Intelligent monitoring instruments and methods for multi-parameter fusion analysis in natural gas pipeline operation

Abstract

This invention belongs to the field of natural gas pipeline operation and maintenance, and specifically discloses an intelligent monitoring instrument for multi-parameter fusion analysis during natural gas pipeline operation. A sensor module is installed on the natural gas pipeline to acquire raw data and signal strength. A preprocessing module preprocesses the raw data to obtain preprocessed data, and then uses an adaptive compression algorithm to obtain compressed data. A fusion module uses a multi-source data fusion algorithm on the compressed data to obtain fusion features. A diagnosis module inputs the fusion features into a fault diagnosis model based on a machine learning algorithm; when a fault is diagnosed, it outputs the fault type. A positioning module acquires the location of the sensor module, geographic information data, and historical operating data; based on the signal strength, sensor module location, geographic information data, historical operating data, and fault type, it uses a positioning algorithm to determine the fault location. An early warning module generates a fault report based on the fault location and fault type, issues an early warning signal, and uploads the fault report.

Description

Technical Field

[0001] This invention relates to the field of natural gas pipeline operation and maintenance, and in particular to an intelligent monitoring instrument for multi-parameter fusion analysis during natural gas pipeline operation. Background Technology

[0002] Traditional natural gas pipeline monitoring systems primarily rely on single sensors for parameter monitoring, such as pressure, temperature, or flow sensors. However, the performance of these sensors is often limited by environmental conditions; for example, sensor accuracy degrades due to temperature changes or long-term use, leading to insufficient accuracy in the monitoring data. Furthermore, the monitoring range of a single sensor is limited, failing to comprehensively reflect the overall operating status of the pipeline. For instance, a pressure sensor can only monitor pressure changes but cannot identify internal corrosion or leakage risks. These limitations make it difficult for traditional monitoring systems to provide reliable early warning and diagnostic capabilities in complex environments.

[0003] Existing natural gas pipeline monitoring systems typically rely on local data storage and periodic manual retrieval, resulting in inefficient data processing and slow response times. For example, sensor-collected data requires manual reading or recording on simple local storage devices, failing to enable real-time transmission and remote monitoring. Furthermore, traditional systems have limited data processing capabilities, hindering comprehensive analysis of multi-source data and leading to lower accuracy in fault diagnosis. This inefficient data processing method not only increases maintenance costs but may also cause delays in data handling, missing optimal fault response opportunities.

[0004] Natural gas pipelines are typically laid in complex geographical environments or extreme climatic conditions, such as mountainous areas, deserts, or cold regions. These environmental factors pose significant challenges to the stability and reliability of monitoring systems. For example, extreme temperatures can degrade sensor performance, and severe weather can affect the stability of data transmission. Furthermore, the complexity of the terrain along the pipeline route makes sensor deployment and maintenance difficult, further limiting the applicability of traditional monitoring systems. Under these conditions, existing monitoring technologies are insufficient to meet the requirements for the long-term safe operation of pipelines. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide an intelligent monitoring instrument for multi-parameter fusion analysis during the operation of natural gas pipelines, in order to solve at least one of the above-mentioned technical problems.

[0006] To achieve the above objectives, firstly, an intelligent monitoring instrument for multi-parameter fusion analysis during natural gas pipeline operation is provided, comprising:

[0007] The sensor module, installed on the natural gas pipeline, is used to acquire raw data and signal strength.

[0008] The preprocessing module is used to preprocess the original data to obtain preprocessed data, and to apply an adaptive compression algorithm to the preprocessed data to obtain compressed data.

[0009] The fusion module is used to obtain fusion features from the compressed data using a multi-source data fusion algorithm;

[0010] The diagnostic module is used to input the fused features into a fault diagnosis model based on machine learning algorithms, and output the fault type when a fault is diagnosed.

[0011] The positioning module is used to acquire the location of the sensor module, geographic information data near the natural gas pipeline, and historical operation data of the natural gas pipeline, and to obtain the fault location using a positioning algorithm based on the signal strength, the location of the sensor module, the geographic information data, the historical operation data, and the fault type.

[0012] The early warning module is used to generate a fault report based on the fault location and the fault type, issue an early warning signal, and upload the fault report.

[0013] Secondly, a smart monitoring method for multi-parameter fusion analysis in natural gas pipeline operation is provided, including the following steps:

[0014] Acquire raw data and signal strength;

[0015] The original data is preprocessed to obtain preprocessed data, and the preprocessed data is then compressed using an adaptive compression algorithm.

[0016] The compressed data is processed using a multi-source data fusion algorithm to obtain fusion features;

[0017] The fused features are input into a fault diagnosis model based on a machine learning algorithm. When a fault is diagnosed, the fault type is output.

[0018] The location of the sensor module, geographic information data near the natural gas pipeline, and historical operation data of the natural gas pipeline are acquired. Based on the signal strength, the location of the sensor module, the geographic information data, the historical operation data, and the fault type, a positioning algorithm is used to obtain the fault location.

[0019] A fault report is generated based on the fault location and fault type, an early warning signal is issued, and the fault report is uploaded.

[0020] The above technical solution has the following beneficial technical effects:

[0021] The intelligent monitoring instrument for multi-parameter fusion analysis in the operation of this natural gas pipeline achieves multi-dimensional data acquisition through the sensor module. Combined with the adaptive compression algorithm of the preprocessing module, it reduces data transmission and storage costs while ensuring the integrity of key information. The fusion module adopts a multi-source data fusion algorithm to effectively improve the spatiotemporal correlation and feature representation capabilities of the monitoring data, overcoming the limitations of single-parameter analysis. The diagnostic module, based on a fault diagnosis model constructed using machine learning algorithms, significantly improves the accuracy of fault identification and early warning capabilities through high-dimensional analysis of fused features. The positioning module achieves precise spatial positioning and source tracing analysis of fault points by fusing multi-dimensional parameters of signal strength, geographic information, and historical operating data. Finally, through the automated fault report generation and early warning signal linkage mechanism of the early warning module, a closed-loop monitoring system covering the entire process from data acquisition, feature analysis, fault diagnosis to location tracking is constructed. This significantly improves the operational safety level of the natural gas pipeline, reduces the cost of manual inspection and the risk of leakage accidents, and provides intelligent support for pipeline maintenance decisions. Attached Figure Description

[0022] The accompanying drawings are provided to better understand the invention and are not intended to unduly limit the scope of the invention. Wherein:

[0023] Figure 1 This is a structural block diagram of an intelligent monitoring instrument for multi-parameter fusion analysis during the operation of a natural gas pipeline, according to an embodiment of the present invention.

[0024] Figure 2 This is a schematic diagram of the structure of an intelligent monitoring instrument for multi-parameter fusion analysis during the operation of a natural gas pipeline, according to an embodiment of the present invention.

[0025] Figure 3 This is a structural block diagram of the preprocessing module in an embodiment of the present invention;

[0026] Figure 4 This is a structural block diagram of the fusion module in an embodiment of the present invention;

[0027] Figure 5 This is a structural block diagram of the diagnostic module in an embodiment of the present invention;

[0028] Figure 6 This is a structural block diagram of the positioning module in an embodiment of the present invention;

[0029] Figure 7 This is a structural block diagram of the early warning module in an embodiment of the present invention;

[0030] Figure 8 This is a flowchart of an intelligent monitoring method for multi-parameter fusion analysis in the operation of a natural gas pipeline, according to an embodiment of the present invention.

[0031] Figure 9 This is a schematic diagram of the structure of a computer system according to an embodiment of the present invention. Detailed Implementation

[0032] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of the present invention, including various details to aid understanding. These details should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0033] Example 1

[0034] like Figures 1 to 2 As shown in the figure, this embodiment provides an intelligent monitoring instrument for multi-parameter fusion analysis during natural gas pipeline operation, including:

[0035] The sensor module, installed on the natural gas pipeline, is used to acquire raw data and signal strength.

[0036] The preprocessing module is used to preprocess the original data to obtain preprocessed data, and to apply an adaptive compression algorithm to the preprocessed data to obtain compressed data.

[0037] The fusion module is used to obtain fusion features from the compressed data using a multi-source data fusion algorithm;

[0038] The diagnostic module is used to input the fused features into a fault diagnosis model based on machine learning algorithms, and output the fault type when a fault is diagnosed.

[0039] The positioning module is used to acquire the location of the sensor module, geographic information data near the natural gas pipeline, and historical operation data of the natural gas pipeline, and to obtain the fault location using a positioning algorithm based on the signal strength, the location of the sensor module, the geographic information data, the historical operation data, and the fault type.

[0040] The early warning module is used to generate a fault report based on the fault location and the fault type, issue an early warning signal, and upload the fault report.

[0041] Specifically, the preprocessing module uses an adaptive compression algorithm to reduce redundant data transmission and lower network costs for operators while ensuring the integrity of key information. The fusion module integrates compressed data from different sensors using a multi-source data fusion algorithm, effectively improving the accuracy of subsequent fault diagnosis. The diagnosis module uses machine learning algorithms for fault diagnosis with high efficiency and accuracy. The positioning module uses the signal strength, sensor module location, geographic information data, historical operating data, and fault type to locate faults, effectively improving positioning accuracy. Geographic information data refers to the geographic information data of the natural gas pipeline itself and its surrounding environment, specifically including the following data objects: First, the geographic data of the pipeline itself, covering the three-dimensional coordinates of the pipeline (including latitude, longitude, and altitude), pipeline burial depth parameters, pipeline route topology, pipe segment material and diameter parameters, and coordinates of key nodes (e.g., valve stations, pressurization stations); second, the geographic data of the surrounding environment, involving geological structure information (e.g., soil type, rock layer distribution), surface features (e.g., rivers, mountains, roads, buildings), and sensitive area markings (e.g., residential areas, ecological protection areas).

[0042] Specifically, the adaptive compression algorithm dynamically selects the compression method based on the data characteristics of the preprocessed data. For example, Principal Component Analysis (PCA) is used for dimensionality reduction compression of steady-state temperature data (low frequency). Lossless compression is used for pressure mutation data (high frequency) to ensure that key information is not lost.

[0043] Specifically, the multi-source data fusion algorithm includes: a weighted average algorithm, which assigns weights based on sensor accuracy (e.g., 0.7 for pressure sensors and 0.3 for flow sensors); or a Kalman filter algorithm, which iteratively optimizes the estimated values ​​through state equations and observation equations to eliminate noise interference and perform data fusion.

[0044] Specifically, the positioning algorithm may include triangulation, distance calculation, machine learning algorithm-generated models, or direct annotation of geographic information data.

[0045] Specifically, the signal strength refers to the amplitude of the physical signal detected by each sensor in the sensor module. For example, the signal strength of an acoustic sensor is used to locate leak points. The signal strength of a gas sensor reflects changes in the concentration of gas within the pipeline. For instance, in the event of a natural gas leak, the signal strength detected by the gas sensor reflects the concentration of the leaking gas, helping to locate the leak point. The signal strength of a fiber optic sensor reflects the attenuation of the light signal in the optical fiber. When a pipeline ruptures or deforms, changes in the signal strength of the fiber optic sensor can detect the location of the damage. The signal strength of a gas quality sensor is related to changes in the gas flow rate within the pipeline. The sensor can sense changes in gas flow rate and use changes in signal strength to determine whether the pipeline is blocked or has an abnormal flow rate. The signal strength of a radar sensor reflects the distance and position of objects around the pipeline. Radar sensors can be used to detect settlement, displacement, or obstacles near the pipeline; changes in signal strength can be used to determine whether there is external interference or potential pipeline damage. The signal strength of an infrared thermal imaging sensor reflects changes in the surface temperature of the pipeline. When a pipeline leaks or experiences an anomaly, the thermal imaging sensor can locate the heat source or leak point by monitoring changes in temperature difference.

[0046] Specifically, the sensor module may include any combination of acoustic sensors, gas sensors, fiber optic sensors, gas quality sensors (e.g., methane concentration sensors or carbon dioxide concentration sensors), radar sensors, infrared thermal imaging sensors, and environmental radiation sensors.

[0047] Specifically, the acoustic sensor is preferably an ultrasonic sensor, installed on the outer surface or inside the pipeline. The ultrasonic sensor is used to detect abnormal sounds or structural problems such as cracks or bubbles inside the natural gas pipeline, and is particularly suitable for detecting sound fluctuations caused by tiny cracks or corrosion in the pipeline wall. By analyzing the frequency and propagation mode of the sound waves using Fourier transform, wavelet transform, time-frequency decomposition, or support vector machine, potential hazards that may lead to leaks or malfunctions can be detected.

[0048] Specifically, the gas sensor employs electronic nose technology and is installed around natural gas pipelines, particularly at pipeline outlets or in areas with a high risk of gas leaks. The electronic nose consists of multiple sub-gas sensors that analyze the combined characteristics of the gases to effectively identify minute changes in the gas, especially variations in trace amounts of harmful gases (e.g., hydrogen sulfide, ammonia, etc.) within the natural gas composition. This electronic nose technology exhibits high sensitivity in gas leak detection, capable of identifying subtle gas changes and providing more accurate monitoring data.

[0049] Specifically, the fiber optic sensor is installed on the outer surface of the natural gas pipeline. By utilizing the changes in the response of optical fibers under different environmental conditions (e.g., temperature, strain), it can monitor physical parameters such as strain and temperature on the outside of the pipeline. The internal state (e.g., temperature, pressure) of the pipeline can be indirectly inferred and monitored through a combination of physical models, regression algorithms, or finite element analysis. In particular, the use of distributed fiber optic sensing technology enables real-time detection of multiple points along the pipeline, offering high spatial resolution and accuracy. The fiber optic sensor is especially useful for long-distance monitoring and in high-risk areas (e.g., mountainous regions, complex terrain). The fiber optic sensor is connected to a data acquisition device via optical fiber, transmitting data in real-time to a central processing system for analysis. The physical model refers to a mathematical model established based on the principles of materials mechanics and thermodynamics, relating external physical quantities (e.g., temperature, strain) to internal states (e.g., temperature, pressure). For example, changes in pressure inside the pipeline cause changes in strain on the pipeline surface; by measuring the strain on the pipeline surface, changes in internal pressure can be inferred. The regression algorithm refers to establishing a regression model based on historical data (e.g., linear regression, support vector regression), which uses external measurement data to estimate the internal state. Machine learning algorithms can be used to train a model based on historical data to predict the internal state of a pipe. Finite Element Analysis (FEA) refers to simulating the physical behavior of a pipe by combining surface strain, temperature, and other data obtained from fiber optic sensors with a finite element model to calculate the internal temperature, pressure, and other conditions of the pipe. Finite Element Analysis can accurately simulate the response of a pipe under different conditions, thus providing a theoretical basis for predicting its internal state.

[0050] Specifically, the gas quality sensor is installed inside the pipeline, preferably in a section of the pipeline with high flow velocity and large pressure differential. The gas quality sensor is used to monitor the gas quality (e.g., methane concentration or oxygen concentration) within the natural gas pipeline in real time. It is highly sensitive to fluctuations in gas composition and is suitable for use in pipeline environments with high flow rates and large pressure differentials. Compared to traditional gas analyzers, the gas quality sensor offers greater adaptability and accuracy.

[0051] Specifically, the radar sensors are installed on the outside of the pipeline, particularly at critical sections (bridge sections, tunnel sections, landslide areas), to monitor pipeline deformation or displacement, providing real-time monitoring, especially during natural disasters such as earthquakes and landslides. By emitting electromagnetic waves and receiving changes in reflected waves, the radar sensors can detect minute changes in the ground or pipeline structure, providing early warnings of potential deformation or collapse. The radar sensors are connected wirelessly or via wired connection to a central processing system to transmit deformation data.

[0052] Specifically, the infrared thermal imaging sensor is installed on the surface of the natural gas pipeline, especially at leak points, corroded areas, or sections with a high risk of gas leakage. The infrared thermal imaging sensor is used to detect the surface temperature distribution of the pipeline and identify any abnormal temperature points (e.g., leak points). By analyzing temperature changes, abnormal heat flow in the pipeline can be detected in advance, thereby identifying potential fault areas (e.g., gas leaks, corrosion, blockages, etc.). The infrared thermal imaging sensor is connected to the central processing system via a data cable to transmit real-time temperature change data.

[0053] Specifically, the environmental radiation sensor is installed in the environment surrounding the natural gas pipeline, particularly in areas near radiation sources such as nuclear power plants and chemical plants. The environmental radiation sensor is used to detect the radiation level in the environment surrounding the natural gas pipeline. In some special environments (e.g., near a nuclear power plant), changes in environmental radiation levels indicate potential safety hazards, especially when the pipeline encounters external impacts, damage, or other unexpected events, providing additional safety monitoring data. The environmental radiation sensor is connected wirelessly or via wired connection to a central processing system to transmit radiation data for real-time analysis.

[0054] Specifically, such as Figure 3 As shown, the preprocessing module specifically includes:

[0055] The first preprocessing unit is used to remove duplicate data from the original data to obtain the first data.

[0056] The second preprocessing unit is used to remove outliers from the first data using an anomaly detection algorithm to obtain the second data.

[0057] The filtering unit is used to obtain the historical second data of the previous time period, generate a threshold interval based on the historical second data using a threshold generation algorithm, and remove values ​​that do not belong to the threshold interval from the second data to obtain preprocessed data.

[0058] A compression unit is used to apply an adaptive compression algorithm to the preprocessed data to obtain compressed data.

[0059] The adaptive compression algorithms include an adaptive selection algorithm based on the magnitude of change, an adaptive compression algorithm based on compression error, and an adaptive compression algorithm based on machine learning.

[0060] Specifically, in the first preprocessing unit, the original data is traversed, duplicate data is marked, and redundant duplicate data is deleted, retaining only one copy of each data point, thereby ensuring data accuracy and consistency. The deduplication algorithm uses fields such as the timestamp of data acquisition to ensure that each record is unique.

[0061] Specifically, in the second preprocessing unit, the anomaly detection algorithm includes Z-score detection and interquartile range (IMR) method. Outliers are data points that are significantly different from the normal data pattern, and they originate from erroneous operations or system malfunctions. Taking Z-score detection as an example, the calculation formula is as follows:

[0062]

[0063] In the formula, z represents the Z-value, x is the data point to be tested, μ is the mean of the data, and σ is the standard deviation of the data. If |z| > 3, the data point is considered an outlier.

[0064] Specifically, in the second preprocessing unit, outliers are either deleted or filled. Filling includes mean filling or nearest neighbor filling. Mean filling involves calculating the average of other data and filling the outlier with the average. Missing values ​​or data with inconsistent formats are also corrected by filling or deleting.

[0065] Specifically, in the filtering unit, the mean or standard deviation of the historical second data is calculated, and then a threshold interval is set according to the threshold or standard deviation of the historical second data. The threshold generation algorithm is a dynamic threshold filtering method. The maximum value of the threshold interval is the sum of the mean and the preset multiplier standard deviation, and the minimum value of the threshold interval is the difference between the mean and the preset multiplier standard deviation. The preset multiplier is preferably [1, 3].

[0066] Specifically, executing the adaptive selection algorithm based on the magnitude of change includes the following steps:

[0067] Calculate the magnitude of change of each data point in the preprocessed data;

[0068] The preprocessed data whose variation exceeds a preset compression threshold is compressed using Huffman coding to obtain high-frequency data;

[0069] The preprocessed data whose variation amplitude is less than or equal to a preset compression threshold is compressed using a principal component analysis algorithm or a wavelet transform algorithm to obtain low-frequency data. The compressed data includes the high-frequency data and the low-frequency data.

[0070] Specifically, the variation amplitude is calculated using variance, and the adaptive selection algorithm based on the variation amplitude is as follows:

[0071]

[0072] In the formula, X represents the magnitude of change, Variance(X) represents the result of data feature extraction, ε represents the compression threshold, A represents compression method A (e.g., principal component analysis algorithm or wavelet transform algorithm), and B represents compression method B (e.g., Huffman coding).

[0073] Executing the adaptive compression algorithm based on compression error includes the following steps:

[0074] The preprocessed data is pre-compressed to obtain pre-compressed data;

[0075] The compression error is calculated based on the pre-compressed data and the pre-processed data.

[0076] When the compression error is less than the preset error threshold, the compression ratio is increased to update the pre-compressed data until the compression error equals the preset error threshold, at which point the pre-compressed data is output as compressed data.

[0077] When the compression error exceeds the preset error threshold, the compression ratio is reduced to update the pre-compressed data until the compression error equals the preset error threshold. Then, the pre-compressed data at this point is output as compressed data.

[0078] Specifically, the preprocessed data is represented as X = {X1, X2, ..., X...} n}, pre-compressed data X c ={X c1 X c2 …, X cn The formula for calculating the compression error is as follows:

[0079]

[0080] In the formula, Error(X, X c ) indicates compression error.

[0081] Specifically, when employing the aforementioned adaptive compression algorithm based on compression error, as much information as possible is retained during the compression process while controlling the compression error. For data such as temperature and pressure, quantization error can be used to control the compression accuracy. Error is controlled by dynamically adjusting the compression ratio, achieving high compression efficiency. This approach is suitable for data computation that balances compression ratio and data accuracy, and can adaptively adjust the compression strategy under different data characteristics.

[0082] Specifically, the adaptive compression algorithm also includes machine learning-based adaptive compression. It trains a machine learning model (e.g., decision tree model, random forest model, support vector machine, gradient boosting machine, etc.) using historical preprocessed data to predict a suitable compression method based on the characteristics of different sensor data. For example, given a dataset X, the machine learning model f(X) predicts which compression method should be used: Method(X) = f(X). This allows for adaptive adjustment of the compression strategy based on the training data, making it particularly suitable for complex and variable data environments. Gradient Boosting Machine (GBM) is an ensemble learning method based on decision trees that improves model performance by progressively reducing prediction errors. GBM achieves a strong classifier by weighted training of multiple weak classifiers (usually decision trees), progressively adjusting model parameters to minimize residuals.

[0083] Specifically, the compressed data employs an adaptive data transmission strategy during transmission. This strategy includes dynamic transmission control based on network conditions and hierarchical transmission with distributed storage. The dynamic transmission control based on network conditions dynamically adjusts the frequency and amount of data transmission according to the signal quality, bandwidth, and latency fluctuations of the wireless network. When network quality is good, higher frequency data can be transmitted; when network quality is poor, the data sampling frequency is reduced or important data is selectively transmitted. The hierarchical transmission and distributed storage employ a hierarchical transmission mechanism, prioritizing the transmission of monitoring data with high real-time requirements (e.g., pressure, temperature, flow rate, pipeline vibration data, gas pressure fluctuations), while more stable data (e.g., flow rate, gas composition, pipeline wall thickness or corrosion monitoring data, gas composition within the pipeline) can be cached or transmitted with a delay. Furthermore, some data can be pre-stored in the distributed storage system. When network conditions are poor, the system can choose to read data from local storage as needed, ensuring data continuity and real-time performance.

[0084] Specifically, such as Figure 4 As shown, the fusion module specifically includes:

[0085] A time alignment unit is used to process the compressed data using timestamp interpolation to obtain time-series data;

[0086] A normalization unit is used to process the time-series data using the Z-value method to obtain normalized data.

[0087] An extraction unit is used to extract features from the standardized data to obtain several key features;

[0088] The fusion unit is used to process several of the key features using a Kalman filter algorithm or a weighted average algorithm to obtain fused features.

[0089] Specifically, in the time-series alignment unit, the timestamp interpolation method determines a unified time series as the target for aligning compressed data from all sources by setting a time interval (e.g., 1 minute). For compressed data from different sensors, interpolation is performed on the same time series based on the timestamp and data value. Taking linear interpolation as an example, given time points t1 and t2, and corresponding data values ​​y1 and y2, the calculation formula for the interpolation result y at time point t between t1 and t2 is as follows:

[0090]

[0091] Specifically, in the standardization unit, the Z-value method converts the time-series data into a standard normal distribution with a mean of 0 and a standard deviation of 1, thereby eliminating the dimensional influence between data from different sensors and making time-series data from different sources comparable. For each source of time-series data, the mean μ and standard deviation σ are calculated, where x represents the time-series data. The formula for calculating the standardized data is as follows:

[0092]

[0093] In the formula, Z represents standardized data.

[0094] Specifically, in the fusion unit, the Kalman filter algorithm is used as an example. The system state is continuously estimated through two steps: prediction and updating. First, a state transition equation is established based on the standardized data to predict the state at the next time step. Then, the covariance of the state estimate is predicted based on the state transition equation. The Kalman gain is calculated based on the predicted covariance and the observation noise covariance. The state estimate is updated based on the observed and predicted values. Finally, the covariance of the estimated state is updated to obtain the fused features.

[0095] Specifically, the weighted average algorithm works by assigning weights to each key feature according to its importance, and then summing multiple key features in a weighted manner to obtain a fused feature.

[0096] The intelligent monitoring instrument for multi-parameter fusion analysis in natural gas pipeline operation, as described in this invention, effectively improves the processing accuracy of monitoring data and the accuracy of fault detection through the design of the fusion module. The time-series alignment unit uses timestamp interpolation to process compressed data, resolving the issue of inconsistent data acquisition times and ensuring time alignment of multi-source data, enabling subsequent analysis to be based on accurate time-series information. The standardization unit uses the Z-value method to standardize the time-series data, eliminating dimensional differences between different monitoring parameters and improving data comparability and analytical consistency. The extraction unit extracts features from the standardized data, effectively extracting key features from a large amount of data, reducing interference from irrelevant information, and thus improving the efficiency of subsequent analysis. Finally, the fusion unit processes key features using Kalman filtering or weighted average algorithms, achieving comprehensive analysis of multi-dimensional information, resulting in more stable and accurate fused features, further improving the accuracy of fault identification and prediction. Overall, the above technical solution, through multi-level data processing and feature fusion, provides more accurate data support and decision-making basis for real-time monitoring of natural gas pipelines, enhancing the safety and reliability of pipeline operation.

[0097] Specifically, such as Figure 5 As shown, the diagnostic module specifically includes:

[0098] A random forest unit is used to input the fused features into a pre-trained random forest model to obtain a first diagnostic result;

[0099] A support vector machine unit is used to input the fused features into a pre-trained support vector machine model to obtain a second diagnostic result;

[0100] Long Short-Term Memory (LSTM) network units are used to input the fused features into a pre-trained LSM network model to obtain a third diagnostic result;

[0101] The diagnostic unit is used to obtain the fault type by using a multi-model optimization algorithm based on the first diagnostic result, the second diagnostic result, and the third diagnostic result.

[0102] The advantages of the above technical solution are as follows: Traditional fault diagnosis models typically rely on a single algorithm (e.g., only random forest or support vector machine) for decision-making, making it difficult to simultaneously analyze the static features and dynamic temporal characteristics of pipeline data. This solution achieves a breakthrough through a heterogeneous model collaborative decision-making mechanism. Random forest units excel at capturing nonlinear correlations of high-dimensional fusion features (e.g., spatial distribution anomalies in multi-sensor data), support vector machine units enhance the robustness of classification boundaries in small-sample scenarios (e.g., distinguishing edge samples of rare fault types), and long short-term memory network units analyze the dynamic evolution of fault features through temporal modeling capabilities (e.g., the propagation delay effect of pressure fluctuations). By fusing the diagnostic results of the three through a multi-model optimization algorithm, the system can not only suppress misjudgments caused by data distribution bias or noise interference from a single model (e.g., overfitting of random forest to transient anomalies, and lag response of LSTM to long-term trends), but also extract complementary diagnostic evidence in cases of multiple concurrent faults or complex operating conditions (e.g., pipe wall corrosion superimposed on pressure transients), thereby improving the interpretability and discrimination accuracy of fault types.

[0103] Specifically, the random forest unit, based on the random forest algorithm, is used to handle multi-classification tasks, helping the system determine the normal and abnormal states of the pipeline. By constructing multiple decision trees and performing voting decisions, it can handle complex and high-dimensional data, making it suitable for various sensor data analysis. During training, it learns from the input fused features (e.g., temperature, pressure, or flow rate) to ultimately obtain a first diagnostic result (i.e., the classification result of the fault type, such as pipeline leakage or abnormal pressure).

[0104] Specifically, the random forest unit includes:

[0105] The training set is divided into sub-units, which are used to build several random forest training sets based on the fusion features;

[0106] A decision tree building subunit is used to build several random forest decision trees based on several random forest training sets using the random forest algorithm. The number of random forest decision trees is the same as the number of random forest training sets.

[0107] A decision subunit is used to generate several decision results based on several of the random forest decision trees;

[0108] The diagnostic subunit is used to vote on several decision results using a voting algorithm to obtain the first diagnostic result.

[0109] Specifically, the support vector machine (SVM) unit performs binary classification tasks based on the SVM algorithm, thereby helping the system determine whether a specific type of fault, such as a leak, exists. The SVM unit constructs an optimal classification hyperplane to segment the data into different categories. During training, by finding the hyperplane that minimizes misclassification, it can effectively handle nonlinear data and accurately determine the fault type when real-time data is input, outputting a binary classification result, such as whether a pipe leak has occurred.

[0110] Specifically, the Long Short-Term Memory (LSTM) network unit is used to analyze the temporal data in the fused features, process the changes in sensor data in the pipeline over time, and identify long-term dependencies in the data. It handles long-term dependencies through built-in memory units and gating mechanisms, learning during training how to recognize fluctuations in data such as temperature, pressure, and flow rate. This is used to identify sudden or abnormal fluctuations in pipeline operation, such as predicting potential faults like thermal expansion or partial blockage when temperature continues to rise and flow rate decreases. The output includes the fault type (e.g., pipeline leak, abnormal temperature, flow fluctuation, etc.) and possible causes (e.g., equipment aging, pipeline leak, or changes in the external environment). Based on this information, the system accurately diagnoses faults, provides targeted maintenance suggestions, and supports subsequent fault location and repair work.

[0111] Specifically, the outputs of the random forest unit, the support vector machine unit, and the long short-term memory network unit are not entirely the same. Typically, the outputs will differ, especially in complex systems where multiple models may produce different fault diagnosis results. The multi-model optimization algorithm uses voting algorithms, confidence level assessment methods, rule-based post-processing methods, model fusion methods, or priority setting methods to filter the multiple outputs and obtain accurate results.

[0112] Specifically, the voting algorithm is used for ensemble learning, such as classification problems. A majority voting method can be employed, selecting the most frequently occurring fault type from the first, second, and third diagnostic results as the final output fault type. For example, if two models predict "pipe leak" and one predicts "temperature anomaly," the final result is "pipe leak." A weighted voting method can also be used, assigning different weights to the first, second, and third diagnostic results, considering the weight of different diagnostic results during voting. For example, the diagnostic result output by a unit with higher accuracy has a higher weight, thus having a greater impact on the final result.

[0113] Specifically, the confidence level determination method is used to attach a confidence level value when generating the first, second, and third diagnostic results, indicating the accuracy of the diagnostic result. Diagnostic results with higher confidence levels are more reliable. When the first, second, and third diagnostic results are different, the diagnostic result with the higher confidence level is selected as the fault type output by comparing the confidence levels. For example, if the confidence level for the first diagnostic result (pipeline leak) is 0.9, the confidence level for the second diagnostic result (leakage) is 0.7, and the confidence level for the third diagnostic result (temperature anomaly) is 0.8, then the first diagnostic result is selected.

[0114] Specifically, the rule-based post-processing method introduces a rule engine or expert system to post-process the first diagnostic result, the second diagnostic result, and the third diagnostic result, and makes a final judgment based on contextual information, historical data, or other known rules. For example, when the first diagnostic result, the second diagnostic result, and the third diagnostic result are inconsistent, but there are diagnostic results that are the same as historical data, the final decision is made based on the historical database.

[0115] Specifically, the model fusion method uses model fusion technology to perform a weighted sum or average of the first, second, and third diagnostic results to obtain a comprehensive prediction result as the fault type output. For example, using a weighted average, the first, second, and third diagnostic results are weighted according to their respective weights to obtain the final diagnostic result. This can reduce the bias of a single model and improve the robustness of the system. For example, if the first diagnostic result is a pipe leak, the second diagnostic result is an abnormal temperature, and the third diagnostic result is an abnormal pressure, the output result is a comprehensive result based on the performance weights of each diagnostic result (e.g., accuracy or training results).

[0116] Specifically, the priority setting is used to detect specific types of faults (e.g., long short-term memory network units are used for timing anomaly detection). In this case, the specific type of fault is only referred to the corresponding unit, and the diagnostic results of other units are not considered.

[0117] Furthermore, in an alternative embodiment, the multi-model optimization algorithm can also employ a stacking method. The stacking method uses the outputs of multiple models as new features input to another learning model for training, thereby obtaining the final prediction result. First, the outputs of multiple base models are used as features input to a secondary model (e.g., logistic regression, neural network, etc.) for comprehensive judgment. The secondary model then makes the final decision based on these output features. The stacking method can better utilize the advantages of different models, especially showing better performance in complex problems.

[0118] Furthermore, in other alternative embodiments, the diagnostic module may specifically include: an Extreme Learning Machine (ELM) unit, used to input the fused features into a pre-trained ELM model to obtain a first diagnostic result; a Graph Convolutional Network (GCN) unit, used to input the fused features into a pre-trained GCN model to obtain a second diagnostic result; a Transformer Network (TCN) unit, used to input the fused features into a pre-trained Transformer Network model to obtain a third diagnostic result; and a diagnostic unit, used to obtain the fault type by employing a multi-model optimization algorithm based on the first diagnostic result, the second diagnostic result, and the third diagnostic result.

[0119] Extreme Learning Machine (ELM) is a novel type of single-hidden-layer feedforward neural network (SLFN). Unlike traditional neural networks, ELM training does not require backpropagation; instead, it randomly initializes hidden layer weights and trains the output layer weights using the least squares method. ELM is computationally efficient and possesses strong generalization ability, performing exceptionally well when handling large-scale datasets. ELM can efficiently classify and diagnose fault characteristics, making it particularly suitable for real-time monitoring systems with massive amounts of data.

[0120] Graph Convolutional Networks (GCNs) are deep learning models based on graph-structured data. They can handle non-Euclidean data and are particularly suitable for learning from graph-structured data. For complex network systems like natural gas pipelines, GCNs can represent the relationships between pipeline nodes (e.g., sensor nodes) through graph structures and extract local features of nodes through convolutional operations for fault diagnosis. By performing convolutional operations on graph data, GCNs can adaptively extract more structured features, improving the accuracy of fault diagnosis. They are highly scalable and can handle large-scale node and connection data. In natural gas pipeline fault diagnosis, GCNs can perform joint analysis of data from different sensor nodes based on the pipeline's topology, thereby improving the accuracy and stability of fault detection.

[0121] The Transformer network (with self-attention mechanism) is a model based on self-attention. Compared to traditional RNNs or LSTMs, the Transformer can efficiently capture long-term dependencies in sequential data through a global self-attention mechanism and can process data in parallel, thus significantly improving computational efficiency. The Transformer can process multiple input sequences simultaneously, making it suitable for parallel processing of multi-dimensional sensor data in natural gas pipeline systems. Through its self-attention mechanism, the Transformer can efficiently process time-series data from multi-dimensional sensors, making it suitable for long-term pipeline monitoring data analysis and improving the accuracy and speed of fault prediction.

[0122] Specifically, such as Figure 6 As shown, the positioning module specifically includes:

[0123] The data acquisition unit is used to acquire the location, geographic information data, and historical operation data of the sensor module;

[0124] The first positioning unit is used to obtain a first position by using a distance calculation method based on the signal strength and the position of the sensor module;

[0125] The second positioning unit is used to obtain the second position by using triangulation based on the position and fault type of the sensor module.

[0126] The third positioning unit is used to train a fault location model based on the historical operation data, input the fault type into the fault location model, and obtain the third position.

[0127] The fourth positioning unit is used to obtain a fourth location based on the geographic information data, the location of the sensor module, and the fault type using a map annotation method;

[0128] The position output unit is used to output the fault position according to any one or more combinations of the first position, the second position, the third position, or the fourth position.

[0129] Specifically, the geographic information data is obtained from a geographic information system, the location of the sensor module is obtained according to the design drawings, and the historical operation data is obtained from a cloud storage.

[0130] Specifically, in the first positioning unit, the distance between the first position (i.e., the fault position) and the position of the sensor module is calculated using a distance calculation method to obtain the first position. The formula for calculating the distance d between the first position and the position of the sensor module is as follows:

[0131] d = vt;

[0132] In the formula, v is the signal propagation speed, which is obtained based on the signal strength; t is the signal propagation time, which is obtained by calculating the delay.

[0133] Specifically, in the second positioning unit, the sensors are arranged uniformly by default, the nature of the fault (e.g., leakage or temperature fluctuation) is known, and the formula for the triangulation method is as follows:

[0134]

[0135] Specifically, x1, y1 and x2, y2 are the known coordinates of the two sensors, d1 and d2 are the distances between the fault and the sensors, and Position is the location where the fault occurred, i.e., the second position.

[0136] Specifically, the working process of the third positioning unit is as follows:

[0137] Historical operational data is one of the key inputs for the third-level fault location unit to predict fault locations. This data includes, but is not limited to, real-time monitoring data such as gas flow rate, pressure, and temperature in the pipeline, which is continuously collected and stored through sensor modules. In addition to real-time data, historical operational data also includes records of past faults, such as fault type, time of occurrence, location, and corrective measures taken. Furthermore, geographical information and environmental factors (such as geology and climate change) of the pipeline's location are also important historical data. By acquiring and storing this historical data, the system can provide rich background information for fault location.

[0138] The construction of the fault location model relies on historical fault cases and their corresponding pipeline locations. First, the system cleans and preprocesses the historical data, removing noise or irrelevant data and standardizing useful data. Then, the system uses machine learning algorithms (such as support vector machines, random forests, and neural networks) to train the fault location model. The model's input includes sensor data (such as gas flow rate, temperature, and pressure), pipeline location data, historical fault types, and their corresponding locations. Through training with machine learning algorithms, the model can identify the relationship between fault types and pipeline locations, providing a basis for future fault prediction.

[0139] Once the system detects a potential fault event, the third localization unit acquires fault type data (e.g., gas leak, pipeline rupture) in real time and inputs it into the pre-trained fault localization model. By passing the correlation information between the fault type and historical data to the model, the system can calculate the possible fault location (i.e., the third location) using algorithms. For example, if the fault type is a gas leak, the model will predict the area most likely to leak based on the specific location of the pipeline in historical leak events and the signal characteristics of the sensors.

[0140] To improve the accuracy of fault location, the third location unit does not rely solely on a single type of historical data, but rather integrates multiple data sources. This data includes not only sensor measurements of temperature, pressure, and flow rate, but also environmental factors such as geology and climate in the pipeline's operating area. By combining this multi-dimensional data, the model can more comprehensively simulate pipeline behavior under specific conditions, improving the accuracy of fault location prediction.

[0141] After the fault type is input and calculated by the location model, the third location unit outputs the predicted fault location, i.e., the third location. This location is represented by geographical coordinates (such as latitude and longitude) or distance from the pipeline. This location output provides maintenance personnel with the specific location of the fault, facilitating rapid location of the fault area, reducing maintenance response time, and ensuring efficient system operation.

[0142] Specifically, the fourth positioning unit directly marks the area where the fault occurred on the map through the geographic information system, GC = f(Sensor Position, Data Points), where Sensor Position is the mark on the map based on the position of the sensor module, Data Points is the operating status data recorded by the sensor, and GC is the location of the fault after calculation and fusion, i.e., the fourth location.

[0143] In some embodiments, the specific process for obtaining the fourth position is as follows:

[0144] The operation of the fourth positioning unit is based on three types of basic data inputs: First, geographic information data, which includes electronic map information of the area where the target natural gas pipeline is located, specifically covering spatial reference data such as topography, road distribution, pipeline route, and coordinate grid system, providing a geographic coordinate framework for subsequent location labeling; second, the location information of the sensor modules, each sensor module is recorded with precise geographic coordinates (longitude, latitude, and altitude if necessary) during installation, which are obtained through calibration using a global satellite navigation system (such as GPS, BeiDou) or a ground coordinate system to form a sensor location dataset; and finally, fault type data, which is identified and generated by the sensor modules or the front-end data analysis unit, such as classification labels for different fault modes such as abnormal pressure, sudden flow changes, and excessive temperature.

[0145] The fourth positioning unit incorporates a Geographic Information System (GIS) module. This module first loads pre-processed geographic information data to construct an electronic map interface containing pipeline routes and sensor distribution points. Based on spatial database technology, the system maps pipeline segments to geographic coordinates and marks the actual locations of each sensor module on the map using specialized symbols, forming a visualized sensor network distribution map. When fault type data is received, the GIS module automatically activates the annotation process and enters fault area positioning mode.

[0146] The specific implementation steps for sensor module position labeling are as follows: First, convert the physical installation coordinates (latitude and longitude) of each sensor into map projection coordinates supported by the GIS system (such as Gauss-Kruger projection or UTM projection) to ensure coordinate system consistency; second, dynamically mark the locations of fault-related sensors on the electronic map with specific icons (such as red dots). If multiple sensors trigger fault signals, mark the locations of all related sensors simultaneously to form a sensor position set; finally, the system automatically generates a sensor position labeling layer, which is superimposed on the basic geographic information layer and supports interactive operations such as zooming and panning, facilitating subsequent data fusion and analysis.

[0147] The operational status data (Data Points) recorded by the sensors include time-series multi-parameter monitoring values ​​(such as pressure, flow rate, temperature, vibration frequency, etc.) and abnormal data points at the time of fault occurrence. The fourth positioning unit first performs spatiotemporal registration on the raw data, attaching a corresponding acquisition timestamp and sensor coordinate information to each data point; secondly, based on the fault type, it extracts feature data segments (such as waveform abrupt changes before and after the fault, duration of parameter exceeding limits, etc.) to form a fault feature dataset with geographic coordinates; finally, through data interpolation or spatial kriging algorithms, the discrete sensor data points are expanded into a continuous spatial distribution field, such as generating a pressure anomaly distribution cloud map, providing a data foundation for spatial location of the fault.

[0148] The fourth positioning unit uses a preset fusion function GC = f(Sensor Position, Data Points) to calculate the fault location. This function is based on a spatial analysis algorithm, and its specific implementation logic is as follows: First, the corresponding analysis model is selected according to the fault type. For example, a pipeline fluid dynamics model is used for pressure change faults, and a mechanical vibration propagation model is used for vibration abnormality faults. Second, the sensor location set is used as a spatial constraint condition. Combined with the spatiotemporal distribution pattern of fault feature data, the most likely area where the fault occurred is inferred by using the least squares method or Kalman filter algorithm. Finally, the geographic coordinates (GC) of the fault location are calculated. These coordinates include the position accuracy confidence interval, which is used to characterize the reliability of the positioning result.

[0149] In the geographic information system interface, the fourth positioning unit marks the calculated fourth location (GC) on the electronic map with a special symbol (such as a flashing yellow triangle). The specific steps are as follows: First, a location error ellipse or buffer area is generated based on the coordinate accuracy confidence interval, and the possible range of the fault location is displayed with different transparency. Second, attribute information such as fault type, occurrence time, and related sensor data are displayed together to form an interactive fault labeling pop-up window. Finally, the system automatically records the labeling history, supports the overlay display and comparative analysis of multiple fault events, and provides the function of exporting location labels (such as KML format files) to facilitate the connection with external geographic information systems or emergency command systems.

[0150] Specifically, the methods for obtaining the first, second, third, and fourth locations can be used in combination. For example, based on the signal strength and the location of the sensor module, this is suitable when signal strength (e.g., pressure, temperature, or flow rate) serves as a clue to a fault. The calibration of the signal strength can provide a preliminary range of the fault location, and then triangulation based on the location of the sensor module can further refine the fault area based on the locations of multiple sensors and the distance to fault detection. The signal strength provides an initial guess for a rough estimate, while triangulation provides a refined spatial estimation. As another example, location based on historical data can be combined with a geographic information system (GIS). The historical data can help the system identify typical operating patterns and trends, thus providing context for real-time data. By combining with a GIS, historical data and the estimated location of the real-time fault point can be combined on a map, providing a more intuitive and accurate fault location. In some special areas, the historical data reveals long-unnoticed patterns, while the GIS can provide specific geographic coordinates, enhancing the accuracy of the location. This system employs a combination of methods. For example, it first determines the approximate location of the fault area using signal strength-based localization, then uses historical data and a geographic information system (GIS) for further precision, and finally confirms the specific fault point through triangulation of sensor deployment locations. This combination allows for the selection of the optimal localization method based on different situations, providing the most accurate results even in complex environments. Therefore, using multiple methods in combination can complement each other's advantages and disadvantages. For instance, signal strength-based localization can quickly provide a preliminary fault range, while triangulation of sensor deployment locations and historical data analysis can further improve accuracy. By combining these methods, the system can more accurately and reliably locate the fault area, thus providing effective support for subsequent maintenance work.

[0151] Specifically, such as Figure 7 As shown, the early warning module specifically includes:

[0152] A label setting unit is used to classify the fault location and the fault type to obtain fault labels;

[0153] The assignment unit is used to assign values ​​and weights to each fault label based on the historical fault database to obtain the fault value and fault weight.

[0154] A scoring unit is used to calculate a fault score by weighting the fault value and the fault weight.

[0155] A rating unit is used to obtain a fault level based on the fault score and a preset risk level;

[0156] A report generation unit is used to generate a fault report based on the fault level, the fault type, and the fault location;

[0157] The early warning unit is used to send early warning signals and fault reports to maintenance personnel based on the fault level.

[0158] Specifically, the tag setting unit functions by classifying and setting fault locations and types in the natural gas pipeline system. The tag setting unit performs structured classification of fault locations and types to generate unique fault tags. The specific process is as follows: First, it receives raw fault information from the detection module, including fault location (e.g., weld A-3 in pipeline segment) and fault type (e.g., corrosion). Then, based on a preset classification rule base, it standardizes and maps fault locations and fault types. Location classification categorizes specific locations hierarchically into major categories such as pipeline segment, valve, and compressor station, and corresponding subcategories (e.g., pipeline segment A). Fault type classification categorizes fault types into major categories such as corrosion, leakage, and pressure anomaly, and associates them with subtypes (e.g., electrochemical corrosion). Finally, it combines the standardized location and type, generating a unique tag in the format "major location category - sub-location category_major type category - sub-type category," for example, "pipeline segment - weld A3_corrosion-electrochemical." In one example, if a gas leak is detected at the flange connection at the C outlet end of the compressor station due to seal aging, the label setting unit parses it as follows: the location category is compressor station, the subcategory is C outlet flange, the type category is leakage, and the subcategory is seal aging. Finally, the label "compressor station-C outlet flange_leakage-seal aging" is generated.

[0159] Specifically, the assignment unit's function is to assign values ​​and weights to different types of fault tags based on a historical fault database. This database is built through years of accumulation and statistical analysis of fault cases during natural gas pipeline operation. The assignment and weighting process quantifies the impact of each fault tag on pipeline operation under different conditions. For example, for a fault tag indicating excessively low pressure, the potential pressure drop caused by the fault can be calculated based on historical data, and a weight can be assigned to it. This embodiment dynamically assigns values ​​to each fault tag, with the weights adaptively adjusted based on historical data and real-time feedback, thereby improving the accuracy of fault assessment.

[0160] In some embodiments, the assignment unit assigns a fault value and weight to each fault label by quantifying historical fault data. The process is as follows: First, it retrieves all records related to the current fault label from the historical fault database to obtain parameters such as occurrence frequency, repair cost, and downtime. Then, based on the fault occurrence frequency and combined with a time decay factor, it calculates the dynamic fault value V using the following formula:

[0161]

[0162] Where α is the time decay factor, T i The historical fault occurrence time is n, where n is the total number of times the record corresponding to this fault tag appears in the historical fault database, and Tc is the current time.

[0163] Finally, based on the degree of impact of the fault, the Analytic Hierarchy Process (AHP) was used to statistically analyze the comprehensive impact score of each label from historical data, and the fault weight value was obtained after normalization.

[0164] As an alternative, the entropy weight method is used to dynamically adjust fault weights. This method dynamically adjusts weights by objectively quantifying the importance of each evaluation indicator based on the distribution characteristics of the data itself, thereby automatically assigning fault weights. In practice, the system first extracts multi-dimensional indicator data (such as frequency of occurrence, repair cost, downtime, and safety risk level) associated with fault tags from the historical fault database, and standardizes this data to eliminate dimensional differences. Then, by analyzing the dispersion of each indicator's data, its information entropy value is calculated. If the data for a certain indicator shows significant differences (e.g., the repair cost of some faults fluctuates greatly), its entropy value is low, indicating that the indicator contains more effective information and should be given a higher weight; conversely, if the indicator data is similar (e.g., the downtime of all faults is similar), the entropy value is high, and the weight is correspondingly reduced. The entire process relies entirely on data objectivity, requires no manual intervention, and when new fault data is added, the system automatically re-evaluates the entropy value and weight of each indicator, reflecting the latest data characteristics in real time. This dynamic mechanism can adapt to changes in the pipeline operating environment, such as increased seasonal corrosion or failure mode migration caused by equipment aging, ensuring that the weight allocation always matches the actual situation, thereby improving the accuracy and timeliness of the early warning model.

[0165] Specifically, the scoring unit's function is to generate a fault score by performing a weighted calculation based on the assigned fault values ​​and fault weights. Using a weighted algorithm (such as weighted average, weighted sum, etc.), the fault values ​​and weights of each fault feature are combined to obtain a comprehensive score. The scoring unit assigns different weights to different features and performs a weighted calculation based on the combined effect of these weights. This algorithm can employ a weighted average method or an improved weighted model to achieve a quantitative assessment of fault risk.

[0166] Specifically, the rating unit compares the fault score with a preset risk level threshold to determine the fault level. The risk level is determined by multiple criteria, such as the severity of the fault type, the likelihood of occurrence, and the scope of its impact. The preset risk levels are divided into several categories, such as minor fault, major fault, critical fault, and extremely critical fault. Based on the fault score calculated by the rating unit, the rating unit can automatically determine the severity of the fault and assign a corresponding fault level.

[0167] Specifically, the report generation unit generates detailed fault reports based on information such as fault level, fault type, and fault location. These reports include, but are not limited to, fault type, fault location, fault impact, and emergency handling suggestions, helping maintenance personnel understand the specifics of the fault and take appropriate action. It can automatically generate detailed reports based on different fault levels, enabling maintenance personnel to quickly locate and resolve faults.

[0168] Specifically, the function of the early warning unit is to send early warning signals and fault reports to maintenance personnel based on the generated fault level. Early warning signals can be delivered via audible and visual alarms, SMS messages, emails, or application notifications, ensuring that maintenance personnel are notified to take appropriate measures immediately.

[0169] The advantages of the above technical solution are as follows: Traditional pipeline fault early warning systems rely on fixed thresholds or linear scoring rules, making it difficult to accurately distinguish the risk differences in complex fault scenarios. This solution, however, uses a dynamic weighting and multi-dimensional scoring coupling mechanism to perform differentiated weight allocation and scoring calculations for fault labels based on a historical fault database. This enables the system to adaptively identify the potential correlation between high-frequency, low-risk faults and low-frequency, high-risk faults. Furthermore, through a non-linear mapping algorithm between fault scores and risk levels, it avoids the limitations of traditional discrete level classification, achieving dynamic assessment of continuous risk probabilities. This design is particularly effective in complex scenarios involving multiple concurrent faults or sudden changes in pipeline operating parameters, accurately capturing transitional states between different risk levels (e.g., the gradual deterioration characteristics when pressure changes are superimposed on localized corrosion), reducing the risk of misjudgment. Simultaneously, the collaborative feedback mechanism between the fault report generation unit and the early warning signal intelligently correlates fault levels with pipeline section operating parameters and maintenance priorities, achieving closed-loop optimization from fault diagnosis to emergency response. This significantly improves the efficiency of maintenance resource scheduling, demonstrating a synergistic effect that traditional single-threshold early warning mechanisms cannot achieve.

[0170] Example 2

[0171] like Figure 8 As shown in the figure, this embodiment provides an intelligent monitoring method for multi-parameter fusion analysis during natural gas pipeline operation, specifically including the following steps:

[0172] S10: Acquire raw data and signal strength;

[0173] S20: Preprocess the original data to obtain preprocessed data, and apply an adaptive compression algorithm to the preprocessed data to obtain compressed data;

[0174] S30: The compressed data is processed using a multi-source data fusion algorithm to obtain fusion features;

[0175] S40: Input the fused features into the fault diagnosis model based on machine learning algorithm, and output the fault type when a fault is diagnosed.

[0176] S50: Obtain the location of the sensor module, geographic information data near the natural gas pipeline, and historical operation data of the natural gas pipeline; and use a positioning algorithm to obtain the fault location based on the signal strength, the location of the sensor module, the geographic information data, the historical operation data, and the fault type.

[0177] S60: Generate a fault report based on the fault location and the fault type, issue a warning signal, and upload the fault report.

[0178] Specifically, step S20 includes the following steps:

[0179] S21: Remove duplicate data from the original data to obtain the first data;

[0180] S22: Use an anomaly detection algorithm to remove outliers from the first data to obtain the second data;

[0181] S23: Obtain the historical second data of the previous time period, generate a threshold interval based on the historical second data using a threshold generation algorithm, remove values ​​that do not belong to the threshold interval from the second data, and obtain preprocessed data;

[0182] S24: The preprocessed data is processed using an adaptive compression algorithm to obtain compressed data;

[0183] The adaptive compression algorithms include an adaptive selection algorithm based on the magnitude of change, an adaptive compression algorithm based on compression error, and an adaptive compression algorithm based on machine learning.

[0184] Specifically, step S30 includes the following steps:

[0185] S31: The compressed data is processed using timestamp interpolation to obtain time-series data;

[0186] S32: The time series data is processed using the Z-value method to obtain standardized data;

[0187] S33: Perform feature extraction on the standardized data to obtain several key features;

[0188] S34: Use the Kalman filter algorithm or the weighted average algorithm to process several of the key features to obtain the fused features.

[0189] Specifically, step S40 includes the following steps:

[0190] S41: Input the fused features into a pre-trained random forest model to obtain the first diagnostic result;

[0191] A support vector machine unit is used to input the fused features into a pre-trained support vector machine model to obtain a second diagnostic result;

[0192] S42: Input the fused features into a pre-trained long short-term memory network model to obtain a third diagnostic result;

[0193] S43: Based on the first diagnostic result, the second diagnostic result, and the third diagnostic result, a multi-model optimization algorithm is used to obtain the fault type.

[0194] Specifically, step S50 includes the following steps:

[0195] S51: Obtain the location, geographic information data, and historical operation data of the sensor module;

[0196] S52: Based on the signal strength and the position of the sensor module, a first position is obtained using a distance calculation method;

[0197] S53: Based on the location and fault type of the sensor module, the second location is obtained using the triangulation method;

[0198] S54: Train a fault location model based on the historical operation data, input the fault type into the fault location model, and obtain the third location;

[0199] S55: Based on the geographic information data, the location of the sensor module, and the fault type, the fourth location is obtained using a map annotation method;

[0200] S56: Output the fault location based on any one or more combinations of the first position, second position, third position, or fourth position.

[0201] Specifically, step S60 includes the following steps:

[0202] S61: Classify the fault location and the fault type to obtain fault labels;

[0203] S62: Assign values ​​and weights to each fault label based on the historical fault database to obtain the fault value and fault weight;

[0204] S63: A fault score is obtained by weighting the fault value and the fault weight;

[0205] S64: Obtain the fault level based on the fault score and the preset risk level;

[0206] S65: Generate a fault report based on the fault level, the fault type, and the fault location;

[0207] S66: Send a warning signal and fault report to the maintenance personnel based on the fault level.

[0208] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0209] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the intelligent monitoring method for multi-parameter fusion analysis during the operation of a natural gas pipeline as described above.

[0210] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. Of course, there are other types of readable storage media, such as quantum memories, graphene memories, etc. It should be noted that the content contained in the computer-readable medium may be appropriately added to or subtracted from the content as required by the legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium may not include electrical carrier signals and telecommunication signals.

[0211] The present invention also provides an electronic device. The electronic device of this invention includes: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the intelligent monitoring method for multi-parameter fusion analysis in natural gas pipeline operation provided by the present invention.

[0212] The following is for reference. Figure 9 It shows a schematic diagram of the structure of a computer system 800 suitable for implementing an electronic device according to embodiments of the present invention. Figure 9 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0213] like Figure 9 As shown, the computer system 800 includes a central processing unit (CPU) 801, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 802 or programs loaded from storage section 808 into random access memory (RAM) 803. The RAM 803 also stores various programs and data required for the operation of the computer system 800. The CPU 801, ROM 802, and RAM 803 are interconnected via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0214] The following components are connected to I / O interface 805: an input section 806 including a keyboard, mouse, etc.; an output section 807 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to I / O interface 805 as needed. A removable medium 811, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 810 as needed so that computer programs read from it can be installed into storage section 808 as needed.

[0215] In particular, according to the embodiments disclosed in this invention, the processes described in the above main step diagrams can be implemented as computer software programs. For example, embodiments of this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the main step diagrams. In the above embodiments, the computer program can be downloaded and installed from a network via communication section 809, and / or installed from removable medium 811. When the computer program is executed by central processing unit 801, it performs the functions defined in the system of this invention.

[0216] It should be noted that the computer-readable medium shown in this invention can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this invention, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0217] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0218] The units described in the embodiments of the present invention can be implemented in software or in hardware. The described units can also be located in a processor; for example, a processor can be described as including a front-end response unit, a receiving unit, and a request unit.

[0219] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. An intelligent monitoring instrument for multi-parameter fusion analysis during natural gas pipeline operation, characterized in that, include: The sensor module, installed on the natural gas pipeline, is used to acquire raw data and signal strength. The preprocessing module is used to preprocess the original data to obtain preprocessed data, and to apply an adaptive compression algorithm to the preprocessed data to obtain compressed data. The fusion module is used to obtain fusion features from the compressed data using a multi-source data fusion algorithm; The diagnostic module is used to input the fused features into a fault diagnosis model based on machine learning algorithms, and output the fault type when a fault is diagnosed. The positioning module is used to acquire the location of the sensor module, geographic information data near the natural gas pipeline, and historical operation data of the natural gas pipeline, and to obtain the fault location using a positioning algorithm based on the signal strength, the location of the sensor module, the geographic information data, the historical operation data, and the fault type. The early warning module is used to generate a fault report based on the fault location and the fault type, issue an early warning signal, and upload the fault report; The preprocessing module specifically includes: The first preprocessing unit is used to remove duplicate data from the original data to obtain the first data. The second preprocessing unit is used to remove outliers from the first data using an anomaly detection algorithm to obtain the second data. The filtering unit is used to obtain the historical second data of the previous time period, generate a threshold interval based on the historical second data using a threshold generation algorithm, and remove values ​​that do not belong to the threshold interval from the second data to obtain preprocessed data. A compression unit is used to apply an adaptive compression algorithm to the preprocessed data to obtain compressed data. The adaptive compression algorithm includes an adaptive selection algorithm based on the magnitude of change, an adaptive compression algorithm based on compression error, and an adaptive compression algorithm based on machine learning. The diagnostic module specifically includes: A random forest unit is used to input the fused features into a pre-trained random forest model to obtain a first diagnostic result; A support vector machine unit is used to input the fused features into a pre-trained support vector machine model to obtain a second diagnostic result; Long Short-Term Memory (LSTM) network units are used to input the fused features into a pre-trained LSM network model to obtain a third diagnostic result; The diagnostic unit is used to obtain the fault type by employing a multi-model optimization algorithm based on the first diagnostic result, the second diagnostic result, and the third diagnostic result; The early warning module specifically includes: A label setting unit is used to classify the fault location and the fault type to obtain fault labels; The assignment unit is used to assign values ​​and weights to each fault label based on the historical fault database to obtain the fault value and fault weight. A scoring unit is used to calculate a fault score by weighting the fault value and the fault weight. The rating unit is used to obtain the fault level based on the fault score and the preset risk level; A report generation unit is used to generate a fault report based on the fault level, the fault type, and the fault location; The early warning unit is used to send early warning signals and fault reports to maintenance personnel based on the fault level. The multi-model optimization algorithm includes any one of the following: voting algorithm, confidence judgment method, rule-based post-processing method, model fusion method, or priority setting method; The model fusion method obtains the final diagnosis result by weighted summation or averaging of the first diagnosis result, the second diagnosis result, and the third diagnosis result, and then determines the fault type based on the final diagnosis result.

2. The intelligent monitoring instrument for multi-parameter fusion analysis in natural gas pipeline operation according to claim 1, characterized in that, The fusion module specifically includes: A time alignment unit is used to process the compressed data using timestamp interpolation to obtain time-series data; A normalization unit is used to process the time-series data using the Z-value method to obtain normalized data. An extraction unit is used to extract features from the standardized data to obtain several key features; The fusion unit is used to process several of the key features using a Kalman filter algorithm or a weighted average algorithm to obtain fused features.

3. The intelligent monitoring instrument for multi-parameter fusion analysis in natural gas pipeline operation according to claim 1, characterized in that, The positioning module specifically includes: The data acquisition unit is used to acquire the location, geographic information data, and historical operation data of the sensor module; The first positioning unit is used to obtain a first position by using a distance calculation method based on the signal strength and the position of the sensor module; The second positioning unit is used to obtain the second position by using triangulation based on the position and fault type of the sensor module. The third positioning unit is used to train a fault location model based on the historical operation data, input the fault type into the fault location model, and obtain the third position. The fourth positioning unit is used to obtain a fourth location based on the geographic information data, the location of the sensor module, and the fault type using a map annotation method; The position output unit is used to output the fault position according to any combination of the first position, the second position, the third position, or the fourth position.

4. The intelligent monitoring instrument for multi-parameter fusion analysis in natural gas pipeline operation according to claim 1, characterized in that, The sensor module includes any combination of acoustic sensors, gas sensors, fiber optic sensors, gas quality sensors, radar sensors, infrared thermal imaging sensors, and environmental radiation sensors.

5. An intelligent monitoring method for multi-parameter fusion analysis in natural gas pipeline operation, characterized in that, The method is based on the intelligent monitoring instrument for multi-parameter fusion analysis in natural gas pipeline operation as described in any one of claims 1-4, and the method includes the following steps: S10: Acquire raw data and signal strength; S20: Preprocess the original data to obtain preprocessed data, and apply an adaptive compression algorithm to the preprocessed data to obtain compressed data; S30: The compressed data is processed using a multi-source data fusion algorithm to obtain fusion features; S40: Input the fused features into the fault diagnosis model based on machine learning algorithm, and output the fault type when a fault is diagnosed. S50: Acquire the location of the sensor module, geographic information data near the natural gas pipeline, and historical operating data of the natural gas pipeline; and use a positioning algorithm to obtain the fault location based on the signal strength, the location of the sensor module, the geographic information data, the historical operating data, and the fault type. S60: Generate a fault report based on the fault location and the fault type, issue a warning signal, and upload the fault report.

Images