A liquefied natural gas sampling and component analysis system

Abstract

The application relates to the technical field of liquefied natural gas analysis, and discloses a liquefied natural gas sampling and component analysis system. A sampling monitoring module of the system collects pipeline fluid parameters and flow states, analyzes flow velocity fluctuation characteristics of a sampling point of a pipe section, calculates a historical opening frequency of a sampling port, identifies an abnormal sampling node, and generates sampling distribution characteristic parameters; a component prediction module locates a component analysis unit corresponding to the abnormality based on the parameters, predicts a potential deviation interval of component data, and generates component deviation distribution characteristics; a path planning module identifies high-deviation-risk storage nodes, analyzes medium distribution density, calculates medium migration emergency degree, plans a transfer path, distributes the medium to low-risk nodes, generates transfer configuration parameters, and compares sampling point component detection frequencies before and after transfer, analyzes data fluctuation characteristics, identifies abnormal change modes, determines a deviation source, and generates traceability characteristic data.

Description

Technical Field

[0001] This invention relates to the field of liquefied natural gas (LNG) analysis technology, specifically to a LNG sampling and component analysis system. Background Technology

[0002] Liquefied natural gas (LNG), as a clean and efficient energy source, continues to increase its share in the global energy mix. Accurate analysis of its components is crucial for trade settlement, safety regulation, and process optimization. However, current LNG sampling and component analysis processes still face numerous technical bottlenecks. In the sampling stage, traditional sampling systems rely heavily on fixed sampling points for data collection, making it difficult to dynamically adapt to the complex flow conditions within pipelines. Flow velocity within pipelines often fluctuates significantly due to factors such as transport pressure and changes in medium composition. Fixed sampling points are prone to insufficient representativeness due to velocity fluctuations, and existing systems lack systematic analysis of the historical opening frequency of each sampling port, making it difficult to quickly identify abnormal sampling points caused by equipment aging or improper operation. This leads to biases in the foundation of subsequent component analysis data.

[0003] In the component analysis stage, existing technologies for managing analysis units are mostly limited to single data detection, lacking in-depth analysis of historical detection records. When component anomalies occur, it is difficult to quickly match the characteristic patterns of historical anomaly events, and it is impossible to effectively assess the data reliability level of different analysis units. This leads to the inability to accurately predict the potential deviation range of component data, affecting the judgment of the true state of liquefied natural gas components. Simultaneously, in the media storage stage, existing systems fail to dynamically identify the risk level of storage nodes based on component deviation characteristics, and the correlation analysis between media distribution density and migration urgency is insufficient. This often results in media accumulation at high-risk deviation nodes, increasing safety hazards and resource waste during storage.

[0004] In terms of deviation tracing, traditional methods rely heavily on manual investigation and lack systematic comparative analysis of changes in the frequency of sampling point detection and fluctuation characteristics of component data before and after transfer. This makes it difficult to accurately identify abnormal component change patterns, resulting in difficulties in determining the source of component data deviations. Consequently, it is impossible to form a closed-loop management system from sampling to tracing, which restricts the refined development of liquefied natural gas analysis technology. Summary of the Invention

[0005] The purpose of this invention is to provide a liquefied natural gas sampling and component analysis system to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides a liquefied natural gas sampling and component analysis system, the system comprising:

[0007] The sampling and monitoring module collects fluid parameters and flow status in the liquefied natural gas pipeline, analyzes the flow velocity fluctuation characteristics of sampling points in different pipe sections, calculates the historical opening frequency of each sampling port, compares the current flow velocity fluctuation characteristics with the historical opening frequency, identifies abnormal sampling nodes, and generates sampling distribution characteristic parameters.

[0008] The component prediction module, based on the sampling distribution characteristic parameters, locates the component analysis unit corresponding to the abnormal sampling node, analyzes the historical detection records of the component analysis unit, matches the characteristic patterns of component abnormal events, evaluates the data reliability level of each analysis unit, predicts the potential deviation range of component data, and generates component deviation distribution characteristics.

[0009] The path planning module, based on the component deviation distribution characteristics, identifies component storage nodes with high deviation risk, analyzes the media distribution density of each storage node, calculates the urgency of media migration, plans media transfer paths, reallocates target media to low-risk storage nodes, and generates media transfer configuration parameters.

[0010] The traceability analysis module, based on the medium transport configuration parameters, compares the component detection frequencies at each sampling point before and after transport, analyzes the fluctuation characteristics of component data, identifies abnormal component change patterns, determines the source of component data deviation, and generates component traceability feature data.

[0011] Preferably, the component prediction module includes:

[0012] The abnormal node matching submodule, based on the sampling distribution characteristic parameters, filters the component detection data of abnormal sampling nodes, analyzes the correlation between the component detection time series and standard physical property parameters, classifies the detection data categories with different degrees of deviation, identifies high-risk component analysis units, and generates an abnormal analysis unit set.

[0013] The credibility level assessment submodule calls the set of anomaly analysis units, extracts the fluctuation characteristics of the detection data of each unit, compares the data patterns of known component anomaly events, calculates the matching degree between the analysis unit and the anomaly event, updates the data credibility level based on the matching results, and generates credibility level assessment indicators.

[0014] The deviation interval prediction submodule, based on the credibility level assessment index, statistically analyzes the detection frequency of high-risk analysis units, extracts the time interval of continuous detection, calculates the fluctuation range of short-term detection results, and generates component deviation distribution characteristics by combining the maximum deviation interval of the component data predicted by the credibility level.

[0015] Preferably, the path planning module includes:

[0016] The risk node identification submodule scans high-risk nodes in the component storage network based on the component deviation distribution characteristics, analyzes the component sensitivity of the node storage medium, filters storage nodes with deviations exceeding the threshold, determines the target node range for media that need to be migrated, and generates a risk node list.

[0017] The migration priority calculation submodule analyzes the media distribution status of the target nodes based on the risk node list, calculates the physical property correlation and transfer frequency of the media between nodes, determines the migration order based on the correlation and transfer frequency, and generates a media migration priority index.

[0018] The transfer path generation submodule analyzes the physical connection topology of the storage nodes based on the media migration priority index, matches the capacity parameters of available transfer devices, plans the shortest feasible path for media migration, updates the access permissions of the target storage nodes, and generates media transfer configuration parameters.

[0019] Preferably, the source tracing analysis module includes:

[0020] The detection frequency comparison submodule, based on the medium transport configuration parameters, obtains the component detection records of each sampling point before and after transport, analyzes the distribution of detection time points and the duration of a single detection, calculates the rate of change of detection frequency per unit time, and generates a detection frequency fluctuation index.

[0021] The abnormal pattern recognition submodule extracts component data that deviates from the standard detection cycle based on the detection frequency fluctuation index, analyzes the distribution pattern of abnormal data, matches the generation patterns of different component deviations, locates the core nodes of component data anomalies, and generates a set of component abnormal patterns.

[0022] The deviation source location submodule, based on the set of component anomaly patterns, traces the original sampling node corresponding to the abnormal data, verifies the status parameters of the sampling equipment, associates the operation records in the medium transport path, determines the final source node of the component deviation, and generates component traceability feature data.

[0023] Preferably, the system further includes:

[0024] The parameter optimization module, based on the component traceability feature data, identifies sampling nodes that continuously generate deviations, adjusts the trigger parameters of the sampling nodes, reallocates the sampling ratio of each pipeline, updates the trigger conditions for component detection, and generates a sampling optimization configuration scheme.

[0025] Preferably, the parameter optimization module includes:

[0026] The sampling node screening submodule, based on the component traceability feature data, counts the frequency of deviation occurrences at each node, calculates the impact weight of the deviation on the overall component analysis, screens sampling nodes whose impact weight exceeds the threshold, and generates a list of high-impact nodes.

[0027] The trigger parameter adjustment submodule, based on the list of high-impact nodes, analyzes historical sampling trigger conditions, evaluates the correlation between trigger conditions and component deviation, reconfigures the sampling trigger time interval and flow rate trigger threshold, and generates a node trigger parameter update table.

[0028] The sampling ratio allocation submodule calculates the percentage change of the medium flow rate in each pipeline based on the node trigger parameter update table, adjusts the sampling frequency ratio of different pipe sections according to the flow rate ratio, and generates a sampling ratio allocation scheme.

[0029] The detection condition update submodule integrates the node trigger parameter update table and the sampling ratio allocation scheme, updates the trigger sensitivity parameters of the component detection equipment, and generates a sampling optimization configuration scheme.

[0030] Preferably, the sampling monitoring module includes:

[0031] The flow velocity fluctuation analysis submodule collects pressure data and temperature gradient of the pipeline fluid, analyzes the flow velocity change curves of continuous sampling points, calculates the flow velocity difference between adjacent sampling points, counts the flow velocity fluctuation frequency within a preset period, and generates flow velocity fluctuation characteristic parameters.

[0032] The sampling frequency statistics submodule retrieves sampling operation records from the historical database based on the flow velocity fluctuation characteristic parameters, analyzes the distribution pattern of sampling operations in different time periods, calculates the concentration of sampling operations per unit time, and generates a sampling frequency concentration index.

[0033] The abnormal node identification submodule, based on the sampling frequency concentration index, compares the current flow velocity fluctuation characteristics with the historical sampling frequency, detects sampling nodes that exceed the normal fluctuation range, marks the abnormal sampling location and the degree of deviation, and generates sampling distribution characteristic parameters.

[0034] Preferably, the trust level assessment submodule performs the following steps:

[0035] The system retrieves the detection records of each unit in the anomaly analysis unit set over the past three years and extracts the extreme values ​​of component concentrations and the time points when they occurred from the records.

[0036] The concentration fluctuation patterns of known component anomalies are matched, the temporal overlap between the unit detection data and the anomaly patterns is calculated, and the data confidence level is set according to the temporal overlap.

[0037] Preferably, the migration priority calculation submodule performs the following steps:

[0038] Obtain the media component type of the nodes in the risk node list, and analyze the physical property compatibility and chemical reaction risk between the types;

[0039] Calculate the safety factor for media exchange between nodes and count the frequency of emergency operations in historical transfer records;

[0040] The migration order is determined based on the safety factor and the frequency of emergency operations.

[0041] Preferably, the deviation source localization submodule performs the following steps:

[0042] By associating the set of abnormal component patterns with the media transfer log, the transfer operation node at the time the abnormal pattern occurred can be traced.

[0043] Verify the equipment calibration records at the corresponding sampling points, analyze the temporal correlation between equipment calibration deviation and component abnormalities, and identify the starting operation node that caused the component deviation.

[0044] Compared with the prior art, the beneficial effects of the present invention are:

[0045] This liquefied natural gas (LNG) sampling and component analysis system achieves dynamic sensing of the fluid state within the pipeline during the sampling phase through multi-module collaborative operation. The sampling monitoring module continuously analyzes the flow velocity fluctuation characteristics at different sampling points in different pipe sections and, combined with the historical opening frequency of each sampling port, can promptly identify abnormal sampling nodes, ensuring the representativeness and reliability of the collected samples and reducing subsequent analysis errors caused by sampling deviations. By generating sampling distribution characteristic parameters, it provides accurate basic data support for the entire analysis process, making subsequent component analysis more closely aligned with actual operating conditions.

[0046] At the component analysis level, the component prediction module accurately locates the component analysis unit corresponding to the abnormal sampling node based on sampling distribution characteristic parameters. Through in-depth analysis of historical detection records, it can effectively match the characteristic patterns of component anomaly events. The assessment of the data reliability level of each analysis unit and the prediction of potential deviation ranges in component data allow analysts to clearly understand the reliability of the data and avoid misjudgments caused by blindly accepting data. The generated component deviation distribution characteristics provide crucial basis for subsequent risk assessment of storage nodes, enabling component analysis to shift from single data detection to a systematic analysis with controllable data quality.

[0047] Based on the component deviation distribution characteristics, the path planning module accurately identifies high-risk component storage nodes. By analyzing the media distribution density of each storage node and considering the urgency of media migration, it formulates a scientifically sound media transfer path. The target media is then redistributed to low-risk storage nodes, optimizing the media storage layout and reducing potential security issues caused by media accumulation at high-risk nodes. This improves the rationality and security of media storage, enabling more efficient utilization of storage resources.

[0048] The source tracing analysis module compares the component detection frequencies at each sampling point before and after transport, deeply analyzing the fluctuation characteristics of component data. This allows for accurate identification of abnormal component change patterns, thus precisely determining the source of component data deviations. This process enables end-to-end tracking from sampling anomalies to deviation tracing, helping technicians to take targeted improvement measures to address deficiencies in sampling, analysis units, or storage processes. This forms a continuously optimized closed-loop management system, driving LNG sampling and component analysis technologies towards greater accuracy and efficiency. The close collaboration between modules ensures the entire system forms an organic whole in data acquisition, analysis, storage optimization, and problem tracing, enhancing the systematic nature and completeness of LNG analysis. Attached Figure Description

[0049] Figure 1 This is a timeline diagram of the liquefied natural gas sampling and component analysis system described in this invention.

[0050] Figure 2 This is a flowchart of the component prediction module.

[0051] Figure 3 A flowchart of the path planning module;

[0052] Figure 4 This is a flowchart of the source tracing analysis module. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] Please see Figure 1 The present invention provides a liquefied natural gas sampling and component analysis system, the system comprising a closed-loop processing flow of a sampling monitoring module, a component prediction module, a path planning module and a source tracing analysis module.

[0055] The sampling and monitoring module collects real-time pipeline fluid pressure data and temperature gradients. The flow velocity fluctuation analysis submodule calculates the velocity difference between adjacent sampling points, and the historical sampling concentration index generated by the sampling frequency statistics submodule is combined with this. The anomaly node identification submodule marks sampling locations that deviate from the normal fluctuation range. The component prediction module receives sampling distribution characteristic parameters. The anomaly node matching submodule classifies the detection data deviation level, the credibility level assessment submodule calculates the matching degree between the analysis unit and historical anomaly events, and the deviation interval prediction submodule generates component deviation distribution characteristics based on the credibility level. The path planning module's risk node identification submodule scans sensitive media nodes in the storage network, the migration priority calculation submodule determines the migration order based on physical property correlation, and the transfer path generation submodule plans the optimal transfer path based on equipment capacity parameters. The source tracing analysis module calculates the rate of change in detection frequency before and after transfer through the detection frequency comparison submodule, the anomaly pattern identification submodule matches the generation pattern of component deviations, and the deviation source location submodule correlates transfer logs to determine the final source node.

[0056] Example 1: See Figure 2 The system revolves around the functionality of the component prediction module. This module consists of three sub-modules working collaboratively: the anomaly node matching sub-module receives the sampling distribution characteristic parameters and immediately initiates a database retrieval program to retrieve the associated component analysis unit dataset marked as anomaly sampling points from the historical repository. The dataset time range is set to the most recent 36 months by default, and the content includes the concentration detection values ​​and corresponding detection timestamps of major components such as methane, ethane, propane, and butane. The system uses a sliding time window algorithm to process the data stream, with the time window length set to 30 minutes. The algorithm scans the concentration time series of each analysis unit in real time. When it identifies a unit where the concentration fluctuation exceeds the baseline value ±5% within three consecutive detection periods, the system automatically classifies the unit into the high-risk detection data category. The criteria for determining the high-risk category are dynamically adjusted according to the component properties, using a strict threshold of ±3% for methane components and a lenient threshold of ±7% for heavy hydrocarbon components.

[0057] When the trust level assessment submodule is activated, it retrieves a predefined feature pattern library of 12 component anomaly events from the core database. The pattern library includes typical anomaly data patterns such as sudden concentration increases, step-down decreases, and sawtooth oscillations, each described by a 128-dimensional feature vector. The subsystem creates a temporary data buffer for each anomaly analysis unit marked as high-risk, extracting the unit's detection data stream from the past 90 days. The data processing employs a standardized noise reduction algorithm to eliminate data disturbances caused by inherent equipment noise. The assessment algorithm calculates the cosine similarity value between the current data stream and the feature vectors of each anomaly pattern, using an overlapping time alignment mechanism. When the similarity calculation result between a unit's data and any anomaly pattern consistently exceeds the 0.85 threshold for more than three detection cycles, the system marks it as a highly matched state. Data reliability is classified into a five-level scale: Level A represents highly reliable data with a match rate below 0.3; Level B represents reliable data with a match rate of 0.3-0.6; Level C represents questionable data with a match rate of 0.6-0.8; Level D represents unreliable data with a match rate of 0.8-0.9; and Level E represents severely anomalous data with a match rate exceeding 0.9. The level assessment results are automatically updated every 24 hours, and Level D and Level E data are specially marked and stored in isolation.

[0058] The deviation interval prediction submodule employs a differentiated processing mechanism based on the credibility level assessment results. For Class C analysis units, the system initiates a time series decomposition algorithm, breaking down the detected data stream into three parts: trend component, periodic component, and random component. The algorithm uses Hodrick-Prescott filtering to extract the trend term, Fourier transform to extract the periodic term, and classifies the remaining part as random fluctuations. In the random component analysis, the three-standard-deviation interval within a 30-day moving window is calculated, and its upper and lower limits are used as short-term prediction boundaries. For Class B units, the analysis window is extended to 60 days, and the standard deviation coefficient is widened to twice the normal value. Class A units only retain a one-standard-deviation interval for monitoring. For Class D and E units, the system skips the real-time analysis stage and directly calls the unit's historical maximum fluctuation record. The algorithm retrieves the concentration range data of the unit since it was included in the monitoring system, taking 120% of the maximum positive deviation value as the upper boundary and 120% of the maximum negative deviation value as the lower boundary to form a conservative prediction interval. The deviation intervals for all units are recalculated hourly and pushed to the path planning module via a distributed message queue.

[0059] Lightweight encryption protocols are used to ensure data integrity during data transmission. When the abnormal node matching submodule transmits data to the trust level assessment submodule, the encapsulation format includes 26 fields such as unit number, timestamp, and component concentration matrix. The deviation interval prediction submodule outputs results using binary encoding, with each frame containing a 128-byte header and a variable-length data body. The header embeds a CRC checksum and version identifier, and the data body stores the floating-point upper and lower limits of the deviation interval for each component. When the overall load of the prediction module exceeds 70%, the system automatically activates a data sampling mechanism, processing only the data streams of the most recently marked high-risk units, while the remaining data enters a delayed processing queue. The module's operating status is monitored in real time via a heartbeat mechanism; if any submodule remains unresponsive for more than 300 milliseconds, a disaster recovery backup process is triggered.

[0060] Example 2: See Figure 3 The system revolves around the functional implementation of the path planning module. This module achieves intelligent decision-making for media transfer through the cascading processing of three sub-modules. Upon startup, the risk node identification sub-module loads component deviation distribution characteristic data to establish a three-dimensional risk assessment model. The first dimension of the model is the storage pressure sensitivity coefficient, calculated from real-time data collected by a pressure sensor array; the second dimension is the temperature gradient influence factor, generated based on the scanning results of an infrared thermal imaging system; and the third dimension is the media component sensitivity index, derived from the detection report of a gas chromatograph. The system performs a three-dimensional parameter scan on each storage node. When the ethane content exceeds 30% and the pressure is below 1.2 MPa, the node is marked as a red warning; when the propane content exceeds 25% and the temperature gradient exceeds 5℃ / m, the node is marked as a yellow warning. The node status is refreshed every 15 minutes, and all warning nodes are aggregated to generate a risk node list.

[0061] The migration priority calculation submodule employs a dynamic weight allocation mechanism to process the risk node list. The module first constructs a media compatibility matrix, with dimensions representing all possible media component types. Matrix element values ​​are calculated using the following formula:

[0062]

[0063] in: This represents the safety factor for components i and j. and These represent the characteristic concentration values ​​of the two components, respectively. The maximum allowable concentration difference of the system. and This is the current temperature reading. The critical reaction temperature is defined as the temperature at which the media combinations have a safety factor below 0.6. Direct contact with these media combinations is automatically prohibited by the system. The subsystem simultaneously accesses the operation log database, retrieving the number of emergency transfer records triggered by each risk node within the past 24 months. An emergency transfer is defined as an unplanned operation with a response time of less than 30 minutes. The system normalizes the safety factor and the number of emergency transfers to generate a priority score from 0 to 100. The scoring algorithm uses a piecewise linear transformation, with the safety factor accounting for 60% of the weight and the number of emergency transfers accounting for 40%. The scores are reordered every 30 minutes, and the top 20% of high-priority node combinations are pushed to the transfer path generation submodule.

[0064] The transfer path generation submodule employs an improved graph theory algorithm to handle the storage network topology. The system models the entire storage facility as a weighted directed graph, where nodes represent storage containers and edges represent connecting pipelines. The edge weights are dynamically calculated using multiple parameters, including the actual pipeline length, the length converted from the number of bends, the length compensated for elevation changes, and the converted value of the currently available pumping capacity. The algorithm maintains a real-time updated equipment status table, recording the current flow margin and maximum operating pressure of each transfer pump. When the planned path needs to pass through multiple relay nodes, the system automatically calculates segmented transfer schemes. Each segmented scheme includes a precise pump start-up and shutdown time sequence, valve opening and closing sequence, and the expected media transfer volume. The path planning process considers contingency response strategies; when an abnormal pressure is detected in a pipeline segment, the backup path calculation process is immediately initiated. All generated transfer path schemes are verified through virtual simulation, where the system uses a computational fluid dynamics model to predict pressure fluctuations and temperature changes during media transfer. Verified schemes are converted into equipment control command sequences, and the access permission identifiers of the relevant storage nodes are updated. Access management employs a role-based access control model, granting differentiated path execution permissions to operators at different levels.

[0065] The system implementation involved several key technical details. The media compatibility matrix was constructed using an incremental update mechanism, automatically expanding the matrix dimensions whenever a new component type was added. A temperature compensation factor was introduced into the safety factor calculation to eliminate assessment bias caused by seasonal temperature differences. The priority scoring algorithm used a dynamically adjusting window to automatically reduce the calculation weight of emergency calls during equipment maintenance. The graph theory algorithm employed a bidirectional search strategy, simultaneously expanding the search range from both the starting and ending points towards the center, significantly improving path planning efficiency. Pipeline weight calculation used a moving average filter to process sensor data, eliminating interference from instantaneous fluctuations. Equipment status monitoring used a publish-subscribe model, with any changes in pumping parameters notifying all relevant calculation modules in real time. The virtual simulation employed multi-resolution modeling technology, using fine meshing for critical pipe sections and simplified models for ordinary pipe sections. The control command generation process included multiple verification steps to ensure millisecond-level time accuracy for each equipment action command. Access control integrated biometric recognition technology, requiring operators to undergo dual authentication via fingerprint and iris scans to execute critical transfer commands.

[0066] The data processing flow features a strict timing control mechanism. The output of the risk node identification submodule is serialized in protobuf format, containing fields such as node number, 3D parameter values, and warning level. The input / output interfaces of the migration priority calculation submodule define strict version compatibility rules to ensure that algorithm upgrades do not affect existing processes. The control command packets generated by the transfer path generation submodule are encrypted using AES-256, with each packet accompanied by a digital signature and timestamp. The system's operating status is displayed in real-time through a distributed monitoring platform, including the number of currently active warning nodes, the progress of ongoing transfer tasks, and key equipment operating parameters. All submodules employ a tiered response strategy for anomaly handling, with corresponding recovery mechanisms for everything from simple data retransmission to complex fault switching. The system design considers degraded operation capabilities under extreme conditions; when some sensors fail, the transfer task can continue based on historical data to estimate current parameters. The task interruption recovery mechanism ensures that any unexpectedly stopped transfer process can resume from the point of interruption, avoiding media waste and equipment idleness. The system maintenance mode supports online hot updates, allowing adjustments to key algorithms and parameters without stopping ongoing transfer tasks.

[0067] Example 3: See Figure 4The focus is on the functionality of the source tracing analysis module. This module precisely pinpoints the source of component deviations through progressive data processing. When the detection frequency comparison submodule starts, it establishes a ±2-hour time window based on the media transfer completion time. The system automatically retrieves component detection records from all sampling points within this window, extracting the timestamp and single-detection duration fields. High-precision clock synchronization technology aligns the timelines of each sampling point, compensating for clock drift between devices. The data processing algorithm calculates the rate of change of detection frequency per unit time, using the following formula:

[0068]

[0069] in: Indicates the rate of change in detection frequency. This represents the average detection interval after transport. It is the average testing interval before transfer. The calibration coefficient configured for the system (default is 0.85). When the calculated value exceeds the threshold of 0.15, the system generates a corresponding detection frequency fluctuation index. This index is associated with the sampling point coordinate information and time distribution characteristics, forming a spatiotemporal composite data packet that is transmitted to downstream modules.

[0070] After receiving the data packet, the anomaly pattern recognition submodule activates the pattern matching engine. The system maintains a recognition library containing eight basic anomaly patterns: Type A (equal-interval missing), Type B (burst dense detection), Type C (random-interval anomaly), Type D (duration drift), Type E (compound frequency disorder), Type F (periodic fluctuation anomaly), Type G (trend shift), and Type H (step change). Each pattern is quantified and described by a 128-dimensional feature vector, including parameters such as time distribution histogram statistics and frequency change gradient. The matching process uses a dynamic time warping algorithm to calculate the minimum warped path distance between the current detected sequence and the baseline sequences of each pattern. When the matching distance between a sequence and the baseline pattern is lower than a set threshold (default 6.5), the system classifies it into the corresponding anomaly category. The matching result is appended with a confidence score, calculated based on the smoothness of the warped path. All identified anomalous data points are labeled with their spatial location and time range, forming a set of component anomaly patterns with weighted coefficients.

[0071] The deviation source localization submodule initiates a reverse tracing mechanism. This mechanism comprises a three-level processing flow: Level 1 tracing starts from the anomaly detection point and searches upstream nodes along the media flow direction. The system calls the pipeline topology database and, combined with the fluid dynamics model, calculates the media transmission path, accurate to each pipeline connection point and valve position. Level 2 operation record association accesses the operation log database, establishing a ±15-minute search window with the anomaly occurrence time as the midpoint. The system extracts operation record fields from all relevant nodes within this time period, including: valve opening adjustment timestamps, pump speed changes, temperature adjustment commands, pressure control actions, and other key parameters. Level 3 equipment status verification calls the calibration records from the equipment management system, focusing on verifying parameters such as the linearity error of the sampling valve, the validity period of the chromatograph calibration certificate, and the zero-point drift value of the pressure sensor. The system establishes a time event chain model, projecting equipment status change points, operation execution points, and anomaly detection points onto a unified time axis. When an anomaly in equipment status is detected earlier than the operation execution time, and the time difference between the two is less than the media transmission time, the system marks the node containing that equipment as a potential deviation source. In the final determination phase, correlation verification is performed, calculating the Pearson correlation coefficient between equipment deviation values ​​and detected anomalies. Nodes with a coefficient exceeding 0.75 are confirmed as the final source nodes. All determination results are encapsulated as component tracing feature data output, with the data structure including a tracing path node sequence, a key evidence index list, and a comprehensive confidence score.

[0072] The technical implementation involves optimization of several core algorithms. The detection frequency analysis employs a multi-scale processing method, configuring differentiated analysis granularity for sampling points of different importance levels: millisecond-level timestamp analysis is used for critical nodes, while second-level analysis is used for ordinary nodes. The dynamic time warping algorithm incorporates constraints, limiting the slope of the warped path to the range of [0.5, 2] to avoid excessive distortion of the time axis. The reverse tracing algorithm uses a heuristic search strategy, prioritizing the expansion of path branches with densely recent operation records. The equipment status verification module integrates a machine learning model; the training data includes the correspondence between historical calibration records and fault cases, and the model automatically identifies the correlation patterns between specific equipment parameter deviations and component anomaly types.

[0073] The system operates with a rigorous data protection mechanism. Detection records are transmitted using timestamped blockchain storage technology to ensure data immutability. Pattern matching is performed in a sandbox environment to prevent abnormal data from contaminating the core system. The reverse tracing process records the complete decision-making path, supporting result backtracking verification. The device status database undergoes hourly consistency checks to eliminate data conflicts. All critical judgment results must be confirmed through a dual-person review mechanism; expert intervention is triggered when review discrepancies exceed allowable limits. Calculation results are output in a standardized message format, including a 40-bit tracing path code, a 128-bit evidence digest, and an 8-bit confidence identifier. System resource allocation employs dynamic priority scheduling; when multiple tracing tasks run concurrently, critical abnormal events with a confidence level higher than 0.8 are prioritized. Anomaly handling mechanisms include timeout retries, path rollback, and data retransmission to ensure functional integrity under extreme conditions.

[0074] Example 4: This section focuses on the functional implementation of the parameter optimization module, which dynamically adjusts the sampling strategy through multi-level data processing. During system operation, the sampling node screening submodule first loads the component traceability feature data and establishes a deviation impact assessment matrix. This matrix includes three assessment dimensions: the deviation magnitude dimension records the percentage difference between the historical maximum deviation value and the standard value for each node; the duration dimension counts the cumulative duration for which the deviation exceeds the threshold; and the impact range dimension calculates the proportion of downstream node chain reactions caused by the deviation of that node. The system processes the actual operating data of an LNG receiving terminal to obtain assessment data for some nodes, as shown in Table 1.

[0075] Table 1: Evaluation data for some nodes are as follows.

[0076]

[0077] The trigger parameter adjustment submodule performs in-depth analysis on the list of high-impact nodes. The system retrieves the sampling records of node N-3098 over the past 180 days to analyze the correlation between its triggering conditions and component deviations. Data shows that in sampling operations triggered by this node when flow velocity fluctuations exceed ±0.8 m / s, 73% of the samples showed methane content deviations. The system adjusts the triggering logic for this node: changing the fixed flow velocity threshold of 2.0 m / s to a dynamic range of 1.6-2.4 m / s, and changing the trigger interval from a fixed 30 seconds to an adaptive algorithm based on the flow velocity change rate. The new algorithm is set to trigger when the flow velocity change rate exceeds 0.2 m / s. 2 The interval was shortened to 15 seconds, below 0.05 m / s. 2 The time was extended to 60 seconds.

[0078] When processing pipeline flow data, the sampling ratio allocation submodule establishes a flow-sampling demand mapping model. The system analyzes real-time monitoring data from 12 main pipelines within the station and finds that when the flow rate of pipeline B-07 reaches 135% of the average, the sampling frequency only increases by 23%. The module recalculates the allocation ratio, increasing the baseline sampling frequency for pipeline B-07 from 8 times per hour to 11 times per hour, while reducing the sampling frequency for the stable flow rate of pipeline E-03 from 10 times to 7 times per hour. The allocation algorithm introduces a flow fluctuation compensation factor; when the instantaneous flow rate change of a pipeline exceeds ±10%, a temporary incremental sampling mechanism is automatically triggered, with the increment proportional to the flow rate change.

[0079] When the detection condition update submodule integrates and optimizes parameters, a multi-level verification process is executed. The new configuration scheme generated by the system contains 247 parameter items, among which key parameters such as the trigger sensitivity of node N-3098 are adjusted from the standard level to the high-sensitivity level, and the sampling priority of pipeline B-07 is increased from level 3 to level 2. The configuration update adopts a rolling effect mechanism, with each batch of parameters first verified in a simulation environment for 72 hours before being deployed to the production system in batches. During the deployment process, 12 key performance indicators are monitored in real time, including sampling coverage, component analysis timeliness, and equipment load balancing. When a 15% decrease in the analysis timeliness of pipeline E-03 is detected, the system automatically rolls back the parameter changes for that pipeline and triggers an expert diagnosis process.

[0080] The technical implementation incorporates several specialized processing mechanisms. Deviation impact assessment employs a sliding time window statistical method, recalculating the score of each node every 24 hours. Trigger parameter adjustment utilizes a genetic algorithm for optimization, with a population size of 50 parameter combinations and 20 iterations to select the optimal solution. Sampling ratio allocation implements dynamic weight adjustment, fine-tuning the ratio coefficient hourly based on the latest traffic data. The configuration update system employs a dual-storage architecture, storing currently effective parameters and backup parameter sets independently, supporting millisecond-level switching. Monitoring metric calculation utilizes a distributed stream processing framework, keeping data latency within 500 milliseconds.

[0081] The system operation and maintenance include comprehensive safeguards. All parameter changes are logged in a complete log, including metadata such as modification time, operator, and impact assessment. Modifications to critical parameters require a three-tiered approval process, and changes exceeding preset risk thresholds must be reviewed by the technical committee. The system automatically generates a weekly parameter optimization report, comparing and analyzing the differences in 12 core operating metrics under new and old configurations. The anomaly handling mechanism includes a parameter conflict detection algorithm that automatically isolates related parameter sets and issues alerts when contradictory configuration items are identified. The system maintenance interface provides a parameter impact simulation function, allowing engineers to preview the potential impact of proposed parameter modifications on system operation.

[0082] The data interaction process follows strict specifications. Node screening results are transmitted in JSON format, including fields such as node ID, 3D evaluation data, and score. Trigger parameter adjustment instructions use a dedicated binary protocol, with each instruction packet including a CRC32 checksum and sequence number. The sampling ratio allocation scheme is published through the enterprise service bus, with subscribers including the sampling execution system and the monitoring system. Configuration update packages use digital signature technology, with the signature key rotating quarterly. All data transmission channels use TLS 1.3 encryption, and biometric authentication is required for critical operations.

[0083] In its practical application at a coastal LNG receiving terminal, the system underwent customized optimization for the special operating conditions during the winter supply guarantee period. When the ambient temperature drops below 5℃, an anti-freezing optimization scheme is automatically activated: the sampling trigger temperature threshold in the northern pipeline area is raised from -10℃ to -5℃, while the sampling interval is shortened by 20%. For the highly volatile operating conditions during typhoon season, the system activates an anti-interference mode: increasing the cutoff frequency of the flow velocity filtering algorithm and increasing the stringency level of outlier removal. These special processing logics are encoded into configurable strategy templates, automatically switching execution based on meteorological warning signals. Operational data shows that the system can adapt to sampling and monitoring needs under different environmental conditions, maintaining the stability of component analysis results. Maintenance personnel can intuitively view the actual effectiveness of parameters at each node through a visual interface, including detailed parameter snapshots of the most recent 10 trigger records. The system supports parameter version retrospective comparison, allowing for the retrieval of complete configuration states from any historical time point for difference analysis.

[0084] Example 5: Functional Implementation of the Focused Sampling Monitoring Module. This module achieves holographic perception and anomaly identification of pipeline fluid status through three collaboratively operating sub-modules. Upon startup, the velocity fluctuation analysis sub-module activates a distributed pressure sensor array, which is deployed at non-equidistant intervals at key nodes of the pipeline. The sampling system employs an adaptive frequency acquisition mode, with a baseline sampling rate of 100 times per second. When the pipeline pressure gradient change exceeds 0.02 MPa per meter, the sampling frequency automatically increases to 500 times per second. The raw pressure data undergoes three-stage filtering: primary filtering eliminates inherent equipment noise, secondary filtering removes pipeline mechanical vibration interference, and tertiary filtering uses a phase correction algorithm to compensate for signal transmission delay. The processed pressure signal generates an instantaneous velocity sequence through a preset conversion model, which integrates fluid dynamics equations and empirical correction coefficients. The system identifies the positional difference between adjacent sampling points, calculates the theoretical transmission delay based on the medium's sound velocity, and aligns the velocity curve using a dynamic time warping algorithm. The fluctuation characteristic calculation includes four core indicators: the maximum velocity difference between adjacent points, the number of peaks within one minute, the average continuous fluctuation period, and the slope coefficient of the cumulative distribution function of the difference. The calculation results are packaged and sent to the downstream module every three minutes.

[0085] The sampling frequency statistics submodule connects to the historical operation database and constructs a three-dimensional histogram analysis model. The model categorizes data sources by day type: weekday data comes from continuous operation records, while holiday data is processed separately. The temperature dimension is set with seven ranges (<-20℃, -20-10℃, -10-0℃, 0-10℃, 10-20℃, 20-30℃, >30℃), and the flow rate dimension is set with five ranges (<1000m³). 3 / h, 1000-2000m 3 / h, 2000-3000m 3 / h, 3000-4000m 3 / h、>4000m 3 / h). The statistical engine scans the historical database hourly, retrieving data from the same period over the past five years for the current time period. The algorithm identifies the cluster centers of sampling operations in three-dimensional space and calculates the sampling operation density value of each cubic unit. The concentration index generation introduces a spatiotemporal weighting factor: the weight coefficient of the pipeline intersection area is doubled in space, and the proportion of peak load periods is increased in time. The system marks high-density clustered areas and abnormally sparse areas. When the operation density of a certain three-dimensional unit exceeds the 90th percentile value of the same period in history, a yellow warning is automatically triggered. If it exceeds the standard for three consecutive statistical periods, it is upgraded to a red warning state. The statistical results are compared with the real-time operation flow to generate a frequency deviation map and mark the coordinates of abnormal hot spots.

[0086] The anomaly node identification submodule integrates a machine learning detection mechanism. The system pre-trains an isolated forest model, with the training dataset containing flow velocity-sampling frequency pairings recorded during twelve consecutive months of normal operation. The feature space is constructed using a 20-dimensional vector: basic parameters (average flow velocity, mean sampling frequency), dynamic parameters (flow velocity change rate, sampling interval variance), correlation parameters (time-domain cross-correlation coefficient between flow velocity and sampling), and statistical parameters (fluctuation amplitude skewness). During model construction, the number of trees is set to 100, and the subsample size for each tree is fixed at 256 sets of data. Upon real-time detection startup, new data points are input into the model after feature extraction to calculate anomaly scores. Anomaly score calculation considers both path depth and node branching features; an alarm response is triggered when the score exceeds a threshold of 0.65. The system maintains an electronic map of the pipeline topology, automatically marking the 3D coordinates of anomaly points. The alarm interface uses a spectral coloring scheme: light yellow indicates minor anomalies with scores of 0.65-0.75, orange corresponds to moderate anomalies of 0.75-0.85, and dark red indicates severe anomalies above 0.85. The anomaly log data table contains four key fields: fault start timestamp, current duration, maximum deviation during the period, and correlation between adjacent points. Each data item is updated and stored in a distributed database in real time.

[0087] The technical implementation incorporates several sophisticated control strategies. The pressure sensor network employs a master-slave redundancy architecture, with two independent acquisition units configured at each measurement point. When the data difference between the master and slave units exceeds 5%, a third-party verification sensor is automatically activated. The flow rate conversion model's parameter calibration utilizes a daily inspection mechanism, performing online calibration using a standard nozzle device every morning. The histogram model establishes a quarterly update mechanism, adjusting the temperature range thresholds in March, June, September, and December each year. The machine learning model implements rolling updates, retaining the latest three months of data and discarding older data during each training iteration to ensure adaptation to gradual system changes. The anomaly detection engine deploys a load balancing system, automatically activating the data downsampling channel when data traffic surges to prevent processing delays.

[0088] A comprehensive data quality control chain is established during system operation. Raw data acquisition is appended with device identification codes and accuracy level labels, and the entire algorithm parameter trajectory is recorded during processing. The historical database implements a dual-active backup mechanism, automatically synchronizing to the disaster recovery center when the primary copy is updated. Anomaly alarms trigger a four-level confirmation process: Level 1: Device self-check to rule out hardware faults; Level 2: Comparison of adjacent data points to confirm anomaly boundaries; Level 3: Verification of equipment maintenance records to determine potential causes; Level 4: Manual review to confirm the final result. Alarm information distribution adopts a tiered push strategy: minor anomalies are sent to the on-duty console, moderate anomalies notify the technical supervisor, and severe alarms go directly to the emergency response team. All operation events are written to the blockchain log, and critical steps require dual-authorization confirmation. The system maintenance interface provides anomaly scenario replay functionality, supporting dynamic parameter replay at any time period.

[0089] An adaptive processing mechanism ensures reliable system operation. During cold seasons (<-15℃), the system automatically reduces the sampling frequency and statistical dimensions, merging temperature ranges into four levels to adapt to equipment characteristics. When encountering typhoon and thunderstorm weather conditions, the flow velocity fluctuation analysis module activates an anti-interference mode: enhanced filtering algorithms suppress lightning strike interference, increasing anomaly detection thresholds to reduce false alarms. For special pipeline cleaning operations, the system temporarily adjusts the isolated forest parameter settings, adding a feature space dimension to incorporate the pig's location parameters. During routine maintenance windows, the system enters a dormant state, reducing the data sampling frequency to one-tenth of the baseline value and suspending alarm notifications. All mode switching automatically follows the scheduling system's instructions, and the switching process does not affect on-the-go data processing. Historical data shows that this solution can stably adapt to various operating conditions from extremely cold environments to hot and humid conditions, maintaining all-weather monitoring capabilities.

[0090] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0091] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A liquefied natural gas sampling and component analysis system, characterized in that, The system includes: The sampling and monitoring module collects fluid parameters and flow status in the liquefied natural gas pipeline, analyzes the flow velocity fluctuation characteristics of sampling points in different pipe sections, calculates the historical opening frequency of each sampling port, compares the current flow velocity fluctuation characteristics with the historical opening frequency, identifies abnormal sampling nodes, and generates sampling distribution characteristic parameters. The component prediction module, based on the sampling distribution characteristic parameters, locates the component analysis unit corresponding to the abnormal sampling node, analyzes the historical detection records of the component analysis unit, matches the characteristic patterns of component abnormal events, evaluates the data reliability level of each analysis unit, predicts the potential deviation range of component data, and generates component deviation distribution characteristics. The path planning module, based on the component deviation distribution characteristics, identifies component storage nodes with high deviation risk, analyzes the media distribution density of each storage node, calculates the urgency of media migration, plans media transfer paths, reallocates target media to low-risk storage nodes, and generates media transfer configuration parameters. The traceability analysis module, based on the medium transport configuration parameters, compares the component detection frequencies at each sampling point before and after transport, analyzes the fluctuation characteristics of component data, identifies abnormal component change patterns, determines the source of component data deviation, and generates component traceability feature data.

2. The liquefied natural gas sampling and component analysis system according to claim 1, characterized in that, The component prediction module includes: The abnormal node matching submodule, based on the sampling distribution characteristic parameters, filters the component detection data of abnormal sampling nodes, analyzes the correlation between the component detection time series and standard physical property parameters, classifies the detection data categories with different degrees of deviation, identifies high-risk component analysis units, and generates an abnormal analysis unit set. The credibility level assessment submodule calls the set of anomaly analysis units, extracts the fluctuation characteristics of the detection data of each unit, compares the data patterns of known component anomaly events, calculates the matching degree between the analysis unit and the anomaly event, updates the data credibility level based on the matching results, and generates credibility level assessment indicators. The deviation interval prediction submodule, based on the credibility level assessment index, statistically analyzes the detection frequency of high-risk analysis units, extracts the time interval of continuous detection, calculates the fluctuation range of short-term detection results, and generates component deviation distribution characteristics by combining the maximum deviation interval of the component data predicted by the credibility level.

3. The liquefied natural gas sampling and component analysis system according to claim 1, characterized in that, The path planning module includes: The risk node identification submodule scans high-risk nodes in the component storage network based on the component deviation distribution characteristics, analyzes the component sensitivity of the node storage medium, filters storage nodes with deviations exceeding the threshold, determines the target node range for media that need to be migrated, and generates a risk node list. The migration priority calculation submodule analyzes the media distribution status of the target nodes based on the risk node list, calculates the physical property correlation and transfer frequency of the media between nodes, determines the migration order based on the correlation and transfer frequency, and generates a media migration priority index. The transfer path generation submodule analyzes the physical connection topology of the storage nodes based on the media migration priority index, matches the capacity parameters of available transfer devices, plans the shortest feasible path for media migration, updates the access permissions of the target storage nodes, and generates media transfer configuration parameters.

4. The liquefied natural gas sampling and component analysis system according to claim 1, characterized in that, The source tracing analysis module includes: The detection frequency comparison submodule, based on the medium transport configuration parameters, obtains the component detection records of each sampling point before and after transport, analyzes the distribution of detection time points and the duration of a single detection, calculates the rate of change of detection frequency per unit time, and generates a detection frequency fluctuation index. The abnormal pattern recognition submodule extracts component data that deviates from the standard detection cycle based on the detection frequency fluctuation index, analyzes the distribution pattern of abnormal data, matches the generation patterns of different component deviations, locates the core nodes of component data anomalies, and generates a set of component abnormal patterns. The deviation source location submodule, based on the set of component anomaly patterns, traces the original sampling node corresponding to the abnormal data, verifies the status parameters of the sampling equipment, associates the operation records in the medium transport path, determines the final source node of the component deviation, and generates component traceability feature data.

5. The liquefied natural gas sampling and component analysis system according to claim 1, characterized in that, The system also includes: The parameter optimization module, based on the component traceability feature data, identifies sampling nodes that continuously generate deviations, adjusts the trigger parameters of the sampling nodes, reallocates the sampling ratio of each pipeline, updates the trigger conditions for component detection, and generates a sampling optimization configuration scheme.

6. The liquefied natural gas sampling and component analysis system according to claim 5, characterized in that, The parameter optimization module includes: The sampling node screening submodule, based on the component traceability feature data, counts the frequency of deviation occurrences at each node, calculates the impact weight of the deviation on the overall component analysis, screens sampling nodes whose impact weight exceeds the threshold, and generates a list of high-impact nodes. The trigger parameter adjustment submodule, based on the list of high-impact nodes, analyzes historical sampling trigger conditions, evaluates the correlation between trigger conditions and component deviation, reconfigures the sampling trigger time interval and flow rate trigger threshold, and generates a node trigger parameter update table. The sampling ratio allocation submodule calculates the percentage change of the medium flow rate in each pipeline based on the node trigger parameter update table, adjusts the sampling frequency ratio of different pipe sections according to the flow rate ratio, and generates a sampling ratio allocation scheme. The detection condition update submodule integrates the node trigger parameter update table and the sampling ratio allocation scheme, updates the trigger sensitivity parameters of the component detection equipment, and generates a sampling optimization configuration scheme.

7. The liquefied natural gas sampling and component analysis system according to claim 1, characterized in that, The sampling and monitoring module includes: The flow velocity fluctuation analysis submodule collects pressure data and temperature gradient of the pipeline fluid, analyzes the flow velocity change curves of continuous sampling points, calculates the flow velocity difference between adjacent sampling points, counts the flow velocity fluctuation frequency within a preset period, and generates flow velocity fluctuation characteristic parameters. The sampling frequency statistics submodule retrieves sampling operation records from the historical database based on the flow velocity fluctuation characteristic parameters, analyzes the distribution pattern of sampling operations in different time periods, calculates the concentration of sampling operations per unit time, and generates a sampling frequency concentration index. The abnormal node identification submodule, based on the sampling frequency concentration index, compares the current flow velocity fluctuation characteristics with the historical sampling frequency, detects sampling nodes that exceed the normal fluctuation range, marks the abnormal sampling location and the degree of deviation, and generates sampling distribution characteristic parameters.

8. The liquefied natural gas sampling and component analysis system according to claim 2, characterized in that, The steps executed by the trust level assessment submodule include: The system retrieves the detection records of each unit in the anomaly analysis unit set over the past three years and extracts the extreme values ​​of component concentrations and the time points when they occurred from the records. The concentration fluctuation patterns of known component anomalies are matched, the temporal overlap between the unit detection data and the anomaly patterns is calculated, and the data confidence level is set according to the temporal overlap.

9. The liquefied natural gas sampling and component analysis system according to claim 3, characterized in that, The migration priority calculation submodule execution steps include: Obtain the media component type of the nodes in the risk node list, and analyze the physical property compatibility and chemical reaction risk between the types; Calculate the safety factor for media exchange between nodes and count the frequency of emergency operations in historical transfer records; The migration order is determined based on the safety factor and the frequency of emergency operations.

10. The liquefied natural gas sampling and component analysis system according to claim 4, characterized in that, The steps executed by the deviation source localization submodule include: By associating the set of abnormal component patterns with the media transfer log, the transfer operation node at the time the abnormal pattern occurred can be traced. Verify the equipment calibration records at the corresponding sampling points, analyze the temporal correlation between equipment calibration deviation and component abnormalities, and identify the starting operation node that caused the component deviation.

Images