A gas safety monitoring and early warning device and method

Abstract

The present invention discloses a gas safety monitoring and early warning device and method, which relate to the technical field of gas safety monitoring, and solve the problems of strong restrictiveness of the gas safety monitoring device, high gas monitoring and early warning error rate, and difficulty in operation and supervision by management personnel; the monitoring and early warning device comprises a gas detector, a safety shut-off valve, a sensor network, a communication channel, an intelligent gateway, and a data master station; an alarm threshold is generated through an abnormal gas usage model, so as to analyze, evaluate, and monitor the gas usage of each application scenario; the gas transmission and distribution pipeline network fault point is located through the pipeline network monitoring model and the alarm threshold, and the management personnel are displayed and warned, thereby improving the adaptability of the gas safety monitoring device in various application scenarios, effectively reducing the gas monitoring and early warning error rate, and simplifying the management personnel's operation and supervision process for the gas transmission and distribution pipeline network.

Description

Technical Field

[0001] The present invention relates to the technical field of gas safety monitoring, and in particular to a gas safety monitoring and early warning device and method. Background Art

[0002] With the development of society and the improvement of people's living standards, gas has become an important energy source in people's daily lives. However, gas safety accidents such as gas leaks and gas explosions occur from time to time, posing a serious threat to people's lives and property. Existing technologies are unable to timely evaluate the health of gas system pipelines, the state deterministic indicators of equipment in the system, response characteristics (large valve opening (or switching valve) torque, long valve opening (or switching valve) time), the ability to collect gas environment characteristic parameters, and the establishment of data models and data processing methods.

[0003] Patent No. CN2019100627491 discloses a method and system for safety monitoring and early warning of urban gas transmission and distribution pipeline networks based on subsidence monitoring. The method includes: integrating historical data of urban ground subsidence and distribution and historical fault data of urban gas transmission and distribution pipeline networks, and using multi-factor evaluation models and GIS spatial analysis technology to divide key areas of gas transmission and distribution pipeline network subsidence; selecting key locations in key areas as subsidence monitoring points, installing pipeline operation parameter collection equipment, and constructing a monitoring network for key areas of urban gas transmission and distribution pipeline network subsidence, and collecting pipeline operation parameters of each subsidence monitoring point in real time; transmitting pipeline operation parameters and pipeline attribute parameters to a background server in real time through wireless communication; receiving, storing and preprocessing pipeline operation parameters and pipeline attribute parameters in real time through the background server, and using a pipeline settlement and damage prediction model to determine the current safety risk level of pipelines at each subsidence monitoring point based on pipeline operation parameters and pipeline attribute parameters, and predicting and alarming pipeline failures.

[0004] Patent No. CN2022114524859 discloses a gas safety warning method and Internet of Things system based on smart gas. The method is executed by a smart gas safety management platform of the Internet of Things system for smart gas gas safety warning, including: determining a gas monitoring object from at least one gas device based on gas usage data; determining an initial object based on the gas monitoring object and the gas usage threshold; determining a target object based on the historical gas data of the initial object; and sending gas safety warning information to gas users corresponding to the target object.

[0005] However, the above patents cannot simultaneously monitor the health status of the gas transmission and distribution network and the safe gas use behavior of users in real time, and are prone to gas monitoring and early warning errors and operational supervision difficulties due to their inability to adapt to different gas application scenarios. Summary of the invention

[0006] The purpose of the present invention is to provide a gas safety monitoring and early warning device and method, which utilizes a specially designed deep learning network model and the air tightness detection principle of the transmission and distribution pipeline to monitor and evaluate the entire gas usage scenario and quickly locate the fault points of the transmission and distribution pipeline network, so as to improve the adaptability of the gas safety monitoring device in various application scenarios, effectively reduce the gas monitoring and early warning error rate, and simplify the operation and supervision process of the gas transmission and distribution pipeline network for management personnel.

[0007] The present invention utilizes the following technical solutions:

[0008] A gas safety monitoring and early warning device includes a gas detector, a safety shut-off valve, a sensor network, a communication channel, an intelligent gateway and a data master station; wherein,

[0009] Gas detectors, used to detect gas concentrations in the gas transmission and distribution network and on-site environment respectively;

[0010] Safety shut-off valve, used for automatic shut-off, manual reset, switch position detection and opening adjustment of gas valves;

[0011] The sensor network is used to collect the massive data generated by the gas transmission and distribution network in different gas application scenarios; the sensor network includes pressure sensors, temperature sensors, optical fiber sensors and acceleration sensors; the massive data includes pressure data, temperature data and gas flow;

[0012] Communication channel, used to transmit the massive data collected by the sensor network to the data master station; communication channels include local channels and remote channels; local channels include FSK, LORA or BLE, etc.; remote channels include LTE or NB_Iot, etc.;

[0013] The data master station is used to process and analyze the massive data received to determine and monitor the operating status of the gas transmission and distribution network; the operating status includes leakage failure, maintenance standby and normal operation;

[0014] The intelligent gateway is used to control the safety shut-off valve according to the data processing and analysis results, and send the monitoring information and early warning information to the data master station.

[0015] Preferably, the data master station includes a data processing module, an analysis and judgment module, a fault location module and an intelligent early warning module; wherein the data processing module is used to pre-process the acquired massive data; the analysis and judgment module is used to analyze and judge the pre-processed massive data according to the preset abnormal gas consumption model and gas application scenario to generate an alarm threshold and a corresponding control signal; the fault location module locates the fault point of the gas transmission and distribution pipeline network according to the alarm threshold and the pipeline monitoring model; the intelligent early warning module displays and warns the management personnel according to the alarm threshold and the location of the fault point of the gas transmission and distribution pipeline network; the gas application scenarios include household scenarios, industrial scenarios and commercial scenarios.

[0016] Preferably, the data processing module includes a data blocking unit, a data cleaning unit, a data classification unit, a data aggregation unit and a data storage unit; the data blocking unit compresses and divides the received massive data using a compression blocking algorithm to form a number of data blocks; the data cleaning unit fills missing values ​​and removes noise from the data blocks using a data filling algorithm and a data denoising algorithm; the data classification unit classifies the data blocks that have completed data cleaning using a clustering algorithm; the data aggregation unit fuses the classified data blocks according to the gas application scenario using a time series algorithm; and the data storage unit stores the fused data blocks using a metadata library.

[0017] Preferably, the analysis and judgment module includes a data analysis unit and an abnormality judgment unit; the data analysis unit uses a preset abnormal gas usage model to analyze and evaluate the fused data block according to the gas application scenario to generate a warning threshold; the abnormality judgment unit judges the warning threshold according to the real-time threshold corresponding to the gas application scenario to form a corresponding control signal.

[0018] Preferably, the abnormal gas usage model first uses the scenario verification function to perform quantitative verification on the stored data blocks to obtain the quantitative verification results of the data blocks, thereby obtaining the feature extraction times of the corresponding data blocks in each gas application scenario:

[0019]

[0020] Among them, A() represents the quantitative verification function, Q represents the data block, F() represents the scene verification function, i represents the data block number, n represents the total number of data blocks, α represents the feature weight coefficient, B() represents the occupied space of the data block, and B(Q) i represents the space occupied by the i-th data block, E() represents the storage time of the data block, and E(Q) i represents the storage time of the i-th data block, M represents the quantization upper limit, 1 represents that the data block is subjected to feature extraction N times, and -1 represents that the data block is subjected to feature extraction 2N times;

[0021] Then, according to the quantitative verification results, the data blocks of the home scene are input into the feature extraction layer for feature extraction N times, and the data blocks of the industrial scene and the commercial scene are input into the feature extraction layer for feature extraction 2N times to obtain the corresponding eigenvalues ​​respectively; the feature extraction layer includes 2 3X3 convolutional layers and 3 average pooling layers; then the obtained eigenvalues ​​are respectively input into the two network branches of the iterative training layer to extract the weight matrix; the first network branch includes 2 3X1 upsampling layers, 2 1X3 upsampling layers, 3 3X3 maximum pooling layers, 2 normalization layers and Tanh activation function, and the second ... The network branch includes three 3X3 convolutional layers, two 3X3 average pooling layers, one 5X5 maximum pooling layer, one normalization layer, four residual blocks and a RELU activation function. Finally, the weight matrices output by the two network branches are concatenated and fused using a depth separation connector with uniform sampling offset. At the same time, the concatenated weight matrices are sequentially passed through three average pooling layers, two fully connected layers, a temporal attention mechanism and a SoftMax activation function to generate the warning thresholds corresponding to each gas application scenario, and the airtightness threshold is used to correct the warning thresholds according to the gas application scenarios.

[0022] Preferably, the pipeline network monitoring model first associates the gas transmission direction, gas consumption point, gas consumption and city map to construct a distribution architecture map of the gas transmission and distribution pipeline network:

[0023]

[0024] Among them, H represents the distribution architecture map, j represents the sequence number of the city map area, β represents the total number of city map areas, G() represents the map construction function, d() represents the gas point marking function, r() represents the gas transmission direction marking function, k() represents the gas volume marking function, θ represents the digitized city map, q represents the city map depth value, represents the depth of the jth area of ​​the digitized city map, IM() represents the Infomap function, ⊙ represents the Hadamard outer product, represents multiple vector addition, represents the Hadamard inner product;

[0025] The distribution architecture map and the stored data blocks are then input into the input ends of the two network branches of the twin network respectively to extract the feature values ​​of the gas transmission and distribution network and the data blocks respectively; the input end of the first network branch includes 2 3X3 convolutional layers, 1 3X3 maximum pooling layer and 2 normalization layers, and the input end of the second network branch includes 2 3X3 convolutional layers, 2 3X3 maximum pooling layers, 1 3X3 average pooling layer and 1 normalization layer; then the extracted feature values ​​are input into the learning ends of the two network branches for iterative training to obtain the weight matrix; the learning end of the first network branch includes 2 combined convolution blocks consisting of 1X3 convolutional layers and 3X1 convolutional layers , 2 3X3 upsampling layers, 2 3X3 downsampling layers, 2 Inception blocks, 1 inverted residual block and SoftMax activation function. The learning end of the second network branch includes 2 1X1 convolutional layers, 2 3X3 upsampling layers, 2 3X3 downsampling layers, 2 residual blocks and RELU activation function. Finally, the weight matrices output by the learning ends of the two network branches are concatenated and fused using a separation connector. At the same time, the concatenated weight matrices are passed through 2 average pooling layers, 3 fully connected layers and ELU activation function in turn, and the weight matrix is ​​updated and verified using the loss function composed of the airtightness threshold and the temporal memory network:

[0026]

[0027] Among them, L() represents the loss function, Y represents the weight matrix, γ represents the number of columns in the weight matrix, z represents the total number of columns in the weight matrix, σ represents the error correction coefficient, and Y γ Indicates the current number of columns in the weight matrix;

[0028] This enables real-time monitoring of the gas transmission and distribution network, and marking and locating fault points in the gas transmission and distribution network.

[0029] Preferably, the intelligent early warning module includes a display unit, an early warning unit and a tracing unit; the display unit displays the marked and located gas transmission and distribution pipeline fault point in the form of a chart or report using a touch screen; the early warning unit warns the management personnel using text messages and alarms based on the comparison result of the warning threshold and the real-time threshold and the marked and located gas transmission and distribution pipeline fault point; the tracing unit traces the data of the gas transmission and distribution pipeline fault point based on the stored data blocks.

[0030] Preferably, the intelligent gateway includes a signal transmission module and a status indication module; the signal transmission module utilizes a signal transceiver circuit to transmit the generated control signal to the safety shut-off valve to complete automatic shut-off, manual reset, switch position detection and opening adjustment operations of the safety shut-off valve; the status indication module indicates the operating status of the gas transmission and distribution network to users and managers based on the monitoring results of the gas transmission and distribution network.

[0031] Preferably, the intelligent gateway is a stand-alone device or an embedded device.

[0032] Also included is a gas safety monitoring and early warning method, comprising the following steps:

[0033] S1: Detect the gas concentration in the gas transmission and distribution network and on-site environment through the gas detector installed on the user side;

[0034] S2: Collecting massive amounts of data generated in different gas application scenarios through sensor networks;

[0035] S3: The data master station processes and analyzes the massive amounts of data collected, and determines and monitors the operating status of the gas transmission and distribution network;

[0036] S4: Locate and warn the fault points of the gas transmission and distribution network based on the processing and analysis results, and control and adjust the safety shut-off valve through the intelligent gateway;

[0037] S5: The monitoring information and early warning information of the gas transmission and distribution network are displayed on the user side and the management personnel side, and sent to the data master station for storage.

[0038] The present invention generates warning thresholds through abnormal gas usage models, thereby analyzing and evaluating the gas usage and monitoring abnormalities in various application scenarios; locates the fault points of the gas transmission and distribution pipeline network through the pipeline network monitoring model and the warning thresholds, and displays and warns the management personnel, thereby improving the adaptability of the gas safety monitoring device in various application scenarios, effectively reducing the gas monitoring and warning error rate, and simplifying the management personnel's operation and supervision process of the gas transmission and distribution pipeline network. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or related technologies, the drawings required for use in the embodiments or related technical descriptions are briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0040] Figure 1 This is a schematic diagram of a gas safety monitoring and early warning device;

[0041] Figure 2 This is a principle block diagram of the gas safety monitoring and early warning method. DETAILED DESCRIPTION

[0042] The present invention is described in detail below with reference to the accompanying drawings and embodiments:

[0043] like Figure 1As shown, a gas safety monitoring and early warning device according to the present invention includes a gas detector, a safety shut-off valve, a sensor network, a communication channel, an intelligent gateway and a data master station; wherein,

[0044] Gas detectors, used to detect gas concentrations in the gas transmission and distribution network and on-site environment respectively;

[0045] Safety shut-off valve, used for automatic shut-off, manual reset, switch position detection and opening adjustment of gas valves;

[0046] The sensor network is used to collect the massive data generated by the gas transmission and distribution network in different gas application scenarios; the sensor network includes pressure sensors, temperature sensors, optical fiber sensors and acceleration sensors; the massive data includes pressure data, temperature data and gas flow;

[0047] Communication channel, used to transmit the massive data collected by the sensor network to the data master station; communication channels include local channels and remote channels; local channels include FSK, LORA or BLE, etc.; remote channels include LTE or NB_Iot, etc.;

[0048] The data master station is used to process and analyze the massive data received to determine and monitor the operating status of the gas transmission and distribution network; the operating status includes leakage failure, maintenance standby and normal operation;

[0049] In this embodiment, the data master station includes a data processing module, an analysis and judgment module, a fault location module and an intelligent early warning module; wherein the data processing module is used to pre-process the acquired massive data; the analysis and judgment module is used to analyze and judge the pre-processed massive data according to the preset abnormal gas consumption model and the gas application scenario to generate a warning threshold and a corresponding control signal; the fault location module locates the fault point of the gas transmission and distribution network according to the warning threshold and the pipeline monitoring model; the intelligent early warning module displays and warns the management personnel according to the warning threshold and the location of the fault point of the gas transmission and distribution network; the gas application scenarios include household scenarios, industrial scenarios and commercial scenarios;

[0050] In the present invention, the data processing module includes a data block unit, a data cleaning unit, a data classification unit, a data aggregation unit and a data storage unit; the data block unit compresses and divides the received mass data by using a compression block algorithm to form a plurality of data blocks; the data cleaning unit fills missing values ​​and removes noise from the data blocks by using a data filling algorithm and a data denoising algorithm; the data classification unit classifies the data blocks that have completed data cleaning by using a clustering algorithm; the data aggregation unit fuses the classified data blocks according to the gas application scenario by using a time series algorithm; the data storage unit stores the fused data blocks by using a metadata database;

[0051] In the present invention, the analysis and judgment module includes a data analysis unit and an abnormality judgment unit; the data analysis unit uses a preset abnormal gas usage model to analyze and evaluate the fused data block according to the gas application scenario to generate a warning threshold; the abnormality judgment unit performs abnormality judgment according to the warning threshold and real-time threshold corresponding to the gas application scenario to form a corresponding control signal;

[0052] In the present invention, the intelligent early warning module includes a display unit, an early warning unit and a tracing unit; the display unit displays the marked and located gas transmission and distribution network fault point in the form of a chart or a report using a touch screen; the early warning unit warns the management personnel using a text message and an alarm according to the comparison result between the warning threshold and the real-time threshold and the marked and located gas transmission and distribution network fault point; the tracing unit performs data tracing of the gas transmission and distribution network fault point according to the stored data block;

[0053] Intelligent gateway, used to control the safety shut-off valve according to the data processing and analysis results, and send monitoring information and early warning information to the data master station;

[0054] In this embodiment, the intelligent gateway includes a signal transmission module and a status indication module; the signal transmission module uses a signal transceiver circuit to transmit the generated control signal to the safety shut-off valve to complete the automatic shut-off, manual reset, switch position detection and opening adjustment operations of the safety shut-off valve; the status indication module indicates the operating status of the gas transmission and distribution network to users and managers based on the monitoring results of the gas transmission and distribution network;

[0055] In the present invention, the intelligent gateway is a stand-alone device or an embedded device;

[0056] Furthermore, the abnormal gas usage model first uses the scenario verification function to perform quantitative verification on the stored data blocks to obtain the quantitative verification results of the data blocks, thereby obtaining the feature extraction times of the corresponding data blocks in each gas application scenario:

[0057]

[0058] Among them, A() represents the quantitative verification function, Q represents the data block, F() represents the scene verification function, i represents the data block number, n represents the total number of data blocks, α represents the feature weight coefficient, B() represents the occupied space of the data block, and B(Q) i represents the space occupied by the i-th data block, E() represents the storage time of the data block, and E(Q) i represents the storage time of the i-th data block, M represents the quantization upper limit, 1 represents that the data block is subjected to feature extraction N times, and -1 represents that the data block is subjected to feature extraction 2N times;

[0059] Then, according to the quantitative verification results, the data blocks of the home scene are input into the feature extraction layer for feature extraction N times, and the data blocks of the industrial scene and the commercial scene are input into the feature extraction layer for feature extraction 2N times to obtain the corresponding eigenvalues ​​respectively; the feature extraction layer includes 2 3X3 convolutional layers and 3 average pooling layers; then the obtained eigenvalues ​​are respectively input into the two network branches of the iterative training layer to extract the weight matrix; the first network branch includes 2 3X1 upsampling layers, 2 1X3 upsampling layers, 3 3X3 maximum pooling layers, 2 normalization layers and Tanh activation function, and the second ... The network branch includes three 3X3 convolutional layers, two 3X3 average pooling layers, one 5X5 maximum pooling layer, one normalization layer, four residual blocks and RELU activation function; finally, the weight matrices output by the two network branches are spliced ​​and fused using a depth separation connector with uniform sampling offset. At the same time, the spliced ​​and fused weight matrices are sequentially passed through three average pooling layers, two fully connected layers, a temporal attention mechanism and a SoftMax activation function to generate the warning thresholds corresponding to each gas application scenario, and the airtightness threshold is used to correct the warning threshold according to the gas application scenario;

[0060] In this embodiment, the working principle of the abnormal gas usage model is:

[0061] First, the gas consumption data blocks in different scenarios are quantitatively verified and input into the feature extraction layer: two 3X3 convolutional layers and three average pooling layers are used to extract features and obtain initial feature values; after obtaining the initial feature values, they are respectively input into the two network branches of the iterative training layer; among them, the content of the gas consumption data block includes gas consumption object information (name, address, etc.), gas consumption time information (start time, end time, gas consumption, etc.), gas consumption quality information (gas consumption quality index, gas consumption pressure, gas consumption temperature, etc.), gas consumption metering information (installation location, model, number, calibration cycle, etc. of metering equipment), gas consumption equipment information (type, specification, quantity, installation location, etc. of gas consumption equipment);

[0062] Then, the two network branches extract weight matrices respectively. The first network branch uses 2 3X1 upsampling layers, 2 1X3 upsampling layers, 3 3X3 maximum pooling layers, 2 normalization layers and Tanh activation function, and the second network branch uses 3 3X3 convolutional layers, 2 3X3 average pooling layers, 1 5X5 maximum pooling layer, 1 normalization layer, 4 residual blocks and RELU activation function to extract the weight matrix. The weight matrices output by the two network branches are spliced ​​and fused through a depth separation connector with uniform sampling offset;

[0063] Finally, the weight matrix after splicing and fusion is passed through three average pooling layers, two fully connected layers, a temporal attention mechanism, and a SoftMax activation function in turn to generate the warning threshold for the corresponding scenario. The real-time change value of the gas pressure monitored in real time is compared with the warning threshold for the corresponding scenario to determine whether there is abnormal gas consumption. If abnormal, the early warning system is triggered for processing.

[0064] Furthermore, the pipeline network monitoring model first associates the gas transmission direction, gas consumption point, gas consumption and city map to construct a distribution architecture map of the gas transmission and distribution network:

[0065]

[0066] Among them, H represents the distribution architecture map, j represents the sequence number of the city map area, β represents the total number of city map areas, G() represents the map construction function, d() represents the gas point marking function, r() represents the gas transmission direction marking function, k() represents the gas volume marking function, θ represents the digitized city map, q represents the city map depth value, represents the depth of the jth area of ​​the digitized city map, IM() represents the Infomap function, ⊙ represents the Hadamard outer product, represents multiple vector addition, represents the Hadamard inner product;

[0067] The distribution architecture map and the stored data blocks are then input into the input ends of the two network branches of the twin network respectively to extract the feature values ​​of the gas transmission and distribution network and the data blocks respectively; the input end of the first network branch includes 2 3X3 convolutional layers, 1 3X3 maximum pooling layer and 2 normalization layers, and the input end of the second network branch includes 2 3X3 convolutional layers, 2 3X3 maximum pooling layers, 1 3X3 average pooling layer and 1 normalization layer; then the extracted feature values ​​are input into the learning ends of the two network branches for iterative training to obtain the weight matrix; the learning end of the first network branch includes 2 combined convolution blocks consisting of 1X3 convolutional layers and 3X1 convolutional layers , 2 3X3 upsampling layers, 2 3X3 downsampling layers, 2 Inception blocks, 1 inverted residual block and SoftMax activation function. The learning end of the second network branch includes 2 1X1 convolutional layers, 2 3X3 upsampling layers, 2 3X3 downsampling layers, 2 residual blocks and RELU activation function. Finally, the weight matrices output by the learning ends of the two network branches are concatenated and fused using a separation connector. At the same time, the concatenated weight matrices are passed through 2 average pooling layers, 3 fully connected layers and ELU activation function in turn, and the weight matrix is ​​updated and verified using the loss function composed of the airtightness threshold and the temporal memory network:

[0068]

[0069] Among them, L() represents the loss function, Y represents the weight matrix, γ represents the number of columns in the weight matrix, z represents the total number of columns in the weight matrix, σ represents the error correction coefficient, and Y γ Indicates the current number of columns in the weight matrix;

[0070] Thus, the gas transmission and distribution network can be monitored in real time, and the fault points of the gas transmission and distribution network can be marked and located;

[0071] In this embodiment, the distribution architecture includes the length, diameter, material and thickness of the pipelines used in the transmission and distribution network, as well as the number, type and adjustment range of regulating valves; the scale, pressure range, transportation distance and area of ​​the pressure regulating station, etc.;

[0072] In this embodiment, the working principle of the pipe network monitoring model is:

[0073] First, the gas transmission direction, gas consumption point, gas consumption and city map are associated to construct a distribution architecture map of the gas transmission and distribution network, and the distribution architecture map and the stored data blocks are respectively input into the input ends of the two network branches of the twin network to obtain the initial feature values; among them, the first input end uses 2 3X3 convolutional layers, 1 3X3 maximum pooling layer and 2 normalization layers to extract the features of the gas transmission and distribution network, and the second input end uses 2 3X3 convolutional layers, 2 3X3 maximum pooling layers, 1 3X3 average pooling layer and 1 normalization layer to extract the features of the data block. After obtaining the feature values, they are respectively input into the learning ends of the two network branches;

[0074] Then, the learning ends of the two network branches are iteratively trained respectively. The first network branch uses 2 combined convolution blocks consisting of 1X3 convolution layers and 3X1 convolution layers, 2 3X3 upsampling layers, 2 3X3 downsampling layers, 2 Inception blocks, 1 inverted residual block and SoftMax activation function, and the second network branch uses 2 1X1 convolution layers, 2 3X3 upsampling layers, 2 3X3 downsampling layers, 2 residual blocks and RELU activation function to extract the weight matrix; the weight matrices output by the learning ends of the two network branches are spliced ​​and fused through the separation connector;

[0075] Finally, the weight matrix after splicing and fusion is passed through two average pooling layers, three fully connected layers and ELU activation function in turn for real-time monitoring and fault point marking and positioning; by monitoring the real-time data of the gas transmission and distribution network, possible faults are predicted and warned in advance to ensure the safe use of gas;

[0076] In the present invention, the pipeline network monitoring model also includes an update feedback layer; the weight matrix of the pipeline network monitoring model, the air tightness threshold obtained by the Bernoulli equation is continuously updated and verified in the update feedback layer using a temporal memory network;

[0077] Among them, the conditions for obtaining the airtightness threshold are: under the condition of valve closure, the gas transmission and distribution pipeline network is judged to be sealed by the real-time change value of the gas transmission and distribution pipeline pressure within a certain period of time is not greater than the airtightness threshold, and the airtightness threshold is set according to the relevant parameters in the Bernoulli equation:

[0078]

[0079] Wherein, P1 represents the pressure inside the gas transmission and distribution pipeline, P2 represents the atmospheric pressure, ρ1 represents the gas density, ρ2 represents the atmospheric density, v1 represents the gas transmission speed inside the pipeline at the leaking location, and v2 represents the gas leakage speed outside the pipeline at the leaking location;

[0080] The pressure difference in the gas distribution pipeline is expressed as a time function:

[0081] P(t)=P1-P2 (5)

[0082] Where P(t) represents the pressure difference function;

[0083] Assuming that the velocity inside the pipe at the leak is 0, the gas transmission velocity is expressed as a time function:

[0084]

[0085] Where v(t) represents the gas transfer function;

[0086] Assume that at a certain time t, the pressure value P(t) in the sealed pipe is:

[0087] P1v1=P2v1+P2v(t)S (7)

[0088] dP(t)v1=P2v(t)Sdt (8)

[0089]

[0090] Where S represents the area of ​​the pipeline leakage, and c2 represents the proportionality coefficient;

[0091] Integrate both sides of the formula and rearrange to get:

[0092] P(t)=a(bt) 2 (10)

[0093] Among them, a represents the area coefficient of the pipeline leakage, and b represents the internal and external pressure balance time of the pipeline leakage.

[0094] like Figure 2 As shown, it also includes a gas safety monitoring and early warning method, comprising the following steps:

[0095] S1: Detect the gas concentration in the gas transmission and distribution network and on-site environment through the gas detector installed on the user side;

[0096] S2: Collecting massive amounts of data generated in different gas application scenarios through sensor networks;

[0097] S3: The data master station processes and analyzes the massive amounts of data collected, and determines and monitors the operating status of the gas transmission and distribution network;

[0098] S4: Locate and warn the fault points of the gas transmission and distribution network based on the processing and analysis results, and control and adjust the safety shut-off valve through the intelligent gateway;

[0099] S5: The monitoring information and early warning information of the gas transmission and distribution network are displayed on the user side and the management personnel side, and sent to the data master station for storage.

[0100] In this embodiment, FSK, LORA, BLE, LTE, NB_Iot, compression block algorithm, data filling algorithm, data denoising algorithm, clustering algorithm, metadata database, SMS message and alarm are all existing technologies and will not be described in detail here.

[0101] Example:

[0102] First, deploy gas detectors on the user side and use them to detect the gas concentration of the gas transmission and distribution network and the on-site environment; collect pressure data, temperature data, gas flow, gas flow rate and other massive data generated by the gas transmission and distribution network in different gas application scenarios (home, industrial and commercial scenarios, etc.) through the pressure sensors, temperature sensors, optical fiber sensors and acceleration sensors of the sensor network; at the same time, transmit the massive data to the data master station through local channels including but not limited to FSK, LORA or BLE;

[0103] Then, the data processing module of the data master station uses a compression block algorithm to compress and divide the received massive data to form several data blocks, and uses a data filling algorithm and a data denoising algorithm to fill missing values ​​and remove noise from the data blocks, and then uses a clustering algorithm to classify the data blocks that have completed data cleaning, and then uses a time series algorithm to fuse the classified data blocks according to the gas application scenario, and uses the metadata database to store the fused data blocks; then the data analysis unit of the analysis and judgment module uses a preset abnormal gas consumption model to analyze and evaluate the fused data blocks according to the gas application scenario to generate a warning threshold; the abnormal judgment unit uses the real-time threshold corresponding to the gas application scenario to judge the warning threshold to form a corresponding control signal;

[0104] Subsequently, the early warning unit of the intelligent early warning module uses short messages and alarms to warn the management personnel based on the comparison results of the warning threshold and the real-time threshold and the marked and located gas transmission and distribution network fault points; and the display unit uses the touch screen to display the marked and located gas transmission and distribution network fault points in the form of charts or reports; at the same time, the tracing unit performs data tracing on the gas transmission and distribution network fault points based on the stored data blocks; then, the signal transmission module of the intelligent gateway uses the signal transceiver circuit to transmit the generated control signal to the safety shut-off valve to complete the automatic shut-off, manual reset, switch position detection and opening adjustment operations of the safety shut-off valve; in addition, the status indication module indicates the operating status of the gas transmission and distribution network to users and management personnel based on the monitoring results of the gas transmission and distribution network;

[0105] Finally, the monitoring information and early warning information of the gas transmission and distribution network are displayed on the user side and the management staff side, and sent to the data master station for storage.

Claims

1. A gas safety monitoring and early warning device, characterized in that: It includes gas detectors, safety shut-off valves, sensor networks, communication channels, intelligent gateways and data master stations; among them, Gas detectors, used to detect gas concentrations in the gas transmission and distribution network and on-site environment respectively; Safety shut-off valve, used for automatic shut-off, manual reset, switch position detection and opening adjustment of gas valves; The sensor network is used to collect the massive data generated by the gas transmission and distribution network in different gas application scenarios; the sensor network includes pressure sensors, temperature sensors, optical fiber sensors and acceleration sensors; the massive data includes pressure data, temperature data and gas flow; Communication channel, used to transmit the massive data collected by the sensor network to the data master station; the communication channel includes local channel and remote channel; the local channel includes FSK, LORA or BLE; the remote channel includes LTE or NB_Iot; The data master station is used to process and analyze the massive data received to determine and monitor the operating status of the gas transmission and distribution network; the operating status includes leakage failure, maintenance standby and normal operation; Intelligent gateway, used to control the safety shut-off valve according to the data processing and analysis results, and send monitoring information and early warning information to the data master station; The data master station includes a data processing module, an analysis and judgment module, a fault location module and an intelligent early warning module; wherein the data processing module is used to pre-process the acquired massive data; the analysis and judgment module analyzes and judges the pre-processed massive data according to the preset abnormal gas consumption model and gas application scenario to generate a warning threshold and a corresponding control signal; the fault location module locates the fault point of the gas transmission and distribution network according to the warning threshold and the pipeline monitoring model; the intelligent early warning module displays and warns the management personnel according to the warning threshold and the location of the fault point of the gas transmission and distribution network; the gas application scenarios include household scenarios, commercial scenarios and industrial scenarios; The data processing module includes a data block unit, a data cleaning unit, a data classification unit, a data aggregation unit and a data storage unit; the data block unit compresses and divides the received mass data using a compression block algorithm to form a number of data blocks; the data cleaning unit fills missing values ​​and removes noise from the data blocks using a data filling algorithm and a data denoising algorithm; the data classification unit classifies the data blocks after data cleaning using a clustering algorithm; the data aggregation unit fuses the classified data blocks according to the gas application scenario using a time series algorithm; the data storage unit stores the fused data blocks using a metadata database; The abnormal gas usage model first uses the scenario verification function to perform quantitative verification on the stored data blocks to obtain the quantitative verification results of the data blocks, thereby obtaining the feature extraction times of the corresponding data blocks in each gas application scenario: Among them, A() represents the quantitative verification function, Q represents the data block, F() represents the scene verification function, i represents the data block number, n represents the total number of data blocks, α represents the feature weight coefficient, B() represents the occupied space of the data block, and B(Q) i represents the space occupied by the i-th data block, E() represents the storage time of the data block, and E(Q) i represents the storage time of the i-th data block, M represents the quantization upper limit, 1 represents that the data block is subjected to feature extraction N times, and -1 represents that the data block is subjected to feature extraction 2N times; Then, according to the quantitative verification results, the data blocks of the home scene are input into the feature extraction layer for feature extraction N times, and the data blocks of the industrial scene and the commercial scene are input into the feature extraction layer for feature extraction 2N times to obtain the corresponding eigenvalues ​​respectively; the feature extraction layer includes 2 3X3 convolutional layers and 3 average pooling layers; then the obtained eigenvalues ​​are respectively input into the two network branches of the iterative training layer to extract the weight matrix; the first network branch includes 2 3X1 upsampling layers, 2 1X3 upsampling layers, 3 3X3 maximum pooling layers, 2 normalization layers and Tanh activation function, and the second ... The network branch includes three 3X3 convolutional layers, two 3X3 average pooling layers, one 5X5 maximum pooling layer, one normalization layer, four residual blocks and RELU activation function; finally, the weight matrices output by the two network branches are spliced ​​and fused using a depth separation connector with uniform sampling offset. At the same time, the spliced ​​and fused weight matrices are sequentially passed through three average pooling layers, two fully connected layers, the temporal attention mechanism and the SoftMax activation function to generate the warning thresholds corresponding to each gas application scenario, and the airtightness threshold is used to correct the warning threshold according to the gas application scenario.

2. The gas safety monitoring and early warning device according to claim 1 is characterized in that: The analysis and judgment module includes a data analysis unit and an abnormality judgment unit; the data analysis unit uses a preset abnormal gas usage model to analyze and evaluate the fused data block according to the gas application scenario to generate a warning threshold; The abnormality judgment unit judges the warning threshold according to the real-time threshold corresponding to the gas application scenario to form a corresponding control signal.

3. The gas safety monitoring and early warning device according to claim 1 is characterized in that: The pipeline network monitoring model first associates the gas transmission direction, gas consumption point, gas consumption and city map to construct a distribution architecture map of the gas transmission and distribution network: Among them, H represents the distribution architecture map, j represents the sequence number of the city map area, β represents the total number of city map areas, G() represents the map construction function, d() represents the gas point marking function, r() represents the gas transmission direction marking function, k() represents the gas volume marking function, θ represents the digitized city map, q represents the city map depth value, represents the depth of the jth area of ​​the digitized city map, IM() represents the Infomap function, ⊙ represents the Hadamard outer product, represents multiple vector addition, represents the Hadamard inner product; The distribution architecture map and the stored data blocks are then input into the input ends of the two network branches of the twin network respectively to extract the feature values ​​of the gas transmission and distribution network and the data blocks respectively; the input end of the first network branch includes 2 3X3 convolutional layers, 1 3X3 maximum pooling layer and 2 normalization layers, and the input end of the second network branch includes 2 3X3 convolutional layers, 2 3X3 maximum pooling layers, 1 3X3 average pooling layer and 1 normalization layer; then the extracted feature values ​​are input into the learning ends of the two network branches for iterative training to obtain the weight matrix; the learning end of the first network branch includes 2 combined convolution blocks consisting of 1X3 convolutional layers and 3X1 convolutional layers , 2 3X3 upsampling layers, 2 3X3 downsampling layers, 2 Inception blocks, 1 inverted residual block and SoftMax activation function. The learning end of the second network branch includes 2 1X1 convolutional layers, 2 3X3 upsampling layers, 2 3X3 downsampling layers, 2 residual blocks and RELU activation function. Finally, the weight matrices output by the learning ends of the two network branches are concatenated and fused using a separation connector. At the same time, the concatenated weight matrices are passed through 2 average pooling layers, 3 fully connected layers and ELU activation function in turn, and the weight matrix is ​​updated and verified using the loss function composed of the airtightness threshold and the temporal memory network: Among them, L() represents the loss function, P(t) represents the pressure difference function at the current moment, Y represents the weight matrix, γ represents the number of columns in the weight matrix, z represents the total number of columns in the weight matrix, σ represents the error correction coefficient, and Y γ represents the current number of columns of the weight matrix, and P(t-1) represents the pressure difference function at the previous moment; This enables real-time monitoring of the gas transmission and distribution network, and marking and locating fault points in the gas transmission and distribution network.

4. The gas safety monitoring and early warning device according to claim 1 is characterized in that: The intelligent early warning module includes a display unit, an early warning unit and a tracing unit; the display unit displays the marked and located gas transmission and distribution network fault points in the form of charts or reports using a touch screen; The early warning unit uses SMS messages and alarms to warn management personnel based on the comparison results of the warning threshold and the real-time threshold and the marked and located gas transmission and distribution network fault points; The tracing unit performs data tracing on the fault points of the gas transmission and distribution network based on the stored data blocks.

5. The gas safety monitoring and early warning device according to claim 1 is characterized in that: The intelligent gateway includes a signal transmission module and a status indication module; the signal transmission module uses a signal transceiver circuit to transmit the generated control signal to the safety shut-off valve to complete the automatic shut-off, manual reset, switch position detection and opening adjustment operations of the safety shut-off valve; the status indication module indicates the operating status of the gas transmission and distribution pipeline network to users and managers based on the monitoring results of the gas transmission and distribution pipeline network.

6. The gas safety monitoring and early warning device according to claim 1 is characterized in that: The intelligent gateway is a stand-alone device or an embedded device.

7. A gas safety monitoring and early warning method, applied to the safety monitoring and early warning device according to any one of claims 1 to 6, characterized in that: The following steps are involved: S1: Detect the gas concentration in the gas transmission and distribution network and on-site environment through the gas detector installed on the user side; S2: Collecting massive amounts of data generated in different gas application scenarios through sensor networks; S3: The data master station processes and analyzes the massive amounts of data collected, and determines and monitors the operating status of the gas transmission and distribution network; S4: Locate and warn the fault points of the gas transmission and distribution network based on the processing and analysis results, and control and adjust the safety shut-off valve through the intelligent gateway; S5: The monitoring information and early warning information of the gas transmission and distribution network are displayed on the user side and the management personnel side, and sent to the data master station for storage.

Images