An intelligent unmanned LNG point supply station monitoring system

Abstract

The present invention discloses an intelligent unmanned LNG point supply station monitoring system, which relates to the field of liquefied natural gas technology. It solves the problem that when a number of on-site monitoring devices remotely transmit real-time data to a remote monitoring end, the real-time data missing rate is high due to the influence of network status and transmission distance, thereby causing a high fault error rate of the LNG point supply station and low fault detection efficiency of the LNG point supply station; the LNG point supply station monitoring system comprises a monitoring center and a number of on-site monitoring devices; a special mark is used as the unique identifier of the LNG point supply station in the monitoring center to reduce the fault error rate of the LNG point supply station and retransmit the missing data; a multimodal fault prediction model is used to predict LNG leakage on a lossless data table to generate an LNG leakage prediction result, and then the corresponding accident emergency equipment is controlled to perform emergency control on the LNG point supply station, so as to improve the fault detection efficiency of the LNG point supply station.

Description

Technical Field

[0001] The present invention relates to the technical field of liquefied natural gas, and in particular to an intelligent unmanned LNG point supply station monitoring system. Background Art

[0002] In recent years, natural gas has gradually become a key component of my country's primary energy structure, with the country's natural gas extraction and utilization levels increasing year by year, both in terms of quantity and quality. Natural gas extraction, as the source of the natural gas industry, is of undeniable importance. However, most natural gas distribution stations in my country still rely on manual inspections, which can lead to problems such as human oversight and seriously impact the safe operation of LNG distribution stations.

[0003] LNG (liquefied natural gas) is a clean, efficient, and environmentally friendly energy source that has been widely used. LNG point-of-sale stations are important production and distribution bases for LNG. With their diverse locations, diverse product offerings, and rapid response times, they are a crucial addition to meeting market demand. Real-time monitoring is essential to ensure safe operation at these stations.

[0004] Patent No. CN2017110203190 discloses a remote monitoring system for the LNG industry chain, including a server, hardware, and software. The hardware includes a memory, and the software includes applications stored in the memory. The applications include functional modules that can implement the following functions: user management, geographic information, operation and maintenance quality control, real-time monitoring, statistical reports, and dispatch and distribution. The communication protocol between the server and the hardware device is a request-response mode protocol between the communication requester and the communication responder. The communication protocol data packet consists of a header, a prefix, a data segment, a checksum, and a tail. The prefix includes an instruction number, a data packet number, and a unique device number. The technical solution of the above application combines intelligent hardware with software applications to achieve intelligent, real-time monitoring and deployment of the LNG industry chain counties, achieving the optimal solution and saving manpower and time.

[0005] Patent No. CN2015105040527 discloses a remote monitoring system for LNG vessels. The system consists of ship-side equipment and a shore-side data center. The ship-side equipment primarily includes a display processor, a video monitor, and a 3G / 4G / VPN router, while the shore-side data center primarily includes multiple functional servers, multiple switches, and a VPN router. This system utilizes an IoT network to connect with client devices, forming an interconnected LNG vessel remote monitoring system. The ship-side equipment and shore-side data center together form an information monitoring system, which implements all of the system's monitoring functions. Its implementation methods include: 1) data collection; 2) analysis and transmission; 3) classification and preprocessing; 4) targeted processing; and 5) real-time adjustment. The collected data is transmitted back to the system platform via the network, enabling dynamic monitoring of the operating status and parameters of the vessel's equipment. This allows for timely analysis and identification of equipment faults, enabling timely and effective repair and maintenance of the equipment, and ultimately achieving information-based management of the operating vessel.

[0006] Although the above patents all utilize remote monitoring systems to remotely monitor the LNG industry chain (including point supply stations), when several on-site monitoring devices remotely transmit real-time data to the remote monitoring end, the real-time data loss rate is high due to the influence of network status and transmission distance, resulting in a high fault error rate and low fault detection efficiency of the LNG point supply station. Summary of the Invention

[0007] The purpose of the present invention is to provide an intelligent unmanned LNG point supply station monitoring system, which can reduce the fault error rate of the LNG point supply station by using a special marking identifier as the unique identifier of the LNG point supply station in the monitoring center; use a multimodal fault prediction model to predict LNG leakage on a non-destructive data table to generate an LNG leakage prediction result, and then control the corresponding accident emergency equipment to perform emergency control on the LNG point supply station; so as to improve the fault detection efficiency of the LNG point supply station.

[0008] The present invention utilizes the following technical solutions:

[0009] An intelligent unmanned LNG point supply station monitoring system includes a monitoring center and several on-site monitoring devices, including:

[0010] On-site monitoring equipment includes data acquisition equipment, a control and analysis center, accident emergency equipment, and a data transmission module. The on-site monitoring equipment collects and processes environmental video data, natural gas flow data, and natural gas concentration data of the LNG point supply station through the data acquisition equipment to form a lossless data table. The control and analysis center uses a multimodal fault prediction model to predict LNG leakage on the lossless data table to generate an LNG leakage prediction result, thereby controlling the corresponding accident emergency equipment to perform emergency control on the LNG point supply station. The data transmission module transmits the LNG leakage prediction result, environmental video data, natural gas flow data, and natural gas concentration data to the monitoring center.

[0011] The monitoring center includes a storage management unit and a remote monitoring unit. The monitoring center stores and manages environmental video data, natural gas flow data, and natural gas concentration data transmitted by several on-site monitoring devices through the storage management unit, and monitors the working status of the on-site monitoring devices in real time through the remote monitoring unit.

[0012] Preferably, the storage management unit is used to store environmental video data, natural gas flow data and natural gas concentration data transmitted by each on-site monitoring device, and manage the location information and alarm events of each LNG supply station;

[0013] The remote monitoring unit is used to monitor the working status of the on-site monitoring equipment in real time and display it through the remote terminal device.

[0014] Preferably, the data acquisition equipment is used to collect and process environmental video data, natural gas flow data, and natural gas concentration data of each LNG supply station to form a lossless data table;

[0015] The control and analysis center is used to predict LNG leakage based on the non-destructive data table to generate LNG leakage prediction results and emergency control signals, thereby controlling the corresponding emergency equipment to perform emergency control on the LNG supply station;

[0016] Accident emergency equipment, used to perform emergency control of the LNG point supply station based on the emergency control signal from the control and analysis center;

[0017] Data transmission module, used to transmit LNG leakage prediction results and emergency control signals to accident emergency equipment, and to transmit environmental video data, natural gas flow data, and natural gas concentration data to the monitoring center;

[0018] Data acquisition equipment includes pressure transmitters, temperature sensors, combustible gas detectors, liquid level sensors, natural gas flow meters and network cameras; accident emergency equipment includes valve opening converters; data transmission protocols include Modbus protocol and FTP protocol.

[0019] Preferably, the steps of collecting and processing the environmental video data, natural gas flow data and natural gas concentration data by the data acquisition device are as follows:

[0020] S1: The network camera is used to collect environmental video data of the LNG point supply station, and the natural gas flow data of the LNG point supply station is obtained in real time through the natural gas flow meter. The combustible gas detector is also used to detect the natural gas concentration data in the atmospheric environment of the LNG point supply station.

[0021] S2: The data acquisition device uses a data segmentation algorithm to divide the acquired environmental video data, natural gas flow data, and natural gas concentration data into several data blocks according to preset time intervals;

[0022] S3: The data acquisition device uses a data classification algorithm to cluster the information in each data block based on the tag identifier generated by the basic information of each LNG supply station to obtain several small data blocks;

[0023] S4: The data collection device uses a data association algorithm to associate and normalize the contents of the small data blocks according to the identification of each LNG supply station to construct a complete data table for each LNG supply station;

[0024] S5: The data acquisition device uses an encryption compression algorithm to encrypt and losslessly compress the complete data table to form a lossless data table.

[0025] Preferably, the control and analysis center uses a multimodal fault prediction model to perform LNG leakage prediction on the non-destructive data table to generate an LNG leakage prediction result in the following steps:

[0026] A: The multimodal fault prediction model uses a data separation layer to divide the lossless data table into image and video data and natural gas data. It then extracts and enhances the image and video data frame by frame to obtain an enhanced image frame set.

[0027] B: Using the first extraction branch of the feature extraction layer to perform feature extraction on the enhanced image frame set to obtain image frame feature values; using the second extraction branch of the feature extraction layer to perform feature extraction on the natural gas data to obtain natural gas feature values;

[0028] C: Using the target detection branch of the iterative learning layer, the extracted image frame feature values are iteratively trained several times to obtain an image weight matrix. Using the fault prediction branch of the iterative learning layer, the extracted natural gas feature values are iteratively trained several times to obtain a flow-concentration weight matrix.

[0029] D: The prediction and recognition layer inputs the real-time acquired image video data and natural gas data into the image weight matrix and flow-concentration weight matrix respectively to generate LNG leak prediction results.

[0030] Preferably, step A includes the following steps:

[0031] A1: Decrypt and decompress the lossless data table using the decryption and decompression algorithms;

[0032] A2: Classify the contents of the decompressed lossless data table using a clustering algorithm based on data type. Data types include video format and numerical format.

[0033] A3: Sort the video content in the lossless data table by time sequence and markers to obtain image and video data. Use a data cleaning algorithm to remove noise and outliers from the numerical content in the lossless data table to obtain natural gas data.

[0034] A4: Use a video frame extraction algorithm to convert the image video data into frame-by-frame images, and use an image enhancement algorithm to perform contrast adjustment, brightness adjustment, edge enhancement, noise removal, histogram equalization, and smoothing and sharpening on the frame-by-frame images to obtain an enhanced image frame set.

[0035] Preferably, the steps for obtaining the marking identification of the LNG point supply station are:

[0036] a: Obtain the latitude and longitude address, network address and station serial number of each LNG supply station;

[0037] The station serial number is obtained by calculating the business license number, production safety license number and special equipment license number of each LNG supply station;

[0038] b: Convert the latitude and longitude addresses into binary coordinate codes, and convert the network address into ternary network codes;

[0039] c: extracting the binary longitude code from the binary coordinate code and concatenating it with the ternary network code to generate a first concatenated code; at the same time, extracting the binary latitude code from the binary coordinate code and concatenating it with the site serial number to generate a second concatenated code;

[0040] d: Performing a Hadamard operation on the first concatenated code and the site serial number to generate a first identification code; and performing a Kronecker operation on the second concatenated code and the ternary network code to generate a second identification code;

[0041] e: Perform bilinear interpolation operation on the first identification code and the second identification code to obtain the marking identification of each LNG supply station.

[0042] The present invention reduces the fault error rate of the LNG point supply station by using a special marking identifier as the unique identifier of the LNG point supply station in the monitoring center; uses a multimodal fault prediction model to predict LNG leakage on a non-destructive data table to generate an LNG leakage prediction result, and then controls the corresponding accident emergency equipment to perform emergency control on the LNG point supply station, thereby improving the fault detection efficiency of the LNG point supply station. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0044] Figure 1 This is the principle block diagram of the LNG point supply station monitoring system;

[0045] Figure 2 This is the principle block diagram of the multi-modal fault prediction model;

[0046] Figure 3 A flowchart for obtaining a tag identifier. DETAILED DESCRIPTION

[0047] The present invention is described in detail below with reference to the accompanying drawings and embodiments:

[0048] like Figure 1-Figure 3 As shown, the present invention provides an intelligent unmanned LNG point supply station monitoring system, a monitoring center and several on-site monitoring devices, wherein:

[0049] On-site monitoring equipment includes data acquisition equipment, a control and analysis center, accident emergency equipment, and a data transmission module. The on-site monitoring equipment collects and processes environmental video data, natural gas flow data, and natural gas concentration data of the LNG point supply station through the data acquisition equipment to form a lossless data table. The control and analysis center uses a multimodal fault prediction model to predict LNG leakage on the lossless data table to generate an LNG leakage prediction result, thereby controlling the corresponding accident emergency equipment to perform emergency control on the LNG point supply station. The data transmission module transmits the LNG leakage prediction result, environmental video data, natural gas flow data, and natural gas concentration data to the monitoring center.

[0050] The monitoring center includes a storage management unit and a remote monitoring unit. The monitoring center stores and manages environmental video data, natural gas flow data, and natural gas concentration data transmitted by several on-site monitoring devices through the storage management unit, and monitors the working status of the on-site monitoring devices in real time through the remote monitoring unit.

[0051] In the present invention, the storage management unit is used to store the environmental video data, natural gas flow data and natural gas concentration data transmitted by each on-site monitoring equipment, and manage the location information and alarm events of each LNG supply station;

[0052] Remote monitoring unit, used to monitor the working status of on-site monitoring equipment in real time and display it through remote terminal equipment;

[0053] In this embodiment, the monitoring center also includes a communication server, a database server, an FTP server, and a Web server; the monitoring center has functions such as real-time data monitoring, map information management, alarm event management, data query and report management, user information management, and system operation logs, which are conventional technologies in this field and will not be described in detail here;

[0054] In the present invention, the on-site monitoring equipment includes data acquisition equipment, a control and analysis center, accident emergency equipment and a data transmission module;

[0055] Data acquisition equipment, used to collect and process environmental video data, natural gas flow data, and natural gas concentration data from each LNG supply station to form a lossless data table;

[0056] The control and analysis center is used to predict LNG leakage based on the non-destructive data table to generate LNG leakage prediction results and emergency control signals, thereby controlling the corresponding emergency equipment to perform emergency control on the LNG supply station;

[0057] Accident emergency equipment, used to perform emergency control of the LNG point supply station based on the emergency control signal from the control and analysis center;

[0058] Data transmission module, used to transmit LNG leakage prediction results and emergency control signals to accident emergency equipment, and to transmit environmental video data, natural gas flow data, and natural gas concentration data to the monitoring center;

[0059] Data acquisition equipment includes pressure transmitters, temperature sensors, combustible gas detectors, liquid level sensors, natural gas flow meters, and network cameras; accident emergency equipment includes valve opening converters; data transmission protocols include Modbus and FTP protocols;

[0060] In this embodiment, the on-site monitoring equipment is connected via an RS-485 bus; the data transmission module is a hybrid structure of C / S and B / S; the natural gas flow meter is used to measure instantaneous flow, daily flow, and cumulative flow data. The model of the natural gas flow meter is LWJ-RTU / V1, which supports the Modbus communication protocol; the model of the network camera is CS-C2C-31WFR;

[0061] In the present invention, the steps of collecting and processing environmental video data, natural gas flow data, and natural gas concentration data by the data acquisition device are as follows:

[0062] S1: The data acquisition device collects environmental video data of the LNG point supply station by calling a network camera, and obtains natural gas flow data of the LNG point supply station in real time through a natural gas flow meter. It also detects natural gas concentration data in the atmospheric environment of the LNG point supply station through a combustible gas detector.

[0063] S2: The data acquisition device uses a data segmentation algorithm to divide the acquired environmental video data, natural gas flow data, and natural gas concentration data into several data blocks according to preset time intervals;

[0064] S3: The data acquisition device uses a data classification algorithm to cluster the information in each data block based on the tag identifier generated by the basic information of each LNG supply station to obtain several small data blocks;

[0065] S4: The data collection device uses a data association algorithm to associate and normalize the contents of the small data blocks according to the identification of each LNG supply station to construct a complete data table for each LNG supply station;

[0066] S5: The data acquisition device uses an encryption compression algorithm to encrypt and losslessly compress the complete data table to form a lossless data table;

[0067] In the present invention, the control and analysis center uses a multimodal fault prediction model to perform LNG leakage prediction on the non-destructive data table to generate the LNG leakage prediction result in the following steps:

[0068] A: The multimodal fault prediction model uses a data separation layer to divide the lossless data table into image and video data and natural gas data. It then extracts and enhances the image and video data frame by frame to obtain an enhanced image frame set.

[0069] In this embodiment, step A includes the following steps:

[0070] A1: The control and analysis center uses decryption and decompression algorithms to decrypt and decompress the lossless data table;

[0071] A2: The control and analysis center uses a clustering algorithm to classify the contents of the decompressed lossless data table based on data type. Data types include video format and numerical format.

[0072] A3: The control and analysis center sorts the video content in the lossless data table according to time series and tag identifiers to obtain image and video data. It also uses a data cleaning algorithm to remove noise and outliers from the numerical content in the lossless data table to obtain natural gas data.

[0073] A4: The control and analysis center uses a video frame extraction algorithm to convert the image video data into frame-by-frame images. It then uses an image enhancement algorithm to perform contrast adjustment, brightness adjustment, edge enhancement, noise removal, histogram equalization, and smoothing and sharpening on the frame-by-frame images to obtain an enhanced image frame set.

[0074] B: The multimodal fault prediction model uses the first extraction branch of the feature extraction layer to extract features from the enhanced image frame set to obtain image frame feature values; and uses the second extraction branch of the feature extraction layer to extract features from the natural gas data to obtain natural gas feature values;

[0075] In this embodiment, the first extraction branch uses two 3×3 convolutional layers, three 5×5 convolutional layers, two convolutional blocks consisting of a 3×1 convolutional layer and a 1×3 convolutional layer, and three 3×3 average pooling layers to perform feature extraction on the enhanced image frame set to obtain image frame feature values, thereby obtaining real-time on-site environmental characteristics and on-site equipment characteristics of each LNG supply station;

[0076] The second extraction branch uses three 3×3 convolutional layers, four 5×5 convolutional layers, and one cascaded 5×5 max pooling layer to extract features from the natural gas data to obtain natural gas feature values, thereby analyzing the relationship between the natural gas flow rate in the LNG supply station and the natural gas concentration in the atmospheric environment;

[0077] C: The multimodal fault prediction model uses the target detection branch of the iterative learning layer to iteratively train the extracted image frame feature values several times to obtain an image weight matrix. The fault prediction branch of the iterative learning layer also iteratively trains the extracted natural gas feature values several times to obtain a flow-concentration weight matrix.

[0078] In this embodiment, the target detection branch uses two parallel LETTERbox modules, three serial CBS modules, two C3K2 modules, three parallel SPPF modules, two serial C2PSA modules, and two BCE activation functions to perform several iterative training on the extracted image frame feature values to obtain an image weight matrix. This allows for real-time tracking of suspicious individuals in the on-site environment when a LNG point supply station fails.

[0079] The fault prediction branch uses two serially connected 3×3 convolutional layers, two 5×5 adaptive pooling layers, two InceptionV2 blocks, two InceptionV4 blocks, four 3×3 max pooling layers, and a ReLU activation function to iteratively train the extracted natural gas features to obtain a flow-concentration weight matrix. This allows the network to determine whether a natural gas leak has occurred at the LNG point-of-sale station based on the relationship between the natural gas flow rate at the station and the natural gas concentration in the atmosphere.

[0080] D: The prediction and recognition layer inputs the real-time acquired image video data and natural gas data into the image weight matrix and flow-concentration weight matrix respectively to generate LNG leak prediction results;

[0081] In this embodiment, the prediction and recognition layer also uses three shuffle blocks, four 1×1 convolutional layers, three fully connected layers, two 7×7 adaptive pooling layers, and a feedback optimization function to correct the LNG leak prediction results of the image weight matrix and the flow-concentration weight matrix;

[0082]

[0083] Among them, Loss represents the feedback optimization function, Model 1 Represents the image weight matrix, Model 2 represents the flow-concentration weight matrix, σ represents the number of iterative training layers, G represents the quantization function, μ1 represents the calculation accuracy coefficient, B σ represents the iterative training loss function of each layer, μ2 represents the preset optimization weight coefficient, E σ Indicates the optimization efficiency of each layer’s iterative training;

[0084] In the present invention, the steps for obtaining the marking identity of the LNG point supply station are as follows:

[0085] a: Obtain the latitude and longitude address, network address and station serial number of each LNG supply station;

[0086] The station serial number is obtained by calculating the business license number, production safety license number and special equipment license number of each LNG supply station;

[0087] b: Convert the latitude and longitude addresses into binary coordinate codes, and convert the network address into ternary network codes;

[0088] c: extracting the binary longitude code from the binary coordinate code and concatenating it with the ternary network code to generate a first concatenated code; at the same time, extracting the binary latitude code from the binary coordinate code and concatenating it with the site serial number to generate a second concatenated code;

[0089] d: Performing a Hadamard operation on the first concatenated code and the site serial number to generate a first identification code; and performing a Kronecker operation on the second concatenated code and the ternary network code to generate a second identification code;

[0090] e: Perform bilinear interpolation operation on the first identification code and the second identification code to obtain the marking identification of each LNG supply station.

[0091] Example:

[0092] The workflow of the LNG point supply station monitoring system is as follows: the on-site monitoring equipment collects and processes the environmental video data, natural gas flow data and natural gas concentration data of the LNG point supply station through data acquisition equipment (pressure transmitter, temperature sensor, combustible gas detector, liquid level sensor, natural gas flow meter and network camera) to form a lossless data table: the environmental video data of the LNG point supply station is collected by calling the network camera, and the natural gas flow data of the LNG point supply station is obtained in real time through the natural gas flow meter; the natural gas concentration data in the atmospheric environment of the LNG point supply station is also detected by the combustible gas detector; the data acquisition equipment uses The data segmentation algorithm divides the acquired environmental video data, natural gas flow data, and natural gas concentration data into several data blocks based on preset time intervals. The data acquisition equipment uses a data classification algorithm to cluster the information within each data block based on the tags generated by the basic information of each LNG supply station to obtain several small data blocks. The data acquisition equipment uses a data association algorithm to associate and normalize the content within the small data blocks based on the tags of each LNG supply station to construct a complete data table for each LNG supply station. The data acquisition equipment uses an encryption and compression algorithm to encrypt and losslessly compress the complete data table to form a lossless data table.

[0093] The control and analysis center uses a multimodal fault prediction model to predict LNG leakage on the lossless data table to generate LNG leakage prediction results, and then controls the corresponding accident emergency equipment (valve opening converter) to perform emergency control on the LNG point supply station: the multimodal fault prediction model uses a data separation layer to divide the lossless data table into image and video data and natural gas data, and extracts and enhances the image and video data frame by frame to obtain an enhanced image frame set: the lossless data table is decrypted and decompressed using a decryption algorithm and a decompression algorithm; the content of the decompressed lossless data table is classified according to the data type using a clustering algorithm; the data type includes video format and numerical format; the video format content in the lossless data table is sorted according to time series and tag identifiers to obtain image and video data; and the numerical format content in the lossless data table is removed from noise and outliers using a data cleaning algorithm to obtain natural gas data; the image and video data is converted into frame-by-frame images using a video frame extraction algorithm, and the frame-by-frame images are subjected to contrast adjustment, brightness adjustment, edge enhancement, noise removal, histogram equalization, and smoothing and sharpening using an image enhancement algorithm to obtain an enhanced image frame set;

[0094] Obtain the latitude and longitude addresses, network addresses, and site serial numbers of each LNG point supply station; the site serial number is obtained by calculating the business license number, production safety license number, and special equipment license number of each LNG point supply station; convert the latitude and longitude addresses into binary coordinate codes, and convert the network address into a ternary network code; extract the binary longitude code from the binary coordinate code and concatenate it with the ternary network code to generate a first concatenation code; simultaneously extract the binary latitude code from the binary coordinate code and concatenate it with the site serial number to generate a second concatenation code; perform a Hadamard operation on the first concatenation code and the site serial number to generate a first identification code; simultaneously perform a Kronecker operation on the second concatenation code and the ternary network code to generate a second identification code; perform a bilinear interpolation operation on the first identification code and the second identification code to obtain the marking identification of each LNG point supply station;

[0095] The first extraction branch of the feature extraction layer is used to extract features from the enhanced image frame set to obtain image frame feature values. The second extraction branch of the feature extraction layer is used to extract features from the natural gas data to obtain natural gas feature values. The target detection branch of the iterative learning layer is used to iteratively train the extracted image frame feature values several times to obtain an image weight matrix. The fault prediction branch of the iterative learning layer is used to iteratively train the extracted natural gas feature values several times to obtain a flow-concentration weight matrix. The prediction and recognition layer inputs the real-time image video data and natural gas data into the image weight matrix and the flow-concentration weight matrix, respectively, to generate an LNG leak prediction result.

[0096] The data transmission module uses the data transmission protocol (Modbus protocol and FTP protocol) to transmit LNG leakage prediction results, environmental video data, natural gas flow data and natural gas concentration data to the monitoring center; at the same time, the storage management unit of the monitoring center stores the environmental video data, natural gas flow data and natural gas concentration data transmitted by each on-site monitoring equipment, and manages the location information and alarm events of each LNG supply station; the remote monitoring unit of the monitoring center monitors the working status of the on-site monitoring equipment in real time and displays it through the remote terminal equipment.

Claims

1. An intelligent unmanned LNG point supply station monitoring system, characterized by: It includes a monitoring center and several on-site monitoring devices, including: On-site monitoring equipment, including data acquisition equipment, control and analysis center, accident emergency equipment and data transmission modules; Data acquisition equipment: This equipment collects environmental video data from LNG point supply stations by calling network cameras, and obtains natural gas flow data from LNG point supply stations in real time through natural gas flow meters. Combustible gas detectors detect natural gas concentration data in the atmospheric environment of LNG point supply stations. Data segmentation algorithms are used to divide the acquired data into several data blocks based on preset time intervals. Data classification algorithms are used to cluster the information within each data block based on the tags generated by the basic information of each LNG point supply station to obtain several small data blocks. Data association algorithms are used to associate and normalize the contents of the small data blocks based on the tags of each LNG point supply station, and a complete data table for each LNG point supply station is constructed. The complete data table is encrypted and losslessly compressed using an encryption and compression algorithm to form a lossless data table. The control and analysis center uses a multimodal fault prediction model: using the data separation layer, the lossless data table is divided into image and video data and natural gas data, and the image and video data are extracted and enhanced frame by frame to obtain an enhanced image frame set; the first extraction branch of the feature extraction layer is used to extract features from the enhanced image frame set to obtain image frame feature values; the second extraction branch of the feature extraction layer is used to extract features from the natural gas data to obtain natural gas feature values; the target detection branch of the iterative learning layer is used to iteratively train the extracted image frame feature values several times to obtain an image weight matrix; the fault prediction branch of the iterative learning layer is used to iteratively train the extracted natural gas feature values several times to obtain a flow-concentration weight matrix; the prediction and recognition layer inputs the real-time image and video data and natural gas data into the image weight matrix and the flow-concentration weight matrix, respectively, to generate an LNG leak prediction result; Data transmission module: transmits LNG leakage prediction results, environmental video data, natural gas flow data and natural gas concentration data to the monitoring center; Monitoring center: includes a storage management unit and a remote monitoring unit; the storage management unit stores and manages environmental video data, natural gas flow data, and natural gas concentration data transmitted by several on-site monitoring devices, and the remote monitoring unit monitors the working status of the on-site monitoring devices in real time.

2. The intelligent unmanned LNG point supply station monitoring system according to claim 1 is characterized by: The data acquisition equipment is used to collect and process environmental video data, natural gas flow data and natural gas concentration data of each LNG supply station to form a lossless data table; The control and analysis center is used to predict LNG leakage based on the non-destructive data table to generate LNG leakage prediction results and emergency control signals, thereby controlling the corresponding emergency equipment to perform emergency control on the LNG supply station; Accident emergency equipment, used to perform emergency control of the LNG point supply station based on the emergency control signal from the control and analysis center; Data transmission module, used to transmit LNG leakage prediction results and emergency control signals to accident emergency equipment, and to transmit environmental video data, natural gas flow data, and natural gas concentration data to the monitoring center; Data acquisition equipment includes pressure transmitters, temperature sensors, combustible gas detectors, liquid level sensors, natural gas flow meters and network cameras; accident emergency equipment includes valve opening converters; data transmission protocols include Modbus protocol and FTP protocol.

3. The intelligent unmanned LNG point supply station monitoring system according to claim 1 is characterized by: The storage management unit is used to store environmental video data, natural gas flow data and natural gas concentration data transmitted by each on-site monitoring device, and manage the location information and alarm events of each LNG supply station; The remote monitoring unit is used to monitor the working status of the on-site monitoring equipment in real time and display it through the remote terminal device.

4. The intelligent unmanned LNG point supply station monitoring system according to claim 1 is characterized by: The control and analysis center uses a decryption algorithm and a decompression algorithm to decrypt and decompress the lossless data table; the control and analysis center uses a clustering algorithm to classify the contents of the decompressed lossless data table according to the data type; the data type includes video format and numerical format; the control and analysis center sorts the video format contents in the lossless data table according to time sequence and tag identifiers to obtain image video data; The numerical format content in the lossless data table is then cleaned using a data cleaning algorithm to remove noise and outliers to obtain natural gas data. The control and analysis center uses a video frame extraction algorithm to convert the image video data into frame-by-frame images, and uses an image enhancement algorithm to perform contrast adjustment, brightness adjustment, edge enhancement, noise removal, histogram equalization, and smoothing and sharpening on the frame-by-frame images to obtain an enhanced image frame set.

5. The intelligent unmanned LNG point supply station monitoring system according to claim 1 is characterized by: The steps for obtaining the marking identity of the LNG point supply station are as follows: a: Obtain the latitude and longitude address, network address and station serial number of each LNG supply station; The station serial number is obtained by calculating the business license number, production safety license number and special equipment license number of each LNG supply station; b: Convert the latitude and longitude addresses into binary coordinate codes, and convert the network address into ternary network codes; c: extracting the binary longitude code from the binary coordinate code and concatenating it with the ternary network code to generate a first concatenated code; at the same time, extracting the binary latitude code from the binary coordinate code and concatenating it with the site serial number to generate a second concatenated code; d: Performing a Hadamard operation on the first concatenated code and the site serial number to generate a first identification code; and performing a Kronecker operation on the second concatenated code and the ternary network code to generate a second identification code; e: Perform bilinear interpolation operation on the first identification code and the second identification code to obtain the marking identification of each LNG supply station.

Images