A method and system for online monitoring of cathodic protection corrosion rate of buried pipelines

Abstract

This invention belongs to the field of buried pipeline corrosion monitoring technology. It provides an online monitoring method and system for the cathodic protection corrosion rate of buried pipelines, comprising: deploying a multi-dimensional collaborative monitoring sensor array; collecting sensor data to assess the health status of the buried pipeline; if the buried pipeline is healthy, dividing the monitoring area into several corrosion analysis units; modeling the corrosion rate of each corrosion analysis unit; compiling historical corrosion rate sequences of the corrosion analysis units and training a multi-layer LSTM network model to predict the corrosion rate, obtaining a predicted corrosion rate sequence; estimating and analyzing the predicted wall thickness of the buried pipeline within each corrosion analysis unit at the predicted time period; if the dispersion is high, screening for high-risk corrosion units; analyzing the predicted wall thickness at the end of the high-risk corrosion units or corrosion analysis units to determine whether to trigger potentiostat regulation; if triggered, regulating the output current of the potentiostat.

Description

Technical Field

[0001] This invention belongs to the field of buried pipeline corrosion monitoring technology, specifically a method and system for online monitoring of cathodic protection corrosion rate of buried pipelines. Background Technology

[0002] Buried pipelines, as vital infrastructure for energy and material transmission, play a crucial role in fields such as oil, natural gas, and chemicals. However, due to their long-term underground burial, buried pipelines are subject to various factors, including complex soil environments, chemical corrosion, and electrochemical corrosion, making them highly susceptible to corrosion. Corrosion not only leads to thinning of the pipeline walls and reduced strength but can also cause serious safety accidents such as pipeline leaks and explosions, posing a significant threat to people's lives, property, and the ecological environment. Therefore, cathodic protection of buried pipelines, accurate monitoring and effective assessment of corrosion status, and timely implementation of protective measures are of great practical significance.

[0003] Existing methods focus only on electrochemical parameters or environmental factors, lacking multi-dimensional data fusion analysis. This makes it difficult to comprehensively reflect the corrosion mechanisms of pipelines in complex environments. For example, the impact of soil stress changes on corrosion rates is often overlooked, leading to discrepancies between monitoring results and actual corrosion conditions. Distributed sensor networks face challenges such as data synchronization difficulties and poor transmission stability. Traditional wired transmission methods are costly and susceptible to environmental interference, while wireless transmission suffers severe signal attenuation over long distances and in complex terrain, resulting in data loss or delays and affecting real-time performance.

[0004] In existing technologies, simple threshold-based early warning mechanisms cannot capture the dynamic trend of corrosion rate changes, making it difficult to achieve early warning. Corrosion rate prediction based on physical models does not fully consider the spatiotemporal heterogeneity of the soil environment and lacks the ability to deeply mine historical data, resulting in insufficient long-term prediction accuracy. The parameters of the potentiostat in cathodic protection systems are usually set in a fixed manner and cannot be dynamically adjusted according to the real-time corrosion status, which can easily lead to insufficient or excessive protection and shorten the service life of pipelines. Existing monitoring technologies often focus on monitoring and analyzing the local corrosion status of pipelines, lacking a comprehensive assessment and risk warning mechanism for the corrosion status of the entire monitoring area. In practical applications, if only local corrosion is considered while ignoring the overall understanding of the corrosion status of the entire monitoring area, it is difficult to formulate scientific and reasonable protection strategies and effectively prevent and control pipeline corrosion accidents.

[0005] To address the above problems, this invention proposes an online monitoring method and system for the cathodic protection corrosion rate of buried pipelines. Summary of the Invention

[0006] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0007] The technical solution adopted by this invention to solve the technical problem is: an online monitoring method for the cathodic protection corrosion rate of buried pipelines, comprising:

[0008] The sensor data within the monitoring area is used to assess the health status of the buried pipeline. If the buried pipeline is healthy, the monitoring area is divided into several corrosion analysis units by gridding. The unit feature vector of each corrosion analysis unit is calculated, and a corrosion rate model is constructed for the corrosion analysis unit based on the unit feature vector.

[0009] Specifically, a multi-dimensional collaborative monitoring sensor array is acquired, and the multi-dimensional collaborative monitoring sensor array is divided into several sensor groups according to spatial distribution. Each sensor group has only one sensor of each type. The sensor data collected by the sensor group is acquired and normalized to obtain normalized sensor data.

[0010] A flag value is set for each sensor group and its initial value is 0. The analytic hierarchy process is used to assign weights to the normalized sensor data. Based on the assigned weights, the normalized sensor data is weighted and summed to obtain the comprehensive anomaly index of the sensor group. If the comprehensive anomaly index is greater than or equal to the anomaly index threshold, the flag value is incremented once. If it is less than the anomaly index threshold, the flag value is reset to 0.

[0011] Obtain the flag value of the sensor group. If the flag value is greater than or equal to the preset judgment stability value, it is determined that the sensor data collected by the sensor group is abnormal, and the sensor group is marked as an abnormal sensor group.

[0012] If no abnormal sensor group is found in the multi-dimensional collaborative monitoring sensor array, the buried pipeline is judged to be in good condition.

[0013] If the buried pipeline is in good condition, the monitoring area is divided into corrosion analysis units using the variable density grid method.

[0014] The sensor data within the corrosion analysis unit are integrated into the original feature vector of the sensor group. The Euclidean distance from the sensor group to the center of the corrosion analysis unit is calculated. Based on the original feature vectors of all sensor groups within the corrosion analysis unit, combined with the calculated Euclidean distance, the unit feature vector of the corrosion analysis unit is calculated using the distance-weighted average method.

[0015] Based on the obtained unit feature vector, the uniform corrosion rate of the corrosion analysis unit is calculated according to the core formula of Faraday's law as the basic model. The soil corrosivity function is embedded in the basic model to modify the basic model and obtain the corrosion rate model of the corrosion analysis unit.

[0016] The historical corrosion rate sequence of the corrosion analysis unit is obtained based on the corrosion rate model. A multi-layer LSTM network model is trained based on the historical corrosion rate sequence to predict the corrosion rate and obtain the predicted corrosion rate sequence.

[0017] Specifically, based on the corrosion rate model, the corrosion rate of the corrosion analysis unit at each time point during the historical monitoring period is calculated and integrated. Outliers and missing values ​​are detected and processed for all corrosion rates to obtain a historical corrosion rate sequence.

[0018] The historical corrosion rate sequence is normalized to obtain the normalized corrosion rate sequence. A multi-layer LSTM network model is constructed and trained using the normalized corrosion rate sequence. Starting from the current time, the trained multi-layer LSTM network model is used to predict the corrosion rate of the corrosion analysis unit in a rolling manner. The predicted values ​​are then inversely normalized and integrated to obtain the predicted corrosion rate sequence of the corrosion analysis unit within the prediction period.

[0019] By integrating historical corrosion rate sequences and predicted corrosion rate sequences, the predicted wall thickness of buried pipelines in each corrosion analysis unit at the end of the predicted time period is estimated and the dispersion is analyzed. If the dispersion is high, risk corrosion units are divided, and the predicted wall thickness at the end of the risk corrosion units or corrosion analysis units is analyzed to determine whether the potentiostat is triggered. If triggered, the output current of the potentiostat is adjusted.

[0020] Specifically, the initial wall thickness of the buried pipeline is obtained, and the data is processed by combining the historical corrosion rate sequence and the predicted corrosion rate sequence to calculate the predicted wall thickness of the buried pipeline at the end of the predicted time period. The calculation result is marked as the predicted wall thickness at the end of the time period.

[0021] The predicted wall thicknesses at the endpoints of all corrosion analysis units are integrated into a single predicted wall thickness set. The average absolute deviation of the predicted wall thickness set is calculated. If the average absolute deviation is greater than the deviation threshold, risk corrosion units are divided within the corrosion analysis units. The predicted wall thicknesses at the endpoints of all risk corrosion units are integrated into a single risk predicted wall thickness set. Data processing is performed on the risk predicted wall thickness set to obtain the risk predicted wall thickness. If the average absolute deviation is less than or equal to the deviation threshold, data processing is performed directly on the predicted wall thickness set to obtain the average predicted wall thickness.

[0022] If a risky corrosion unit exists and the predicted wall thickness meets the preset wall thickness condition, or the average predicted wall thickness meets the preset wall thickness condition, the potentiostat is triggered for adjustment.

[0023] The risk corrosion unit is divided as follows: the predicted wall thickness of the endpoint of all corrosion analysis units is integrated into a predicted wall thickness set. The average absolute deviation of the predicted wall thickness set is calculated. If the average absolute deviation is greater than the deviation threshold, the predicted wall thickness set is normalized. The normalized predicted wall thickness set is used as a feature vector and input into the K-means algorithm to cluster into two clusters. The predicted wall thickness at the center of the two clusters is compared. The corrosion analysis unit corresponding to the cluster with the smaller predicted wall thickness at the center of the cluster is marked as a risk corrosion unit.

[0024] If there are risky corrosion units, the predicted corrosion rates of all risky corrosion units at the next time point from the current time point are processed to obtain the controlled predicted rate. If there are no risky corrosion units, the predicted corrosion rates of all corrosion analysis units at the next time point from the current time point are processed to obtain the controlled predicted rate.

[0025] The target potential is dynamically adjusted by combining the control and prediction rate. The potential deviation is defined. Based on the PID algorithm, the potential deviation is used as the input and the output current of the potentiostat is used as the output. The output current of the potentiostat is calculated and the output of the potentiostat is dynamically adjusted according to the output current of the potentiostat.

[0026] An online monitoring system for cathodic protection corrosion rate of buried pipelines includes the following modules:

[0027] Data acquisition module: Deploys a multi-dimensional collaborative monitoring sensor array to collect sensor data within the monitoring area;

[0028] Rate analysis module: Acquire sensor data to assess the health status of buried pipelines. If the buried pipeline is in good condition, the monitoring area is divided into several corrosion analysis units by gridding. The unit feature vector of each corrosion analysis unit is calculated, and a corrosion rate model is constructed for the corrosion analysis unit based on the unit feature vector.

[0029] Corrosion prediction module: Based on the corrosion rate model, the historical corrosion rate sequence of the corrosion analysis unit is obtained. A multi-layer LSTM network model is trained based on the historical corrosion rate sequence to predict the corrosion rate and obtain the predicted corrosion rate sequence.

[0030] Current control module: Integrates historical corrosion rate sequence and predicted corrosion rate sequence, estimates and analyzes the dispersion of the predicted wall thickness of buried pipelines in each corrosion analysis unit at the end of the predicted time period. If the dispersion is high, risk corrosion units are divided, and the predicted wall thickness at the end of the risk corrosion unit or corrosion analysis unit is analyzed to determine whether the potentiostat control is triggered. If triggered, the output current of the potentiostat is controlled.

[0031] The beneficial effects of the present invention are as follows:

[0032] 1. This invention deploys a multi-dimensional collaborative monitoring sensor array to comprehensively collect data and accurately assess the health status of buried pipelines. It can promptly detect potential problems, perform grid-based division of pipelines in good condition and subsequent analysis, accurately locate corrosion risks in different areas, and use a multi-layer LSTM network model to predict corrosion rates, significantly improving prediction accuracy. This allows managers to understand pipeline corrosion in advance, rationally arrange maintenance plans, reduce production shutdowns and maintenance caused by sudden corrosion accidents, reduce economic losses, and ensure the continuity of energy transmission.

[0033] 2. This invention estimates the final predicted wall thickness based on historical and predicted corrosion rate sequences and analyzes the dispersion to screen out high-risk corrosion units. It can accurately focus on high-risk areas and determine whether to trigger potentiostat adjustment by analyzing the final predicted wall thickness. This enables intelligent and precise adjustment of protective measures, avoiding resource waste caused by blind adjustment. At the same time, it effectively enhances the cathodic protection effect, extends the service life of buried pipelines, reduces safety risks caused by pipeline corrosion, and provides a solid guarantee for the stable operation of energy transportation systems. Attached Figure Description

[0034] The invention will now be further described with reference to the accompanying drawings.

[0035] Figure 1 This is a flowchart illustrating the steps of an online monitoring method for cathodic protection corrosion rate of buried pipelines according to an embodiment of the present invention.

[0036] Figure 2 This is a system module architecture diagram of an online monitoring system for cathodic protection corrosion rate of buried pipelines according to an embodiment of the present invention. Detailed Implementation

[0037] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0038] Example 1

[0039] Please see Figure 1 As shown in the embodiment of the present invention, an online monitoring method for cathodic protection corrosion rate of buried pipelines includes the following steps:

[0040] S1: Deploy a multi-dimensional collaborative monitoring sensor array to collect sensor data;

[0041] A multi-dimensional collaborative monitoring sensor array is deployed within the monitoring area, including electrochemical sensors, environmental factor sensors, and stress monitoring units. Sensor data is collected during the monitoring period through the multi-dimensional collaborative monitoring sensor array.

[0042] Among them, the electrochemical sensor measures electrochemical data based on a three-electrode system, and the electrochemical data includes the polarization resistance of the buried pipeline surface. Open circuit potential on the surface of buried pipelines and the polarization potential of the buried pipeline surface In the three-electrode system, the working electrode is the buried pipeline body, the reference electrode is a saturated copper sulfate electrode (CSE), and the auxiliary electrode is a platinum electrode. The environmental factor sensor uses an integrated soil parameter sensor to collect soil corrosion parameters, including soil moisture (h), soil pH, and soil resistivity. The stress monitoring unit uses a fiber Bragg grating stress sensor to measure stress parameters, including the circumferential stress of the buried pipeline. Axial stress of buried pipelines ;

[0043] The sensor data includes electrochemical data, soil corrosion parameters, and stress parameters;

[0044] The monitoring period refers to a time period during which the cathodic protection corrosion rate of buried pipelines is monitored.

[0045] The sensor spacing is determined by a variable density grid layout method. The sensor spacing is then used to deploy the multi-dimensional collaborative monitoring sensor array. A three-layer protective structure is used to ensure the long-term stable operation of the sensors. The multi-dimensional collaborative monitoring sensor array is evenly divided into several sensor groups according to the spatial distribution. The sensor groups are numbered according to the spatial order. Spatially adjacent sensor groups have adjacent numbers. Each sensor group has one and only one sensor of each type.

[0046] Each sensor group is connected to an edge node, which is equipped with a multi-core processor and a real-time operating system, as well as a digital lock-in amplifier to enhance the acquisition capability of weak electrochemical signals.

[0047] The sensor data collected by the edge nodes is preprocessed. First, a moving average filter is used to remove random noise. Then, an IIR band-stop filter is used to perform power frequency notch filtering to eliminate interference from the power grid. Finally, characteristic parameters including root mean square value and kurtosis value are calculated to determine the validity and stability of the sensor data.

[0048] The preprocessed sensor data is transmitted using a hybrid star and mesh networking mode. The sensor data is first transmitted over short distances via a LoRa wireless link, and then transmitted through spatially adjacent edge nodes. The sensor data is aggregated to the regional gateway and transmitted to the cloud server through multiple edge nodes. Clock synchronization is achieved through the NTP network time protocol and the GPS timing module. The GPS timing module, combined with a hardware timer, timestamps the sensor data to ensure the time consistency of all sensor data collected at the same time.

[0049] It should be noted that precise clock synchronization makes data collected by sensors at different locations and of different types comparable in terms of time, providing an accurate time reference for subsequent monitoring and analysis.

[0050] S2: Assess the health status of buried pipelines based on sensor data. If the buried pipeline is healthy, divide the monitoring area into several corrosion analysis units by gridding, calculate the unit feature vector of each corrosion analysis unit, and model the corrosion rate of the corrosion analysis unit based on the unit feature vector.

[0051] Real-time acquisition of sensor data collected by a multi-dimensional collaborative monitoring sensor array, and analysis of the corrosion status of buried pipelines based on the sensor data;

[0052] Specifically, based on the sensor group with sequence number i, the open-circuit potential of the buried pipeline surface collected by the sensor group is... Circumferential stress of buried pipelines Axial stress of buried pipelines Z-score positive normalization was performed separately, and the polarization resistance of the buried pipeline surface was also analyzed. Perform reverse normalization to obtain normalized sensor data, including open-circuit potential mapping values. Circumferential stress mapping value Axial stress mapping value and polarization resistance mapping value ;

[0053] Set a flag value for each sensor group The initial value is assigned to 0, where i represents the sensor group number. The analytic hierarchy process (AHP) is used to assign weights to the normalized sensor data. Based on these weights, the normalized sensor data is weighted and summed to obtain the comprehensive anomaly index of the sensor group. The comprehensive anomaly index is compared with the anomaly index threshold. If the comprehensive anomaly index is greater than or equal to the anomaly index threshold, the flag value is set. Perform an increment operation; if the overall anomaly index is less than the anomaly index threshold, adjust the flag value. Return to 0;

[0054] If the flag value If the value is greater than or equal to the preset stability value, it is determined that the sensor data collected by the sensor group is abnormal, and the sensor group is marked as an abnormal sensor group.

[0055] For abnormal sensor groups, two abnormal sensor groups with adjacent serial numbers are grouped into the same sensor group set. The number of abnormal sensor groups in each sensor group set is obtained and compared with the judgment stability value. If the number of abnormal sensor groups is less than the judgment stability value, the reason why the sensor group in the sensor group set is marked as an abnormal sensor group is that the sensor is faulty. A fault repair signal is generated, and the faulty sensor is located according to the serial number of the abnormal sensor group.

[0056] If the number of abnormal sensor groups is greater than or equal to the judgment stability value, then the reason why the sensor group in the sensor group set is marked as an abnormal sensor group is that the buried pipeline has severe corrosion. A corrosion processing signal is generated, and the area where the buried pipeline has severe corrosion is located according to the sequence number of the abnormal sensor group.

[0057] Based on the generated fault repair signal or corrosion treatment signal, staff are arranged to carry out sensor fault repair and buried pipeline corrosion treatment. For abnormal sensor groups after sensor fault repair or buried pipeline corrosion treatment, the abnormal sensor group mark is cleared, so that the abnormal sensor group returns to the sensor group when it was not marked as an abnormal sensor.

[0058] If there are no abnormal sensor groups in the multi-dimensional collaborative monitoring sensor array, the buried pipeline is judged to be in good condition. The monitoring area is divided into L×W corrosion analysis units, and each corrosion analysis unit contains the same number of sensor groups.

[0059] For each corrosion analysis unit, based on the k-th sensor group within the j-th corrosion analysis unit, the sensor data is integrated into the original feature vector of the sensor group. The Euclidean distance from the sensor group to the center of the corrosion analysis unit is calculated. Based on the original feature vectors of all sensor groups within the corrosion analysis unit, and combined with the calculated Euclidean distances, the unit feature vector of the corrosion analysis unit is calculated using the distance-weighted average method. ;

[0060] ;

[0061] Based on the obtained unit feature vector, the corrosion rate of the corrosion analysis unit is modeled in combination with physical laws.

[0062] Specifically, according to the core formula of Faraday's law, the uniform corrosion rate of the j-th corrosion analysis unit... It can be represented as: ;

[0063] in, The corrosion current density is represented by the Stern-Geary formula, where M represents the atomic weight of the metal and n represents the valence of the metal. This indicates the density of the steel used in buried pipelines. The stress correction factor is expressed by the following formula: ;

[0064] in, By fitting the constant load tensile test, the accelerating effect of stress on corrosion was reflected. This represents the yield strength of the steel used in buried pipelines, indicating the stress value at which the steel begins to undergo significant plastic deformation. This represents the equivalent stress, which comprehensively measures the effects of circumferential and axial stresses on steel. ;

[0065] Using uniform corrosion rate as the basic model, soil corrosion parameters are obtained from the unit feature vector, and a three-factor correction function including soil moisture, soil resistivity, and soil pH is constructed. The soil corrosion function is labeled as a soil corrosion function. The soil corrosion function is embedded into the basic model to modify the basic model and obtain the corrosion rate model of the corrosion analysis unit.

[0066] It should be noted that the purpose of this step is to calculate a comprehensive anomaly index based on normalized sensor data and compare it with a threshold to achieve early warning of corrosion anomalies. Through cluster analysis of adjacent abnormal sensor groups, sensor failures and pipeline corrosion can be distinguished, the problem area can be accurately located, the monitoring area can be gridded when there are no anomalies, a corrosion rate model can be constructed by combining physical laws, and stress and soil parameter correction can be integrated to provide mechanistic model support for prediction, realizing the transformation from data collection to knowledge discovery.

[0067] S3: Based on the corrosion rate model, the historical corrosion rate sequence of the corrosion analysis unit is sorted out and a multi-layer LSTM network model is trained to predict the corrosion rate, and the predicted corrosion rate sequence is obtained.

[0068] Based on the obtained corrosion rate model, the corrosion rate of the corrosion analysis unit is calculated in real time. ,in, Representing the current time point, based on any corrosion analysis unit, the time period starting from the time point when the corrosion analysis unit was marked as the corrosion analysis unit and ending at the current time point is marked as the historical monitoring period. The corrosion rate of the corrosion analysis unit at each time point within the historical monitoring period is integrated. All corrosion rates are subjected to outlier detection and processing according to the time sequence, and missing values ​​are supplemented by linear interpolation to obtain the historical corrosion rate sequence. ;

[0069] ;

[0070] in, Indicates a point in time within the historical monitoring period. There are a total of R+1 time points during the historical monitoring period;

[0071] Z-Score normalization was used to normalize the historical corrosion rate sequence, transforming it into a dimensionless sequence with a mean of 0 and a standard deviation of 1, thus eliminating the influence of dimensional differences on model training and obtaining a normalized corrosion rate sequence.

[0072] Based on the temporal characteristics of the normalized corrosion rate sequence, a multi-layer LSTM network model is constructed to predict the corrosion rate of the corrosion analysis unit.

[0073] Specifically, the multi-layer LSTM network model adopts a two-layer LSTM structure: the first layer is configured with 64 neurons and sequence output is enabled to extract deep features in the time dimension; the second layer is set with 32 neurons to focus on key information; a Dropout regularization layer is connected after each layer with a dropout rate of 0.2. By randomly dropping some neuron connections, overfitting is suppressed; the output end maps the hidden state of the LSTM layer to a one-dimensional space through a fully connected layer to generate a single-step erosion rate prediction value.

[0074] Training samples are generated using a sliding time window technique. The sliding window length is set, and the normalized corrosion rate sequence is converted into sample pairs. The sample pairs are divided into a dataset, a validation set, and a test set according to time order. The training set is used for parameter learning, the validation set is used to monitor overfitting, and the test set is used to evaluate generalization ability, ensuring that the distribution of sample pairs conforms to the time dependence characteristics of time series data.

[0075] Mean squared error (MSE) is chosen as the loss function to quantify the average squared deviation between the predicted and true values, which is suitable for continuous value prediction tasks of erosion rate. The Adam optimizer is used to dynamically adjust the learning rate to balance training speed and convergence accuracy. It combines the advantages of AdaGrad's adaptive gradient and RMSProp's moving average to effectively handle non-stationary data. During training, the gradient of the loss function with respect to all trainable parameters is calculated through the backpropagation algorithm, and the optimizer updates the parameters according to the gradient direction. After each training round, the validation set loss is calculated. If the loss does not decrease for 10 consecutive rounds, the early stopping mechanism is triggered to terminate the training, avoid invalid iterations and prevent model overfitting.

[0076] Starting from the current time, the corrosion rate of the corrosion analysis unit is predicted in a rolling manner using the trained multi-layer LSTM network model.

[0077] Specifically, a time-series sample of the sliding window length is extracted from the normalized corrosion rate sequence, with the current time point as the endpoint. This time-series sample is input into a multilayer LSTM network model to generate a normalized predicted value for the next time point. The predicted value is then added to the time-series sample according to the time sequence, and the corrosion rate at the earliest time point in the time-series sample is removed, forming a new time-series sample, which is then input into the multilayer LSTM network model to achieve dynamic prediction every time step. This dynamic prediction is repeated time-step to predict the corrosion rate of the corrosion analysis unit within a preset prediction period. The obtained predicted values ​​are then denormalized to restore the dimensions, and integrated to obtain the predicted corrosion rate sequence of the corrosion analysis unit within the prediction period. ;

[0078] ;

[0079] in, Indicates a point in time within the forecast period. U represents the number of time points within the prediction period;

[0080] It should be noted that the purpose of this step is to use historical corrosion rate sequences to construct a multi-layer LSTM network model to achieve time series prediction. Training samples are generated through a sliding window, and the model is optimized by combining an early stop mechanism to avoid overfitting. This achieves an upgrade from a mechanistic model to a data-driven prediction model, rolling prediction of future corrosion rates, providing time-dimensional decision-making basis for preventive maintenance, and improving the pipeline's full life cycle management capabilities.

[0081] S4: Integrate historical corrosion rate sequences and predicted corrosion rate sequences, estimate and analyze the dispersion of the predicted wall thickness of buried pipelines in each corrosion analysis unit at the end of the predicted time period. If the dispersion is high, screen high-risk corrosion units, analyze the predicted wall thickness at the end of the high-risk corrosion units or corrosion analysis units, and determine whether the potentiostat is triggered. If triggered, adjust the output current of the potentiostat.

[0082] Obtain the initial wall thickness of buried pipelines By combining historical corrosion rate sequences and predicted corrosion rate sequences, the predicted wall thickness of buried pipelines at the end of the prediction period is calculated, and the calculation results are marked as the predicted wall thickness at the end of the prediction period. The formula is: ;

[0083] in, This indicates the duration of the interval between adjacent time points, and the duration of the interval between all adjacent time points is the same.

[0084] Obtain the endpoint predicted wall thickness of all corrosion analysis units, integrate the endpoint predicted wall thickness of all corrosion analysis units into a predicted wall thickness set, calculate the mean absolute deviation (MAD) of the predicted wall thickness set, and compare it with the deviation threshold.

[0085] If the average absolute deviation of the predicted wall thickness set is greater than the deviation threshold, it indicates that the dispersion of the predicted wall thickness at the end of all corrosion analysis units is high, and the deviation of the predicted wall thickness at the end of different corrosion analysis units is large.

[0086] The predicted wall thickness set is Z-score normalized. The normalized predicted wall thickness set is then used as a feature vector input into the K-means algorithm. Specifically, two cluster centers are randomly initialized in the normalized predicted wall thickness set. The Euclidean distance from each endpoint predicted wall thickness to the two cluster centers is calculated. The endpoint predicted wall thickness is assigned to the cluster with the closer distance. After the assignment is completed, the center of each cluster is recalculated, which is the mean of all endpoint predicted wall thicknesses in the cluster. The Euclidean distance from each endpoint predicted wall thickness to the two cluster centers is calculated and assigned repeatedly until the cluster centers no longer change.

[0087] Comparing the predicted wall thickness at the endpoints of the two cluster centers, the corrosion analysis units corresponding to the predicted wall thickness at the endpoints of the clusters with smaller predicted wall thicknesses at the cluster centers are marked as high-risk corrosion units. All high-risk corrosion units are then analyzed at time points. The predicted corrosion rates are summed and averaged to obtain the controlled predicted rate. The predicted wall thickness of the endpoint of all risk corrosion units is integrated into a set of predicted wall thicknesses. The predicted wall thicknesses are then summed and averaged to obtain the predicted wall thickness.

[0088] If the average absolute deviation of the predicted wall thickness set is less than or equal to the deviation threshold, it indicates that the dispersion of the predicted wall thickness at the endpoint of all corrosion analysis units is low, and the deviation of the predicted wall thickness at the endpoint of different corrosion analysis units is small. Therefore, all corrosion analysis units are considered at time points... The predicted corrosion rates are summed and averaged to obtain the controlled predicted rate. The average predicted wall thickness is obtained by summing and averaging the predicted wall thickness set.

[0089] If a set of predicted wall thicknesses exists and the predicted wall thickness is greater than the preset standard wall thickness value, or if a set of predicted wall thicknesses does not exist and the average predicted wall thickness is greater than the preset standard wall thickness value, the potentiostat will be activated.

[0090] For the risk prediction wall thickness set or predicted wall thickness set that triggers potentiostat regulation, obtain the corresponding regulation prediction rate, and dynamically adjust the target potential based on the regulation prediction rate. The higher the corrosion rate, the more negative the target potential should be to enhance the protection strength, but it should not exceed the hydrogen evolution threshold of -1.2V.

[0091] Define potential deviation ; ;

[0092] in This represents the polarization potential on the surface of the buried pipeline collected at time point t.

[0093] Based on the PID algorithm, using potential deviation As input, the output current of the potentiostat is used. The output should include the calculation formula: ;

[0094] in, Represents a time variable. Represents the proportionality coefficient. Represents the integral coefficient. Represents the differential coefficient;

[0095] Based on the calculated output current of the potentiostat The output of the potentiostat is dynamically adjusted;

[0096] It should be noted that the purpose of this step is to divide the risk corrosion units using MAD and K-means algorithms, and to dynamically adjust the output current of the potentiostat based on the PID algorithm to achieve adaptive control of cathodic protection. This step integrates monitoring, prediction and regulation in a closed loop to form an intelligent operation and maintenance system of perception, analysis, decision-making and execution, maximizing the protection effect while avoiding the risk of hydrogen evolution and extending the service life of the pipeline.

[0097] The technical solution of this invention is as follows: a multi-dimensional collaborative monitoring sensor array is deployed to collect sensor data. The health status of the buried pipeline is assessed based on the sensor data. If the buried pipeline is healthy, the monitoring area is divided into several corrosion analysis units in a grid. The unit feature vector of each corrosion analysis unit is calculated, and the corrosion rate of the corrosion analysis unit is modeled based on the unit feature vector. Based on the corrosion rate model, the historical corrosion rate sequence of the corrosion analysis unit is organized, and a multi-layer LSTM network model is trained to predict the corrosion rate, resulting in a predicted corrosion rate sequence. The historical corrosion rate sequence and the predicted corrosion rate sequence are integrated to estimate and analyze the dispersion of the predicted wall thickness of the buried pipeline in each corrosion analysis unit at the end of the predicted time period. If the dispersion is high, risk corrosion units are screened. The predicted wall thickness at the end of the risk corrosion unit or the corrosion analysis unit is analyzed to determine whether the potentiostat is triggered. If triggered, the output current of the potentiostat is adjusted.

[0098] Example 2

[0099] like Figure 2 As shown in the embodiment of the present invention, an online monitoring system for cathodic protection corrosion rate of buried pipelines includes the following modules:

[0100] Data acquisition module: Deploys a multi-dimensional collaborative monitoring sensor array to collect sensor data within the monitoring area;

[0101] Rate analysis module: Acquire sensor data to assess the health status of buried pipelines. If the buried pipeline is in good condition, the monitoring area is divided into several corrosion analysis units by gridding. The unit feature vector of each corrosion analysis unit is calculated, and a corrosion rate model is constructed for the corrosion analysis unit based on the unit feature vector.

[0102] Corrosion prediction module: Based on the corrosion rate model, the historical corrosion rate sequence of the corrosion analysis unit is obtained. A multi-layer LSTM network model is trained based on the historical corrosion rate sequence to predict the corrosion rate and obtain the predicted corrosion rate sequence.

[0103] Current control module: Integrates historical corrosion rate sequence and predicted corrosion rate sequence, estimates and analyzes the dispersion of the predicted wall thickness of buried pipelines in each corrosion analysis unit at the end of the predicted time period. If the dispersion is high, risk corrosion units are divided, and the predicted wall thickness at the end of the risk corrosion unit or corrosion analysis unit is analyzed to determine whether the potentiostat control is triggered. If triggered, the output current of the potentiostat is controlled.

[0104] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for online monitoring of cathodic protection corrosion rate of buried pipelines, characterized in that: include: The sensor data within the monitoring area is used to assess the health status of the buried pipeline. If the buried pipeline is healthy, the monitoring area is divided into several corrosion analysis units by gridding. The unit feature vector of each corrosion analysis unit is calculated, and a corrosion rate model is constructed for the corrosion analysis unit based on the unit feature vector. The historical corrosion rate sequence of the corrosion analysis unit is obtained based on the corrosion rate model. A multi-layer LSTM network model is trained based on the historical corrosion rate sequence to predict the corrosion rate and obtain the predicted corrosion rate sequence. The predicted corrosion rate sequence is obtained as follows: Based on the corrosion rate model, the corrosion rate of the corrosion analysis unit at each time point during the historical monitoring period is calculated and integrated to obtain the historical corrosion rate sequence. The historical corrosion rate sequence is normalized to obtain the normalized corrosion rate sequence. The multilayer LSTM network model is trained using the normalized corrosion rate sequence. Starting from the current time, the trained multilayer LSTM network model is used to make rolling predictions of the corrosion rate of the corrosion analysis unit. The predicted values ​​are then reverse-normalized and integrated to obtain the predicted corrosion rate sequence of the corrosion analysis unit within the prediction period. By integrating historical corrosion rate sequences and predicted corrosion rate sequences, the predicted wall thickness of buried pipelines in each corrosion analysis unit at the end of the predicted time period is estimated and the dispersion is analyzed. If the dispersion is high, risk corrosion units are divided, and the predicted wall thickness at the end of the risk corrosion units or corrosion analysis units is analyzed to determine whether the potentiostat is triggered. If triggered, the output current of the potentiostat is adjusted. Obtain the initial wall thickness of buried pipelines By combining historical corrosion rate sequences and predicted corrosion rate sequences, the predicted wall thickness of buried pipelines at the end of the prediction period is calculated, and the calculation results are marked as the predicted wall thickness at the end of the prediction period. The formula is: ； in, This indicates the corrosion rate of the corrosion analysis unit at a specific time point within the historical monitoring period. Indicates a point in time within the historical monitoring period. This indicates the corrosion rate of the corrosion analysis unit at a given time point within the prediction period. Indicates a point in time within the predicted period; in, This indicates the duration of the interval between adjacent time points, and the duration of the interval between all adjacent time points is the same.

2. The method for online monitoring of cathodic protection corrosion rate of buried pipelines according to claim 1, characterized in that: The method for assessing the health status of the buried pipeline is as follows: A multi-dimensional collaborative monitoring sensor array is acquired and divided into several sensor groups according to spatial distribution. Each sensor group contains only one sensor of each type. The flag value of the sensor group is acquired, and the sensor group corresponding to the flag value that is greater than or equal to the preset judgment stability value is marked as an abnormal sensor group. If no abnormal sensor group is found in the multi-dimensional collaborative monitoring sensor array, the buried pipeline is judged to be in good condition.

3. The method for online monitoring of cathodic protection corrosion rate of buried pipelines according to claim 2, characterized in that: The flag value is calculated as follows: Acquire sensor data collected by the sensor group and normalize it to obtain normalized sensor data; A flag value is set for each sensor group and initialized to 0. The analytic hierarchy process (AHP) is used to assign weights to the normalized sensor data and perform weighted summation to obtain the comprehensive anomaly index of the sensor group. If the comprehensive anomaly index is greater than or equal to the anomaly index threshold, the flag value is incremented once; if it is less than the anomaly index threshold, the flag value is reset to 0.

4. The method for online monitoring of cathodic protection corrosion rate of buried pipelines according to claim 1, characterized in that: The corrosion rate model is constructed as follows: The monitoring area was divided into corrosion analysis units using a variable density grid method; The sensor data within the corrosion analysis unit are integrated into the original feature vector of the sensor group, and combined with the calculated Euclidean distance from the sensor group to the center of the corrosion analysis unit, the unit feature vector of the corrosion analysis unit is calculated using the distance-weighted average method. Based on the unit feature vector, the uniform corrosion rate of the corrosion analysis unit is calculated according to the core formula of Faraday's law as the basic model. The basic model is then modified by embedding the constructed soil corrosivity function into it, thus obtaining the corrosion rate model of the corrosion analysis unit.

5. The method for online monitoring of cathodic protection corrosion rate of buried pipelines according to claim 1, characterized in that: The method for determining whether potentiostat control is triggered is as follows: The predicted wall thickness of the endpoint of all corrosion analysis units is obtained and integrated into a set of predicted wall thicknesses. The average absolute deviation of the set of predicted wall thicknesses is calculated. If the average absolute deviation is greater than the deviation threshold, the corrosion analysis units are divided into risk corrosion units. The predicted wall thickness of the endpoint of all risk corrosion units is integrated into a set of risk predicted wall thicknesses. The risk predicted wall thicknesses are processed to obtain the risk predicted wall thickness. If the average absolute deviation is less than or equal to the deviation threshold, the predicted wall thicknesses are processed directly to obtain the average predicted wall thickness. If a risky corrosion unit exists and the predicted wall thickness meets the preset wall thickness condition, or the average predicted wall thickness meets the preset wall thickness condition, the potentiostat is triggered for adjustment.

6. The method for online monitoring of cathodic protection corrosion rate of buried pipelines according to claim 5, characterized in that: The method for obtaining the predicted wall thickness at the endpoint is as follows: The initial wall thickness of the buried pipeline is obtained. The data is processed by combining the historical corrosion rate sequence and the predicted corrosion rate sequence to calculate the predicted wall thickness of the buried pipeline at the end of the predicted time period. The calculation result is marked as the predicted wall thickness at the end of the time period.

7. The method for online monitoring of cathodic protection corrosion rate of buried pipelines according to claim 5, characterized in that: The risk corrosion unit is divided as follows: The predicted wall thicknesses of all corrosion analysis units are obtained and integrated into a predicted wall thickness set. The average absolute deviation of the predicted wall thickness set is calculated. If the average absolute deviation is greater than the deviation threshold, the predicted wall thickness set is normalized. The normalized predicted wall thickness set is used as a feature vector and input into the K-means algorithm to cluster into two clusters. The predicted wall thicknesses of the center of the two clusters are compared, and the corrosion analysis units in the cluster with the smaller predicted wall thickness at the center of the cluster are marked as risk corrosion units.

8. The method for online monitoring of cathodic protection corrosion rate of buried pipelines according to claim 1, characterized in that: The method for regulating the output current of the potentiostat is as follows: If there are risky corrosion units, the predicted corrosion rates of all risky corrosion units at the next time point from the current time point are processed to obtain the controlled predicted rate. If there are no risky corrosion units, the predicted corrosion rates of all corrosion analysis units at the next time point from the current time point are processed to obtain the controlled predicted rate. The target potential is dynamically adjusted by combining the control and prediction rate. The potential deviation is defined. Based on the PID algorithm, the potential deviation is used as the input and the output current of the potentiostat is used as the output. The output current of the potentiostat is calculated and the output of the potentiostat is dynamically adjusted according to the output current of the potentiostat.

9. An online monitoring system for cathodic protection corrosion rate of buried pipelines, the system being used to implement the monitoring method as described in any one of claims 1-8, characterized in that: Includes the following modules: Data acquisition module: Deploys a multi-dimensional collaborative monitoring sensor array to collect sensor data within the monitoring area; Rate analysis module: Acquire sensor data to assess the health status of buried pipelines. If the buried pipeline is in good condition, the monitoring area is divided into several corrosion analysis units by gridding. The unit feature vector of each corrosion analysis unit is calculated, and a corrosion rate model is constructed for the corrosion analysis unit based on the unit feature vector. Corrosion prediction module: Based on the corrosion rate model, the historical corrosion rate sequence of the corrosion analysis unit is obtained. A multi-layer LSTM network model is trained based on the historical corrosion rate sequence to predict the corrosion rate and obtain the predicted corrosion rate sequence. Current control module: Integrates historical corrosion rate sequence and predicted corrosion rate sequence, estimates and analyzes the dispersion of the predicted wall thickness of buried pipelines in each corrosion analysis unit at the end of the predicted time period. If the dispersion is high, risk corrosion units are divided, and the predicted wall thickness at the end of the risk corrosion unit or corrosion analysis unit is analyzed to determine whether the potentiostat control is triggered. If triggered, the output current of the potentiostat is controlled.

Images