Gas supply pipeline fault data acquisition system based on SCADA system

Abstract

This invention relates to the field of industrial automation and control technology, specifically to a gas pipeline fault data acquisition system based on a SCADA system. The system includes a fault correlation matrix construction module, a real-time anomaly data judgment module, a single anomaly data processing module, a multi-type anomaly data analysis module, a comprehensive analysis module, and a management database. The system cleans historical data, constructs a matrix, and verifies its accuracy through the fault correlation matrix construction module. The real-time anomaly data judgment module judges anomalies according to set rules. The single and multi-type anomaly data analysis modules process single and multi-type anomaly data respectively. The comprehensive analysis module uses a multivariate parameter correlation model for early warning and integrates a matrix to push high-confidence fault types. It can analyze the inherent relationship between historical faults and monitoring data, effectively improving the accuracy of fault diagnosis, quickly locating fault points, increasing fault handling efficiency, and preventing fault occurrence and reducing the scope of fault impact.

Description

Technical Field

[0001] This invention relates to the field of industrial automation and control technology, and more specifically, to a gas supply pipeline fault data acquisition system based on SCADA system. Background Technology

[0002] With the acceleration of urbanization, natural gas, as a clean and efficient energy source, is being used more and more widely in residential life and industrial production. As the "arteries" for natural gas transportation, gas supply pipelines are constantly expanding their coverage and becoming increasingly complex in scale.

[0003] However, natural gas is a flammable and explosive gas. Once a gas pipeline malfunctions, such as a pipeline leak or equipment damage, it will not only lead to the interruption of gas supply and affect the normal operation of society, but may also cause serious safety accidents such as fires and explosions, posing a huge threat to people's lives, property and the ecological environment. At the same time, gas supply pipelines are in a complex geographical environment and operating conditions for a long time, and are affected by a variety of factors such as soil corrosion, external damage, temperature changes and equipment aging. The probability of malfunctions cannot be ignored. Therefore, timely and accurate acquisition of fault data of gas supply pipelines is crucial to ensuring the safe operation of pipelines.

[0004] However, existing technologies still have shortcomings. For example, Chinese patent application number 202110010962.5 discloses a method, device, and computer storage medium for determining faults in gathering and transportation pipelines. This solution determines the operating status of each node in the gathering and transportation pipeline based on the first state parameter, the second state parameter, and the operating status rules of each node. When there is a faulty node in the gathering and transportation pipeline that is in a faulty state, the fault information of the gathering and transportation pipeline is determined based on the node information of the faulty node and the network model of the gathering and transportation pipeline. This can identify the fault in the gathering and transportation pipeline, reduce the fault investigation time, and improve the efficiency of fault handling.

[0005] However, this scheme still has the following shortcomings: First, the scheme mainly relies on simple comparison of node status parameters, such as comparing the difference of calculated parameters with the threshold, and referring to the node pressure comparison table to judge the fault. It is difficult to deeply explore the complex internal relationship between the fault and various types of data. It may lack the accuracy in grasping the internal relationship between the fault and the data, which greatly reduces the accuracy of fault diagnosis.

[0006] Second, this solution can only determine information such as the fault node after the fault occurs, and cannot prevent the fault in advance. Once the fault occurs, it is also difficult to quickly judge the severity of the fault and the possible scope of its impact, which is not conducive to reducing the losses and impacts caused by the fault. Summary of the Invention

[0007] To overcome the shortcomings in the prior art, embodiments of the present invention provide a gas supply pipeline fault data acquisition system based on a SCADA system, which can effectively solve the problems involved in the prior art.

[0008] The objective of this invention can be achieved through the following technical solution: This invention provides a gas supply pipeline fault data acquisition system based on SCADA system, including: a fault correlation matrix construction module, used to construct a fault correlation matrix and verify the accuracy based on historical fault types and monitoring data of each type.

[0009] The real-time anomaly data determination module is used to set fault judgment rules, determine anomaly data, and record it.

[0010] The single anomaly data processing module is used to index the correlation information in the fault correlation matrix according to the fault type, generate a priority list of correlation data and filter the high-priority correlation data types, match the data at the corresponding monitoring points, and thus verify the changing trend of the correlation data.

[0011] The multi-type anomaly data analysis module is used to construct multi-dimensional correlation vectors, generate a fault type priority list by calculating the mean of correlation strength, and then match monitoring points for cluster analysis to delineate high-probability fault areas.

[0012] The comprehensive analysis module is used to calculate the comprehensive failure probability through a multivariate parameter correlation model, trigger early warnings, present the failure distribution by combining a heat map, and calculate the comprehensive priority by integrating the failure correlation matrix, thereby pushing out failure types with high support and high confidence.

[0013] The management database is used to store system operation-related data such as historical fault information, various monitoring data, fault correlation matrices, and abnormal data records.

[0014] Preferably, the specific operation steps of the fault correlation matrix construction module include: S1. Extracting historical fault information under different time periods and different operating conditions from the long-term operation records of the gas supply pipeline. The fault information includes the specific time of the fault occurrence, the fault type, and various types of data collected by each monitoring point before and after the fault occurrence. The data of the historical fault information is cleaned, and the continuous data is converted into a standard normal distribution with a mean of 0 and a standard deviation of 1.

[0015] S2. Classify and statistically analyze the historical fault data to identify various fault types. Use each fault type as the row of a matrix and each type of monitoring data as the column of the matrix to construct an initial fault correlation matrix framework. At this time, the values ​​of the matrix cells are all empty.

[0016] S3. For each fault type and each type of monitoring data combination, calculate the correlation strength between the two, and fill the correlation strength into the matrix cell to complete the fault correlation matrix construction. Then, select a validation set from historical fault data and use the fault correlation matrix to verify the prediction accuracy of the fault type.

[0017] Preferably, the specific analysis method of step S3 is as follows: convert the fault type into a binary variable, mark the occurrence of the fault as 1 and the absence of the fault as 0, sort the monitoring data and assign a level, calculate the level difference between the fault binary variable and the monitoring data, traverse the level sequence of the fault binary variable and the monitoring data, calculate the correlation strength between each fault type and the monitoring data type according to the Spearman rank correlation coefficient formula, and fill the calculated correlation strength value into the cell corresponding to the fault correlation matrix.

[0018] A subset of data is randomly selected from historical fault data as a validation set. The fault cases in the validation set are judged using the constructed fault association matrix. By finding the fault type and association strength corresponding to the data in the validation set in the matrix, the possible fault types are predicted. The association strength values ​​in the matrix cells are adjusted according to the prediction accuracy of the fault types until the prediction accuracy of the fault types reaches the standard value.

[0019] Preferably, the specific analysis method of the real-time abnormal data judgment module is as follows: for various types of data collected in real time by the corresponding monitoring equipment at each monitoring point of the gas supply pipeline, the upper and lower limits of the normal operating value range corresponding to each type of data are preset, the upper limit of the change rate of each type of data is set, and the time period for judgment is also set.

[0020] Once the real-time value of a single or multiple types of monitoring data is acquired, it is compared with the pre-set upper and lower limits of the normal operating range to determine whether the real-time value is less than the lower limit or greater than the upper limit. If so, the rate of change of the single or multiple types of monitoring data within the time period is further calculated.

[0021] If the rate of change of one or more types of monitoring data is greater than the upper limit of its corresponding rate of change, it is determined that the data has initially experienced a data anomaly fault, and the type of abnormal data, monitoring point location, real-time value, timestamp of the anomaly occurrence, and anomaly level are recorded.

[0022] Preferably, the specific analysis method for the associated data priority list is as follows: using the type of abnormal data as an index, locate the corresponding column in the fault association matrix, and extract the association strength value of all fault types under that column.

[0023] Set a correlation strength threshold, iterate through the extracted correlation strength values, and filter out fault types and other corresponding correlation data types whose correlation strength with abnormal data exceeds the threshold.

[0024] Sort the filtered correlation strength values ​​from highest to lowest absolute value, and then sort the corresponding fault types and correlation data types to generate a priority list of correlation data.

[0025] Preferably, the specific operation method of the single abnormal data processing module is as follows: filter out high-priority associated data types according to the associated data priority list, and match the monitoring point location, device model and communication address corresponding to the high-priority associated data types within the set range according to the filtered high-priority associated data types.

[0026] Using the system's communication protocol, the monitoring points and devices with the highest correlation strength in the list of associated data are determined according to the priority list of associated data, and the current values ​​of the high-priority associated data and the trend curves for the nearby time period are obtained from them.

[0027] The high-priority correlated data and the original abnormal data are organized to form a dataset containing abnormal data points, correlated data points, monitoring point locations, and timestamps.

[0028] The high-priority associated data is validated to exclude invalid data caused by sensor failure or communication interruption. The trend of change is verified to be consistent with expectations by combining the association direction recorded in the fault association matrix.

[0029] Preferably, the correlation direction between each data type and the fault type recorded in the fault correlation matrix includes positive correlation and negative correlation. The positive correlation indicates that the data anomaly and the fault occurrence show the same trend, and the negative correlation indicates that the data anomaly and the fault occurrence show the opposite trend.

[0030] Preferably, the specific analysis method for the fault type priority list is as follows: locate the corresponding column of each abnormal data type in the fault association matrix, generate a multi-column association strength set, and construct a multi-dimensional association vector between abnormal data types and fault types.

[0031] Calculate the average correlation strength between fault types and each abnormal data type. Based on the set global correlation strength threshold, filter fault types whose average correlation strength is greater than the global correlation strength threshold, sort them from high to low according to the average correlation strength, and generate a fault type priority list.

[0032] Preferably, the specific analysis method of the multi-type abnormal data analysis module is as follows: match the associated data types corresponding to each fault type in the fault type priority list, and then determine the monitoring points corresponding to these associated data types, which are recorded as each abnormal data monitoring point.

[0033] Using a geographic information system, cluster analysis is performed on the spatial distribution of each abnormal data monitoring point. The midpoint of the line connecting the abnormal data monitoring points is taken as the potential leakage point, and a high-probability fault area is delineated.

[0034] Preferably, the specific operation method of the comprehensive analysis module is as follows: start the multivariate parameter correlation model, analyze the combination and change characteristics of various parameters, calculate the comprehensive fault probability, and trigger a fault warning when the comprehensive fault probability is greater than the set fault probability threshold.

[0035] The spatial distribution of each abnormal data monitoring point and its associated monitoring points is displayed by overlaying them using a geographic information system, and the probability distribution of the fault is displayed by a heat map. The associated monitoring points refer to other monitoring points that have data association, spatial association, fault logic association, or other relationships with each abnormal data monitoring point.

[0036] By integrating a multivariate parameter association model and a fault association matrix, when more than one abnormal data is triggered simultaneously, the system automatically matches the parameter association rules of the corresponding fault type in the fault association matrix. For each matched fault type, the system calculates the comprehensive priority based on the average association strength combined with the parameter weights, and pushes fault types with support ≥80% and confidence ≥90% to the system.

[0037] Compared with the prior art, the present invention has the following beneficial effects: First, the present invention constructs a fault correlation matrix to show the correlation between fault types and monitoring data, and uses multidimensional correlation vectors and the mean of correlation strength to generate a fault type priority list, which can more comprehensively and accurately grasp the intrinsic relationship between faults and data, and effectively improve the accuracy of fault diagnosis.

[0038] Second, this invention monitors data in real time and identifies anomalies. It also conducts in-depth analysis of anomalous data, including single and multiple types of anomalous data, to clarify the direction of troubleshooting and delineate the fault area. This helps to quickly locate the fault point and improve the efficiency of fault handling.

[0039] Third, this invention provides early warning by combining a multivariate parameter correlation model to calculate the comprehensive failure probability, and integrates a failure correlation matrix to calculate the comprehensive priority and push the failure type, which can prevent failures from occurring in advance and reduce the scope of failure impact. Attached Figure Description

[0040] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is a system module connection diagram of the present invention.

[0042] Figure 2 A flowchart for constructing the fault correlation matrix.

[0043] Figure 3 This is a flowchart for a single abnormal data processing module. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] Please see Figure 1 As shown, the gas supply pipeline fault data acquisition system based on SCADA system includes a fault correlation matrix construction module, a real-time abnormal data judgment module, a single abnormal data processing module, a multi-type abnormal data analysis module, a comprehensive analysis module, and a management database.

[0046] The management database and fault correlation matrix construction module, real-time abnormal data judgment module, single abnormal data processing module, multi-type abnormal data analysis module, and comprehensive analysis module are connected. The comprehensive analysis module is connected to the single abnormal data processing module and the multi-type abnormal data analysis module. The real-time abnormal data judgment module is connected to the single abnormal data processing module and the multi-type abnormal data analysis module. The fault correlation matrix construction module is connected to the single abnormal data processing module and the multi-type abnormal data analysis module.

[0047] The fault correlation matrix construction module is used to construct a fault correlation matrix and verify its accuracy based on historical fault types and monitoring data of each type.

[0048] Please see Figure 2 As shown, the specific operation steps of the fault correlation matrix construction module include: S1. Extracting historical fault information under different time periods and operating conditions from the long-term operation records of the gas supply pipeline. The fault information includes the specific time of the fault occurrence, the fault type, and various data collected by each monitoring point before and after the fault occurrence. The historical fault information data is cleaned, and the continuous data is converted into a standard normal distribution with a mean of 0 and a standard deviation of 1. This helps to unify the dimensions and scale of the data. Different monitoring data may have different units and value ranges. Standardization can eliminate the influence of dimensions, make different types of data comparable, and facilitate subsequent data analysis and model construction.

[0049] It should be noted that the data cleaning mainly deals with missing values, outliers, and duplicate values ​​in the data. If the proportion of missing values ​​in the dataset is small and the distribution of missing values ​​is relatively random, records containing missing values ​​can be directly deleted. If there are many missing values, deletion may lead to a significant reduction in the amount of data. In this case, mean filling, median filling, or mode filling can be used.

[0050] The Z-score method is used to identify outliers. The Z-score of each data point is calculated by dividing the difference between the data point and the mean by the standard deviation. Data points with an absolute Z-score greater than 3 are considered outliers. For the identified outliers, it is determined whether to delete them or correct them based on their proportion in the dataset.

[0051] Check if there are any completely identical records in the dataset (i.e., all fields have the same value). If duplicate records are found, they need to be deleted to ensure the uniqueness of the data.

[0052] It should be noted that the specific analytical method for converting continuous data into a standard normal distribution with a mean of 0 and a standard deviation of 1 is as follows: For each continuous data field, calculate the mean and standard deviation of all data in that field. Then, for each data point in that field, subtract the mean from the value of that data point and divide by the standard deviation to obtain the standardized data value.

[0053] S2. Classify and statistically analyze the historical fault data to identify various fault types. Use each fault type as the row of a matrix and each type of monitoring data as the column of the matrix to construct an initial fault correlation matrix framework. At this point, all matrix cell values ​​are empty. With fault types as rows and monitoring data as columns, a clear matrix structure is formed, making the relationship between different types of faults and their related monitoring data immediately apparent. The matrix structure makes the calculation and filling of correlation strength more orderly and convenient, facilitating the mining of potential correlations between data through matrix operations and analysis, and providing strong support for the construction of fault prediction models.

[0054] S3. For each fault type and each type of monitoring data combination, calculate the correlation strength between the two, and fill the correlation strength into the matrix cell to complete the fault correlation matrix construction. Select a validation set from historical fault data and use the fault correlation matrix to verify the prediction accuracy of the fault type. By representing the correlation strength with specific values, the influence of different monitoring data on different fault types can be evaluated more accurately, helping to determine which monitoring data are most critical for fault diagnosis and prediction.

[0055] The specific analysis method of step S3 is as follows: the fault type is converted into a binary variable, with fault occurrence marked as 1 and non-occurrence marked as 0; the monitoring data is sorted and assigned a level; the level difference between the fault binary variable and the monitoring data is calculated; the level sequence of the fault binary variable and the monitoring data is traversed; the correlation strength between each fault type and the monitoring data type is calculated according to the Spearman rank correlation coefficient formula; and the calculated correlation strength value is filled into the corresponding cell of the fault correlation matrix. Converting the fault type into a binary variable, with 1 representing fault occurrence and 0 representing non-occurrence, allows the fault information to be presented in a concise and clear manner, which can more intuitively reflect the relationship between the fault state and other data. The Spearman rank correlation coefficient has relatively low requirements for the distribution of data and is suitable for this kind of graded data. It can effectively capture the nonlinear relationship between data, thereby more comprehensively reflecting the intrinsic connection between faults and monitoring data.

[0056] It should be noted that the formula for the Spearman rank correlation coefficient is as follows: Where i is the number of the i-th observation record, i = 1, 2, ..., n, and n is the number of observation records, d i It is the rank difference of the i-th observation record. The Spearman rank correlation coefficient ranges from -1 to 1. When r s When r is close to 1, it indicates a positive correlation between the fault type and the monitoring data; that is, the larger the value of the monitoring data, the higher the probability of the fault occurring. s When the value is close to -1, it indicates a negative correlation, meaning the larger the value of the monitored data, the lower the probability of a fault occurring; when r s When the value is close to 0, it indicates that there is almost no correlation between the two.

[0057] A subset of historical fault data is randomly selected as the validation set. The fault cases in the validation set are judged using the constructed fault association matrix. By finding the fault type and association strength corresponding to the data in the validation set in the matrix, the possible fault types are predicted. The association strength values ​​in the matrix cells are adjusted according to the prediction accuracy of the fault types until the prediction accuracy of the fault types reaches the standard value. By adjusting the association strength values, the model can better fit the data and gradually improve the model's predictive ability, enabling it to more accurately predict the fault types of gas supply pipelines.

[0058] It should be noted that the specific analysis method for the prediction accuracy of the fault type is as follows: the number of samples that the model correctly predicts as a certain fault type and the actual situation is also that fault type is denoted as true positive (TP); the number of samples that the model predicts as a certain fault type but the actual situation is not that fault type is denoted as false positive (FP); the number of samples that the model correctly predicts as a fault type other than that fault type and the actual situation is indeed not that fault type is denoted as true negative (TN); and the number of samples that the model predicts as a fault type other than that fault type but the actual situation is that fault type is denoted as false negative (FN). A confusion matrix is ​​constructed using a binary classification problem (fault presence and fault absence). See Table 1 for details. Table 1 lists representative confusion matrices.

[0059] Table 1. Confusion Matrix

[0060] Actual fault Actual non-fault Predicting Faults TP FP Predicting non-faults FN TN

[0061] For a single fault type, the formula for calculating its prediction accuracy is as follows: For multiple fault types, the true positives, false positives, true negatives, and false negatives of all fault types are summarized, and then the accuracy is calculated according to the formula for the single fault type mentioned above.

[0062] The real-time anomaly data determination module is used to set fault judgment rules, determine anomaly data, and record it.

[0063] The specific analysis method of the real-time abnormal data judgment module is as follows: For various types of data collected in real time by the corresponding monitoring equipment at each monitoring point of the gas supply pipeline, the upper and lower limits of the normal operating value range corresponding to each type of data are preset, and the upper limit of the change rate of each type of data and the time period used for judgment are also set; this helps to clearly define the normal operating range of the data collected by the monitoring equipment at each monitoring point of the gas supply pipeline.

[0064] Once the real-time value of a single or multiple types of monitoring data is acquired, it is compared with the pre-set upper and lower limits of the normal operating range to determine whether the real-time value is less than the lower limit or greater than the upper limit. If so, the rate of change of the single or multiple types of monitoring data within the time period is further calculated. Focusing on the rate of change can provide early warning of some abnormal situations that may develop into serious faults, making it easier to take timely measures for investigation and handling, and avoid further deterioration of the fault.

[0065] It should be noted that the specific analysis method for the rate of change of the single or multiple types of monitoring data within the calculation time period is as follows: Within a defined time period, find the starting data point, whose data value is denoted as y1 and whose corresponding timestamp is t1; and the data value of the ending data point is denoted as y2 and whose corresponding timestamp is t2. Then, apply the formula... To calculate the rate of change, t2-t1 represents the duration of the time period, and the unit must be consistent with the data time unit.

[0066] If the rate of change of one or more types of monitoring data is greater than the upper limit of its corresponding rate of change, it is determined that the data has initially experienced a data anomaly fault. The type of abnormal data, the location of the monitoring point, the real-time value, the timestamp of the anomaly occurrence, and the anomaly level are recorded. Detailed anomaly information recording facilitates fault tracking and statistics, and helps to improve the reliability and stability of the entire system.

[0067] The single anomaly data processing module is used to index the correlation information in the fault correlation matrix according to the fault type, generate a priority list of correlation data and filter the high-priority correlation data types, match the data at the corresponding monitoring points, and thus verify the changing trend of the correlation data.

[0068] It should be noted that the specific analytical method for verifying the changing trend of the associated data is as follows: calculate the mean and standard deviation of the associated data in different time periods. If the mean changes significantly in a certain time period, or the standard deviation increases, it indicates that the changing trend of the data has changed.

[0069] The least squares method is used to fit a straight line, and the slope of the line is calculated. A positive slope indicates that the data is on an upward trend, a negative slope indicates that the data is on a downward trend, and a slope of zero indicates that the data has no obvious trend.

[0070] Please see Figure 3 As shown, the specific analysis method of the associated data priority list is as follows: using the type of abnormal data as an index, locate the corresponding column in the fault association matrix, and extract the association strength value of all fault types under that column; it can quickly establish a connection between the abnormal data monitored in real time and the historical fault data, and use the fault association information accumulated in the historical data to provide a strong basis for the analysis of the current abnormal situation.

[0071] Set a correlation strength threshold, iterate through the extracted correlation strength values, and filter out the fault types and other related data types whose correlation strength with the abnormal data exceeds the threshold. Extract the correlation strength values ​​of all fault types under this column so that the analysis process can take into account all possible fault types and their correlation degree related to the abnormal data type, and avoid missing important information.

[0072] Sort the filtered correlation strength values ​​from highest to lowest, and then sort the corresponding fault types and related data types to generate a priority list of related data. This can highlight key information that is strongly correlated with the current abnormal data, focusing on the fault types and data types that are most likely to be related to the current abnormal situation, reducing unnecessary interference information, and improving the efficiency and accuracy of fault diagnosis.

[0073] Please refer to Table 2 for details. Table 2 lists the priority list of related data, which consists of representative data.

[0074] Table 2. List of Priorities for Related Data

[0075] Fault type Associated fault types Correlation strength Pipeline leak flow 0.8 Valve malfunction temperature 0.6

[0076] The specific operation method of the single abnormal data processing module is as follows: based on the priority list of associated data, high priority associated data types are filtered out, and according to the filtered high priority associated data types, the monitoring point location, equipment model and communication address corresponding to the high priority associated data types within the set range are matched; it can accurately locate the specific location and equipment information of the relevant data, and improve the targeting and efficiency of data collection.

[0077] Using the system's communication protocol, the monitoring points and devices with the highest correlation strength in the list of associated data are determined according to the priority list. The current value of the high-priority associated data and the trend curve for the nearby time period are obtained from them. The trend curve can be used to observe the data change pattern and discover potential abnormal patterns or trends, which helps to more accurately judge the development trend of the fault.

[0078] The high-priority correlated data and the original abnormal data are organized to form a dataset containing abnormal data points, correlated data points, monitoring point locations, and timestamps. The dataset provides a comprehensive and systematic information foundation for subsequent analysis and processing, making it easier to find the inherent connections and patterns between data, and helping to understand the mechanism and process of fault occurrence more deeply.

[0079] Please refer to Table 3 for details. Table 3 lists a dataset consisting of representative data.

[0080] Table 3. Dataset

[0081]

[0082] The high-priority associated data is validated to exclude invalid data caused by sensor failure or communication interruption. The trend of change is verified to match expectations by combining the association direction recorded in the fault association matrix. Further confirmation of the correlation between data and fault helps to discover potential problems and anomalies, and improves the comprehensiveness and accuracy of fault diagnosis.

[0083] It should be noted that the specific analytical method for verifying whether the trend of change meets expectations is as follows: Select a moving average period, organize the high-priority correlated data into time series data, calculate the moving average of each period according to the selected moving average period, plot the moving average of each period on a chart in chronological order to form a moving average curve, determine the expected trend of change of high-priority correlated data based on the correlation direction recorded in the fault correlation matrix and the current fault situation, if there is a positive correlation and the fault is aggravating, the expected moving average should show an upward trend, if there is a negative correlation and the fault is easing, the expected moving average should show a downward trend, quantify the trend of change by calculating the slope of the moving average curve, compare the positive or negative slope with the expected trend of change, and determine whether it meets expectations.

[0084] The fault correlation matrix records the correlation direction between each data type and the fault type, including positive correlation and negative correlation. Positive correlation indicates that data anomalies and fault occurrences show the same trend, while negative correlation indicates that data anomalies and fault occurrences show the opposite trend. Clarifying the correlation direction helps to more accurately judge the fault. When data anomalies are detected, the possible direction of the fault can be accurately known based on the characteristics of positive or negative correlation.

[0085] The multi-type anomaly data analysis module is used to construct multi-dimensional correlation vectors, generate a fault type priority list by calculating the mean of correlation strength, and then match monitoring points for cluster analysis to delineate high-probability fault areas.

[0086] The specific analysis method for the fault type priority list is as follows: locate the corresponding column of each abnormal data type in the fault association matrix, generate a set of multiple columns of association strength, and construct a multi-dimensional association vector between abnormal data types and fault types; the multi-dimensional association vector comprehensively considers the association between various abnormal data and faults, captures information from multiple dimensions, and makes fault judgment more accurate and comprehensive.

[0087] It should be noted that the dimension of the multidimensional correlation vector is equal to the number of fault types. For each abnormal data type, the values ​​in its correlation strength set are filled into the corresponding positions of the vector in the order of fault types. The correlation vectors corresponding to each abnormal data type are combined together to form a multidimensional correlation vector matrix.

[0088] For example, if the detected abnormal data types are "pressure anomaly" and "flow anomaly", then the corresponding columns are located and the association strength sets are extracted: the association strength set corresponding to "pressure anomaly" is [0.8, 0.3, 0.6], and the association strength set corresponding to "flow anomaly" is [0.2, 0.7, 0.4]. The constructed multidimensional association vector matrix is ​​as follows:

[0089] The system calculates the average correlation strength between fault types and various abnormal data types. Based on a set global correlation strength threshold, it filters fault types whose average correlation strength is greater than the global correlation strength threshold and sorts them from high to low according to the average correlation strength, generating a fault type priority list. This allows the system to focus on key fault types that are closely related to abnormal data, avoiding blindly checking among many possible fault types and concentrating efforts on the most likely faults, thus improving fault troubleshooting efficiency.

[0090] The specific analysis method of the multi-type abnormal data analysis module is as follows: match the associated data types corresponding to each fault type in the fault type priority list, and then determine the monitoring points corresponding to these associated data types, which are recorded as each abnormal data monitoring point; this can accurately locate monitoring points related to potential faults, provide clear targets for subsequent data collection and analysis, and avoid ineffective monitoring of irrelevant monitoring points.

[0091] By using geographic information system cluster analysis to analyze the spatial distribution of each abnormal data monitoring point, and taking the midpoint of the line connecting the abnormal data monitoring points as the potential leakage point, a high-probability fault area is delineated. This can quickly pinpoint the approximate range of possible faults, focusing fault investigation from scattered monitoring points to a specific area, greatly reducing the workload and time required for investigation.

[0092] It should be noted that the specific analysis method for delineating high-probability fault areas is as follows: collect the geographic coordinate information of all abnormal data monitoring points, including longitude and latitude, as well as related abnormal data values, import the data into the geographic information system software, and create a layer containing the location and attribute information of the monitoring points.

[0093] A clustering analysis algorithm is run to divide the abnormal data monitoring points into different clusters. For each cluster, the average coordinates of all monitoring points in the cluster are calculated to obtain the coordinates of the midpoint. The midpoint is regarded as a candidate location of potential leakage points and marked on the map. A circle is drawn with a set radius centered on each potential leakage point, and the area inside the circle is defined as a high-probability fault area.

[0094] The comprehensive analysis module is used to calculate the comprehensive failure probability through a multivariate parameter correlation model, trigger early warnings, present the failure distribution by combining a heat map, and calculate the comprehensive priority by integrating the failure correlation matrix, thereby pushing out failure types with high support and high confidence.

[0095] The specific operation method of the comprehensive analysis module is as follows: start the multivariate parameter correlation model, analyze the combination and change characteristics of various parameters, calculate the comprehensive failure probability, and trigger a fault warning when the comprehensive failure probability is greater than the set failure probability threshold; this helps to intervene in the early stage of the failure, prevent the further development and expansion of the failure, and reduce the impact of the failure on the system and the possible losses.

[0096] It should be noted that the specific analysis method for the comprehensive failure probability is as follows: the multivariate parameter correlation model is trained using historical failure data, and the historical failure data is divided into training set and test set. During the training process, the loss function of the model is minimized or the evaluation index of the model is maximized to train the model well. The trained model is evaluated using the test set, and various performance indicators of the model are calculated. The model is optimized by adjusting the parameters until the model reaches satisfactory performance.

[0097] During system operation, the current values ​​of various parameters are acquired in real time and standardized according to the data preprocessing method. The standardized real-time parameter data is then input into the trained multivariate parameter association model. The model analyzes the combined change characteristics of the parameters and outputs the comprehensive fault probability. The comprehensive fault probability usually ranges from 0 to 1, where 0 indicates that the system has no fault and 1 indicates that the system is definitely in a fault state.

[0098] The spatial distribution of each abnormal data monitoring point and its associated monitoring points is displayed by overlaying a geographic information system, and the probability distribution of the fault is displayed by a heat map. The associated monitoring points refer to other monitoring points that have data association, spatial association, fault logic association and other relationships with each abnormal data monitoring point. This makes it easy to quickly understand the operating status of the entire gas supply pipeline system and locate key areas where faults may occur.

[0099] By integrating a multivariate parameter association model with a fault association matrix, when more than one abnormal data point is triggered simultaneously, the system automatically matches the parameter association rules of the corresponding fault type in the fault association matrix. For each matched fault type, a comprehensive priority is calculated based on the average association strength combined with parameter weights. Fault types with a support of ≥80% and a confidence of ≥90% are then pushed to the system. This ensures that the pushed fault types have high credibility and reliability, avoids wasting time and effort on low-probability fault types, further improves the efficiency and accuracy of fault handling, and helps to quickly restore the normal operation of gas supply pipelines.

[0100] It should be noted that the support refers to the frequency of association between a certain fault type and multiple concurrently occurring abnormal data. This is achieved by counting the number N instances in historical fault information where a specific combination of multiple abnormal data occurs simultaneously and corresponds to a certain fault type. 共同出现 The total number N instances of this specific combination of anomalous data in historical data. 异常组合出现 ,pass The support level for this fault type is calculated.

[0101] The confidence level represents the reliability of inferring the occurrence of a certain fault type when a specific combination of multiple abnormal data occurs, while also counting the total number N instances of that fault type occurring in historical data.故障总出现 ,pass The confidence level for this fault type is calculated.

[0102] The management database is used to store system operation-related data such as historical fault information, various monitoring data, fault correlation matrices, and abnormal data records.

[0103] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention, which are still covered within the protection scope of the present invention.

Claims

1. A gas supply pipeline fault data acquisition system based on SCADA system, characterized in that, include: The fault correlation matrix construction module is used to construct a fault correlation matrix and verify its accuracy based on historical fault types and monitoring data of each type. The real-time anomaly data determination module is used to set fault judgment rules, judge anomaly data, and record it. The single anomaly data processing module is used to index the correlation information in the fault correlation matrix according to the fault type, generate a priority list of correlation data and filter the high priority correlation data types, match the data at the corresponding monitoring point, and thus verify the trend of correlation data changes. The multi-type anomaly data analysis module is used to construct multi-dimensional correlation vectors, generate a fault type priority list by calculating the mean correlation strength, and then match monitoring points for cluster analysis to delineate high-probability fault areas. The comprehensive analysis module is used to calculate the comprehensive failure probability through a multivariate parameter correlation model, trigger early warnings, present the failure distribution by combining a heat map, and calculate the comprehensive priority by integrating the failure correlation matrix, thereby pushing high support and high confidence failure types. The management database is used to store system operation-related data such as historical fault information, various monitoring data, fault correlation matrices, and abnormal data records. The specific analysis method for the priority list of associated data is as follows: using the type of abnormal data as an index, locate the corresponding column in the fault association matrix, and extract the association strength value of all fault types under that column; Set a correlation strength threshold, iterate through the extracted correlation strength values, and filter out the fault types and other corresponding correlation data types whose correlation strength with abnormal data exceeds the threshold; sort the filtered correlation strength values ​​and their corresponding fault types and correlation data types from high to low according to the absolute value of the correlation strength, and generate a priority list of correlation data. The specific analysis method for the fault type priority list is as follows: locate the corresponding column of each abnormal data type in the fault association matrix, generate a multi-column association strength set, and construct a multi-dimensional association vector between abnormal data types and fault types; calculate the average association strength between fault types and each abnormal data type, filter each fault type whose average association strength is greater than the global association strength threshold according to the set global association strength threshold, sort them from high to low according to the average association strength, and generate a fault type priority list.

2. The gas supply pipeline fault data acquisition system based on SCADA system according to claim 1, characterized in that, The specific operation steps of the fault correlation matrix construction module include: S1. Extracting historical fault information under different time periods and operating conditions from the long-term operation records of the gas supply pipeline. The fault information includes the specific time of the fault occurrence, the fault type, and various data collected by each monitoring point before and after the fault occurrence. The historical fault information data is cleaned, and the continuous data is converted into a standard normal distribution with a mean of 0 and a standard deviation of 1. S2. Classifying and statistically analyzing the historical fault information data to identify various fault types. Using each fault type as the row of a matrix and each type of monitoring data as the column of the matrix, an initial fault correlation matrix framework is constructed. At this time, the matrix cell values ​​are all empty. S3. For each combination of fault type and each type of monitoring data, the correlation strength between them is calculated, and the correlation strength is filled into the matrix cell to complete the construction of the fault correlation matrix. A validation set is selected from the historical fault data, and the fault correlation matrix is ​​used to verify the prediction accuracy of the fault type.

3. The gas supply pipeline fault data acquisition system based on SCADA system according to claim 2, characterized in that, The specific analysis method for step S3 is as follows: The fault type is converted into a binary variable, with fault occurrence marked as 1 and non-fault occurrence marked as 0. The monitoring data is sorted and assigned a level. The level difference between the fault binary variable and the monitoring data is calculated. The fault binary variable and the monitoring data level sequence are traversed. The correlation strength between each fault type and the monitoring data type is calculated according to the Spearman rank correlation coefficient formula. The calculated correlation strength value is filled into the corresponding cell of the fault correlation matrix. A subset of data is randomly selected from historical fault data as a validation set. The fault cases in the validation set are judged using the constructed fault association matrix. By finding the fault type and association strength corresponding to the data in the validation set in the matrix, the possible fault types are predicted. The association strength values ​​in the matrix cells are adjusted according to the prediction accuracy of the fault types until the prediction accuracy of the fault types reaches the standard value.

4. The gas supply pipeline fault data acquisition system based on SCADA system according to claim 1, characterized in that, The specific analysis method of the real-time anomaly data determination module is as follows: For the various data collected in real time by the corresponding monitoring equipment at each monitoring point of the gas supply pipeline, the upper and lower limits of the normal operating value range of each type of data are preset, as well as the upper limit of the rate of change of each type of data and the time period for judgment are set. Once the real-time value of a single or multiple types of monitoring data is acquired, it is compared with the pre-set upper and lower limits of the normal operating value range to determine whether the real-time value is less than the lower limit or greater than the upper limit. If so, the rate of change of the single or multiple types of monitoring data within the time period is further calculated. If the rate of change of one or more types of monitoring data is greater than the upper limit of its corresponding rate of change, it is determined that the data has initially experienced a data anomaly fault, and the type of abnormal data, monitoring point location, real-time value, timestamp of the anomaly occurrence, and anomaly level are recorded.

5. The gas supply pipeline fault data acquisition system based on SCADA system according to claim 4, characterized in that, The specific operation method of the single abnormal data processing module is as follows: Based on the priority list of associated data, high-priority associated data types are filtered out, and based on the filtered high-priority associated data types, the monitoring point location, device model, and communication address corresponding to the high-priority associated data types within the set range are matched. Using the system's communication protocol, the monitoring points and devices with the highest correlation strength in the list of associated data are determined according to the priority list of associated data, and the current values ​​of the high-priority associated data and the trend curves in the nearby time period are obtained from them; The high-priority correlated data and the original abnormal data are organized to form a dataset containing abnormal data points, correlated data points, monitoring point locations, and timestamps; The high-priority associated data is validated to exclude invalid data caused by sensor failure or communication interruption. The trend of change is verified to be consistent with expectations by combining the association direction recorded in the fault association matrix.

6. The gas supply pipeline fault data acquisition system based on SCADA system according to claim 5, characterized in that, The correlation direction between each data type and the fault type recorded in the fault correlation matrix includes positive correlation and negative correlation. Positive correlation indicates that data anomalies and fault occurrences show the same trend, while negative correlation indicates that data anomalies and fault occurrences show the opposite trend.

7. The gas supply pipeline fault data acquisition system based on SCADA system according to claim 6, characterized in that, The specific analysis method of the multi-type abnormal data analysis module is as follows: Match the associated data types corresponding to each fault type in the fault type priority list, and then determine the monitoring points corresponding to these associated data types, which are recorded as each abnormal data monitoring point; Using a geographic information system, cluster analysis is performed on the spatial distribution of each abnormal data monitoring point. The midpoint of the line connecting the abnormal data monitoring points is taken as the potential leakage point, and a high-probability fault area is delineated.

8. The gas supply pipeline fault data acquisition system based on SCADA system according to claim 1, characterized in that, The specific operation method of the comprehensive analysis module is as follows: The multivariate parameter correlation model is activated to analyze the combined change characteristics of various parameters and calculate the comprehensive failure probability. When the comprehensive failure probability is greater than the set failure probability threshold, a failure warning is triggered. The spatial distribution of each abnormal data monitoring point and its associated monitoring points is overlaid and displayed by a geographic information system, and the fault probability distribution is displayed by a heat map. The associated monitoring points refer to other monitoring points that have data association, spatial association, fault logic association and other relationships with each abnormal data monitoring point. By integrating a multivariate parameter association model and a fault association matrix, when more than one abnormal data is triggered simultaneously, the system automatically matches the parameter association rules of the corresponding fault type in the fault association matrix. For each matched fault type, the system calculates the comprehensive priority based on the average association strength combined with the parameter weights, and pushes fault types with support ≥80% and confidence ≥90% to the system.

Images