temperature, which often delayed the effective timing of fire alarms or fire control. The present invention improves the timeliness of fire identification by predicting leakage in advance, dynamically monitoring multiple parameters such as temperature and images, including timely detection and warning in various stages such as before the fire occurs, at the initial stage of the fire, and during the outbreak of the fire;

[0019] In the improved solution, the present invention classifies and identifies the areas and types of fire occurrences in the liquefied hydrocarbon tank farm, which is not only targeted but also provides different alarm levels, thus reflecting the safety situation of the fire in the liquefied hydrocarbon tank farm in a more efficient and clearer manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic diagram of the composition of the fire identification system for the liquefied hydrocarbon tank farm in the embodiment of the present invention;

[0021] Figure 2 It is a flowchart of the fire identification system for the liquefied hydrocarbon tank farm in the embodiment of the present invention to identify a fire;

[0022] Figure 3 It is a schematic diagram of the neural network model in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] The present invention will be further described in detail below in conjunction with the embodiments with reference to the drawings.

[0024] Embodiment 1

[0025] As Figure 1 shown is a preferred embodiment of the fire identification system for the liquefied hydrocarbon tank farm of the present invention. The fire identification system for the liquefied hydrocarbon tank farm includes: a camera, a flame detector, a combustible gas detector, a temperature sensing optical fiber sensor, a vibration optical fiber sensor, a pressure fiber Bragg grating sensor, a video data acquisition module, a detector data acquisition module, an optical fiber sensor data acquisition module, a data processing module, an intelligent determination module, a display output module, and an alarm output module. The video data acquisition module acquires the video image data of the camera; the detector data acquisition module acquires the data of the flame detector and the combustible gas detector; the optical fiber sensor data module acquires the temperature data of the temperature sensing optical fiber sensor, the vibration data of the vibration optical fiber sensor, and the pressure data of the pressure fiber Bragg grating sensor; the video data acquisition module, the detector data acquisition module, and the optical fiber sensor data acquisition module establish a multi-sensor information database with the data they respectively acquire, and at the same time perform preliminary data processing. For the data processed by the data processing module, one is transmitted to the intelligent determination module for determination to identify a fire; the second is transmitted to the display output module, and then the processed data is displayed in real time through the client; the third is that when the data processed by the data processing module is within the risk value range, it is transmitted to the alarm output module to prompt the risk.

[0026] The present invention uses the information and data of multi-sensors and multi-parameters to automatically analyze and determine under a certain neural network algorithm to complete the required decision-making and determination for the information processing process.

[0027] First, each data acquisition module (including the video data acquisition module, the detector data acquisition module, and the fiber optic sensor data acquisition module) is connected to each data detection device (including cameras, flame detectors, combustible gas detectors, temperature sensing fiber optic sensors, vibration fiber optic sensors, and pressure fiber Bragg grating sensors). The data collected by the data detection devices is orderly sent to the data acquisition module through their respective communication ports according to the set sampling time and sampling frequency. The detector data acquisition module is connected to the flame detector and the combustible gas detector, and the real-time data of the detector is collected in real time through the AI analog interface; among them, the liquefied hydrocarbon tank farm fire recognition system in the present invention can receive the analog signal of the flame detector or directly receive the alarm data of the flame detector from the site, and the data is more comprehensive. The fiber optic sensor acquisition module is connected to the temperature / vibration fiber optic sensor and the pressure fiber optic sensor to obtain temperature, vibration, and pressure data in real time. The data processing module in the present invention can receive the real-time dynamic temperature alarm system of the temperature sensing fiber optic sensor or perform real-time and dynamic data analysis and recognition on the temperature data to maximize the value of the temperature data and improve the accuracy and timeliness of the alarm. The video data acquisition module is connected to the camera, and image information is obtained in real time through the video data acquisition module; the video data acquisition module separately collects the video data of the tank top, the tank body, and the valve area. The number and location of the cameras set in the tank group are determined according to the size and layout of the tank group. The cameras set in the tank group are used to monitor the tank body and the bottom valve group, and the cameras set on the tank top are used to monitor the tank top and the top valve group; in addition, the cameras are mainly used to judge and identify whether a fire has occurred, predict the size and area of the fire, and identify and analyze the situation in the valve area. The cameras are set in positions with a large detection range and few obstructions to improve the monitoring efficiency and the degree of security prevention.

[0028] Then, each data acquisition module transmits the collected information to the data processing module. The data processing module performs preliminary data processing by establishing a multi-sensor information database and converts it into an input signal that can be recognized and processed by a neural network. Finally, the intelligent determination module makes an identification and determination based on this multi-sensor information database to determine whether a fire has occurred. For example, comprehensive determination is made according to different flame sensitivity coefficients and flame dynamic alarm values, the temperature values and temperature rise values of the temperature sensing fiber optic sensors, combined with the pressure and temperature values of the tank group.

[0029] The alarm output module of the liquefied hydrocarbon tank farm fire recognition system in the present invention is used to output the result determined by the system, which can be output through a communication protocol interface or a dry contact interface, and at the same time has the function of sound and light alarm output of a sound and light alarm. The display output module can display the collected data related to the real-time fire alarm through the client.

[0030] Embodiment 2

[0031] The classification and identification measures for different fire occurrence situations of liquefied hydrocarbon tank farms are as follows:

[0032] Type 1: Fire occurs around the liquefied hydrocarbon tank. The temperature around the liquefied hydrocarbon tank rises, and the heat radiation and heat convection of the flame have a greater impact on the surface of the spherical tank. The heat radiation causes the internal of the tank to be in a gasification state, and a series of thermal response changes will occur in the pressure and the temperature of the tank wall, which easily leads to the rupture of the tank body and may result in the following situations: First, the strength of adjacent storage tanks decreases, the tank wall tears, and the liquid in the tank leaks, further expanding the fire; Second, the medium in the tank volatilizes, the pressure in the tank increases, and a physical explosion occurs. Therefore, temperature-sensitive optical fiber sensors need to be installed on the tank body to detect the temperature of the tank body. When the temperature of the temperature-sensitive optical fiber sensor is too high, a high-temperature alarm will be triggered. At the same time, the pressure change of the tank body also needs to be detected. The method is to directly paste the pressure fiber Bragg grating sensor along the axial and circumferential directions of the tank body on the outer wall of the tank. Under the action of the internal medium pressure of the tank body, the tank body deforms, thereby causing a change in the wavelength of the pressure fiber Bragg grating sensor pasted on its surface; The pressure change of the tank body is characterized by the change in the wavelength of the fiber Bragg grating caused by the axial and circumferential deformation of the tank body, realizing the measurement and monitoring of the pressure of the liquefied hydrocarbon tank.

[0033] Type 2: Fire occurs in the liquefied hydrocarbon tank itself. The main reason for the fire in the liquefied hydrocarbon tank itself is the fire caused by the leakage of the tank wall. Therefore, temperature-sensitive optical fiber sensors are installed on the tank wall.

[0034] For the cases of fire occurring in the tank body and its surroundings in types 1 and 2, the present invention introduces a camera-assisted recognition mode to improve the detection accuracy. First, the liquefied hydrocarbon tank area is divided into several alarm areas; then the optical fiber sensor data acquisition module collects the temperature of the tank body in real time and conducts zoning positioning of the alarm areas. When the temperature in the alarm area of the tank body is too high or the temperature rise is abnormal, the optical fiber sensor acquisition module converts the real-time temperature information to the data processing module, and the data processing module conducts preliminary analysis and processing on the alarm temperature value and the alarm area; next, the data processing module triggers an interlock signal, and then issues an instruction to the nearest camera through the video data acquisition module for video-assisted determination; finally, the camera collects video signals of the temperature abnormal area, collects the status of the alarm area of the tank body in real time, and conducts image analysis and determination on the positioned temperature abnormal block. The image analysis and determination are as follows: The camera conducts intelligent analysis on the parameters of smoke, flame, and brightness of the video image. When the image intelligent recognition algorithm recognizes abnormal flame, smoke, and brightness, the collected information is transmitted to the intelligent determination module. The intelligent determination module establishes a multi-parameter model of the tank body posture. In the model, intelligent model analysis is conducted on the temperature, flame, brightness, and smoke of the tank body to obtain the abnormal coefficient of the tank body posture; the probability of fire is determined through the abnormal coefficient of the posture. The intelligent determination module extracts the historical picture data of this tank body (the historical photo database of the tank body is updated regularly to ensure the latest historical data), conducts intelligent comparison on the changed part, and analyzes the deviation change range and degree.

[0035] The division of the above alarm areas is that the data processing module divides each tank area into N areas, and each area is marked with a specific position number by the data processing module, corresponding to the preset positions set for M cameras respectively, as shown in Table 1.

[0036] Table 1 Camera preset positions of liquefied hydrocarbon tank area

[0037] 1 Partition A of Tank Group 1 Preset Position 001 of Camera No. 1 2 Partition B of Tank Group 1 Preset Position 002 of Camera No. 1 3 Partition C of Tank Group 1 Preset Position 001 of Camera No. 2 4 Partition D of Tank Group 1 Preset Position 002 of Camera No. 2 5 Partition E of Tank Group 1 Preset Position 001 of Camera No. 3 6 Partition A of Tank Group 2 Preset Position 004 of Camera No. 2 7 Partition B of Tank Group 2 Preset Position 005 of Camera No. 2 8 Partition C of Tank Group 2 Preset Position 002 of Camera No. 3 9 Partition D of Tank Group 2 Preset Position 001 of Camera No. 4 10 Partition E of Tank Group 2 Preset Position 003 of Camera No. 3 … … …

[0038] Type 3: Fire is caused by leakage at easily leakable locations such as valves of liquefied hydrocarbon tanks. The prerequisite for a leakage fire is the leakage of combustible liquefied hydrocarbon. When leakage occurs, it will trigger changes in temperature fiber optic sensors, vibration fiber optic sensors, and combustible gas detectors. Through the detector data acquisition module and fiber optic sensor data acquisition module, the leakage status and scope of the liquefied hydrocarbon tank can be obtained in real-time and dynamically. Specifically, liquefied hydrocarbon leakage may occur at the connection parts of the inlet and outlet material pipelines and sewage pipelines at the bottom of the tank with the tank bottom, valves, and flange plates. The substances in a conventional liquefied hydrocarbon tank are gaseous at normal temperature and liquid under pressure. When leakage occurs, the temperature will change at the nearby positions of the leakage. Once liquefied hydrocarbon leaks from the bottom of the tank, a large amount of liquefied hydrocarbon leaks and vaporizes by absorbing heat, reducing the temperature of the surrounding air; when liquefied hydrocarbon leaks and changes from liquid phase to gas phase, it needs to absorb a large amount of heat, causing the moisture in the air to condense, and forming a drifting white mist-like vapor cloud belt on the ground within its diffusion range; at the same time, the leaked fluid medium will impact the sealing surface and generate elastic waves. Therefore, temperature fiber optic sensors and vibration fiber optic sensors are set in easily leakable areas such as valves, so as to detect sudden temperature changes through temperature fiber optic sensors and detect high-frequency vibration signals through vibration fiber optic sensors; for the large amount of "white mist" generated during leakage, it can be judged by the video of the camera, and whether the "white mist" is a leakage and the leakage scope can be identified through image recognition, and the leakage probability value is output. At the same time, combustible gas detectors are conventionally set on the top and bottom of the tank to detect the leakage value of combustible gas in the easily leakable area in real-time.

[0039] As shown in Table 2, flame detectors are set at easily leakable safety valves, tank root valves, pipe connections, etc., and different leakage alarm areas are divided for the top and bottom areas of the tank to correspond to the preset positions of the flame detectors; the present invention uses a pan-tilt flame detector, and the flame detector is aimed at the easily leakable position for detection. When a leakage anomaly occurs, at this time, the pan-tilt of the flame detector rotates to collect flame signals near the leakage alarm area. Conventional flame detectors usually only report two states: fire alarm and non-fire alarm. We use different numerical values of the flame detector to judge the probability of the fire state, and through data learning such as superposing different numerical values and video judgment, an alarm value more suitable for the environmental area of this tank farm is formed. For example, when a leakage alarm is detected, if the flame detector detects a flame, the leakage fire alarm of the tank group is immediately triggered; when no leakage alarm is detected, and the flame detector detects the early warning value F1 of the flame detector, a fire early warning is triggered; when no leakage alarm is detected, and the flame detector detects the alarm value F2 of the flame detector, a fire alarm is triggered.

[0040] Table 2 Preset Positions of Flame Detectors

[0041] 1 Root Valve of Tank Group 1 Preset Position 001 of Flame Detector No. 1 2 Pipe Connection of Tank Group 1 Preset Position 002 of Flame Detector No. 1 3 Bottom Safety Accessories of Tank Group 1 Preset Position 003 of Flame Detector No. 1 4 Safety Valve on Top of Tank of Tank Group 1 Preset Position 001 of Flame Detector No. 2 5 Top Connection of Tank Top of Tank Group 1 Preset Position 002 of Flame Detector No. 2 6 Safety Accessories on Top of Tank of Tank Group 1 Preset Position 003 of Flame Detector No. 2 … … …

[0042] Table 3 Preset Positions of Cameras in Easily Leakable Areas of Liquefied Hydrocarbon Tank Farms

[0043] 1 Root Valve of Tank Group 1 Preset Position 001 of Camera No. 5 2 Pipe Connection of Tank Group 1 Preset Position 002 of Camera No. 5 3 Bottom Safety Accessories of Tank Group 1 Preset Position 003 of Camera No. 5 4 Safety Valve on Top of Tank of Tank Group 1 Preset Position 001 of Camera No. 6 5 Top Connection of Tank Top of Tank Group 1 Preset Position 002 of Camera No. 6 6 Top Connection of Tank Top of Tank Group 1 Preset Position 003 of Camera No. 6 … … …

[0044] For the case of leakage fire in the tank battery, the present invention also introduces a mode of camera-assisted recognition to improve the accuracy of detection. When using the liquefied hydrocarbon tank farm fire recognition and auxiliary determination system for fire monitoring, the pan-tilt camera collects video signals from the area with abnormal parameters, and in real time collects the status of the alarm areas such as the valve groups of the tank battery that are prone to leakage, and conducts image analysis and determination on the alarm areas. The arrangement positions of the cameras are shown in Table 3. Intelligent analysis is carried out on the smoke, flame and brightness parameters of the video images. When the image intelligent recognition algorithm recognizes abnormal flames, smoke and brightness, the collected information is transmitted to the intelligent determination module.

[0045] Example 3

[0046] The detection values of the above-mentioned temperature-sensitive optical fiber sensors, vibration optical fiber sensors, pressure fiber Bragg grating sensors, pressure fiber Bragg grating sensors, flame detectors, video image output values, and alarm values of combustible gas detectors are output to the intelligent determination module through the data processing module. The intelligent determination module establishes a neural network model, mainly for the present invention to establish a neural network determination model based on the detection values of temperature-sensitive optical fiber sensors, vibration optical fiber sensors, pressure fiber Bragg grating sensors, flame detector detection values, intelligent recognition output values of camera video images, and alarm values of combustible gas detectors, extracts characteristic parameters, and verifies the effectiveness and reliability of multi-parameter joint fire determination through experimental data, and explores a new direction for reducing the false alarm rate of fire alarms and improving the accuracy of fire detection. Figure 2 A multi-sensor and multi-parameter fire determination model for data processing using a neural network consists of three parts: sensor measurement and extraction of characteristic parameters, signal processing by the neural network, and recognition and alarm.

[0047] The intelligent determination module of the present invention establishes a neural network model. The neural network model has strong environmental adaptability, learning ability, fault tolerance ability and parallel processing ability, making the signal processing process closer to the thinking activity of the human brain. When the input-output relationship cannot be described by a specific function expression, the neural network intelligent algorithm can simulate the internal connection between the input and output through learning, training and simulation to achieve an intelligent effect.

[0048] The neural network model of the present invention is divided into three layers, namely the input layer, the hidden layer, and the output layer. Among them, the output value of the video image, the detection value of the temperature-sensitive optical fiber sensor, the detection value of the vibration optical fiber sensor, the detection value of the pressure optical fiber grating sensor, the detection value of the flame detector, and the alarm value of the combustible gas detector are selected as monitoring parameters, so as to construct a neural network fire determination model. The parameter values of the sensors are normalized respectively as N variables input to the neural network. At the same time, the output vector is selected as the encoding corresponding to 7 types of fire alarm determinations, such as Figure 3 as shown.

[0049] Regarding the input values of the input layer. The present invention selects the detection value of the temperature-sensitive optical fiber sensor, the detection value of the vibration optical fiber sensor, the detection value of the pressure optical fiber grating sensor, the detection value of the flame detector, the alarm value of the combustible gas detector, and the output value of the video image, six typical fire characteristic parameters, classifies and grades the parameters of each sensor, and normalizes each value at the same time. The percentage after normalization is used as the input value, and the percentage value of each parameter represents the probability of a fire. It mainly includes the following:

[0050] Detection value of the temperature-sensitive optical fiber sensor: high temperature alarm T1 of the tank body, high temperature alarm T2 of the tank top / tank bottom, low temperature alarm T3 of the tank top / tank bottom, temperature rise rate T4 of the tank body, temperature change rate T5 of the tank top / tank bottom;

[0051] Detection value of the vibration optical fiber sensor S;

[0052] Detection value of the pressure optical fiber grating sensor: high pressure alarm P1 of the tank body, very high pressure alarm P2 of the tank body;

[0053] Detection value of the flame detector: early warning value F1 of the flame detector, alarm value F2 of the flame detector;

[0054] Alarm value of the combustible gas detector: high alarm G1 of the combustible gas detector, very high alarm G2 of the combustible gas detector;

[0055] Output value of the video image: flame detection value C1 of the camera, smoke detection value C2 of the camera, leakage detection value C3 of the camera.

[0056] The output values of the output layer include 7 types, namely non-fire, early warning of fire outside the tank group, alarm of fire outside the tank group, early warning of tank body fire, alarm of tank body fire, early warning of leakage fire in the tank group, alarm of leakage fire in the tank group.

[0057] The neural network in the present invention is just one type of deep learning. The number of input and output layers can be increased or decreased according to the acquisition requirements. Different types of transfer algorithms can also be selected. The neural network is also a rough way, and the specific implementation can be refined and extended according to different requirements. The most important thing in the design of the neural network is to determine the optimal number of neurons in the hidden layer. At present, the determination of the number of hidden layer nodes can be achieved by comparing and training networks with different numbers of neurons in the hidden layer to find a relatively good one as the number of neurons in the hidden layer. It is more reasonable to select M hidden layer node numbers, and the learning and training of the network can be realized through learning samples. The selected neural network algorithm can better describe the correlation between various parameters of fire detection by selecting appropriate transfer functions, probability-based, and statistical transfer functions, etc., but it cannot accurately describe the correlation and causality between model parameters.

Claims

1. A liquefied hydrocarbon tank farm fire recognition system, characterized in that: It includes a camera, a flame detector, a combustible gas detector, a temperature-sensing optical fiber sensor, a vibration optical fiber sensor, a pressure fiber grating sensor, a video data acquisition module, a detector data acquisition module, an optical fiber sensor data acquisition module, a data processing module, and an intelligent judgment module; the video data acquisition module is connected to the camera and acquires the video image data captured by the camera; the detector data acquisition module is connected to the flame detector and the combustible gas detector and acquires the data of the flame detector and the combustible gas detector; the optical fiber sensor data module acquires the temperature data of the temperature-sensing optical fiber sensor, the vibration data of the vibration optical fiber sensor, and the pressure data of the pressure fiber grating sensor. The video data acquisition module, the detector data acquisition module, and the optical fiber sensor data acquisition module are all connected to the data processing module and transmit the data they respectively acquire to the data processing module for processing. The data processing module is connected to the intelligent judgment module, and the intelligent judgment module performs fire recognition based on the data processed by the data processing module.

2. The liquefied hydrocarbon tank farm fire recognition system according to claim 1, wherein: The temperature-sensing optical fiber sensors are arranged on the outer wall and valves of the liquefied hydrocarbon tanks in the liquefied hydrocarbon tank farm; the vibration optical fiber sensors are arranged at the pipe joints and near the valves in the liquefied hydrocarbon tank farm; the pressure fiber grating sensors are pasted on the outer wall of the liquefied hydrocarbon tanks in the liquefied hydrocarbon tank farm; the flame detectors are arranged at the pipe joints and near the valves in the liquefied hydrocarbon tank farm; the combustible gas detectors are arranged at the top and bottom of the liquefied hydrocarbon tanks.

3. The liquefied hydrocarbon tank farm fire recognition system according to claim 2, wherein: It further includes a display output module connected to the data processing module, and the display output module displays the data processed by the data processing module in real time through the client.

4. The liquefied hydrocarbon tank farm fire recognition system according to claim 3, characterized in that: It further includes an alarm output module connected to the intelligent judgment module, and the alarm output module outputs the system judgment result through the communication protocol interface and / or the dry contact interface.

5. The liquefied hydrocarbon tank farm fire recognition system according to claim 4, wherein: The alarm output module further includes an alarm device, and the alarm device is an audible and visual alarm.

6. The liquefied hydrocarbon tank farm fire recognition system according to claim 5, wherein: When the temperature of the temperature-sensing optical fiber sensor on the outer wall of the liquefied hydrocarbon tank is too high and triggers a high-temperature alarm, the nearby camera is set to: collect video signals in the high-temperature alarm area.

7. The liquefied hydrocarbon tank farm fire recognition system according to claim 6, characterized in that: The video data acquisition module performs intelligent analysis on the video image for parameters such as smoke, flame, and brightness. When flames, smoke, and abnormal brightness are recognized, the abnormal information collected is then transmitted to the intelligent judgment module. The intelligent judgment module establishes a multi-parameter model of the tank body state. In the model, intelligent model analysis is performed on the temperature, flame, brightness, and smoke of the tank body to obtain the abnormal coefficient of the tank body state.

8. The liquefied hydrocarbon tank farm fire recognition system according to claim 5, wherein: Temperature-sensing optical fiber sensors, vibration optical fiber sensors, combustible gas detectors, and cameras are arranged in the areas where the liquefied hydrocarbon tanks are prone to leakage. When the temperature-sensing optical fiber sensors, vibration optical fiber sensors, combustible gas detectors, and cameras trigger a leakage alarm, the nearby flame detector is set to: rotate to the leakage alarm area to collect flame signals.

9. The liquefied hydrocarbon tank farm fire recognition system according to claim 1, wherein: The intelligent determination module has a neural network model established. The input values of the input layer of the neural network model are: the detection value of the temperature-sensitive optical fiber sensor, the detection value of the vibration optical fiber sensor, the detection value of the pressure fiber grating sensor, the detection value of the flame detector, the alarm value of the combustible gas detector, and the video image output value; the output values of the output layer are: non-fire, fire warning for the periphery of the tank group, fire alarm for the periphery of the tank group, fire warning for the tank body, fire alarm for the tank body, fire warning for the leakage fire of the tank group, and fire alarm for the leakage fire of the tank group.

Images