Industrial gas plant operation monitoring method and system based on industrial cloud computing

Abstract

The application provides an industrial gas plant monitoring method and system based on industrial cloud computing, and relates to the technical field of industrial automation and data management. The method comprises the following steps: acquiring real-time monitoring data through multiple sensors arranged at preset positions at multiple time points in a monitoring period; acquiring real-time positioning information at multiple time points in the monitoring period; acquiring historical monitoring data in multiple historical monitoring periods; training a gas pipeline leakage prediction model according to the historical monitoring data to obtain a trained gas pipeline leakage prediction model; processing the real-time monitoring data according to the trained gas pipeline leakage prediction model to obtain a real-time predicted leakage probability and a real-time predicted leakage severity coefficient; and determining a monitoring result according to the real-time positioning information, the real-time predicted leakage probability and the real-time predicted leakage severity coefficient. According to the application, the accuracy and effectiveness of industrial gas plant monitoring can be improved.

Description

Technical Field

[0001] This invention relates to the field of industrial automation and data management technology, and in particular to an industrial gas plant monitoring method and system based on industrial cloud computing. Background Technology

[0002] In related technologies, such monitoring systems are mostly independently deployed localized monitoring systems, which suffer from problems such as data silos, limited intelligent alarm capabilities, poor mobile access experience, and difficulties in historical data analysis and predictive maintenance. In other words, the related technologies are unable to provide early warnings of potential risk situations and are unable to improve the accuracy of plant monitoring.

[0003] The information disclosed in the background section of this application is intended only to enhance the understanding of the general background of this application and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0004] This invention provides an industrial gas plant management monitoring method and system based on industrial cloud computing, which can solve the technical problems that related technologies are unable to provide early warnings of possible risk situations and improve the accuracy of plant management monitoring.

[0005] According to a first aspect of the present invention, an industrial gas plant management monitoring method based on industrial cloud computing is provided, comprising: At multiple moments during the monitoring cycle, real-time monitoring data is acquired through various sensors set at preset locations. The real-time monitoring data includes: equipment monitoring data, monitoring valve data, pressure data, local ambient temperature, real-time gas concentration, and real-time pipeline flow rate. Real-time location information is acquired at multiple points during the monitoring period; Acquire historical monitoring data from multiple historical monitoring periods, wherein the historical monitoring data includes: historical equipment monitoring data, historical monitoring valve data, historical pressure data, historical local ambient temperature, historical gas concentration, and historical pipeline path flow rate; Based on the historical monitoring data, the gas pipeline leakage prediction model is trained to obtain the trained gas pipeline leakage prediction model. The real-time monitoring data is processed according to the trained gas pipeline leakage prediction model to obtain the real-time predicted leakage probability and the real-time predicted leakage severity coefficient. The monitoring results are determined based on the real-time location information, the real-time predicted leakage probability, and the real-time predicted leakage severity coefficient.

[0006] According to the present invention, a gas pipeline leakage prediction model is trained based on the historical monitoring data to obtain a trained gas pipeline leakage prediction model, including: Based on the historical monitoring data, obtain the historical baseline monitoring data for the historical normal monitoring period in which no abnormalities occurred; Based on the historical benchmark monitoring data, obtain the historical theoretical expected flow for multiple historical monitoring periods; Based on the historical gas concentration, determine the rate of change of historical gas concentration; Based on the historical theoretical expected flow rate, the historical pipeline path flow rate, the historical equipment monitoring data, the historical monitoring valve data, the historical pressure data, and the historical local ambient temperature, the historical comprehensive anomaly results and the historical comprehensive anomaly coefficient are determined; Obtain historical actual leakage results for the predicted time corresponding to multiple moments in multiple historical monitoring periods; Based on the historical gas concentration, determine the historical actual leakage severity coefficients for the predicted time corresponding to multiple moments in multiple historical monitoring cycles; Historical monitoring data were processed based on a gas pipeline leakage prediction model to obtain the leakage probability and leakage severity coefficient of historical samples. Based on the historical actual leakage results, the historical actual leakage severity coefficient, the historical sample leakage probability, the historical sample leakage severity coefficient, the historical comprehensive anomaly results, the historical comprehensive anomaly coefficient, the historical gas concentration change rate, and the historical gas concentration, the training loss function of the gas pipeline leakage prediction model is determined. The gas pipeline leakage prediction model is trained using the training loss function of the gas pipeline leakage prediction model to obtain the trained gas pipeline leakage prediction model.

[0007] According to the present invention, based on the historical theoretical expected flow rate, the historical pipeline path flow rate, the historical equipment monitoring data, the historical monitoring valve data, the historical pressure data, and the historical local ambient temperature, the historical comprehensive anomaly result and the historical comprehensive anomaly coefficient are determined, including: Based on the historical equipment monitoring data, determine the abnormal equipment monitoring results and the abnormal equipment monitoring coefficient; Based on the historical valve monitoring data, determine the abnormal valve monitoring results; Based on the historical pressure data, determine the abnormal pressure monitoring results and the pressure monitoring abnormality coefficient; Based on the historical local ambient temperature, determine the abnormal temperature monitoring results and the abnormal temperature monitoring coefficient; Based on the historical theoretical expected flow and the historical pipeline path flow, determine the abnormal flow monitoring results and the abnormal flow monitoring coefficient; Based on the abnormal flow monitoring results, the abnormal flow monitoring coefficient, the abnormal equipment monitoring results, the abnormal equipment monitoring coefficient, the abnormal valve monitoring results, the abnormal pressure monitoring results, the abnormal pressure monitoring coefficient, the abnormal temperature monitoring results, and the abnormal temperature monitoring coefficient, the historical comprehensive abnormal results and the historical comprehensive abnormal coefficient are determined.

[0008] According to the present invention, determining the historical actual leakage severity coefficient for a predicted time corresponding to multiple moments in multiple historical monitoring cycles based on the historical gas concentration includes: according to the formula:

[0009] Determine the historical actual leakage severity coefficient for the predicted time corresponding to the i-th time in the k-th historical monitoring cycle. Where max is the function for finding the maximum value. The historical gas concentration at the predicted time corresponds to the i-th time point in the k-th historical monitoring cycle. To preset the gas concentration threshold, Let be the historical gas concentration at the i-th moment of the k-th historical monitoring cycle.

[0010] According to the present invention, the training loss function of the gas pipeline leakage prediction model is determined based on the historical actual leakage results, the historical actual leakage severity coefficient, the historical sample leakage probability, the historical sample leakage severity coefficient, the historical comprehensive anomaly results, the historical comprehensive anomaly coefficient, the historical gas concentration change rate, and the historical gas concentration, including: according to the formula:

[0011] Determine the training loss function for the gas pipeline leakage prediction model. Where, if is a conditional function, Let be the historical gas concentration change rate at the ith moment of the k-th historical monitoring period. To preset the threshold for the rate of change of gas concentration, This represents the historical comprehensive anomaly result at the i-th moment of the k-th historical monitoring period. Let i be the predicted leakage probability of the historical sample at the i-th moment corresponding to the k-th historical monitoring period. This represents the historical actual leakage result at the predicted time corresponding to the i-th time point in the k-th historical monitoring cycle. This represents the historical actual leakage severity coefficient at the predicted time corresponding to the i-th time point in the k-th historical monitoring cycle. The leakage severity coefficient of the historical sample at the predicted time corresponding to the i-th time of the k-th historical monitoring period is given. Let be the historical comprehensive anomaly coefficient at the ith moment of the kth historical monitoring period. To preset the gas concentration threshold, Let be the historical gas concentration at the i-th moment of the k-th historical monitoring period, where K is the number of historical monitoring periods (k≤K), n is the number of moments in the historical monitoring period (i≤n), and k, K, i, and n are all positive integers.

[0012] According to the present invention, the monitoring result is determined based on the real-time location information, the real-time predicted leakage probability, and the real-time predicted leakage severity coefficient, including: Based on the real-time location information, determine the real-time personnel location information and the real-time vehicle location information; Based on the real-time predicted leakage probability and the real-time predicted leakage severity coefficient, the real-time early warning level and real-time early warning location are determined. The monitoring results are determined based on the real-time personnel location information, the real-time vehicle location information, the real-time warning level, and the real-time warning location.

[0013] According to the present invention, determining the real-time early warning level and real-time early warning location based on the real-time predicted leakage probability and the real-time predicted leakage severity coefficient includes: The real-time early warning result is determined based on the real-time predicted leakage probability and the preset real-time predicted leakage probability threshold. The location of the real-time warning is determined based on the location of the gas pipeline segment where the real-time warning result is greater than 0; Based on the real-time predicted leakage severity coefficient, the real-time early warning severity result is determined; The real-time warning level is determined based on the real-time warning results and the real-time warning severity results.

[0014] According to a second aspect of the present invention, an industrial gas plant monitoring system based on industrial cloud computing is provided, comprising: The sensing data module is used to acquire real-time monitoring data at multiple moments during the monitoring cycle through various sensors set at preset locations. The real-time monitoring data includes: equipment monitoring data, monitoring valve data, pressure data, local ambient temperature, real-time gas concentration, and real-time pipeline flow rate. The location information module is used to acquire real-time location information at multiple points in the monitoring period; The historical data module is used to acquire historical monitoring data from multiple historical monitoring periods. The historical monitoring data includes: historical equipment monitoring data, historical monitoring valve data, historical pressure data, historical local ambient temperature, historical gas concentration, and historical pipeline flow rate. The model training module is used to train the gas pipeline leakage prediction model based on the historical monitoring data to obtain the trained gas pipeline leakage prediction model. The model prediction module is used to process the real-time monitoring data according to the trained gas pipeline leakage prediction model to obtain the real-time predicted leakage probability and the real-time predicted leakage severity coefficient. The monitoring results module is used to determine the monitoring results based on the real-time location information, the real-time predicted leakage probability, and the real-time predicted leakage severity coefficient.

[0015] Technical Effects: According to this invention, real-time monitoring data of the factory's operating environment, as well as real-time location information of personnel and vehicles, can be acquired through various sensors and monitoring systems. A gas pipeline leak prediction model is trained using historical monitoring data to obtain a trained gas pipeline leak prediction model. Based on this trained model, real-time monitoring data is processed to predict whether a gas leak will occur and the severity of the leak, obtaining a real-time predicted leak probability and a real-time predicted leak severity coefficient. Furthermore, based on real-time location information, the real-time predicted leak probability, and the real-time predicted leak severity coefficient, corresponding early warning information and monitoring results are generated, improving the accuracy and effectiveness of industrial gas plant monitoring. When determining the historical actual leak severity coefficient, the historical actual leak severity coefficient can be determined based on historical gas concentrations at multiple times corresponding to prediction times across multiple historical monitoring periods. During the calculation process, the relative difference between historical gas concentrations and preset gas concentration thresholds can accurately analyze and determine the historical actual leak severity coefficient under conditions where a gas leak exists. Furthermore, the increase in historical gas concentrations can accurately analyze the historical actual leak severity coefficient under conditions where a gas leak is not determined, improving the comprehensiveness and accuracy of the historical actual leak severity coefficient. When determining the training loss function for a gas pipeline leak prediction model, it can be based on historical actual leak results, historical actual leak severity coefficients, historical sample leak probabilities, historical sample leak severity coefficients, historical comprehensive anomaly results, historical comprehensive anomaly coefficients, historical gas concentration change rates, and historical gas concentrations. During the calculation process, the influence of historical actual leak results, historical actual leak severity coefficients, historical gas concentration change rates, and historical gas concentrations on leak probabilities and leak severity can be used to determine the impact of these data on the errors of historical sample leak probabilities and historical sample leak severity coefficients. Based on this impact and the errors of historical sample leak probabilities and historical sample leak severity coefficients, the training loss function can be set, thereby improving the accuracy of the gas pipeline leak prediction model in predicting gas leak severity and gas leak probability, and more effectively enhancing the precision of the gas pipeline leak prediction model.

[0016] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Other features and aspects of the invention will become clearer from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art 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 embodiments can be obtained based on these drawings without creative effort.

[0018] Figure 1 A schematic flowchart of an industrial gas plant monitoring method based on industrial cloud computing according to an embodiment of the present invention is shown as an example.

[0019] Figure 2 A schematic diagram of a training gas pipeline leakage prediction model according to an embodiment of the present invention is shown as an example.

[0020] Figure 3 An exemplary schematic diagram illustrating the determination of monitoring results according to an embodiment of the present invention is shown.

[0021] Figure 4 A block diagram of an industrial gas plant monitoring system based on industrial cloud computing according to an embodiment of the present invention is shown as an example. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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.

[0023] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0024] Figure 1 An exemplary flowchart of an industrial gas plant monitoring method based on industrial cloud computing according to an embodiment of the present invention is shown, the method comprising: Step S1: At multiple moments during the monitoring cycle, real-time monitoring data is acquired through various sensors set at preset locations. The real-time monitoring data includes: equipment monitoring data, monitoring valve data, pressure data, local ambient temperature, real-time gas concentration, and real-time pipeline flow rate. Step S2: Obtain real-time location information at multiple points during the monitoring period; Step S3: Obtain historical monitoring data from multiple historical monitoring cycles, wherein the historical monitoring data includes: historical equipment monitoring data, historical monitoring valve data, historical pressure data, historical local ambient temperature, historical gas concentration, and historical pipeline path flow rate; Step S4: Based on the historical monitoring data, train the gas pipeline leakage prediction model to obtain the trained gas pipeline leakage prediction model. Step S5: Process the real-time monitoring data according to the trained gas pipeline leakage prediction model to obtain the real-time predicted leakage probability and the real-time predicted leakage severity coefficient. Step S6: Determine the monitoring result based on the real-time location information, the real-time predicted leakage probability, and the real-time predicted leakage severity coefficient.

[0025] The industrial gas plant monitoring method based on industrial cloud computing according to an embodiment of the present invention can acquire real-time monitoring data of the operating environment in the factory, as well as real-time location information of personnel and vehicles, through various sensors and monitoring systems. It can also train a gas pipeline leakage prediction model using historical monitoring data to obtain a trained gas pipeline leakage prediction model. Based on the trained model, real-time monitoring data is processed to predict whether a gas leak will occur and the severity of the leak, obtaining a real-time predicted leak probability and a real-time predicted leak severity coefficient. Furthermore, based on real-time location information, the real-time predicted leak probability, and the real-time predicted leak severity coefficient, corresponding early warning information and monitoring results are generated, thereby improving the accuracy and effectiveness of industrial gas plant monitoring.

[0026] According to an embodiment of the present invention, in step S1, real-time monitoring data is acquired at multiple times during the monitoring cycle by using various sensors set at preset locations. The real-time monitoring data includes: equipment monitoring data, monitoring valve data, pressure data, local ambient temperature, real-time gas concentration, and real-time flow rate of pipeline path.

[0027] For example, the real-time concentration of a specific gas can be read by gas detectors deployed near a specific gas pipeline; the real-time flow rate of the pipeline can be read by flow meters (such as mass flow meters, vortex flow meters, or Coriolis flow meters) installed downstream of the pipeline; equipment monitoring data (such as equipment vibration data, equipment current data, and equipment temperature data) can be obtained by sensors installed on corresponding equipment (such as pumps and compressors) of the specific gas pipeline; valve monitoring data (such as valve opening and closing time, valve position change frequency, etc.) can be obtained by valve position transmitters and actuator status feedback and timers installed on corresponding valves (such as VMB valve bodies) of the specific gas pipeline; local ambient temperature can be obtained by distributed temperature sensors set in the non-heat-traced area of ​​the specific gas pipeline; and pressure data of the pipeline can be obtained by pressure sensors set in the middle of the straight section of the specific gas pipeline.

[0028] According to one embodiment of the present invention, in step S2, real-time positioning information is acquired at multiple moments during the monitoring period.

[0029] For example, the LocalSense® system (factory location management system) can obtain high-precision real-time location and tracking of personnel, vehicles and assets within the factory area, and acquire real-time location information.

[0030] According to an embodiment of the present invention, in step S3, historical monitoring data from multiple historical monitoring cycles are acquired, wherein the historical monitoring data includes: historical equipment monitoring data, historical monitoring valve data, historical pressure data, historical local ambient temperature, historical gas concentration, and historical pipeline path flow rate.

[0031] For example, the system will store the collected real-time monitoring data in a database and obtain historical monitoring data of a specific gas pipeline monitored in the current monitoring cycle from past historical monitoring cycles through the database.

[0032] According to one embodiment of the present invention, in step S4, the gas pipeline leakage prediction model is trained based on the historical monitoring data to obtain the trained gas pipeline leakage prediction model.

[0033] Figure 2 A schematic diagram of a training gas pipeline leakage prediction model according to an embodiment of the present invention is shown as an example.

[0034] According to an embodiment of the present invention, step S4 includes: Step S41: Based on the historical monitoring data, obtain the historical baseline monitoring data for the historical normal monitoring period in which no abnormal conditions have occurred; Step S42: Based on the historical benchmark monitoring data, obtain the historical theoretical expected flow rate for multiple historical monitoring periods; Step S43: Determine the rate of change of historical gas concentration based on the historical gas concentration; Step S44: Based on the historical theoretical expected flow rate, the historical pipeline path flow rate, the historical equipment monitoring data, the historical monitoring valve data, the historical pressure data, and the historical local ambient temperature, determine the historical comprehensive anomaly result and the historical comprehensive anomaly coefficient; Step S45: Obtain the historical actual leakage results for the predicted time corresponding to multiple times in multiple historical monitoring cycles; Step S46: Based on the historical gas concentration, determine the historical actual leakage severity coefficients for the predicted time corresponding to multiple moments in multiple historical monitoring cycles; Step S47: Process historical monitoring data according to the gas pipeline leakage prediction model to obtain the leakage probability and leakage severity coefficient of historical samples. Step S48: Determine the training loss function of the gas pipeline leakage prediction model based on the historical actual leakage results, the historical actual leakage severity coefficient, the historical sample leakage probability, the historical sample leakage severity coefficient, the historical comprehensive anomaly results, the historical comprehensive anomaly coefficient, the historical gas concentration change rate, and the historical gas concentration. Step S49: Train the gas pipeline leakage prediction model according to the training loss function of the gas pipeline leakage prediction model to obtain the trained gas pipeline leakage prediction model.

[0035] For example, acquire historical baseline monitoring data for normal monitoring cycles where no abnormalities occurred (e.g., confirming no gas leaks and normal equipment operation); based on the historical baseline monitoring data, acquire historical theoretical expected flow rates for multiple historical monitoring cycles. For example, based on historical baseline monitoring data when no abnormalities occurred, acquire historical baseline flow rates for the same or similar processes during the same historical monitoring cycle; determine the historical theoretical expected flow rates for multiple historical monitoring cycles based on the average of multiple historical baseline flow rates; assess whether there are abnormalities and the degree of abnormality in flow rates, equipment, valves, pressure, and ambient temperature based on historical theoretical expected flow rates, historical pipeline path flow rates, historical equipment monitoring data, historical monitoring valve data, historical pressure data, and historical local ambient temperature, and determine the historical comprehensive abnormality result and historical comprehensive abnormality coefficient; acquire historical actual leakage results for predicted times corresponding to multiple moments in multiple historical monitoring cycles. For example, if the first moment of the first historical monitoring cycle is 8:00 AM on March 10th, then the predicted time corresponding to that moment is 8:30 AM on March 10th. Determine whether a leak occurred at that predicted time. If a leak occurred, the historical actual leakage result is 1; otherwise, the historical actual leakage result is 1. 0; Based on historical gas concentrations, assess the severity of leakage at the predicted time, and determine the historical actual leakage severity coefficients for the predicted time corresponding to multiple times in multiple historical monitoring periods; Process historical monitoring data according to the gas pipeline leakage prediction model to obtain historical sample leakage probabilities and historical sample leakage severity coefficients. The gas pipeline leakage prediction model is a type of neural network model, including: data preprocessing and input layer, feature extraction layer, feature fusion layer, and decision and output layer. Train the gas pipeline leakage prediction model with historical data so that it can predict the probability and severity of gas leakage, generating historical sample leakage probabilities and historical sample leakage severity coefficients for the predicted time corresponding to multiple times in historical monitoring periods; Based on historical actual leakage results, historical actual leakage severity coefficients, historical sample leakage probabilities, historical sample leakage severity coefficients, historical comprehensive anomaly results, historical comprehensive anomaly coefficients, historical gas concentration change rates, and historical gas concentrations, determine the training loss function of the gas pipeline leakage prediction model; Train the gas pipeline leakage prediction model according to the training loss function to obtain the trained gas pipeline leakage prediction model.

[0036] According to one embodiment of the present invention, step S44 includes; Step S441: Based on the historical equipment monitoring data, determine the equipment monitoring anomaly results and the equipment monitoring anomaly coefficient; Step S442: Determine the abnormal valve monitoring results based on the historical valve monitoring data; Step S443: Based on the historical pressure data, determine the pressure monitoring anomaly results and the pressure monitoring anomaly coefficient; Step S444: Based on the historical local ambient temperature, determine the temperature monitoring anomaly result and the temperature monitoring anomaly coefficient; Step S445: Based on the historical theoretical expected flow rate and the historical pipeline path flow rate, determine the abnormal flow monitoring results and the abnormal flow monitoring coefficient. Step S446: Based on the abnormal flow monitoring results, the abnormal flow monitoring coefficient, the abnormal equipment monitoring results, the abnormal equipment monitoring coefficient, the abnormal valve monitoring results, the abnormal pressure monitoring results, the abnormal pressure monitoring coefficient, the abnormal temperature monitoring results, and the abnormal temperature monitoring coefficient, determine the historical comprehensive abnormal results and the historical comprehensive abnormal coefficient.

[0037] For example, if the current, vibration, and temperature of a device all show a slow upward trend relative to the baseline values ​​(which can be determined based on the average value of the device under long-term normal operating conditions), a slow increase means that the increase within the corresponding ten-minute time window exceeds twice the standard deviation of its corresponding multiple baseline values. A slow increase implies an increase in the load on the device, and one reason for the increased load is the presence of undetected minor leaks downstream. The device monitoring anomaly result is then 1; conversely, it is 0. The final device monitoring anomaly result is determined by summing and averaging the device monitoring anomaly results of all devices monitoring a specific gas pipeline in the current monitoring cycle. The device monitoring anomaly coefficient can be determined based on the sum of the relative differences between the device's current, vibration, and temperature and its baseline values. To determine the final equipment monitoring anomaly coefficient, the anomaly coefficients of all equipment monitoring the specific gas pipeline under the current monitoring cycle are summed and averaged. Based on the monitored valve data, valve monitoring anomalies are determined. For example, if the valve exhibits frequent small opening adjustments within the corresponding ten-minute time window (the number of changes in the valve position feedback signal (0-100%) per unit time is counted; under normal steady state, this should be 0. Continuous, high-frequency small jumps (e.g., more than 5 changes within 10 seconds) are considered abnormal, usually indicating PID controller instability or valve mechanical jamming; prolonged switching action time (if the historical average opening time is 10 seconds, and the current three consecutive times exceed 15 seconds, a warning is triggered; this usually indicates aging of the drive mechanism (e.g., decreased motor torque, weak spring) or valve internal leakage); or slight fluctuations in the feedback signal (the variance or standard deviation of continuously sampled valve position feedback data is calculated). The variance calculation is as follows: under normal steady-state conditions, the variance should be close to 0. If the variance exceeds a threshold (e.g., 0.2%FS), it is considered a fluctuation. In this case, the valve monitoring anomaly result is 1; otherwise, it is 0. The final valve monitoring anomaly result is determined by summing and averaging the valve monitoring anomaly results of all valves in the specific gas pipeline monitored in the current monitoring cycle. Based on the pressure data, the pressure monitoring anomaly result and pressure monitoring anomaly coefficient are determined. For example, based on the pressure data, the pressure-time curve is determined. If the pressure-time curve shows a continuous downward trend that excludes normal consumption (the slope of the pressure-time curve within the time window is less than 0, and the absolute value of the slope is greater than a preset threshold, which can be set as the average of the slopes of multiple pressure-time curves during normal equipment operation plus twice the standard deviation), it may indicate the presence of a small leak. In this case, the pressure monitoring anomaly result is 1; otherwise, it is 0. The pressure monitoring anomaly coefficient can be determined based on... The determination is made based on the local ambient temperature; the abnormal temperature monitoring result and the abnormal temperature monitoring coefficient are determined. For example, some gases absorb heat due to adiabatic expansion when leaking, leading to a decrease in local temperature. If an abnormally low temperature point appears in the non-heat-traced area of ​​a pipeline section, the abnormal temperature monitoring result is 1; otherwise, the abnormal temperature monitoring result is 0. The abnormal temperature monitoring coefficient can be determined based on... After determination, the reference temperature can be determined based on the average temperature of the same location under normal operating conditions. Based on historical theoretical expected flow and historical pipeline path flow, determine the abnormal flow monitoring results and flow monitoring anomaly coefficient. For example, if the historical pipeline path flow is less than 0.95 times the historical theoretical expected flow (the lost flow is most likely due to leakage into the environment through tiny leaks in the pipes or joints), the flow monitoring anomaly result is 1; otherwise, the flow monitoring anomaly result is 0. The flow monitoring anomaly coefficient is determined. The larger the flow monitoring anomaly coefficient, the more abnormal flow is missing. The historical comprehensive anomaly result is determined by summing the flow monitoring anomaly results, equipment monitoring anomaly results, valve monitoring anomaly results, pressure monitoring anomaly results, and temperature monitoring anomaly results. The historical comprehensive anomaly coefficient is determined by summing the flow monitoring anomaly coefficient, equipment monitoring anomaly coefficient, pressure monitoring anomaly coefficient, and temperature monitoring anomaly coefficient.

[0038] According to an embodiment of the present invention, step S46 includes: determining the historical actual leakage severity coefficient of the predicted time corresponding to the i-th time of the k-th historical monitoring cycle according to formula (1). , (1) Where max is the function for finding the maximum value. The historical gas concentration at the predicted time corresponds to the i-th time point in the k-th historical monitoring cycle. To preset the gas concentration threshold, Let be the historical gas concentration at the i-th moment of the k-th historical monitoring cycle.

[0039] According to one embodiment of the present invention, To preset the gas concentration threshold, it can be set according to the monitoring thresholds stipulated by the industry or the country. "Time" indicates that a gas leak is predicted to occur at the i-th moment of the k-th historical monitoring cycle, and represents the historical actual leak severity coefficient. The value is , This represents the relative difference between the historical gas concentration and the preset gas concentration threshold at the predicted time corresponding to the i-th time of the k-th historical monitoring period. The larger the value, the more gas is leaking and the more serious the leak. When, it indicates that the predicted time corresponding to the i-th moment of the k-th historical monitoring cycle is not determined to have a gas leak, and the historical actual leak severity coefficient. The value is , This represents the increase in the historical gas concentration at the predicted time corresponding to the i-th time in the k-th historical monitoring period compared to the historical gas concentration at the i-th time. The higher the value, the more abnormally high the gas concentration is. Although it may not exceed the preset gas concentration threshold at that moment, there is a possibility of further leakage. Indicates taking The maximum value of 1, and the above process of taking the maximum value, can make No greater than That is, the historical actual leakage severity coefficient under the condition that the existence of gas leakage is not determined will not be greater than the historical actual leakage severity coefficient under the condition that the existence of gas leakage is determined.

[0040] In this way, the historical actual leakage severity coefficients corresponding to multiple moments in multiple historical monitoring cycles can be determined based on historical gas concentrations. During the calculation process, the historical actual leakage severity coefficients under gas leakage conditions can be accurately analyzed and determined based on the relative difference between historical gas concentrations and preset gas concentration thresholds. The historical actual leakage severity coefficients under gas leakage conditions can be accurately analyzed based on the increase in historical gas concentrations, thereby improving the comprehensiveness and accuracy of the historical actual leakage severity coefficients.

[0041] According to an embodiment of the present invention, step S48 includes: determining the training loss function of the gas pipeline leakage prediction model according to formula (2). , (2) Where if is a conditional function. Let be the historical gas concentration change rate at the ith moment of the k-th historical monitoring period. To preset the threshold for the rate of change of gas concentration, This represents the historical comprehensive anomaly result at the i-th moment of the k-th historical monitoring period. Let i be the predicted leakage probability of the historical sample at the i-th moment corresponding to the k-th historical monitoring period. This represents the historical actual leakage result at the predicted time corresponding to the i-th time point in the k-th historical monitoring cycle. This represents the historical actual leakage severity coefficient at the predicted time corresponding to the i-th time point in the k-th historical monitoring cycle. The leakage severity coefficient of the historical sample at the predicted time corresponding to the i-th time of the k-th historical monitoring period is given. Let be the historical comprehensive anomaly coefficient at the ith moment of the kth historical monitoring period. To preset the gas concentration threshold, Let be the historical gas concentration at the i-th moment of the k-th historical monitoring period, where K is the number of historical monitoring periods (k≤K), n is the number of moments in the historical monitoring period (i≤n), and k, K, i, and n are all positive integers.

[0042] According to an embodiment of the present invention, in formula (2), the condition function The value includes the following two cases, when the following conditions are met: In the case where the historical gas concentration change rate at time i in the k-th historical monitoring cycle is greater than a preset gas concentration change rate threshold, the preset gas concentration change rate threshold can be set based on the average normal gas concentration change rate during the same process operation at time i in the k-th historical monitoring cycle. The value of the conditional function is... , The larger the value, the greater the historical gas concentration change rate, the greater the probability of an impending leak, and the greater the probability of a leak in the historical sample. When this condition is not met... In the case where the historical gas concentration change rate at time i in the k-th historical monitoring period is less than or equal to the preset gas concentration change rate threshold, the value of the conditional function is... , The larger the value, the smaller the historical gas concentration change rate, the lower the probability of an impending leak, and the lower the probability of a leak in the historical sample. This represents the historical comprehensive anomaly result at the i-th moment of the k-th historical monitoring period. The larger the value, the more abnormal situations that could lead to leakage are predicted at time ith in the k-th historical monitoring period, and the greater the probability of leakage in the historical samples. Indicates when hour, The size is positively correlated with the historical sample leakage probability, when hour, The magnitude of the error is negatively correlated with the historical sample leakage probability, while the historical comprehensive anomaly result is positively correlated with the historical sample leakage probability.

[0043] According to one embodiment of the present invention, The error between the historical sample leakage probability and the historical actual leakage result at the predicted time corresponding to the i-th time of the k-th historical monitoring cycle is calculated using... right A weighted summation is performed to obtain the first training loss function. During the training process, the first training loss function is reduced to improve the accuracy of the gas pipeline leakage prediction model in predicting the probability of gas leakage, thereby improving the accuracy of the gas pipeline leakage prediction model.

[0044] According to one embodiment of the present invention, Let be the historical comprehensive anomaly coefficient at the ith moment of the kth historical monitoring period. The larger the value, the more severe the abnormal situation indicating a possible leak at time ith in the k-th historical monitoring period, and the more severe the potential leak. A higher historical sample leak severity coefficient indicates a more serious leak. This represents the relative difference between the historical gas concentration at time i in the k-th historical monitoring period and the preset gas concentration threshold. The larger this ratio, the higher the historical gas concentration at time i in the k-th historical monitoring period, and the more severe the leakage will be. A higher historical sample leakage severity coefficient indicates a higher historical gas concentration. express and The larger the data, the larger the historical sample leakage severity coefficient, and the greater the impact on the error of the historical sample leakage severity coefficient.

[0045] According to one embodiment of the present invention, The error between the historical sample leakage severity coefficient at the predicted time corresponding to the i-th moment of the k-th historical monitoring cycle and the historical actual leakage severity coefficient at the predicted time corresponding to the i-th moment of the k-th historical monitoring cycle is calculated using... right A weighted summation is performed to obtain the second training loss function. During the training process, the second training loss function is reduced, which improves the accuracy of the gas pipeline leak prediction model in predicting the severity of gas leaks, thereby improving the accuracy of the gas pipeline leak prediction model.

[0046] According to one embodiment of the present invention, This indicates that the training loss function of the gas pipeline leakage prediction model is determined by summing the first training loss function and the second training loss function.

[0047] In this way, the training loss function of the gas pipeline leak prediction model can be determined based on historical actual leak results, historical actual leak severity coefficients, historical sample leak probabilities, historical sample leak severity coefficients, historical comprehensive anomaly results, historical comprehensive anomaly coefficients, historical gas concentration change rates, and historical gas concentrations. During the calculation process, the impact of the historical actual leak results, historical actual leak severity coefficients, historical gas concentration change rates, and historical gas concentrations on the leak probability and leak severity can be used to determine the error of the above data on the historical sample leak probability and historical sample leak severity coefficients. Based on this impact and the error of the historical sample leak probability and historical sample leak severity coefficients, the training loss function can be set to improve the accuracy of the gas pipeline leak prediction model in predicting the severity and probability of gas leaks, thus improving the accuracy of the gas pipeline leak prediction model in a more targeted manner.

[0048] According to one embodiment of the present invention, in step S5, the real-time monitoring data is processed according to the trained gas pipeline leakage prediction model to obtain the real-time predicted leakage probability and the real-time predicted leakage severity coefficient.

[0049] For example, real-time monitoring data can be processed based on a trained gas pipeline leak prediction model to obtain the real-time predicted leak probability and the real-time predicted leak severity coefficient 30 minutes later.

[0050] According to an embodiment of the present invention, in step S6, the monitoring result is determined based on the real-time location information, the real-time predicted leakage probability, and the real-time predicted leakage severity coefficient.

[0051] Figure 3 An exemplary schematic diagram illustrating the determination of monitoring results according to an embodiment of the present invention is shown.

[0052] According to an embodiment of the present invention, step S6 includes: Step S61: Determine real-time personnel location information and real-time vehicle location information based on the real-time location information; Step S62: Determine the real-time early warning level and real-time early warning location based on the real-time predicted leakage probability and the real-time predicted leakage severity coefficient; Step S63: Determine the monitoring result based on the real-time personnel positioning information, the real-time vehicle positioning information, the real-time warning level, and the real-time warning location.

[0053] For example, based on the real-time location information obtained from the factory location management system, the real-time personnel location information and real-time vehicle location information within the factory can be determined; based on the real-time predicted leakage probability and the real-time predicted leakage severity coefficient, the real-time warning level and real-time warning location can be determined; based on the real-time personnel location information, real-time vehicle location information, real-time warning level, and real-time warning location, the monitoring results can be determined, such as issuing warnings of the corresponding level to personnel and vehicles near the real-time warning location based on the real-time personnel location information and real-time vehicle location information.

[0054] According to an embodiment of the present invention, step S62 includes: Step S621: Determine the real-time early warning result based on the real-time predicted leakage probability and the preset real-time predicted leakage probability threshold; Step S622: Determine the real-time warning location based on the location of the gas pipeline segment with a real-time warning result greater than 0; Step S623: Determine the real-time warning severity result based on the real-time predicted leakage severity coefficient; Step S624: Determine the real-time warning level based on the real-time warning result and the real-time warning severity result.

[0055] For example, when the real-time predicted leakage probability is less than a preset real-time predicted leakage probability threshold (which can be set to 0.3), the real-time warning result is 0. When the real-time predicted leakage probability is greater than or equal to the preset real-time predicted leakage probability threshold, if the real-time predicted leakage probability is in the range of 0.3-0.39, the real-time warning result is 1; if the real-time predicted leakage probability is in the range of 0.4-0.49, the real-time warning result is 2, and so on. Gas pipelines with real-time warning results greater than 0 indicate that a leak may occur. The real-time warning location is determined based on the location of the gas pipeline segment with a real-time warning result greater than 0. The real-time warning severity result is determined based on the real-time predicted leakage severity coefficient. For example, when the real-time predicted leakage severity coefficient is less than 0.3, the real-time warning severity result is 1; when the real-time predicted leakage severity coefficient is in the range of 0.3-0.59, the real-time warning severity result is 3; when the real-time predicted leakage severity coefficient is in the range of 0.6-0.99, the real-time warning severity result is 5; and when the real-time predicted leakage severity coefficient is greater than 1, the real-time warning severity result is 7. The real-time warning level is determined by multiplying the real-time warning result and the real-time warning severity result.

[0056] The industrial gas plant monitoring method based on industrial cloud computing according to an embodiment of the present invention can acquire real-time monitoring data of the operating environment in the factory, as well as real-time location information of personnel and vehicles, through various sensors and monitoring systems. It can also train a gas pipeline leakage prediction model using historical monitoring data to obtain a trained gas pipeline leakage prediction model. Based on the trained model, real-time monitoring data is processed to predict whether a gas leak will occur and the severity of the leak, obtaining a real-time predicted leak probability and a real-time predicted leak severity coefficient. Furthermore, based on real-time location information, the real-time predicted leak probability, and the real-time predicted leak severity coefficient, corresponding early warning information and monitoring results are generated, thereby improving the accuracy and effectiveness of industrial gas plant monitoring. When determining the historical actual leakage severity coefficient, the historical actual leakage severity coefficient at the predicted time corresponding to multiple moments in multiple historical monitoring cycles can be determined based on historical gas concentrations. During the calculation process, the historical actual leakage severity coefficient under the condition of gas leakage can be accurately analyzed and determined based on the relative difference between historical gas concentrations and preset gas concentration thresholds. The historical actual leakage severity coefficient under the condition of no gas leakage can be accurately analyzed based on the increase in historical gas concentrations, thereby improving the comprehensiveness and accuracy of the historical actual leakage severity coefficient. When determining the training loss function for a gas pipeline leak prediction model, it can be based on historical actual leak results, historical actual leak severity coefficients, historical sample leak probabilities, historical sample leak severity coefficients, historical comprehensive anomaly results, historical comprehensive anomaly coefficients, historical gas concentration change rates, and historical gas concentrations. During the calculation process, the influence of historical actual leak results, historical actual leak severity coefficients, historical gas concentration change rates, and historical gas concentrations on leak probabilities and leak severity can be used to determine the impact of these data on the errors of historical sample leak probabilities and historical sample leak severity coefficients. Based on this impact and the errors of historical sample leak probabilities and historical sample leak severity coefficients, the training loss function can be set, thereby improving the accuracy of the gas pipeline leak prediction model in predicting gas leak severity and gas leak probability, and more effectively enhancing the precision of the gas pipeline leak prediction model.

[0057] Figure 4 An exemplary block diagram of an industrial gas plant monitoring system based on industrial cloud computing according to an embodiment of the present invention is shown, the system comprising: The sensing data module is used to acquire real-time monitoring data at multiple moments during the monitoring cycle through various sensors set at preset locations. The real-time monitoring data includes: equipment monitoring data, monitoring valve data, pressure data, local ambient temperature, real-time gas concentration, and real-time pipeline flow rate. The location information module is used to acquire real-time location information at multiple points in the monitoring period; The historical data module is used to acquire historical monitoring data from multiple historical monitoring periods. The historical monitoring data includes: historical equipment monitoring data, historical monitoring valve data, historical pressure data, historical local ambient temperature, historical gas concentration, and historical pipeline flow rate. The model training module is used to train the gas pipeline leakage prediction model based on the historical monitoring data to obtain the trained gas pipeline leakage prediction model. The model prediction module is used to process the real-time monitoring data according to the trained gas pipeline leakage prediction model to obtain the real-time predicted leakage probability and the real-time predicted leakage severity coefficient. The monitoring results module is used to determine the monitoring results based on the real-time location information, the real-time predicted leakage probability, and the real-time predicted leakage severity coefficient.

[0058] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0059] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are merely examples and do not limit the present invention. The objectives of the present invention have been fully and effectively achieved. The functions and structural principles of the present invention have been demonstrated and explained in the embodiments, and any variations or modifications may be made to the implementation of the present invention without departing from the stated principles.

Claims

1. A method for monitoring industrial gas plant operations based on industrial cloud computing, characterized in that, include: At multiple moments during the monitoring cycle, real-time monitoring data is acquired through various sensors set at preset locations. The real-time monitoring data includes: equipment monitoring data, monitoring valve data, pressure data, local ambient temperature, real-time gas concentration, and real-time pipeline flow rate. Real-time location information is acquired at multiple points during the monitoring period; Acquire historical monitoring data from multiple historical monitoring periods, wherein the historical monitoring data includes: historical equipment monitoring data, historical monitoring valve data, historical pressure data, historical local ambient temperature, historical gas concentration, and historical pipeline path flow rate; Based on the historical monitoring data, the gas pipeline leakage prediction model is trained to obtain the trained gas pipeline leakage prediction model. The real-time monitoring data is processed according to the trained gas pipeline leakage prediction model to obtain the real-time predicted leakage probability and the real-time predicted leakage severity coefficient. The monitoring results are determined based on the real-time location information, the real-time predicted leakage probability, and the real-time predicted leakage severity coefficient.

2. The industrial gas plant monitoring method based on industrial cloud computing according to claim 1, characterized in that, Based on the historical monitoring data, the gas pipeline leakage prediction model is trained to obtain a trained gas pipeline leakage prediction model, including: Based on the historical monitoring data, obtain the historical baseline monitoring data for the historical normal monitoring period in which no abnormalities occurred; Based on the historical benchmark monitoring data, obtain the historical theoretical expected flow for multiple historical monitoring periods; Based on the historical gas concentration, determine the rate of change of historical gas concentration; Based on the historical theoretical expected flow rate, the historical pipeline path flow rate, the historical equipment monitoring data, the historical monitoring valve data, the historical pressure data, and the historical local ambient temperature, the historical comprehensive anomaly results and the historical comprehensive anomaly coefficient are determined; Obtain historical actual leakage results for the predicted time corresponding to multiple moments in multiple historical monitoring periods; Based on the historical gas concentration, determine the historical actual leakage severity coefficients for the predicted time corresponding to multiple moments in multiple historical monitoring cycles; Historical monitoring data were processed based on a gas pipeline leakage prediction model to obtain the leakage probability and leakage severity coefficient of historical samples. Based on the historical actual leakage results, the historical actual leakage severity coefficient, the historical sample leakage probability, the historical sample leakage severity coefficient, the historical comprehensive anomaly results, the historical comprehensive anomaly coefficient, the historical gas concentration change rate, and the historical gas concentration, the training loss function of the gas pipeline leakage prediction model is determined. The gas pipeline leakage prediction model is trained using the training loss function of the gas pipeline leakage prediction model to obtain the trained gas pipeline leakage prediction model.

3. The industrial gas plant monitoring method based on industrial cloud computing according to claim 2, characterized in that, Based on the historical theoretical expected flow rate, the historical pipeline path flow rate, the historical equipment monitoring data, the historical monitoring valve data, the historical pressure data, and the historical local ambient temperature, the historical comprehensive anomaly results and historical comprehensive anomaly coefficients are determined, including: Based on the historical equipment monitoring data, determine the abnormal equipment monitoring results and the abnormal equipment monitoring coefficient; Based on the historical valve monitoring data, determine the abnormal valve monitoring results; Based on the historical pressure data, determine the abnormal pressure monitoring results and the pressure monitoring abnormality coefficient; Based on the historical local ambient temperature, determine the abnormal temperature monitoring results and the abnormal temperature monitoring coefficient; Based on the historical theoretical expected flow and the historical pipeline path flow, determine the abnormal flow monitoring results and the abnormal flow monitoring coefficient; Based on the abnormal flow monitoring results, the abnormal flow monitoring coefficient, the abnormal equipment monitoring results, the abnormal equipment monitoring coefficient, the abnormal valve monitoring results, the abnormal pressure monitoring results, the abnormal pressure monitoring coefficient, the abnormal temperature monitoring results, and the abnormal temperature monitoring coefficient, the historical comprehensive abnormal results and the historical comprehensive abnormal coefficient are determined.

4. The industrial gas plant monitoring method based on industrial cloud computing according to claim 2, characterized in that, Based on the historical gas concentration, determine the historical actual leakage severity coefficients for predicted times corresponding to multiple moments in multiple historical monitoring periods, including: according to the formula: Determine the historical actual leakage severity coefficient for the predicted time corresponding to the i-th time in the k-th historical monitoring cycle. Where max is the function for finding the maximum value. The historical gas concentration at the predicted time corresponds to the i-th time point in the k-th historical monitoring cycle. To preset the gas concentration threshold, Let be the historical gas concentration at the i-th moment of the k-th historical monitoring cycle.

5. The industrial gas plant monitoring method based on industrial cloud computing according to claim 2, characterized in that, Based on the historical actual leakage results, the historical actual leakage severity coefficient, the historical sample leakage probability, the historical sample leakage severity coefficient, the historical comprehensive anomaly results, the historical comprehensive anomaly coefficient, the historical gas concentration change rate, and the historical gas concentration, the training loss function of the gas pipeline leakage prediction model is determined, including: according to the formula: Determine the training loss function for the gas pipeline leakage prediction model. Where, if is a conditional function, Let be the historical gas concentration change rate at the ith moment of the k-th historical monitoring period. To preset the threshold for the rate of change of gas concentration, This represents the historical comprehensive anomaly result at the i-th moment of the k-th historical monitoring period. Let i be the predicted leakage probability of the historical sample at the i-th moment corresponding to the k-th historical monitoring period. This represents the historical actual leakage result at the predicted time corresponding to the i-th time point in the k-th historical monitoring cycle. This represents the historical actual leakage severity coefficient at the predicted time corresponding to the i-th time point in the k-th historical monitoring cycle. The leakage severity coefficient of the historical sample at the predicted time corresponding to the i-th time of the k-th historical monitoring period is given. Let be the historical comprehensive anomaly coefficient at the ith moment of the kth historical monitoring period. To preset the gas concentration threshold, Let be the historical gas concentration at the i-th moment of the k-th historical monitoring period, where K is the number of historical monitoring periods (k≤K), n is the number of moments in the historical monitoring period (i≤n), and k, K, i, and n are all positive integers.

6. The industrial gas plant monitoring method based on industrial cloud computing according to claim 1, characterized in that, Based on the real-time location information, the real-time predicted leakage probability, and the real-time predicted leakage severity coefficient, the monitoring results are determined, including: Based on the real-time location information, determine the real-time personnel location information and the real-time vehicle location information; Based on the real-time predicted leakage probability and the real-time predicted leakage severity coefficient, the real-time early warning level and real-time early warning location are determined. The monitoring results are determined based on the real-time personnel location information, the real-time vehicle location information, the real-time warning level, and the real-time warning location.

7. The industrial gas plant monitoring method based on industrial cloud computing according to claim 6, characterized in that, Based on the real-time predicted leakage probability and the real-time predicted leakage severity coefficient, the real-time early warning level and real-time early warning location are determined, including: The real-time early warning result is determined based on the real-time predicted leakage probability and the preset real-time predicted leakage probability threshold. The location of the real-time warning is determined based on the location of the gas pipeline segment where the real-time warning result is greater than 0; Based on the real-time predicted leakage severity coefficient, the real-time early warning severity result is determined; The real-time warning level is determined based on the real-time warning results and the real-time warning severity results.

8. An industrial gas plant monitoring system based on industrial cloud computing, characterized in that, For performing the method of any one of claims 1-7, comprising: The sensing data module is used to acquire real-time monitoring data at multiple moments during the monitoring cycle through various sensors set at preset locations. The real-time monitoring data includes: equipment monitoring data, monitoring valve data, pressure data, local ambient temperature, real-time gas concentration, and real-time pipeline flow rate. The location information module is used to acquire real-time location information at multiple points in the monitoring period; The historical data module is used to acquire historical monitoring data from multiple historical monitoring periods. The historical monitoring data includes: historical equipment monitoring data, historical monitoring valve data, historical pressure data, historical local ambient temperature, historical gas concentration, and historical pipeline path flow rate. The model training module is used to train the gas pipeline leakage prediction model based on the historical monitoring data to obtain the trained gas pipeline leakage prediction model. The model prediction module is used to process the real-time monitoring data according to the trained gas pipeline leakage prediction model to obtain the real-time predicted leakage probability and the real-time predicted leakage severity coefficient. The monitoring results module is used to determine the monitoring results based on the real-time location information, the real-time predicted leakage probability, and the real-time predicted leakage severity coefficient.

Images