A control method and control system for a pipeline gas safety valve

Abstract

The application discloses a kind of pipeline gas safety valve control method and control system, it is related to gas safety control technical field, comprising the following steps: obtaining the operation data of gas pipeline, constructs the operation state image of gas pipeline, and extracts evolution characteristics, generates operation state characteristic vector;According to operation state characteristic vector, grade division is carried out to operation state, and according to historical operation data, the threshold value of division is dynamically revised, and the corresponding safety valve control decision instruction is output;According to safety valve control decision instruction, drive safety valve to execute open-close action, and the consistency between valve actual action state and control instruction is checked, and valve execution result data is generated;According to valve execution result data rewrite, update operation state feature construction process;The application is through operation state image and closed-loop feedback control, solves the problem that control decision in traditional pipeline gas safety valve control is not strong in adaptability, and execution result lacks feedback.

Description

Technical Field

[0001] This invention relates to the field of gas safety control technology, and more specifically, to a method and control system for controlling a pipeline gas safety valve. Background Technology

[0002] With the widespread application of piped gas systems, gas safety valves, as key safety control components in these systems, directly impact the safety and stability of the gas delivery process. Traditional piped gas safety valve control often faces challenges such as complex operating state changes, difficulty in identifying abnormal states, and delayed control decisions. When the operating state of a gas pipeline evolves, failure to promptly and accurately identify and implement corresponding controls can easily affect the accuracy and responsiveness of the safety valve. Therefore, achieving effective control of gas safety valves under complex operating conditions has become a crucial technical challenge that urgently needs to be addressed.

[0003] For example, the invention patent with publication number CN121232686A discloses an AI-assisted adaptive control method and system for gas safety valves. The method includes: connecting the output of a SEPIC circuit containing continuous and discontinuous current modes to a gas safety valve coil; asynchronously acquiring data from preset electrical and gas safety indicators to obtain electrical and gas safety indicator sequences; using an AI fusion analyzer on an edge MCU to perform interactive feature fusion analysis on the two sequences to determine an adaptive control scheme for the gas safety valve; switching the circuit mode based on this control scheme, and driving the safety valve according to the switched SEPIC circuit. This solves the technical problems of fragmented analysis of electrical parameters and gas safety indicators, single circuit operation mode, and inaccurate control response in traditional gas safety valve control, achieving efficient and reliable adaptive driving of the gas safety valve and improving gas usage safety and control flexibility.

[0004] For example, the invention patent with publication number CN111637267A discloses a method for intelligently opening a gas safety valve. The method includes: obtaining the rated pressure of the gas safety valve; obtaining the rated pressure loss of the gas safety valve; obtaining a first set of parameters related to the actual operation of the gas safety valve, the first set of parameters including at least two of the actual outlet pressure, actual inlet pressure, or actual pressure loss; determining a temporary rated pressure loss based on the rated pressure and the first set of parameters; and determining the open or closed state of the gas safety valve based on the temporary rated pressure loss and the rated pressure loss. This method enables the rapid and accurate opening of the gas safety valve. By determining the opening of the gas safety valve based on the real-time acquisition of the pressure or pressure difference between the gas safety valve's inlet and outlet, the valve can be opened more quickly and accurately.

[0005] The above-disclosed technical solutions have at least the following technical problems:

[0006] Traditional pipeline gas safety valve control methods primarily rely on real-time acquisition of single or limited parameter information for operational status assessment. They lack a holistic mechanism for depicting the evolution of the gas pipeline's operational status, making it difficult to reflect the characteristics of operational status changes over time. This results in limited adaptability of control decisions to complex operational scenarios. Furthermore, after valve execution, traditional gas safety valve control methods typically only focus on the result of the control command issuance, lacking a verification and feedback mechanism for the consistency between the actual safety valve execution status and the control command, making it difficult to form an effective closed loop in the control process.

[0007] To address the above problems, this invention proposes a solution. Summary of the Invention

[0008] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a pipeline gas safety valve control method and control system, which solves the problems of insufficient characterization of operating status, weak adaptability of control decisions, and lack of feedback on execution results in gas safety valve control by using operating status profiling and closed-loop feedback control.

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] A method for controlling a pipeline gas safety valve includes: acquiring operational data of the gas pipeline, constructing a profile of the gas pipeline's operational status, extracting evolutionary features, and generating an operational status feature vector; classifying the operational status into levels based on the operational status feature vector, dynamically correcting the classification threshold based on historical operational data, and outputting corresponding safety valve control decision commands; driving the safety valve to perform opening and closing actions according to the safety valve control decision commands, verifying the consistency between the actual valve action state and the control commands, and generating valve execution result data; and writing back the valve execution result data to update the operational status feature construction process.

[0011] In a preferred technical solution, the extraction of evolutionary features and generation of operating state feature vectors specifically involves: arranging the operating state profiles of gas pipelines obtained within multiple consecutive preset time windows in chronological order to form a temporal evolution sequence of the operating state profiles; mapping the multi-dimensional feature embedding vectors corresponding to the operating state profiles to the same feature space, and generating an evolution trajectory of the operating state in the feature space based on the changing relationship of the multi-dimensional feature embedding vectors within adjacent time windows; outputting the direction and magnitude of change of pressure-related features in the feature space within adjacent time windows based on the evolution trajectory, and mapping the consistency between the pressure change rate and the change direction to the pressure state potential energy gradient to obtain the pressure state potential energy evolution features; and outputting the cumulative effect of the flow rate change intensity within multiple time windows based on the continuous changes of flow rate-related features in the operating state evolution trajectory, and mapping the temporal accumulation result of the flow rate change intensity to the cumulative amount of flow state potential energy. The process involves obtaining the accumulated potential energy characteristics of the flow state; analyzing the change amplitude of the operating state in the feature space before and after valve action based on the correspondence between the valve sub-state model and the operating state profile, and mapping the change amplitude to the valve regulation potential energy release amount to obtain the valve regulation potential energy evolution characteristics; constructing the state transition sequence of the gas pipeline operating state within a continuous time window based on the operating state label corresponding to the operating state profile; analyzing the transition relationship between adjacent operating states in the state transition sequence and outputting the corresponding state transition information; outputting the response lag time between the occurrence of pressure and flow anomalies and the triggering of valve action based on the time alignment relationship between the operating state profile and the valve sub-state model; outputting the operating state regulation coefficient through a weighted method based on the response lag time, combined with the valve action frequency, action amplitude, and duration; and extracting the operating state evolution efficiency characteristics by associating the regulation cost coefficient with the corresponding pressure change intensity and flow change intensity.

[0012] The evolution characteristics of pressure state potential energy, the accumulation characteristics of flow state potential energy, the evolution characteristics of valve regulation potential energy, the amount of state transition information, and the efficiency characteristics of operating state evolution are collected to form a multi-domain evolution feature set. The multi-domain evolution feature set is then uniformly scaled using a nonlinear scaling compression function and ordered according to a preset feature sequence to generate a gas pipeline operating state feature vector.

[0013] In a preferred technical solution, the step of classifying the operating state according to the operating state feature vector is as follows: The operating state feature vector is used as the input set of features to be determined; based on the input set of features to be determined, the value distribution of each feature under different levels is statistically analyzed to form a feature combination interval; within the feature combination interval, the pressure change rate feature and the flow change intensity feature are set as the main constraint features, and the valve action response evolution feature is set as the feedback constraint feature; a multi-dimensional feature joint constraint rule is established based on the main constraint feature and the feedback constraint feature; the input set of features to be determined is aligned feature-by-feature with the feature combination intervals corresponding to each operating state level, and the results of each evolution feature in the input set of features to be determined falling into the same level feature combination interval are selected; the operating state level that satisfies the multi-dimensional joint constraint rule is determined as the initial operating state level of the current time window; based on the confirmed initial operating state level, the distribution of the operating state feature vector of the corresponding level within the historical control cycle is statistically analyzed, and the feature value range that triggers the actual action of the safety valve is used as the feature distribution boundary; based on the feature distribution boundary, an offset correction is performed on the current level boundary interval.

[0014] In a preferred technical solution, the offset correction of the current level boundary interval based on the feature distribution boundary is performed as follows: The feature distribution boundaries under the same operating state level are statistically analyzed, and feature intervals in which each evolutionary feature appears continuously within the level and does not cause level changes are extracted to obtain stable feature boundary segments; all stable feature boundary segments are combined to form a historical reference boundary set; the feature distribution boundaries and the historical reference boundary set are aligned feature by feature to obtain the offset direction and offset amount of the current feature interval relative to the historical stable boundary; the correction direction of each evolutionary feature is determined based on the offset direction, offset amount, and valve execution feedback data; when the execution feedback data is stable and the offset amount is within a set threshold range, the offset direction is corrected along the current offset amount; based on the confirmed correction direction, the stable feature boundary segments in the reference boundary set are used as constraints to perform restricted offset processing on the current level boundary interval; the feature boundaries after constraint correction are recombined to output the dynamic level boundary interval.

[0015] In a preferred technical solution, the step of driving the safety valve to perform opening and closing actions according to the safety valve control decision command, and verifying the consistency between the actual valve action state and the control command to generate valve execution result data, specifically includes: executing actions based on the safety valve control decision command and sending a valve control signal; aligning the actual opening and closing state with the target opening and closing state in time; calculating the difference between the actual opening and closing state and the target opening and closing state to obtain the execution final value deviation; analyzing the changing trends of the actual opening and closing state curve and the target opening and closing state to identify the actions in the valve execution process and generate a deviation identifier; judging based on the deviation identifier, when any deviation exceeds a preset threshold, the valve execution state within the control cycle is judged as deviation execution; statistically analyzing the frequency and duration distribution of deviations in the deviation execution judgment results using a sliding window method, and outputting the valve execution consistency characteristics.

[0016] In a preferred technical solution, the step of statistically analyzing the frequency and duration distribution of deviations in the deviation execution judgment results using a sliding window method and outputting valve execution consistency characteristics is as follows: A deviation signal is constructed based on the difference between the valve control command and the actual execution feedback; based on the analysis of the deviation signal, the positions where the deviation signal exceeds a preset change in the time dimension are identified, forming multiple time segments with relatively stable execution characteristics; statistical descriptive information of each time segment is extracted and combined to form segment-level execution characteristics; based on the segment-level execution characteristics, the valve execution state of each time segment is identified, the valve execution behavior is abstracted into a finite set of states, and arranged in chronological order to form an execution state sequence; the execution state sequence is analyzed, and adjacent time segments continuously in a deviation execution state or abnormal execution state are merged to form a unique... The system identifies and records the start time, end time, and duration of each deviation execution event. It statistically analyzes the time intervals between deviation execution events and adjacent deviation execution events within the control cycle, the percentage of cumulative duration, and the distribution characteristics of the duration of each deviation execution event. The time intervals, percentage of cumulative duration, and distribution characteristics are categorized into two types of quantitative features: one is the deviation execution percentage and persistence, and the other is the deviation execution frequency and clustering characteristics. When the deviation execution percentage is low, the duration is short, and the event interval distribution is stable, the valve execution is considered to have high consistency. When the deviation execution percentage increases, the duration lengthens, or the event interval shows a convergent trend, the valve execution consistency is considered to decrease. Based on the judgment results, a comprehensive evaluation result of the valve execution consistency is formed, and valve execution result data is generated.

[0017] In a preferred technical solution, the step of forming a comprehensive evaluation result of valve execution consistency based on the judgment result and generating valve execution result data is as follows: Based on the comprehensive evaluation result of valve execution consistency, a corresponding execution consistency score is generated for each control cycle, and the score is stored together with the control cycle identifier to form an execution consistency history sequence arranged in chronological order; based on the execution consistency score of the current control cycle and its relative position in the execution consistency history sequence, an execution status classification result is generated; the execution status classification result is matched with the execution final value deviation, deviation identifier, and execution duration deviation within the current control cycle; after completing the execution status matching, the safety valve control decision command, actual valve execution status, execution consistency score, execution status classification result, and corresponding deviation characteristics of the current control cycle are associated and stored to form a valve execution result record, and the valve execution result record is output as valve execution result data.

[0018] In a preferred technical solution, the process of writing back and updating the operational status feature construction based on valve execution result data is as follows: The valve execution result data is parsed, and the execution feedback corresponding to each control cycle is formed into an independent execution feedback data unit, which is then aligned with the valve action response evolution feature. Based on the execution status classification result, the execution result is marked in the valve action response evolution feature, transforming the valve action response evolution feature into valve response feature data with execution result labels. Based on the valve response feature data, adaptive weight updates are performed on the dimensions of the valve response feature data to form an execution feedback-aware operational status feature vector. The execution feedback-aware operational status feature vector is combined with the execution feedback data to adaptively correct the time window length of the current operational status feature.

[0019] In a preferred technical solution, the step of combining the execution feedback-aware operating state feature vector with execution feedback data to adaptively correct the time window length of the current operating state features is as follows: Based on the distribution of execution state labels in the execution feedback-aware operating state feature vector, and combined with the valve execution state classification results recorded in the execution feedback data unit, the stability of the operating state evolution within the time window is obtained; based on the stability of the operating state evolution within the time window, the execution state labels of each feature dimension are statistically analyzed to obtain the execution state distribution; the time window length is determined and corrected based on the execution state distribution; and the continuity and fluctuation trend of the operating state feature evolution within the time window are judged based on the stability of the operating state evolution within the time window. The execution state distribution is obtained. Based on the execution state distribution, when more than a preset number of dimensions remain in a stable execution state label, it is determined that the evolution of the running state features is highly continuous, and the existing time window length is maintained. The corrected time window is bound to the execution feedback-aware running state feature vector as input data for the construction of the running state features in the next control cycle. The running state feature vectors before and after the time window correction are version-identified, labeled with the corresponding control cycle index and valve execution state classification results, and stored in chronological order to form a running state feature evolution record set. Based on the running state feature evolution record set, the execution feedback-aware running state feature vector of the current version and the corrected time window are output to complete the construction and update of the running state features.

[0020] A pipeline gas safety valve control system includes an operating state feature construction module, an operating state level classification and decision-making module, a safety valve execution and consistency verification module, and a feature write-back and update module, with connections between the modules. The operating state feature construction module acquires operating data of the gas pipeline, constructs a profile of the gas pipeline's operating state, and generates an operating state feature vector. The operating state level classification and decision-making module classifies the operating state of the gas pipeline according to the operating state feature vector and outputs safety valve control decision commands. The safety valve execution and consistency verification module drives the safety valve to perform opening and closing actions according to the safety valve control decision commands, verifies the consistency between the actual valve action state and the control commands, and generates valve execution result data. The feature write-back and update module updates the operating state feature construction process by writing back the valve execution result data.

[0021] The technical effects and advantages of the pipeline gas safety valve control method and control system of the present invention are as follows:

[0022] 1. This invention acquires multi-dimensional operational data of gas pipelines and constructs an operational status profile. It introduces time evolution modeling to jointly analyze pressure, flow rate, and valve sub-states within a continuous time window, forming an operational status evolution trajectory. By extracting the potential energy evolution characteristics of pressure, the potential energy accumulation characteristics of flow rate, and the potential energy evolution characteristics of valve regulation, it enables early characterization of pipeline operational status changes and potential risks. This effectively avoids misjudgments caused by relying solely on instantaneous data, significantly improving the comprehensiveness and foresight of operational status identification.

[0023] 2. This invention drives the gas safety valve to perform corresponding opening and closing actions based on the safety valve control decision command, and verifies the consistency between the actual valve action state and the control command during the execution process, generating corresponding valve execution result data. This enables timely identification of deviations between the control command and the execution result. Furthermore, the valve execution result data is written back to the operation state feature construction process to update the subsequent operation state profile and evolution feature extraction, forming an effective closed-loop feedback mechanism in the control process. This effectively reduces control errors and significantly improves the reliability and execution consistency of the gas safety valve control. Attached Figure Description

[0024] Figure 1 This is a schematic flowchart of a pipeline gas safety valve control method according to the present invention.

[0025] Figure 2 This is a schematic diagram of the control system structure of a pipeline gas safety valve control method according to the present invention. Detailed Implementation

[0026] 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.

[0027] Example 1, Figure 1 The present invention provides a method for controlling a pipeline gas safety valve, comprising the following steps:

[0028] S1. Obtain the operation data of the gas pipeline, construct a profile of the gas pipeline operation status, extract evolution features, and generate an operation status feature vector.

[0029] In this embodiment, operational data of the gas pipeline is acquired to construct a profile of the gas pipeline's operational status, as detailed below:

[0030] The gas pipeline's operating data is collected through a data acquisition device. The operating data includes: internal operating pressure data, instantaneous gas flow data, and safety valve operation status data.

[0031] Physical boundary verification is performed on operating pressure data and instantaneous gas flow data based on valve action status data, and abnormal sampled values ​​that exceed the design allowable range are removed. Specifically, the operating pressure data and instantaneous gas flow data obtained at the same sampling time are compared with the corresponding allowable range. When the pressure value or flow value exceeds the physical reachable range defined by the valve action status, it is determined that the sampling point does not meet the physical constraints. The sampled values ​​that do not meet the physical boundary constraints are marked as abnormal sampled values ​​and removed from the subsequent feature construction process.

[0032] For missing data caused by short-term communication interruptions or instantaneous sensor failures, a piecewise linear interpolation algorithm is used to complete the data, forming a structured operational dataset that is continuous in time, consistent in mechanism, and unified in dimension.

[0033] Based on the structured operational dataset, a causal coupled operational status model is constructed, and an operational status profile of the gas pipeline is output.

[0034] In this embodiment, a causal coupled operational status model is constructed based on the structured operational dataset, and an operational status profile of the gas pipeline is output, as follows:

[0035] Within the same time window, sub-state description models are constructed for operating pressure data, instantaneous gas flow data, and safety valve action status data, respectively. The sub-state description models include pressure sub-state model, flow sub-state model, and valve sub-state model.

[0036] The average pressure, maximum pressure, minimum pressure, and pressure variation are calculated using the internal pressure data of the pipeline.

[0037] Based on the pressure fluctuation amplitude and rate of change, the pressure stability is calculated by a weighted method to reflect the smoothness of the pressure state. The pressure statistical characteristics, variation characteristics and stability characteristics are combined to form a pressure state characteristic set, and the pressure sub-state model is output.

[0038] Based on instantaneous gas flow data, the average flow, peak flow, and flow fluctuation range are calculated to characterize the basic transport load level of the pipeline.

[0039] Time series analysis is performed on the flow data to calculate the flow change period and frequency, which is used to identify stable or fluctuating transport states and to obtain the flow change rate and flow fluctuation amplitude.

[0040] Based on the rate of change of traffic and the amplitude of traffic fluctuation, traffic impact features are extracted. Traffic statistical features, change features and impact features are fused to output a traffic sub-state model. Among them, the traffic impact feature is calculated by jointly calculating the rate of change of traffic and the amplitude of traffic fluctuation. When both are in the preset high value range within the same time window, it is determined that there is obvious traffic impact behavior in the window, and the combined value of the weighted product of the two is used as the traffic impact feature.

[0041] State encoding is performed based on the valve action status data of the safety valve, transforming the valve opening, closing and regulation processes into discrete state sequences;

[0042] The valve action state data on the same time axis as the discrete state sequence are aligned with pressure and flow changes for analysis, and the response relationship of valve action to pressure release and flow adjustment is extracted.

[0043] Based on the changes in pressure and flow before and after the valve action is triggered, a set of valve response features is constructed, and a valve sub-state model is output.

[0044] The sub-state models are jointly modeled according to the correlation order of pressure state, flow state, and valve state to obtain a causal coupled operating state model. The pressure change rate is used as the input variable of the flow state model, the flow change trend is used as the driving variable of the valve state model, and the valve action frequency and opening and closing changes are used as feedback terms to write back into the pressure state model and the flow state model. This constructs a multi-sub-state directed coupling structure that includes forward causal action and feedback regulation action, forming a causal coupled operating state model.

[0045] Based on the causal coupling operation state model, the operation state profile of the gas pipeline is output, which is used to characterize the overall operation behavior of the gas pipeline within the time window.

[0046] In this embodiment, based on the causal coupling operation state model, a profile of the gas pipeline operation state is output, as follows:

[0047] Based on the causal coupling operation state model, the mapping relationship between the output pressure change rate and the flow fluctuation amplitude, as well as the response matching degree between the flow change trend and the valve action frequency, are combined to form a cross-substate coupling constraint matrix to describe the mutual constraints between substates. The mapping relationship is used to characterize the constraint strength of the upstream physical state on the amplitude change of the substate variable, and the response matching degree is used to characterize the consistency of the control action in following the substate change trend. Then, using each substate as the matrix row and column index, the constraint coefficient of the mapping relationship is filled into the off-diagonal element positions between the corresponding physical substates, and the response matching degree is filled into the corresponding positions as the weight term between the control-related substates. Finally, a cross-substate coupling constraint matrix that simultaneously contains physical causal constraints and control response constraints is formed.

[0048] Using the cross-substate coupling constraint matrix as a constraint condition, the features of each substate are mapped to the same dimension embedding space, and the feature dimensions are adaptively weighted according to the coupling strength to form a multi-dimensional feature embedding space.

[0049] Based on the multidimensional feature embedding space and combined with the running result data, a set of mapping rules is formed by the DBSCAN clustering method. DBSCAN clusters the embedding vectors and maps each cluster center or cluster range to the running state type, forming a mapping rule between the cluster center and the state type. Each rule corresponds to the vector range of a specific cluster.

[0050] The multidimensional features generated in the current time window are embedded into the vector input mapping rule set, and the running status labels and their feature distribution contours are matched to generate a gas pipeline running status profile that represents the overall running behavior of the gas pipeline within the time window.

[0051] In this embodiment, evolutionary features are extracted to generate a running state feature vector, as detailed below:

[0052] The gas pipeline operation status profiles obtained within multiple consecutive preset time windows are arranged in chronological order to form a time evolution sequence of the operation status profiles;

[0053] The multi-dimensional feature embedding vectors corresponding to the running status profile are mapped to the same feature space, and the evolution trajectory of the running status in the feature space is generated based on the change relationship of the multi-dimensional feature embedding vectors within adjacent time windows.

[0054] Based on the evolution trajectory, the direction and magnitude of change of pressure-related features in the feature space within adjacent time windows are output, and the consistency between the pressure change rate and the change direction is mapped to the pressure state potential energy gradient, which is used to characterize the driving force intensity of the pressure state evolving into a high-risk or low-risk region, thus obtaining the pressure state potential energy evolution features. The pressure-related features include pressure fluctuation amplitude, pressure change rate, and change direction.

[0055] Based on the continuous changes in flow-related characteristics in the operational state evolution trajectory, the cumulative effect of flow change intensity within multiple time windows is output, and the time accumulation result of flow change intensity is mapped to the flow state potential energy accumulation, which is used to characterize the cumulative impact of continuous flow fluctuations on the stability of system operational state, thus obtaining the flow state potential energy accumulation characteristics.

[0056] Based on the correspondence between the valve sub-state model and the operating state profile, the change range of the operating state in the feature space before and after the valve action is analyzed, and the change range is mapped to the valve regulation potential energy release amount, which is used to characterize the system's ability to suppress or regulate the evolution of the operating state through valve action, and obtain the valve regulation potential energy evolution characteristics.

[0057] Based on the operating status labels corresponding to the operating status profiles, construct the state transition sequence of the gas pipeline operating status within a continuous time window;

[0058] Analyze the transition relationship between adjacent running states in the state transition sequence and output the corresponding state transition information, which is used to quantify the information increment carried when the running state undergoes a sudden change or mode reconstruction.

[0059] Based on the time alignment relationship between the operational status profile and the valve sub-state model, the response lag time between the occurrence of abnormal pressure and flow and the triggering of valve action is output.

[0060] Based on the response lag time, combined with the valve action frequency, action amplitude and duration, the operating status adjustment coefficient is output by weighting method;

[0061] By correlating the adjustment cost coefficient with the corresponding pressure change intensity and flow change intensity, the operating state evolution efficiency characteristics are extracted.

[0062] The evolution characteristics of pressure state potential energy, the accumulation characteristics of flow state potential energy, the evolution characteristics of valve regulation potential energy, the amount of state transition information, and the efficiency characteristics of operating state evolution are collected to form a multi-domain evolution characteristic set.

[0063] By using a nonlinear scaling function (hyperbolic tangent scaling function or sigmoid scaling mapping function) to perform a unified scaling mapping on the multi-domain evolutionary feature set, the drastic evolutionary features can gain enhanced discriminative power in the feature space, while suppressing redundant expression of stationary evolutionary features.

[0064] The multi-domain evolution features mapped to a unified scale are combined in an orderly manner according to a preset feature order to generate a gas pipeline operation status feature vector that has the meaning of time evolution mechanism and the ability to distinguish operation status.

[0065] S2, based on the operating status feature vector, classifies the operating status into levels, dynamically corrects the classification threshold based on historical operating data, and outputs the corresponding safety valve control decision command.

[0066] In this embodiment, the operating status is classified into levels based on the operating status feature vector, as follows:

[0067] Use the running state feature vector as the input set of features to be determined;

[0068] Based on the input set of features to be determined, the value distribution of each feature under different levels is statistically analyzed, and the value range of each feature under each level is combined to form the feature combination interval corresponding to each operating state level.

[0069] Within the feature combination range, the pressure change rate feature and the flow rate change intensity feature are set as the main constraint features to characterize pipeline load changes, and the valve action response evolution feature is set as the feedback constraint feature to limit the control execution effect.

[0070] Establish multi-dimensional feature joint constraint rules based on the main constraint features and feedback constraint features;

[0071] Align the input set of features to be determined with the feature combination intervals corresponding to each operating state level feature by feature, and filter the results in which each evolution feature in the input set of features to be determined falls into the same level feature combination interval.

[0072] The operational status level that satisfies the multidimensional joint constraint rules is determined as the initial operational status level for the current time window;

[0073] Based on the initial operating status level, extract the valve historical execution records corresponding to that level, and perform statistical analysis on the execution deviation amplitude and execution stability in the valve execution records. Execution stability is obtained by statistically analyzing the fluctuation of valve actions within a continuous time window, such as the rate of change of valve opening and closing frequency and the rate of change of action amplitude.

[0074] When the execution deviation is within the allowable range and the execution stability index reaches the set threshold, the initial operating status level is confirmed.

[0075] When the execution deviation exceeds the allowable range or the execution stability index falls below the set threshold, the confidence level of the initial operating state level is reduced.

[0076] Based on the confirmed initial operating status level, the distribution of the corresponding operating status feature vector within the historical control cycle is statistically analyzed, and the range of feature values ​​that trigger the actual action of the safety valve is used as the boundary of the feature distribution.

[0077] Based on the feature distribution boundary, perform offset correction on the current level boundary interval.

[0078] It should be noted that when the number of historical valve execution records for the initial operation of the system or the corresponding operating status level does not reach the preset statistical requirements, the level confirmation process based on the valve historical execution records will not be introduced for the time being. Instead, the initial operating status level will be determined directly based on the main constraint features and multi-dimensional joint constraint rules, and the execution deviation and execution stability analysis will be gradually enabled as historical execution data accumulates in subsequent control cycles.

[0079] In this embodiment, based on the feature distribution boundary, an offset correction is performed on the current level boundary interval, as follows:

[0080] Statistical analysis is performed on the feature distribution boundaries under the same operating state level. Feature intervals in which each evolutionary feature appears continuously within the level and does not cause level changes are extracted, and these feature intervals are marked as stable feature boundary segments.

[0081] All stable feature boundary fragments are combined into a historical reference boundary set, which is used to constrain the adjustable range of the level boundary;

[0082] Align the feature distribution boundary with the historical reference boundary set feature by feature to obtain the offset direction and offset amount of the current feature interval relative to the historical stable boundary;

[0083] Based on the offset direction, offset amount, and valve execution feedback data, the correction direction for each evolution feature is determined. The execution feedback data includes execution deviation, execution stability, and response matching degree.

[0084] When the execution feedback data is stable and the offset is within the set threshold range, the offset direction is corrected by following the current offset.

[0085] When the execution feedback data is stable but the offset exceeds the threshold or the execution deviation increases, the offset direction is limited by the historical stability boundary, and only convergent adjustments are allowed within the boundary.

[0086] When the execution feedback data is below the threshold, the offset is prohibited from exceeding the historical stable boundary, and the original level boundary shape is maintained first.

[0087] Based on the confirmed correction direction, the stable feature boundary segments in the reference boundary set are used as constraints to perform restricted offset processing on the current level boundary interval. The offset processing does not exceed the outer range of the historical stable boundary, and the original level boundary shape is maintained first when the execution feedback is unstable, so as to prevent the level boundary from changing abruptly due to a single or short-term anomaly.

[0088] The feature boundaries after constraint correction are recombined to output the dynamic level boundary interval.

[0089] In this embodiment, the corresponding safety valve control decision command is output as follows:

[0090] By mapping the running state feature vector to the dynamically corrected level boundary interval, the level affiliation of each feature is determined, and a set of candidate control actions for the corresponding level is established.

[0091] The candidate control action set is weighted based on the execution feedback data;

[0092] When the valve performs stably and continuously at the same level, increase the priority of this action in the selection process;

[0093] When the valve continuously performs deviation or abnormal deviation, the weight of the corresponding action is reduced, restricting its selection.

[0094] By weighting the data, an optimal set of actions is formed that conforms to hierarchical constraints and takes into account historical consistency, thus ensuring the reliability and feasibility of the control strategy.

[0095] The action path with the best historical execution consistency is selected from the weighted candidate action set and encoded according to the safety valve control protocol to generate standardized control decision instructions.

[0096] S3, drive the safety valve to perform opening and closing actions according to the safety valve control decision command, and verify the consistency between the actual valve action state and the control command, and generate valve execution result data;

[0097] In this embodiment, the safety valve is driven to perform opening and closing actions according to the safety valve control decision command, and the consistency between the actual valve action state and the control command is verified to generate valve execution result data, as follows:

[0098] Based on the safety valve control decision command, the valve control signal is sent to execute actions, including: driving the valve actuator to adjust the valve state to the target open / closed state, recording the valve execution start time, action duration and completion time, forming the valve action original log, and synchronously sampling the current open / closed state collected by the valve real-time sensor with the target state.

[0099] The actual opening and closing state is time-aligned with the target opening and closing state so that the two can be compared under a unified time reference.

[0100] The difference between the actual opening and closing state and the target opening and closing state is calculated to obtain the execution final value deviation, which is used to characterize the degree of deviation between the valve's final state and the target state of the control command.

[0101] Analyze the actual opening and closing state curves and the changing trends of the target opening and closing state, identify the actions (reverse actions, stagnant actions, or overshoot actions) during valve execution, and generate deviation indicators.

[0102] The determination is based on the deviation indicator. When any deviation exceeds the preset threshold, the valve execution status within the control cycle is determined to be deviation execution.

[0103] The frequency and duration of deviations in the deviation judgment results are statistically analyzed using a sliding window method, and the valve execution consistency characteristics are calculated.

[0104] In this embodiment, the actual opening and closing state curves and the target opening and closing trend are analyzed to identify the actions (reverse actions, stagnant actions, or overshoot actions) during valve execution, and deviation indicators are generated, as follows:

[0105] Align the actual opening and closing state curve of the valve with the target opening and closing trend point by point in time, and calculate the deviation between the actual opening and closing and the target opening and closing at each point in time.

[0106] Identify abnormal behaviors (reverse actions, stagnant actions, or overshooting actions) based on deviation values ​​and their changing trends.

[0107] If the deviation sign changes repeatedly within a short period of time and exceeds the set threshold, it is determined to be a reverse action;

[0108] If the deviation remains close to zero for an extended period and the target opening / closing changes significantly, it is determined to be a stalled action.

[0109] If the actual opening and closing exceeds the target opening and closing setting value by a certain range and then reverts, it is judged as an overshoot action;

[0110] The time interval corresponding to each abnormal behavior is marked as a deviation event, and a deviation identification signal is generated for subsequent analysis.

[0111] In this embodiment, the frequency and duration distribution of deviations in the deviation judgment results are statistically analyzed using a sliding window method, and the valve execution consistency characteristics are calculated, as follows:

[0112] Based on the difference between valve control commands and actual execution feedback, a deviation signal is constructed to reflect the valve execution behavior, and the deviation signal is continuously monitored.

[0113] Based on the analysis of the deviation signal, the locations where the deviation signal changes significantly in the time dimension are identified, thereby dividing the continuous operation process into multiple time segments with relatively stable execution characteristics. The analysis of the deviation signal includes the deviation amplitude, the degree of fluctuation, and the trend of change.

[0114] For each time interval, statistical descriptive information that can characterize the valve's execution characteristics is extracted and combined to form interval-level execution characteristics. The statistical descriptive information includes the overall level of deviation within the interval, fluctuation stability, extreme deviation characteristics, change trend, and interval duration.

[0115] Based on the segment-level execution characteristics, the valve execution status of each time segment is identified, and the valve execution behavior is abstracted into a finite set of states including normal execution status and deviation execution status, forming a sequence of execution states arranged in chronological order;

[0116] The execution state sequence is analyzed, and adjacent time segments that are continuously in a deviated execution state or an abnormal execution state are merged to form independent deviated execution events. The start time, end time and duration of each deviated execution event are recorded.

[0117] The frequency and persistence characteristics of deviation execution behavior are characterized by the time interval between deviation execution events and adjacent deviation execution events within the statistical control period, the proportion of cumulative duration, and the distribution characteristics of the duration of each deviation execution event.

[0118] Based on the time interval, the proportion of cumulative duration, and the distribution characteristics of duration, two types of quantitative characteristics are formed. One type is the proportion of deviation execution and the persistence characteristics. By calculating the ratio of the cumulative duration of deviation execution events to the total duration of the control cycle, as well as the mean and dispersion of the duration, it is used to reflect the intensity of the impact of deviation behavior on the overall execution process. The other type is the frequency of deviation execution and the clustering characteristics. By statistically analyzing the number of occurrences of deviation execution events per unit time and the changing trend of event intervals, it is used to determine whether deviation behavior is concentrated or continuously evolving.

[0119] When the percentage of deviations is low, the duration is short, and the event intervals are stable, the valve execution is considered to have high consistency.

[0120] When the percentage of deviations increases, the duration lengthens, or the event intervals show a convergence trend, it is determined that the valve's execution consistency has decreased.

[0121] Based on the judgment results, a comprehensive evaluation result of valve execution consistency is formed, and valve execution result data is generated.

[0122] In this embodiment, based on the judgment result, a comprehensive evaluation result of valve execution consistency is formed, and valve execution result data is generated, as follows:

[0123] Based on the comprehensive evaluation results of valve execution consistency, a corresponding execution consistency score is generated for each control cycle, and the score and control cycle identifier are stored together to form a historical sequence of execution consistency arranged in chronological order.

[0124] Based on the execution consistency score of the current control cycle and its relative position in the execution consistency history sequence, the valve execution performance is classified into states, and execution state classification results are generated.

[0125] When the execution consistency score of multiple consecutive control cycles is within the preset stable range, the current cycle is determined to be in a stable execution state.

[0126] When the consistency score deviates briefly from the stable range but does not occur continuously, the current period is judged as an acceptable deviation execution state.

[0127] When the consistency score falls into the abnormal range for multiple consecutive control cycles, the current cycle is determined to be an abnormal deviation execution state.

[0128] The execution status classification results are matched with the execution final value deviation, deviation identifier, and execution duration deviation within the current control cycle. If the execution final value deviation and deviation identifier appear only in a single control cycle and do not form a continuous record in the historical sequence, they are judged as transient disturbance characteristics. If the execution final value deviation appears repeatedly in multiple adjacent control cycles, it is judged as a persistent deviation characteristic, and the execution status classification results are corrected accordingly.

[0129] After the execution status matching is completed, the safety valve control decision command of the current control cycle, the actual valve execution status, the execution consistency score, the execution status classification result and the corresponding deviation characteristics are associated and stored to form a structured valve execution result record, and the valve execution result record is output as valve execution result data.

[0130] S4, Write back the valve execution result data and update the operation status feature construction process;

[0131] In this embodiment, the process of updating the operational status feature construction is based on the write-back of valve execution result data, as detailed below:

[0132] By analyzing the valve execution result data, the execution feedback corresponding to each control cycle is formed into an independent execution feedback data unit, and aligned with the valve action response evolution characteristics;

[0133] Based on the classification results of the execution status, the execution results are marked in the valve action response evolution characteristics to identify the execution status type corresponding to the response evolution characteristics, so that the valve action response evolution characteristics are transformed into valve response characteristic data with execution result labels;

[0134] Based on valve response feature data, adaptive weight updates are performed on the dimensions of valve response feature data to form an execution feedback-aware operating state feature vector.

[0135] By combining the execution feedback-aware runtime status feature vector with the execution feedback data, the time window length of the current runtime status feature is adaptively adjusted.

[0136] In this embodiment, based on the valve response feature data, an adaptive weight update is performed on the dimensions of the valve response feature data to form an execution feedback-aware operating state feature vector, as detailed below:

[0137] The valve response feature data is assigned initial weights, and the weighted feature vector is used as the state input, the feature weight adjustment is used as the action output, and the execution result label is used as the reward function, thereby training the system to learn the optimal feature weight adjustment strategy.

[0138] For the new execution feedback data received, the prediction error is calculated using the recursive least squares method, and the weights of each feature are dynamically adjusted according to the prediction error to make the weighted feature vector fit the actual execution result. The valve operating condition is also perceived. Specifically, after receiving new execution feedback data, the valve execution result predicted by the added operating status feature vector is compared with the actual execution result, and the difference is calculated as the prediction error.

[0139] When the valve operating conditions change, the automatic feature adjustment method is used. The new operating conditions are taken as the task input, and the initial weights are quickly updated using a small amount of new data, so that the system can quickly adapt to the new operating conditions.

[0140] By continuously iterating the above process, reinforcement learning continuously optimizes the feature weight strategy, RLS maintains real-time fine-tuning, and automatic feature adjustment methods ensure rapid adaptation to new operating conditions, thereby forming a closed-loop adaptive capability and outputting a stable, feedback-aware operating state feature vector.

[0141] In this embodiment, the execution feedback-aware runtime state feature vector is combined with execution feedback data to adaptively adjust the time window length of the current runtime state feature, as follows:

[0142] Based on the distribution of execution state labels in the execution feedback-aware operation state feature vector, and combined with the valve execution state classification results recorded in the execution feedback data unit, the overall stability of the operation state characteristics within the current time window is comprehensively characterized to obtain the time window operation state evolution stability.

[0143] Based on the stability of the operational state evolution within the time window, the execution state labels of each feature dimension within the time window are statistically analyzed to obtain the execution state distribution. The stability of the operational state evolution is an intermediate representation result obtained based on the execution feedback-aware operational state feature vector within the current time window and the valve execution state classification results recorded in the execution feedback data unit. It is obtained by statistically analyzing the distribution of execution state labels corresponding to each feature dimension within the time window, comprehensively reflecting the proportion of stable execution labels, the concentration of deviation execution labels, and the continuity of state labels in the time dimension. It is used to quantitatively describe the stability or fluctuation of the operational state characteristics within the time window.

[0144] The time window length is determined and adjusted based on the execution status distribution;

[0145] When the number of feature dimensions corresponding to the stable execution status label exceeds the preset threshold and remains continuously distributed within the time window, it is determined that the evolution of the running status features is highly continuous, and the existing time window length is maintained.

[0146] When the labels of deviation execution or abnormal deviation execution are concentrated and exceed the preset ratio, it is determined that the fluctuation of the evolution of the operating status characteristics has increased, and the length of the time window is adjusted in a convergent manner.

[0147] The corrected time window is bound to the execution feedback-aware running state feature vector and used as input data for the construction of the running state feature of the next control cycle, so that the time window length determination result can directly participate in the subsequent feature construction.

[0148] Version identification is performed on the operating status feature vectors before and after the time window correction, the corresponding control cycle index and valve execution status classification results are marked, and they are stored in chronological order to form an operating status feature evolution record set;

[0149] Based on the evolution record set of running state features, output the execution feedback-aware running state feature vector of the current version and the corrected time window to complete the construction and update of running state features.

[0150] Example 2, Figure 2 The present invention provides a system for a pipeline gas safety valve control method, comprising an operating state feature construction module, an operating state level classification and decision-making module, a safety valve execution and consistency verification module, and a feature write-back and update module, with connections between the modules;

[0151] The operation status feature construction module is used to acquire the operation data of gas pipelines, construct the operation status profile of gas pipelines, and generate operation status feature vectors.

[0152] The operation status classification and decision module is used to classify the operation status of gas pipelines according to the operation status feature vector and output safety valve control decision commands.

[0153] The safety valve execution and consistency verification module is used to drive the safety valve to perform opening and closing actions according to the safety valve control decision command, and to verify the consistency between the actual valve action state and the control command, and generate valve execution result data;

[0154] The feature write-back and update module is used to write back and update the operation status feature construction process based on the valve execution result data.

[0155] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0156] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0157] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0158] The above description is merely a specific technical solution of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0159] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for controlling a pipeline gas safety valve, characterized in that, include: Obtain operational data of gas pipelines, construct a profile of the gas pipeline's operational status, extract evolutionary features, and generate operational status feature vectors; Based on the operating status feature vector, the operating status is classified into levels, and the classification threshold is dynamically corrected based on historical operating data, and the corresponding safety valve control decision command is output. The safety valve is driven to perform opening and closing actions according to the safety valve control decision command, and the consistency between the actual valve action state and the control command is verified to generate valve execution result data. Write back the valve execution result data and update the operation status feature construction process; The operation status is classified into levels based on the operation status feature vector, as follows: Use the running state feature vector as the input set of features to be determined; Based on the input set of features to be determined, the value distribution of each feature under different level states is statistically analyzed to form feature combination intervals; Within the feature combination range, the pressure change rate feature and the flow change intensity feature are set as the main constraint features, and the valve action response evolution feature is set as the feedback constraint feature. Establish multi-dimensional feature joint constraint rules based on the main constraint features and feedback constraint features; Align the input set of features to be determined with the feature combination intervals corresponding to each operating state level feature by feature, and filter the results in which each evolution feature in the input set of features to be determined falls into the same level feature combination interval. The operational status level that satisfies the multidimensional joint constraint rules is determined as the initial operational status level for the current time window; Based on the confirmed initial operating status level, the distribution of the corresponding operating status feature vector within the historical control cycle is statistically analyzed, and the range of feature values ​​that trigger the actual action of the safety valve is used as the boundary of the feature distribution. Based on the feature distribution boundary, perform offset correction on the current level boundary interval; The offset correction is performed on the current level boundary interval based on the feature distribution boundary, as follows: By statistically analyzing the feature distribution boundaries under the same operating state level, and extracting the feature intervals in which each evolutionary feature appears continuously within the level without causing level changes, stable feature boundary segments are obtained. All stable feature boundary segments are combined into a historical reference boundary set. Align the feature distribution boundary with the historical reference boundary set feature by feature to obtain the offset direction and offset amount of the current feature interval relative to the historical stable boundary; The correction direction for each evolution feature is determined based on the offset direction, offset amount, and valve execution feedback data. When the execution feedback data is stable and the offset is within the set threshold range, the offset direction is corrected by following the current offset. Based on the confirmed correction direction, and using stable feature boundary segments in the reference boundary set as constraints, a restricted offset process is performed on the current level boundary interval. The feature boundaries after constraint correction are recombined to output the dynamic level boundary interval.

2. The method for controlling a pipeline gas safety valve according to claim 1, characterized in that, The extraction of evolutionary features to generate a running state feature vector is as follows: The gas pipeline operation status profiles obtained within multiple consecutive preset time windows are arranged in chronological order to form a time evolution sequence of the operation status profiles; The multi-dimensional feature embedding vectors corresponding to the running status profile are mapped to the same feature space, and the evolution trajectory of the running status in the feature space is generated based on the change relationship of the multi-dimensional feature embedding vectors within adjacent time windows. Based on the evolution trajectory, the direction and magnitude of change of pressure-related features in the feature space within adjacent time windows are output, and the consistency between the pressure change rate and the change direction is mapped to the pressure state potential energy gradient to obtain the pressure state potential energy evolution features. Based on the continuous changes in flow-related features in the operational state evolution trajectory, the cumulative effect of flow change intensity in multiple time windows is output, and the time accumulation result of flow change intensity is mapped to the flow state potential energy accumulation, thus obtaining the flow state potential energy accumulation feature. Based on the correspondence between the valve sub-state model and the operating state profile, the change range of the operating state in the feature space before and after the valve action is analyzed, and the change range is mapped to the valve regulation potential energy release amount to obtain the valve regulation potential energy evolution characteristics. Based on the operating status labels corresponding to the operating status profiles, construct the state transition sequence of the gas pipeline operating status within a continuous time window; Analyze the transition relationships between adjacent running states in the state transition sequence and output the corresponding state transition information. Based on the time alignment relationship between the operational status profile and the valve sub-state model, the response lag time between the occurrence of abnormal pressure and flow and the triggering of valve action is output. Based on the response lag time, combined with the valve action frequency, action amplitude and duration, the operating status adjustment coefficient is output by weighting method; By correlating the adjustment cost coefficient with the corresponding pressure change intensity and flow change intensity, the operating state evolution efficiency characteristics are extracted. The evolution characteristics of pressure state potential energy, the accumulation characteristics of flow state potential energy, the evolution characteristics of valve regulation potential energy, the amount of state transition information, and the efficiency characteristics of operating state evolution are collected to form a multi-domain evolution characteristic set. A nonlinear scaling compression function is used to perform a unified scaling mapping on the multi-domain evolution feature set, and then the features are combined in an ordered manner according to a preset feature order to generate a gas pipeline operation status feature vector.

3. A pipeline gas safety valve control method according to claim 1, characterized in that, The process involves driving the safety valve to perform opening and closing actions according to the safety valve control decision command, verifying the consistency between the actual valve action state and the control command, and generating valve execution result data, as detailed below: Based on the safety valve control decision command, execute the action and send the valve control signal; Time alignment is performed between the actual opening / closing state and the target opening / closing state; The difference between the actual open / closed state and the target open / closed state is calculated to obtain the execution final value deviation. Analyze the actual opening and closing state curves and the changing trends of the target opening and closing state, identify the actions of the valve during the execution process, and generate deviation indicators; The determination is based on the deviation indicator. When any deviation exceeds the preset threshold, the valve execution status within the control cycle is determined to be deviation execution. The frequency and duration distribution of deviations in the deviation judgment results are statistically analyzed using a sliding window method, and the valve execution consistency characteristics are output.

4. A pipeline gas safety valve control method according to claim 3, characterized in that, The method of statistically analyzing the frequency and duration distribution of deviations in the deviation judgment results using a sliding window approach, and outputting the valve execution consistency characteristics, is as follows: A deviation signal is constructed based on the difference between valve control commands and actual execution feedback; Based on the analysis of the deviation signal, the positions where the deviation signal changes beyond the preset time dimension are identified, forming multiple time segments with relatively stable execution characteristics; Extract statistical description information for each time segment and combine them to form segment-level execution features; Based on the segment-level execution characteristics, the valve execution status of each time segment is identified, the valve execution behavior is abstracted into a finite set of states, and arranged in chronological order to form an execution state sequence; The execution state sequence is analyzed, and adjacent time segments that are continuously in a deviated execution state or an abnormal execution state are merged to form independent deviated execution events. The start time, end time and duration of each deviated execution event are recorded. The time interval between deviation execution events and adjacent deviation execution events within the statistical control period, the percentage of cumulative duration, and the distribution characteristics of the duration of each deviation execution event; The time interval, cumulative duration percentage, and distribution characteristics are divided into two categories of quantitative characteristics: one is the percentage and persistence of deviation execution, and the other is the frequency and clustering characteristics of deviation execution. When the percentage of deviations is low, the duration is short, and the event intervals are stable, the valve execution is considered to have high consistency. When the percentage of deviations increases, the duration lengthens, or the event intervals show a convergence trend, it is determined that the valve's execution consistency has decreased. Based on the judgment results, a comprehensive evaluation result of valve execution consistency is formed, and valve execution result data is generated.

5. The method for controlling a pipeline gas safety valve according to claim 4, characterized in that, Based on the judgment results, a comprehensive evaluation result of valve execution consistency is formed, and valve execution result data is generated, as follows: Based on the comprehensive evaluation results of valve execution consistency, a corresponding execution consistency score is generated for each control cycle, and the score and control cycle identifier are stored together to form a historical sequence of execution consistency arranged in chronological order. Based on the execution consistency score of the current control cycle and its relative position in the execution consistency history sequence, an execution status classification result is generated; Match the execution status classification results with the execution final value deviation, deviation identifier, and execution duration deviation within the current control cycle; After the execution status matching is completed, the safety valve control decision command of the current control cycle, the actual valve execution status, the execution consistency score, the execution status classification result and the corresponding deviation characteristics are associated and stored to form a valve execution result record, and the valve execution result record is output as valve execution result data.

6. The method for controlling a pipeline gas safety valve according to claim 1, characterized in that, The process of writing back and updating the operational status feature construction based on valve execution result data is as follows: By analyzing the valve execution result data, the execution feedback corresponding to each control cycle is formed into an independent execution feedback data unit, and aligned with the valve action response evolution characteristics; Based on the classification results of execution status, the execution results are marked in the valve action response evolution characteristics, so that the valve action response evolution characteristics are transformed into valve response characteristic data with execution result labels; Based on valve response feature data, adaptive weight updates are performed on the dimensions of valve response feature data to form an execution feedback-aware operating state feature vector. By combining the execution feedback-aware runtime status feature vector with the execution feedback data, the time window length of the current runtime status feature is adaptively adjusted.

7. A method for controlling a pipeline gas safety valve according to claim 6, characterized in that, The step of combining the execution feedback-aware runtime state feature vector with execution feedback data to adaptively adjust the time window length of the current runtime state features is as follows: Based on the distribution of execution state labels in the execution feedback-aware operation state feature vector, and combined with the valve execution state classification results recorded in the execution feedback data unit, the time window operation state evolution stability is obtained. Based on the stability of the running state evolution within the time window, the execution state labels of each feature dimension within the time window are statistically analyzed to obtain the execution state distribution; The time window length is determined and adjusted based on the execution status distribution; The continuity and fluctuation trend of the evolution of the running state characteristics within the time window are judged based on the stability of the running state evolution within the time window, and the execution state distribution is obtained. Based on the execution state distribution, when more than a preset number of dimensions remain in a stable execution state label, it is determined that the evolution of the running state feature has strong continuity, and the existing time window length is maintained. The corrected time window is bound to the execution feedback-aware operating state feature vector and used as input data for constructing the operating state features of the next control cycle. Version identification is performed on the operating status feature vectors before and after the time window correction, the corresponding control cycle index and valve execution status classification results are marked, and they are stored in chronological order to form an operating status feature evolution record set; Based on the evolution record set of running state features, output the execution feedback-aware running state feature vector of the current version and the corrected time window to complete the construction and update of running state features.

8. A pipeline gas safety valve control system using the pipeline gas safety valve control method as described in any one of claims 1-7, characterized in that, It includes a module for constructing operational status features, a module for classifying and deciding operational status levels, a module for executing safety valves and verifying consistency, and a module for writing back and updating features. These modules are interconnected. The operation status feature construction module is used to acquire the operation data of gas pipelines, construct the operation status profile of gas pipelines, and generate operation status feature vectors. The operation status classification and decision module is used to classify the operation status of gas pipelines according to the operation status feature vector and output safety valve control decision commands. The safety valve execution and consistency verification module is used to drive the safety valve to perform opening and closing actions according to the safety valve control decision command, and to verify the consistency between the actual valve action state and the control command, and generate valve execution result data; The feature write-back and update module is used to write back and update the operation status feature construction process based on the valve execution result data.

Images