Natural gas pipeline pig intelligent real-time positioning system and method

Abstract

The application relates to the technical field of pig intelligent detection, and discloses a natural gas pipeline pig intelligent real-time positioning system and method, which comprises a track matching module, a position prediction module, a position correction module, a speed calculation module and an arrival time prediction module; when the pig intelligent real-time positioning is carried out, the integrity of original data acquisition of the pig under different operating conditions is ensured by adopting multi-source sensing data acquisition and signal preprocessing technology and setting corresponding data acquisition standards for different pipeline environments; meanwhile, the environmental interference error in sensor data acquisition can be detected and corrected in real time by carrying out real-time filtering and feature extraction on multi-source data of inertial measurement, acoustics and pressure, the reliability of the pig positioning data source is ensured, and the positioning error caused by single sensor deviation is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent pig detection, in particular to a natural gas pipeline pig intelligent real-time positioning system and method. BACKGROUND

[0002] Natural gas pipeline, also known as gas pipeline, is a pipeline system for transporting natural gas from the mining site or processing plant to the city gas distribution center or user, which has the advantages of low transportation cost, high safety and low loss. The pig is a special tool for cleaning the pipeline, which is driven by gas, liquid or pipeline transportation medium. It can carry an electromagnetic emission device and a ground receiving instrument to form an electronic tracking system, and can also be configured with other accessories to complete various complex pipeline tasks.

[0003] At present, due to the complex and changeable operation environment of the natural gas pipeline pig, when the pig is positioned in real time, the single sensor data relied on is easily disturbed by the internal environment of the pipeline, and the sensor data collection error cannot be corrected in real time. When the sensor data has measurement deviation, it will cause large calculation error of the pig position, and cannot guarantee the accuracy of the positioning result.

[0004] Therefore, the present application provides a natural gas pipeline pig intelligent real-time positioning system and method to solve the above problems. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a natural gas pipeline pig intelligent real-time positioning system and method to solve the problem of large calculation error of the pig position in the background technology.

[0006] To achieve the above purpose, the present application provides the following technical scheme: A natural gas pipeline pig intelligent real-time positioning system and method, the method comprising the following steps:

[0007] S1, collecting multi-source sensing data of the pig running in the natural gas pipeline to generate pig running original data;

[0008] S2, signal preprocessing and feature extraction are performed on the pig running original data to generate pig running feature data;

[0009] S3, pig running state recognition processing is performed based on the pig running feature data to generate pig running state identification data;

[0010] S4, pig running trajectory matching processing is performed according to the pig running state identification data and pipeline geographic information data to generate pig running trajectory matching data;

[0011] S5, utilize the pig running track matching data to combine pipeline topology data to perform pig running position prediction processing, generate pig running position prediction data;

[0012] S6, based on the pig running position prediction data and real-time sensing data, perform pig position correction processing, generate pig position correction data;

[0013] S7, according to the pig position correction data, perform pig running speed calculation processing, generate pig running speed data;

[0014] S8, combine the pig position correction data and pig running speed data to perform pig arrival time prediction processing, generate pig arrival time prediction data.

[0015] Preferably, the S1 collecting pig multi-source sensing data includes the following operations:

[0016] S11, through the multi-axis inertial measurement unit carried by the pig, collect three-axis acceleration, three-axis angular velocity and three-axis magnetic field intensity data of the pig in the natural gas pipeline, generate pig inertial measurement data;

[0017] S12, through the acoustic sensor carried by the pig, collect sound signal data generated by the friction between the pig and the pipe wall, generate pig acoustic signal data;

[0018] S13, through the pressure sensor carried by the pig, collect the pressure difference data before and after the pig, generate pig pressure difference data;

[0019] S14, through the odometer sensor carried by the pig, collect pig running distance data, generate pig running mileage data;

[0020] S15, through the radio frequency identification reader arranged along the pipeline, collect radio frequency signal data when the pig passes, generate pig radio frequency identification data;

[0021] S16, time stamp alignment and data packaging processing of the pig inertial measurement data, pig acoustic signal data, pig pressure difference data, pig running mileage data and pig radio frequency identification data, generate pig running original data.

[0022] Preferably, the S2 signal preprocessing and feature extraction of the pig running original data includes the following operations:

[0023] S21, wavelet denoising processing and Kalman filtering processing are performed on the pig inertial measurement data, sensor noise and pipeline vibration interference are eliminated, pig inertial measurement filtered data are generated;

[0024] S22. Perform frequency domain transformation and feature extraction on the acoustic signal data of the pigging device, extract the spectral and time domain features of the acoustic signal, and generate acoustic feature data of the pigging device.

[0025] S23. Perform moving average filtering and abrupt change detection processing on the pig pressure difference data to generate pig pressure characteristic data;

[0026] S24. Perform differential processing and cumulative error correction processing on the pigging mileage data to generate pigging mileage correction data.

[0027] S25. Perform signal strength analysis and timestamp matching on the pig radio frequency identification data to generate pig radio frequency positioning data;

[0028] S26. Perform feature-level fusion processing on the pig inertial measurement filter data, pig acoustic feature data, pig pressure feature data, pig mileage correction data and pig radio frequency positioning data to generate pig operation feature data.

[0029] Preferably, the operation status identification process based on the pig operation characteristic data in step S3 includes the following operations:

[0030] S31. Establish a pigging operation status identification model, including five status types: normal operation status, deceleration status, jamming status, rolling status, and reverse motion status.

[0031] S32. Extract the time-domain features, frequency-domain features, and time-frequency-domain features from the pig operation feature data to generate the pig operation feature vector;

[0032] S33. Use the long short-term memory network in the deep learning algorithm to perform time-series pattern recognition processing on the pig operation feature vector to generate pig operation status probability distribution data.

[0033] S34. The Viterbi algorithm is used to perform optimal path search processing on the probability distribution data of the pig's operating status to generate the pig's operating status sequence data. The mathematical expression of the Viterbi algorithm is:

[0034] ;

[0035] in Indicates time In state The probability of the maximum probability path. Indicates from state Transition to state The transition probability, Indicates the state The following observations the emission probability of the object, representing time the observation value.

[0036] Preferably, the pig running trajectory matching processing in S4 includes the following operations:

[0037] S41, acquiring natural gas pipeline geographic information data, including pipeline three-dimensional coordinate data, pipeline elevation data, pipeline curvature data and pipeline feature point data;

[0038] S42, establishing a pipeline geographic information feature database, including elbow position data, tee position data, valve position data and pipe diameter change point data;

[0039] S43, performing feature matching processing on the pig running state identification data and the pipeline geographic information feature database, using a dynamic time warping algorithm to calculate the similarity of the pig running features and the pipeline geographic features, and generating pig-geographic feature matching degree data;

[0040] S44, performing fusion processing on the pig-geographic feature matching degree data and the pig running mileage data based on a particle filtering algorithm, and generating pig running trajectory matching data;

[0041] S45, performing trajectory smoothing processing and confidence assessment on the pig running trajectory matching data, and generating pig running trajectory optimization data.

[0042] Preferably, the pig running position prediction processing in S5 includes the following operations:

[0043] S51, constructing a natural gas pipeline topology structure diagram, including pipeline node data, pipeline segment data, pipeline connection relationship data and pipeline flow direction data;

[0044] S52, establishing a pig running position prediction model based on a hidden Markov model, taking the pipeline topology structure as the state space and the pig running features as the observation space;

[0045] S53, using a forward-backward algorithm to calculate the probability distribution of the pig at each position in the pipeline topology structure, generating pig position probability distribution data, and the mathematical expression of the forward algorithm is:

[0046] ;

[0047] wherein represents the forward probability of being in state at time , represents the transition probability from state to state , representing the state observed emission probability, representing the time observation value;

[0048] S54, adopting a maximum posterior probability estimation method to perform position point estimation processing on the pig position probability distribution data, and generating pig running position prediction data;

[0049] S55, performing uncertainty quantification and error ellipse calculation on the pig running position prediction data, and generating pig position prediction confidence interval data.

[0050] Preferably, the pig position correction processing in S6 includes the following operations:

[0051] S61, acquiring pig sensor data in real time, including inertial measurement data, acoustic data, pressure data and mileage data;

[0052] S62, establishing a pig position correction model based on extended Kalman filtering, taking pig running position prediction data as state prediction value and real-time sensor data as observation value;

[0053] S63, calculating the residual between the state prediction value and the observation value, adopting chi-square test to detect abnormal measurement value, and generating residual detection data;

[0054] S64, using an adaptive filtering algorithm to dynamically adjust filtering parameters according to the residual size, and generating pig position correction data;

[0055] S65, performing smoothing processing and outlier rejection on the pig position correction data, and generating pig position optimization data.

[0056] Preferably, the pig running speed calculation processing in S7 includes the following operations:

[0057] S71, extracting pig position optimization data under adjacent time stamps, calculating pig displacement change, and generating pig displacement data;

[0058] S72, adopting a difference algorithm to calculate pig instantaneous speed, combining with sliding window average processing to generate pig running speed data;

[0059] S73, performing acceleration calculation and jerk analysis on the pig running speed data, and generating pig running dynamic characteristic data;

[0060] S74, establishing a speed-pipe diameter-pressure relationship model, analyzing the correlation between pig running speed and pipeline parameters, and generating pig speed characteristic data;

[0061] S75, probability distribution fitting and abnormal speed early warning are performed on the pig running speed data, and pig speed early warning data is generated.

[0062] Preferably, the pig arrival time prediction processing in S8 includes the following operations:

[0063] S81, key point position data of the natural gas pipeline along the line are acquired, including distribution stations, valve rooms, user access points and emergency cutoff points;

[0064] S82, a pig arrival time prediction model based on deep reinforcement learning is established, considering pipeline topological structure, pig running speed historical data and pipeline pressure change factors;

[0065] S83, the remaining distance and predicted running time of the pig to each key point are calculated, and pig arrival time prediction data are generated;

[0066] S84, a Monte Carlo simulation method is used to analyze the uncertainty of the prediction results, and pig arrival time probability distribution data are generated;

[0067] S85, time deviation compensation and real-time update processing are performed on the pig arrival time prediction data, and pig arrival time optimization data are generated.

[0068] Preferably, the system includes:

[0069] A trajectory matching module matches the pig running state with the pipeline geographic information, including a geographic information database, a feature matching unit, a particle filter unit and a trajectory optimization unit;

[0070] A position prediction module predicts the pig running position, including a topological structure library, a hidden Markov model unit, a probability calculation unit and an uncertainty evaluation unit;

[0071] A position correction module corrects the pig position according to real-time sensing data, including an extended Kalman filter unit, a residual detection unit and an adaptive filter unit;

[0072] A speed calculation module calculates the pig running speed, including a displacement calculation unit, a speed estimation unit and a dynamic characteristic analysis unit;

[0073] An arrival time prediction module predicts the time of the pig arriving at a key point, including a key point database, a reinforcement learning model unit and an uncertainty analysis unit.

[0074] Advantages

[0075] Compared with the prior art, the present application provides a natural gas pipeline pig intelligent real-time positioning system and method, which has the following advantages:

[0076] 1. In the present application, when the pig intelligent real-time positioning is carried out, the integrity of the original data acquisition of the pig under different operating conditions is ensured by adopting multi-source sensing data acquisition and signal preprocessing technology and setting corresponding data acquisition standards for different pipeline environments; at the same time, through real-time filtering and feature extraction of inertial measurement, acoustic, pressure multi-source data, the environmental interference error in sensor data acquisition can be detected and corrected in real time, the reliability of the pig positioning data source is ensured, and the positioning error caused by the deviation of a single sensor is reduced.

[0077] 2. In the present application, when the pig operating state monitoring is carried out, the pig operating state recognition model is established and the pig operating characteristic vector is calculated to judge whether the pig appears abnormal state such as jamming, rolling and reverse motion in real time; the system can timely find the problem of operating track deviation, and when the pig state is abnormal, the risk warning signal can be generated in real time through the operating risk assessment processing, so that the control strategy can be adjusted in time when the pig operates abnormally, and the safety of the pig operation and the accuracy of the track matching are ensured.

[0078] 3. In the present application, when the pig position is determined, the data fusion processing and position correction technology are adopted to synchronously integrate the pig position prediction data, speed data and arrival time evaluation data; the weighted fusion and real-time error correction are carried out according to the confidence degree of different data sources, so that the system can realize the intelligent positioning of multi-source cooperation, reduce the risk of amplification of accumulated error in pig position calculation, and improve the accuracy of positioning result and the overall reliability of the system. BRIEF DESCRIPTION OF DRAWINGS

[0079] Fig. 1 The flowchart of the intelligent real-time positioning method of the natural gas pipeline pig in the present application;

[0080] Fig. 2 The framework diagram of the intelligent real-time positioning system of the natural gas pipeline pig in the present application. DETAILED DESCRIPTION

[0081] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0082] Specific embodiment: please refer to Figs. 1-2 An intelligent real-time positioning system and method of a natural gas pipeline pig, the method comprising the following steps:

[0083] S1, collect multi-source sensing data of the pig when running in the natural gas pipeline, generate pig running original data;

[0084] S2, perform signal preprocessing and feature extraction on the pig running original data, generate pig running feature data;

[0085] S3, perform pig running state recognition processing based on the pig running feature data, generate pig running state identification data;

[0086] S4, perform pig running trajectory matching processing according to the pig running state identification data and pipeline geographic information data, generate pig running trajectory matching data;

[0087] S5, perform pig running position prediction processing by using the pig running trajectory matching data in combination with pipeline topological structure data, generate pig running position prediction data;

[0088] S6, perform pig position correction processing based on the pig running position prediction data and real-time sensing data, generate pig position correction data;

[0089] S7, perform pig running speed calculation processing according to the pig position correction data, generate pig running speed data;

[0090] S8, perform pig arrival time prediction processing in combination with the pig position correction data and pig running speed data, generate pig arrival time prediction data.

[0091] The collection of multi-source sensing data of the pig in S1 includes the following operations:

[0092] S11, collect three-axis acceleration, three-axis angular velocity and three-axis magnetic field intensity data of the pig in the natural gas pipeline by a multi-axis inertial measurement unit carried by the pig, generate pig inertial measurement data;

[0093] S12, collect sound signal data generated by friction between the pig and the pipe wall by an acoustic sensor carried by the pig, generate pig acoustic signal data;

[0094] S13, collect pig front and back pressure difference data by a pressure sensor carried by the pig, generate pig pressure difference data;

[0095] S14, collect pig running distance data by a mileage wheel sensor carried by the pig, generate pig running mileage data;

[0096] S15, collect radio frequency signal data when the pig passes through by a radio frequency identification reader arranged along the pipeline, generate pig radio frequency identification data;

[0097] S16, time stamp alignment and data packaging processing of the pig inertia measurement data, pig acoustic signal data, pig pressure difference data, pig running mileage data and pig radio frequency identification data are performed to generate pig running original data.

[0098] S2, signal preprocessing and feature extraction of the pig running original data include the following operations:

[0099] S21, wavelet denoising processing and Kalman filtering processing are performed on the pig inertia measurement data to eliminate sensor noise and pipeline vibration interference to generate pig inertia measurement filtered data;

[0100] S22, frequency domain transformation and feature extraction are performed on the pig acoustic signal data to extract acoustic signal spectrum features and time domain features to generate pig acoustic feature data;

[0101] S23, moving average filtering and mutation point detection processing are performed on the pig pressure difference data to generate pig pressure feature data;

[0102] S24, difference processing and cumulative error correction processing are performed on the pig running mileage data to generate pig mileage corrected data;

[0103] S25, signal strength analysis and time stamp matching processing are performed on the pig radio frequency identification data to generate pig radio frequency positioning data;

[0104] S26, feature level fusion processing is performed on the pig inertia measurement filtered data, pig acoustic feature data, pig pressure feature data, pig mileage corrected data and pig radio frequency positioning data to generate pig running feature data.

[0105] S3, running state recognition processing based on the pig running feature data includes the following operations:

[0106] S31, a pig running state recognition model is established, including five state types of normal running state, deceleration state, blockage state, rolling state and reverse motion state;

[0107] S32, time domain features, frequency domain features and time-frequency domain features in the pig running feature data are extracted to generate a pig running feature vector;

[0108] S33, a long short-term memory network in a deep learning algorithm is used to perform time series pattern recognition processing on the pig running feature vector to generate pig running state probability distribution data, including the following steps:

[0109] S331, a long short-term memory network architecture is constructed, including an input layer, a hidden layer and an output layer, and the hidden layer contains LSTM units;

[0110] S332. Use the inertial measurement filter data of the pig, the acoustic feature data of the pig, the pressure feature data of the pig, the mileage correction data of the pig, and the radio frequency positioning data of the pig as multi-channel input sequences to train the weight parameters of the LSTM network.

[0111] S333. Optimize network parameters and minimize the state recognition error function through a time-series backpropagation mechanism;

[0112] S334. Use regularization techniques to prevent overfitting and improve the network's ability to generalize to unknown data.

[0113] S335. Output the probability distribution data of the pig's operating status as the network prediction result;

[0114] S34. The Viterbi algorithm is used to perform optimal path search processing on the probability distribution data of the pig's operating status, generating the pig's operating status sequence data. The mathematical expression of the Viterbi algorithm is:

[0115] ;

[0116] in Indicates time In state The probability of the maximum probability path. Indicates from state Transition to state The transition probability, Indicates the state The following observations The probability of launch, Indicates time The observed values.

[0117] The pigging trajectory matching process in S4 includes the following operations:

[0118] S41. Obtain geographic information data of natural gas pipelines, including pipeline three-dimensional coordinate data, pipeline elevation data, pipeline curvature data, and pipeline feature point data;

[0119] S42. Establish a pipeline geographic information feature database, including elbow location data, tee location data, valve location data, and pipe diameter change point data;

[0120] S43. Perform feature matching processing between the pig operation status identification data and the pipeline geographic information feature database, and use the dynamic time warping algorithm to calculate the similarity between the pig operation features and the pipeline geographic features to generate pig-geographic feature matching degree data.

[0121] S44, fuse the pig-geofeature matching degree data and the pig running mileage data based on the particle filter algorithm to generate pig running trajectory matching data, including the following operations:

[0122] S441, initialize a particle set, each particle representing a position and state of the pig;

[0123] S442, predict the next time state of the particle according to a state transition model;

[0124] S443, calculate the weight of each particle according to an observation model, and use the observation data for importance sampling;

[0125] S444, resample the particles to avoid degeneration problems and retain high-weight particles;

[0126] S445, output the particle weighted average as the pig running trajectory matching data;

[0127] S45, perform trajectory smoothing processing and confidence evaluation on the pig running trajectory matching data to generate pig running trajectory optimization data.

[0128] The pig running position prediction processing in S5 includes the following operations:

[0129] S51, construct a natural gas pipeline topology structure diagram, including pipeline node data, pipeline segment data, pipeline connection relationship data and pipeline flow direction data;

[0130] S52, establish a pig running position prediction model based on a hidden Markov model, taking the pipeline topology structure as the state space and the pig running feature as the observation space, including the following operations:

[0131] S521, define the state space of the hidden Markov model, corresponding to the node positions in the pipeline topology structure, the state space is a discrete finite set, and each state corresponds to a specific position node in the pipeline topology;

[0132] S522, define the observation space of the hidden Markov model, corresponding to the pig running feature data, the observation space includes quantized characteristic values of the track parameter, the track gauge parameter and the horizontal shape parameter;

[0133] S523, train the state transition probability matrix and the observation probability matrix based on the forward-backward algorithm, maximize the model likelihood function through iterative updating, specifically including:

[0134] Use the forward probability to calculate the likelihood of the observation sequence;

[0135] Use the backward probability to calculate the posterior probability of state transition;

[0136] Update the model parameters using the expectation-maximization algorithm;

[0137] S524. Optimize model parameters based on the Viterbi algorithm, minimizing prediction error by finding the optimal state sequence, specifically including:

[0138] Calculate the optimal path probability for each state;

[0139] Backtracking to obtain the globally optimal state sequence;

[0140] Adjust the model parameters based on the sequence alignment results;

[0141] S525. Verify the goodness of fit of the model. Use root mean square error and correlation coefficient to evaluate the prediction accuracy of the model and ensure that the model matches the actual operating conditions of the pipeline.

[0142] S53. Calculate the probability distribution of the pig's location at each position in the pipeline topology using the forward-backward algorithm, generating pig location probability distribution data. The mathematical expression for the forward algorithm is:

[0143] ;

[0144] in Indicates time In state The forward probability, Indicates from state Transition to state The transition probability, Indicates the state The following observations The probability of launch, Indicates time Observed values;

[0145] S54. The maximum a posteriori probability estimation method is used to perform location point estimation on the pig location probability distribution data to generate pig operation location prediction data.

[0146] S55. Perform uncertainty quantification and error ellipse calculation on the pig's operating position prediction data to generate pig position prediction confidence interval data.

[0147] The following operations are included in the pig position correction process in S6:

[0148] S61. Real-time acquisition of pig sensor data, including inertial measurement data, acoustic data, pressure data, and mileage data;

[0149] S62. Establish a pig position correction model based on extended Kalman filtering, using the pig's operating position prediction data as the state prediction value and real-time sensor data as the observation value, including the following operations:

[0150] S621. Define the state equation and observation equation. The state variables include the pig position, velocity, and acceleration.

[0151] State equation formula:

[0152] ;

[0153] Observation equation formula:

[0154] ;

[0155] in express Time-state vector express Observation vector at time, Represents the state transition function. Represents the observation function, and These represent process noise and observation noise, respectively.

[0156] S622. Based on the linearization requirements of extended Kalman filtering, perform local linearization approximation on the state equation and observation equation;

[0157] S623. Prediction steps: Calculate the state prediction value and covariance prediction;

[0158] State prediction formula:

[0159] ;

[0160] in This represents the prior state estimate. This represents the posterior state estimate. Represents the control input vector;

[0161] Covariance prediction formula:

[0162] ;

[0163] in This indicates a priori estimation of the covariance. This indicates the posterior estimate of the covariance. Represents the process noise covariance matrix. Represents the state transition Jacobian matrix. This represents the transpose of the state transition Jacobian matrix;

[0164] S624. Update steps: Calculate the Kalman gain, update the state estimate and covariance;

[0165] Kalman gain formula:

[0166] ;

[0167] wherein denotes the Kalman gain, denotes the observation Jacobian matrix, denotes the transpose of the observation Jacobian matrix, denotes the observation noise covariance matrix;

[0168] State update formula:

[0169] ;

[0170] wherein denotes the observation measurement vector, denotes the observation function;

[0171] Covariance update formula:

[0172] ;

[0173] wherein denotes the identity matrix;

[0174] S625, processing the nonlinear error, improving the correction accuracy by iterative extended Kalman filtering;

[0175] S63, calculating the residual between the state prediction value and the observation value, detecting abnormal measurement values by chi-square test, and generating residual detection data;

[0176] S64, dynamically adjusting the filtering parameters according to the residual size by using the adaptive filtering algorithm, and generating the pig position correction data;

[0177] S65, performing smoothing processing and outlier rejection on the pig position correction data, and generating the pig position optimization data.

[0178] The pig running speed calculation processing in S7 includes the following operations:

[0179] S71, extracting the pig position optimization data under the adjacent time stamps, calculating the pig displacement change, and generating the pig displacement data;

[0180] S72, calculating the instantaneous speed of the pig by using the difference algorithm, and combining the sliding window average processing to generate the pig running speed data, including the following steps:

[0181] S721, obtaining the pig position optimization data under the adjacent time stamps, and extracting the position coordinate sequence;

[0182] S722, calculating the displacement difference value between the adjacent time points, and calculating the instantaneous speed by using the first-order forward difference formula:

[0183] ;

[0184] wherein denotes the instantaneous velocity at the time point, and denote the position coordinates at adjacent time points, and denote the corresponding timestamps, denotes the summation index;

[0185] S723, setting the sliding window size parameter , which takes an integer value ranging from 5 to 20;

[0186] S724, performing sliding window averaging processing on the instantaneous velocity sequence, and calculating the average velocity in the window using the following formula:

[0187] ;

[0188] wherein denotes the instantaneous velocity at the time point, denotes the window size, denotes the current time point index;

[0189] S725, processing window boundary data and continuously filling the velocity data using a step-by-step filling method;

[0190] S726, outputting the smoothed pig running speed data;

[0191] S73, performing acceleration calculation and jerk analysis on the pig running speed data to generate pig running dynamic characteristic data;

[0192] S74, establishing a speed-pipe diameter-pressure relationship model to analyze the correlation between pig running speed and pipeline parameters, and generating pig speed characteristic data;

[0193] S75, performing probability distribution fitting and abnormal speed early warning on the pig running speed data to generate pig speed early warning data.

[0194] The pig arrival time prediction processing in S8 includes the following operations:

[0195] S81, obtaining key point position data along the natural gas pipeline, including distribution stations, valve rooms, user access points, and emergency cutoff points;

[0196] S82, establishing a pig arrival time prediction model based on deep reinforcement learning, considering pipeline topological structure, pig running speed historical data, and pipeline pressure change factors;

[0197] S83, calculate the remaining distance of the pig to each key point and the predicted running time, and generate pig arrival time prediction data;

[0198] S84, perform uncertainty analysis on the prediction results using the Monte Carlo simulation method, and generate pig arrival time probability distribution data, including the following operations:

[0199] S841, define random variables such as pig running speed, pipeline pressure change and distance error;

[0200] S842, generate a set of random samples to simulate the pig running process;

[0201] S843, run multiple simulation tests to statistically analyze the distribution of pig arrival time;

[0202] S844, calculate the confidence interval and percentile to quantify the prediction uncertainty;

[0203] S845, output the pig arrival time probability distribution data as the simulation result;

[0204] S85, perform time bias compensation and real-time update processing on the pig arrival time prediction data, and generate pig arrival time optimization data.

[0205] The method further comprises:

[0206] S9, perform running risk assessment processing based on the position correction data, the running speed data and the arrival time prediction data, and generate pig running risk assessment data;

[0207] The running risk assessment processing includes: establishing a risk evaluation index system, calculating the risk level using the fuzzy comprehensive evaluation method, and generating a risk warning signal, including the following operations:

[0208] S91, based on the risk evaluation index system, establish a fuzzy comprehensive evaluation factor set and an evaluation level set , wherein the factor set includes pig running state risk factors, pipeline environment risk factors and equipment reliability risk factors, and the evaluation level set includes four evaluation levels: safe, low risk, medium risk and high risk;

[0209] S92, based on the weight distribution principle of the fuzzy comprehensive evaluation, determine the weight distribution vector of each risk factor to the evaluation target, and the weight value is obtained by combining expert experience judgment and risk index data characteristics, and satisfies the weight normalization condition;

[0210] S93, construct fuzzy relation matrix By single factor evaluation, the membership degree of each risk factor to each evaluation grade is determined, and the fuzzy mapping from risk index to evaluation grade is completed;

[0211] S94, fuzzy synthesis operation is carried out, and the comprehensive evaluation result vector is calculated by using weighted average type fuzzy synthesis operator The calculation formula is:

[0212] ;

[0213] Among them , the comprehensive evaluation result vector is represented by , the weight distribution vector is represented by , and the fuzzy relation matrix is represented by

[0214] S95, based on the comprehensive evaluation result vector , the risk grade is determined by the determination rule of fuzzy comprehensive evaluation, and the corresponding risk early warning signal is generated;

[0215] S10, the position correction data, the running speed data, the arrival time prediction data and the running risk evaluation data are subjected to multi-dimensional data fusion processing, the intelligent real-time positioning data of the pig is generated and output to the pipeline monitoring center, the multi-dimensional data fusion processing is subjected to multi-dimensional data fusion processing by using D-S evidence theory algorithm, including the following steps:

[0216] S101, define the recognition framework, including the pig state and position proposition;

[0217] S102, assign a basic probability assignment function to each evidence source;

[0218] S103, calculate the conflict factor between evidence sources Using the formula:

[0219] ;

[0220] Among them and represent the basic probability assignment of different evidence sources;

[0221] S104, apply Dempster combination rule to fuse the evidence, generate the fused basic probability assignment, use the formula:

[0222] ;

[0223] Among them , the basic probability assignment of proposition ;

[0224] S105, according to the fusion result decision pig intelligent real-time positioning data, output to the pipeline monitoring center.

[0225] The system comprises:

[0226] The trajectory matching module matches the pig running state with the pipeline geographic information, including a geographic information database, a feature matching unit, a particle filter unit and a trajectory optimization unit;

[0227] The position prediction module predicts the pig running position, including a topological structure library, a hidden Markov model unit, a probability calculation unit and an uncertainty evaluation unit;

[0228] The position correction module corrects the pig position according to real-time sensing data, including an extended Kalman filter unit, a residual detection unit and an adaptive filter unit;

[0229] The speed calculation module calculates the pig running speed, including a displacement calculation unit, a speed estimation unit and a dynamic characteristic analysis unit;

[0230] The arrival time prediction module predicts the pig arrival time at key points, including a key point database, a reinforcement learning model unit and an uncertainty analysis unit.

[0231] The operation steps of the natural gas pipeline pig intelligent real-time positioning system and method are as follows:

[0232] Step 1, multi-source sensing data acquisition and preprocessing

[0233] The system first synchronously collects multi-source sensing data during the pig running process through various sensors carried by the pig and radio frequency identification readers arranged along the pipeline. The collected raw data is aligned and encapsulated with time stamps to form a standardized pig running raw data set. Then, wavelet denoising, Kalman filtering and moving average filtering signal processing techniques are used to denoise and filter the raw data, eliminate sensor noise and pipeline vibration interference, extract effective feature information, and generate standardized pig running feature data.

[0234] Step 2, intelligent identification and modeling of running state

[0235] Based on the preprocessed running feature data, the system uses the long short-term memory network in the deep learning algorithm for time series pattern recognition processing. By constructing a network architecture including an input layer, a hidden layer and an output layer, the pig running feature vector is analyzed in time series to generate running state probability distribution data. The state probability distribution is searched for the optimal path using the Viterbi algorithm to establish a pig running state recognition model, accurately identify five running states including normal running, deceleration, blockage, rolling and reverse motion, and generate corresponding state identification data.

[0236] Step three, running track matching and position prediction

[0237] The system matches the identified running state identification data with the pre-established pipeline geographic information feature database, calculates the similarity of the pig running features and the pipeline geographic features using the dynamic time warping algorithm, and generates matching degree data. Based on the particle filtering algorithm, the matching degree data is fused with the pig running mileage data for processing, and pig running track matching data is generated. Combined with the pipeline topology data, a position prediction model is established using a hidden Markov model, the probability distribution of the pig at each position in the pipeline topology is calculated through a forward-backward algorithm, and position prediction data and its confidence interval are generated.

[0238] Step four, real-time position correction and speed calculation

[0239] The system corrects the position prediction results in real time through an extended Kalman filter model, fuses the prediction data with real-time sensor data, calculates the residual error between the state prediction value and the observed value, and detects abnormal measurement values using the chi-square test. Using an adaptive filtering algorithm, the filter parameters are dynamically adjusted according to the residual error, and the optimized pig position correction data is generated. Based on the corrected position data, the pig instantaneous speed is calculated using a difference algorithm, and smooth and reliable running speed data is generated through sliding window averaging processing.

[0240] Step five, arrival time prediction and risk assessment

[0241] Combining the pig position correction data and running speed data, the system establishes an arrival time prediction model based on deep reinforcement learning, calculates the remaining distance and predicted running time of the pig to each key point. The Monte Carlo simulation method is used to analyze the uncertainty of the prediction results, and the arrival time probability distribution data is generated. Based on position, speed and arrival time data, a risk assessment index system is established through fuzzy comprehensive evaluation method, the risk level is calculated and the corresponding risk warning signal is generated.

[0242] Step six, multi-source data fusion and positioning output

[0243] Finally, the system uses the D-S evidence theory algorithm to fuse multiple sources of data, including position correction data, running speed data, arrival time prediction data and risk assessment data. By defining the recognition framework and the basic probability assignment function, the conflict factor between evidence sources is calculated, the Dempster combination rule is applied for evidence fusion, and comprehensive pig intelligent real-time positioning data is generated. The final result is output to the pipeline monitoring center, providing complete and accurate positioning information support for pig running monitoring and dispatching decision-making.

[0244] It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting; it is not intended to exclude myriad other embodiments of the present application that other inventors can develop based on the same general inventive concepts embodied by the described embodiments. That is, although the present application is described in terms of particular embodiments and implementations, it is to be understood that the terminology used is for the purpose of descriptive clarity and that it is intended to be limited only by the words recited in the appended claims. The scope of the present application shall be limited only by the claims.

[0245] While the embodiments of the application have been shown and described herein, it is to be understood that the application is not limited to these embodiments. Rather, numerous modifications are possible without departing from the spirit and scope of the present application as delineated by the claims and their equivalents.

Claims

1. A method for intelligent real-time positioning of a natural gas pipeline pig, characterized in that: The method includes the following steps: S1. Collect multi-source sensor data of the pigging device during operation in the natural gas pipeline and generate raw data of the pigging device operation. S2. Perform signal preprocessing and feature extraction on the raw data of the pig operation to generate pig operation feature data; S3. Based on the pig operation characteristic data, perform pig operation status identification processing to generate pig operation status identification data; S4. Perform pig operation trajectory matching processing based on the pig operation status identification data and pipeline geographic information data to generate pig operation trajectory matching data. S5. Using the matching data of the pig's running trajectory and the pipeline topology data, perform pig running position prediction processing to generate pig running position prediction data. S6. Based on the predicted operating position data of the pig and the real-time sensor data, perform pig position correction processing to generate pig position correction data. S7. Calculate the operating speed of the pig based on the pig position correction data to generate pig operating speed data. S8. Combine the pig position correction data and pig running speed data to perform pig arrival time prediction processing to generate pig arrival time prediction data.

2. The intelligent real-time positioning method for a natural gas pipeline pig according to claim 1, characterized in that: The process of acquiring multi-source sensor data from the pig in S1 includes the following operations: S11. The multi-axis inertial measurement unit mounted on the pig collects triaxial acceleration, triaxial angular velocity and triaxial magnetic field strength data of the pig in the natural gas pipeline, and generates pig inertial measurement data; S12. Acoustic signal data generated by the friction between the pig and the pipe wall is collected by the acoustic sensor mounted on the pig, and the pig acoustic signal data is generated. S13. Collect pressure difference data before and after the pig by the pressure sensor on the pig and generate pig pressure difference data. S14. Collect the pig's running distance data through the mileage wheel sensor mounted on the pig, and generate pig running mileage data; S15. Collect radio frequency signal data of the pig as it passes by through radio frequency identification readers arranged along the pipeline, and generate pig radio frequency identification data. S16. The inertial measurement data, acoustic signal data, pressure difference data, mileage data, and radio frequency identification data of the pig are timestamped and encapsulated to generate the original pig operation data.

3. The intelligent real-time positioning method for a natural gas pipeline pig according to claim 1, characterized in that: The signal preprocessing and feature extraction of the raw data from the pig operation in S2 includes the following operations: S21. Perform wavelet noise reduction and Kalman filtering on the inertial measurement data of the pig to eliminate sensor noise and pipeline vibration interference, and generate pig inertial measurement filtered data. S22. Perform frequency domain transformation and feature extraction on the acoustic signal data of the pigging device, extract the spectral and time domain features of the acoustic signal, and generate acoustic feature data of the pigging device. S23. Perform moving average filtering and abrupt change detection processing on the pig pressure difference data to generate pig pressure characteristic data; S24. Perform differential processing and cumulative error correction processing on the pigging mileage data to generate pigging mileage correction data. S25. Perform signal strength analysis and timestamp matching on the pig radio frequency identification data to generate pig radio frequency positioning data; S26. Perform feature-level fusion processing on the pig inertial measurement filter data, pig acoustic feature data, pig pressure feature data, pig mileage correction data and pig radio frequency positioning data to generate pig operation feature data.

4. The intelligent real-time positioning method for a natural gas pipeline pig according to claim 1, characterized in that: The operation status identification process based on the pig operation characteristic data in S3 includes the following operations: S31. Establish a pigging operation status identification model, including five status types: normal operation status, deceleration status, jamming status, rolling status, and reverse motion status. S32. Extract the time-domain features, frequency-domain features, and time-frequency-domain features from the pig operation feature data to generate the pig operation feature vector; S33. Use the long short-term memory network in the deep learning algorithm to perform time-series pattern recognition processing on the pig operation feature vector to generate pig operation status probability distribution data. S34. The Viterbi algorithm is used to perform optimal path search processing on the probability distribution data of the pig's operating status to generate the pig's operating status sequence data. The mathematical expression of the Viterbi algorithm is: ； in Indicates time In state The probability of the maximum probability path. Indicates from state Transition to state The transition probability, Indicates the state The following observations The probability of launch, Indicates time The observed values.

5. The intelligent real-time positioning method for a natural gas pipeline pig according to claim 1, characterized in that: The pig's trajectory matching process in S4 includes the following operations: S41. Obtain geographic information data of natural gas pipelines, including pipeline three-dimensional coordinate data, pipeline elevation data, pipeline curvature data, and pipeline feature point data; S42. Establish a pipeline geographic information feature database, including elbow location data, tee location data, valve location data, and pipe diameter change point data; S43. Perform feature matching processing on the pig operation status identification data and the pipeline geographic information feature database, and use the dynamic time warping algorithm to calculate the similarity between the pig operation features and the pipeline geographic features to generate pig-geographic feature matching degree data. S44. Based on the particle filter algorithm, the matching degree data of the pig-geographic feature is fused with the pig operation mileage data to generate pig operation trajectory matching data. S45. Perform trajectory smoothing and confidence assessment on the pig's running trajectory matching data to generate optimized pig running trajectory data.

6. The intelligent real-time positioning method for a natural gas pipeline pig according to claim 1, characterized in that: The pig's operating location prediction process in S5 includes the following operations: S51. Construct a natural gas pipeline topology diagram, including pipeline node data, pipeline segment data, pipeline connection relationship data, and pipeline flow direction data; S52. Establish a pig operation location prediction model based on hidden Markov model, taking the pipeline topology as the state space and the pig operation characteristics as the observation space. S53. Calculate the probability distribution of the pig's location at each position in the pipeline topology using the forward-backward algorithm, generating pig location probability distribution data. The mathematical expression for the forward algorithm is: ； in Indicates time In state The forward probability, Indicates from state Transition to state The transition probability, Indicates the state The following observations The probability of launch, Indicates time Observed values; S54. The maximum a posteriori probability estimation method is used to perform location point estimation processing on the pig position probability distribution data to generate pig operation position prediction data. S55. Perform uncertainty quantification and error ellipse calculation on the pig's operating position prediction data to generate pig position prediction confidence interval data.

7. The intelligent real-time positioning method for a natural gas pipeline pig according to claim 1, characterized in that: The pig position correction process in S6 includes the following operations: S61. Real-time acquisition of pig sensor data, including inertial measurement data, acoustic data, pressure data, and mileage data; S62. Establish a pig position correction model based on extended Kalman filter, using the pig operation position prediction data as the state prediction value and real-time sensor data as the observation value. S63. Calculate the residual between the predicted state value and the observed value, use the chi-square test to detect abnormal measurement values, and generate residual detection data; S64. Use an adaptive filtering algorithm to dynamically adjust the filtering parameters according to the residual size to generate pig position correction data; S65. Smooth the pig position correction data and remove outliers to generate pig position optimization data.

8. The intelligent real-time positioning method for a natural gas pipeline pig according to claim 1, characterized in that: The calculation and processing of the pig's operating speed in S7 includes the following operations: S71. Extract the optimized location data of the pig under adjacent timestamps, calculate the change in pig displacement, and generate pig displacement data. S72. The instantaneous speed of the pig is calculated using a differential algorithm, and the pig's operating speed data is generated by combining the sliding window averaging process. S73. Perform acceleration calculation and jerk analysis on the pig's operating speed data to generate dynamic characteristic data of the pig's operation. S74. Establish a speed-pipe diameter-pressure relationship model, analyze the correlation between the pig's operating speed and pipeline parameters, and generate pig speed characteristic data. S75. Perform probability distribution fitting and abnormal speed warning on the pig operation speed data to generate pig speed warning data.

9. The intelligent real-time positioning method for a natural gas pipeline pig according to claim 1, characterized in that: The pig arrival time prediction process in S8 includes the following operations: S81. Obtain location data of key points along the natural gas pipeline, including distribution stations, valve chambers, user access points, and emergency cut-off points; S82. Establish a pipeline pig arrival time prediction model based on deep reinforcement learning, taking into account pipeline topology, historical data of pipeline pig operating speed, and pipeline pressure change factors. S83. Calculate the remaining distance and predicted running time of the pig to each key point, and generate pig arrival time prediction data. S84. The Monte Carlo simulation method is used to perform uncertainty analysis on the prediction results and generate the probability distribution data of the arrival time of the pig; S85. Perform time deviation compensation and real-time update processing on the predicted arrival time data of the pig to generate optimized arrival time data of the pig.

10. A real-time intelligent positioning system for a natural gas pipeline pig, characterized in that: A method for intelligent real-time positioning of a natural gas pipeline pig according to any one of claims 1-9, characterized in that the system comprises: The trajectory matching module matches the pig's operating status with pipeline geographic information, including a geographic information database, feature matching unit, particle filtering unit, and trajectory optimization unit; The location prediction module predicts the operating location of the pig, including a topology library, hidden Markov model units, probability calculation units, and uncertainty assessment units. The position correction module corrects the position of the pig based on real-time sensor data, and includes an extended Kalman filter unit, a residual detection unit, and an adaptive filter unit. The speed calculation module calculates the operating speed of the pig, including a displacement calculation unit, a speed estimation unit, and a dynamic characteristic analysis unit. The arrival time prediction module predicts the arrival time of the pig at key points, and includes a key point database, a reinforcement learning model unit, and an uncertainty analysis unit.

Images