Natural gas station control system fault detection method and device

Abstract

The invention relates to the technical field of natural gas station control, and particularly provides a natural gas station control system fault detection method and device, and the method comprises the steps: collecting the multi-dimensional operation parameters of natural gas in a natural gas station control system in real time; the multi-dimensional operation parameters are preprocessed to generate feature vectors, the feature vectors are input into a reversible neural network model, a reconstruction error is output, a reversible neural network comprises a network architecture composed of a reversible transformation module, and the reversible transformation module is used for learning the multi-dimensional operation parameters of the natural gas station control system under the normal working condition; the reconstruction error is used for representing a multi-dimensional deviation degree between the multi-dimensional operation parameter and an operation parameter of reversible neural network learning under a normal working condition; and when the reconstruction error is greater than a reconstruction error threshold value, calculating a joint confidence coefficient based on the confidence rule base and the reconstruction error, and when the joint confidence coefficient exceeds a confidence coefficient threshold value, generating an interactive report containing a fault type, thereby realizing a purpose of performing fault diagnosis according to the multi-dimensional real-time operation data.

Description

Technical Field

[0001] This application relates to the technical field of natural gas station control, and particularly relates to a method and device for fault detection of a natural gas station control system. Background Art

[0002] As a key node in the urban gas transmission and distribution system, the control system of a natural gas station monitors multi-dimensional parameters such as pressure, temperature, and flow rate in real time to cope with faults such as stuck pressure regulating valves, fouling of heat exchangers, and low-temperature blockage of pipelines.

[0003] Existing fault detection technologies are mainly threshold-based alarm systems. However, due to the induction of faults such as stuck pressure regulating valves and fouling of heat exchangers, they are often the result of the abnormal synergistic action of multi-dimensional operating parameters. Relying solely on threshold detection will result in a large number of missed alarms and false alarms. For example, when a complex fault actually occurs, it may be missed due to the multi-dimensional operating parameters not reaching the threshold synchronously; when a certain operating parameter exceeds the limit accidentally under non-fault conditions, it will trigger a false alarm, resulting in a reduction in the accuracy and reliability of fault detection. Summary of the Invention

[0004] In view of the above problems, this application provides a method and device for fault detection of a natural gas station control system, which realizes the purpose of fault diagnosis based on multi-dimensional real-time operating data through a dual judgment mechanism of a reversible neural network and a belief rule base.

[0005] To achieve the purpose of the present invention, this application provides the following technical solutions: In a first aspect, this application provides a method for fault detection of a natural gas station control system, and the method includes the following steps: Real-time collect multi-dimensional operating parameters of natural gas in the natural gas station control system, where the multi-dimensional operating parameters include pressure values, temperature values, and flow rates; Preprocess the multi-dimensional operating parameters to generate feature vectors, where the preprocessing sequentially includes outlier removal, moving window mean filtering, and normalization; Input the feature vectors into a reversible neural network model to output a reconstruction error, where the reversible neural network includes a network architecture composed of reversible transformation modules, and the affine coupling layer and convolutional layer of the reversible transformation module are used to learn the multi-dimensional operating parameters of the natural gas station control system under normal conditions, and the reconstruction error is used to characterize the multi-dimensional deviation degree between the multi-dimensional operating parameters and the operating parameters learned by the reversible neural network under normal conditions; When the reconstruction error is greater than the reconstruction error threshold, calculate the combined confidence based on the belief rule base and the reconstruction error. When the combined confidence exceeds the confidence threshold, generate an interactive report including the fault type. The belief rule base is constructed based on the mechanism of natural gas station equipment and the fault event case base. The antecedents of the rules in the belief rule base include that the pressure volatility is greater than the first threshold, the temperature volatility is greater than the second threshold, and the flow rate volatility is greater than the third threshold. The consequents of the rules are associated with three types of faults: stuck pressure regulating valve, fouling of heat exchanger, and low-temperature blockage of pipeline. The corresponding relationship between the antecedents of the rules and the faults is as follows: the combination of abnormal pressure fluctuation and abnormal temperature fluctuation corresponds to a stuck pressure regulating valve, the combination of abnormal pressure fluctuation and abnormal flow rate fluctuation corresponds to a low-temperature blockage of the pipeline, and the combination of abnormal temperature fluctuation and abnormal flow rate fluctuation corresponds to fouling of the heat exchanger. The belief rule base also includes the initial rule confidence levels corresponding to the three types of faults, and the confidence threshold takes the initial rule confidence level corresponding to the corresponding fault.

[0006] In a possible implementation manner, the interactive report further includes a warning level, countermeasures, and the timing of implementing the countermeasures. After determining the fault type when the combined confidence exceeds the confidence threshold, the method further includes: Match countermeasures according to the fault type. Specifically, for the fault of a stuck pressure regulating valve, prioritize checking the wear condition of the valve core and performing lubrication measures; for the fault of fouling of the heat exchanger, match starting a chemical cleaning program and checking the gap between heat exchange fins; for the fault of low-temperature blockage of the pipeline, match increasing the power of the tracing system. When the combined confidence is greater than the confidence threshold and less than the first confidence threshold, increase the frequency of collecting the corresponding operating parameters; When the combined confidence is greater than or equal to the first confidence threshold and less than the second confidence threshold, trigger a yellow warning and implement the corresponding countermeasures during the next low-load period; When the combined confidence is greater than or equal to the second confidence threshold, trigger a red warning and implement the corresponding countermeasures.

[0007] In a possible implementation manner, the belief rule base is constructed based on the mechanism of natural gas station equipment and the fault event case base, and further includes: Label the operating parameters with pressure volatility greater than the first threshold, temperature volatility greater than the second threshold, and flow rate volatility greater than the third threshold in the fault event operation data as abnormal pressure fluctuation, abnormal temperature fluctuation, and abnormal flow rate fluctuation respectively, where the fault event operation data is extracted from the fault event case base of the natural gas station control system; According to the mechanism of natural gas station equipment and the proportion of abnormal fluctuations of each operating parameter, conduct expert scoring on abnormal pressure fluctuation, abnormal temperature fluctuation, and abnormal flow rate fluctuation to determine the weight coefficients corresponding to each operating parameter. Based on each of the weight coefficients, use the evidential reasoning algorithm to determine the initial rule confidence degrees of each fault.

[0008] In a possible implementation manner, using the weight coefficients and the evidential reasoning algorithm to determine the initial rule confidence degrees of each fault further includes: Respectively determine the confidence degrees of abnormal pressure fluctuations, abnormal temperature fluctuations, and abnormal flow fluctuations corresponding to each fault, and establish a fault evidence matrix; According to the fault evidence matrix and the weight coefficients, calculate the original rule confidence degrees of each fault by using the weighted summation method; Based on the equipment operating state and expert experience, adjust the original rule confidence degrees to obtain the value ranges of the initial rule confidence degrees of each fault.

[0009] In a possible implementation manner, calculating the combined confidence degree based on the confidence rule base and the reconstruction error further includes: When the volatility matches the corresponding rule antecedent, calculate the feature matching degree based on the initial rule confidence degree, the volatility, and the threshold value matching the volatility, where the volatility includes pressure volatility, temperature volatility, or flow volatility, and the threshold values corresponding to the volatility include the first threshold value, the second threshold value, or the third threshold value, and the feature matching degree μ i = initial rule confidence degree × min(volatility / corresponding threshold value, 1); Quantize the reconstruction error into a basic probability assignment function m(A) = 1 - exp(-λE), where E represents the reconstruction error, λ represents a preset fault sensitivity coefficient, and different values are set according to different faults of the natural gas station control system; Combine the basic probability assignment function m(A) with the feature matching degree μ i Perform weighted fusion to obtain a corrected probability assignment value m'(A) = α·m(A) + (1 - α)·μ i and calculate the combined confidence degree according to the Dempster combination rule:

[0010] where α represents an empirical coefficient, A and B represent any two of the three rule antecedents, C represents any one of the three types of faults, and K is a conflict factor, .

[0011] In a possible implementation manner, where A and B represent any two of the three rule antecedents, and C represents any one of the three types of faults, further includes: When A is abnormal pressure fluctuation and B is abnormal temperature fluctuation, C is the jam of the pressure regulating valve; when A is abnormal pressure fluctuation and B is abnormal flow fluctuation, C is the low-temperature blockage of the pipeline; when A is abnormal temperature fluctuation and B is abnormal flow fluctuation, C is the fouling of the heat exchanger, and the remaining combinations of A and B are classified into the conflict factor K.

[0012] In a possible implementation manner, the method further includes: If the combined confidence level is in [0.7β, β], input the feature vector and the equally-spaced sampling sequence of the multi-dimensional operating parameters within at least the most recent 12 hours into the LSTM prediction model, where β represents the confidence threshold; Perform multi-step prediction through the trained LSTM prediction model, and output the pressure prediction value, temperature prediction value, and flow prediction value at 15-minute intervals within at least the next 3 hours; If the pressure prediction value exceeds the pressure safety threshold, trigger a pre-maintenance alarm for the pressure regulating valve; if the temperature prediction value exceeds the temperature safety threshold, trigger an inspection alarm for the heat exchanger; if the flow prediction value exceeds the flow safety threshold, trigger an inspection alarm for pipeline blockage, where each of the safety thresholds is the pressure, temperature, and flow during a fault-free event at the natural gas station.

[0013] In a possible implementation manner, the training process of the LSTM prediction model includes: Perform interval sampling on the multi-dimensional operating parameters of the fault-free event case library for 30 consecutive days extracted from the natural gas station control system at 15-minute intervals to construct a training sample set; perform preprocessing on the training sample set, including removing outliers, performing mean filtering through a sliding window, and standardization; Configure the network structure, and use the Huber loss function for model training, where the network structure includes an input gate layer, a forget gate layer, and an output gate layer. The input gate layer is used to receive 48 time steps of 12-hour historical window data, and each time step includes 3 feature dimensions of pressure, temperature, and flow. The first forget gate layer includes 64 neurons, the second forget gate layer includes 64 neurons, and the output gate layer includes 3 dense units corresponding to the predictions of pressure, temperature, and flow respectively, and use the Huber loss function for model training.

[0014] In a second aspect, the present application provides a fault detection device for a natural gas station control system, and the device includes: A data acquisition module, configured to collect in real time the multi-dimensional operating parameters of natural gas in the natural gas station control system, where the multi-dimensional operating parameters include pressure values, temperature values, and flow rates; A preprocessing module, configured to perform preprocessing on the multi-dimensional operating parameters to generate a feature vector, where the preprocessing sequentially includes outlier removal, sliding window mean filtering, and standardization; An analysis and processing module for inputting the feature vector into a reversible neural network model and outputting a reconstruction error. The reversible neural network includes a network architecture composed of reversible transformation modules. The affine coupling layer and convolutional layer of the reversible transformation module are used to learn the multi-dimensional operating parameters of the natural gas station control system under normal operating conditions. The reconstruction error is used to characterize the multi-dimensional deviation degree between the multi-dimensional operating parameters and the operating parameters under normal operating conditions learned by the reversible neural network. A report generation module for calculating a joint confidence when the reconstruction error is greater than a reconstruction error threshold based on a belief rule base and the reconstruction error, and generating an interactive report including a fault type when the joint confidence exceeds a confidence threshold. The belief rule base is constructed based on the equipment mechanism of the natural gas station and a fault event case base. The antecedents of the rules in the belief rule base include that the pressure volatility is greater than a first threshold, the temperature volatility is greater than a second threshold, and the flow rate volatility is greater than a third threshold. The consequents of the rules are associated with three types of faults: pressure regulating valve jamming, heat exchanger fouling, and pipeline low-temperature blockage. The corresponding relationship between the antecedents of the rules and the faults is as follows: the combination of abnormal pressure fluctuation and abnormal temperature fluctuation corresponds to pressure regulating valve jamming, the combination of abnormal pressure fluctuation and abnormal flow rate fluctuation corresponds to pipeline low-temperature blockage, and the combination of abnormal temperature fluctuation and abnormal flow rate fluctuation corresponds to heat exchanger fouling. The belief rule base also includes the initial rule confidence levels corresponding to the three types of faults, and the confidence threshold takes the initial rule confidence level corresponding to the fault.

[0015] In a third aspect, the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, it executes the steps of the above-mentioned fault detection method for the natural gas station control system.

[0016] The embodiment of the present application provides a fault detection method for a natural gas station control system. By collecting multi-dimensional operating parameters such as pressure, temperature, flow rate, and valve opening in real time, after preprocessing, they are input into a reversible neural network model, and a reconstruction error that can characterize the multi-dimensional deviation degree between the current operating state and the normal operating conditions is output. When the reconstruction error exceeds the reconstruction error threshold, it is determined that there may be an abnormality in the current operating state, and then the belief rule base is activated for in-depth analysis. Further, the belief rule base adopts a parameter combination analysis strategy to perform multi-parameter joint analysis of pressure-temperature, pressure-flow rate, and temperature-flow rate to calculate the joint confidence. When the joint confidence exceeds the confidence threshold, an interactive report including the fault type is generated. The present application realizes accurate fault diagnosis of the real-time operating state through a dual judgment mechanism of a reversible neural network and a belief rule base. Description of the Drawings

[0017] The drawings are used to provide a further understanding of the present application and constitute a part of the specification. They are used to explain the present application together with the embodiments of the present application and do not constitute a limitation to the present application. Figure 1 Flow chart of the structure of a fault detection method for a natural gas station control system provided by an embodiment of the present application. Specific implementation manners

[0018] To make the objectives, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be clearly and completely described below with reference to the accompanying drawings in the present application. Apparently, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without making creative efforts shall fall within the protection scope of the present application.

[0019] The present invention proposes the following technical solutions and corresponding embodiments.

[0020] Embodiment 1 As Figure 1 shown in the flow chart of the structure of a fault detection method for a natural gas station control system provided by an embodiment of the present application, the embodiment of the present application provides a fault detection method for a natural gas station control system, including the following steps: S100: Real-time collect multi-dimensional operation parameters of natural gas in the natural gas station control system. The multi-dimensional operation parameters include pressure value, temperature value and natural gas flow rate. Among them, the pressure transmitter collects the pressure value, the temperature transmitter collects the temperature value, and the flow meter collects the natural gas flow rate.

[0021] S200: Preprocess the multi-dimensional operation parameters to generate feature vectors, where the preprocessing includes outlier rejection, moving window mean filtering and normalization in sequence.

[0022] Specifically, first use the 3σ criterion to reject outliers in the multi-dimensional operation parameters, and then perform mean filtering through a moving window; finally, the pressure parameter uses the designed pressure value of the station as the normalization reference, and the temperature and flow parameters are normalized based on the mean and standard deviation of the operation data in the recent period (for example, within the recent 30 days) of the natural gas station, and each operation parameter is converted into a dimensionless feature vector. When performing normalization, specifically, the pressure parameter uses the rated working pressure P design of the station design as the normalization reference, and calculate the normalized pressure P 1,norm = P1 / P design and the normalized pressure difference ΔP norm = ΔP / P design, where the pressure difference ΔP between the front and rear ends of the voltage regulator is ΔP = P1 - P2, P1 is the main pipeline pressure value measured in real time by the pressure transmitter, and P2 is the pressure after voltage regulation. This processing not only retains the engineering physical meaning of the pressure parameter but also makes the data of stations with different pressure levels comparable. The temperature parameter calculates the rolling mean μ and standard deviation σ based on the historical data of the last 30 days, and implements Z-score normalization to calculate the normalized temperature t 1,std = (t1 - μ) / σ, where t1 is the natural gas fluid temperature measured in real time by the temperature transmitter, and the rolling mean and standard deviation are updated in a timely manner to track the environmental temperature change. Flow normalization includes first calculating the normalized flow rate L according to the current flow rate and the maximum flow rate j , norm = L j / L max for linear normalization, and then combining the moving standard deviation σ of the flow parameter L to calculate the normalized flow difference ΔL j,std = ΔL j / σ L , where ΔL j = L j - L max , L j is the instantaneous volume flow rate measured by the flow meter, and L max is the maximum throughput designed for the pipeline. The finally generated feature vector is such as [P 1,norm , ΔP norm , t 1,std , L j , norm , ΔL j , std , …], and its dimension is consistent with the number of parameters, which not only meets the input requirements of the neural network but also embeds the state characteristics of the equipment operation through the parameter coupling calculation (such as the retention of the pressure-flow ratio) in the preprocessing process.

[0023] Therefore, the feature vector generated after the above three-level preprocessing not only retains the dynamic correlation characteristics of the original operating parameters but also eliminates the influence of measurement noise.

[0024] S300 inputs the feature vector into the reversible neural network model and outputs the reconstruction error. The reversible neural network includes a network architecture composed of reversible transformation modules. The affine coupling layer and convolutional layer of the reversible transformation module are used to learn the multi-dimensional operating parameters of the natural gas station control system under normal conditions, and the reconstruction error is used to characterize the multi-dimensional deviation degree between the multi-dimensional operating parameters and the operating parameters under normal conditions learned by the reversible neural network.

[0025] The reversible neural network model therein is a deep learning architecture based on a generative model, which is usually used for image synthesis, image reconstruction, and learning of data distribution. It converts complex input data into a simple probability distribution through reversible neural network layers and normalizing flow techniques, thereby achieving efficient generation and inference tasks.

[0026] It should be noted that the construction of the reversible neural network model and its analysis and processing methods include the following steps: First, design a network architecture composed of reversible transformation modules. Each reversible transformation module includes a first affine coupling layer and a reversible 1×1 convolutional layer, where the first affine coupling layer splits the input feature vector into a pressure parameter group, a temperature parameter group, and a flow parameter group.

[0027] Then, learn the non-linear mapping relationship between any two groups of parameters through a fully connected neural network. Then, perform a reversible shuffle on the feature channels through a 1×1 convolutional layer to enhance the feature interaction between parameters in different dimensions.

[0028] Then, in the network training stage, use the multi-dimensional operating parameter dataset under normal conditions to optimize the network parameters through maximum likelihood estimation, so that the model can accurately reconstruct the input feature vector.

[0029] Finally, during fault detection, input the feature vector corresponding to the multi-dimensional operating parameters collected in real time into the trained reversible neural network model, and output the reconstruction error. Through the affine coupling layer of the reversible neural network model, the non-linear relationship between multiple parameters such as pressure, temperature, and flow can be accurately captured. The output reconstruction error not only reflects the degree of abnormality of the overall operating state but also retains the fault feature information of each parameter dimension, improving the accuracy of fault detection.

[0030] S400 When the reconstruction error is greater than the reconstruction error threshold, calculate the joint confidence based on the belief rule base and the reconstruction error. When the joint confidence exceeds the confidence threshold, generate an interactive report containing the fault type.

[0031] The interactive report includes the fault type. The belief rule base is constructed based on the equipment mechanism of the natural gas station and the fault event case base. The antecedents of the rules in the belief rule base include that the pressure volatility is greater than the first threshold, the temperature volatility is greater than the second threshold, and the flow volatility is greater than the third threshold. The consequents of the rules are associated with three types of faults: stuck pressure regulating valve, fouling of heat exchanger, and low-temperature blockage of pipeline. The corresponding relationship between the antecedents of the rules and the faults is: the combination of abnormal pressure fluctuation and abnormal temperature fluctuation corresponds to a stuck pressure regulating valve, the combination of abnormal pressure fluctuation and abnormal flow fluctuation corresponds to a low-temperature blockage of the pipeline, and the combination of abnormal temperature fluctuation and abnormal flow fluctuation corresponds to fouling of the heat exchanger. The belief rule base also includes the initial rule confidence corresponding to the three types of faults, and the confidence threshold is taken as the initial rule confidence corresponding to the fault.

[0032] Among them, a pressure volatility greater than the first threshold indicates abnormal pressure fluctuation, a temperature volatility greater than the second threshold indicates abnormal temperature fluctuation, and a flow rate volatility greater than the third threshold indicates abnormal flow rate fluctuation.

[0033] Exemplarily, when collecting the pressure value of multi-dimensional operating parameters, the sampling frequency is exemplified as 1 Hz, sampling is based on a 60 s sliding window, and the pressure volatility is calculated based on the pressure values collected within the sliding window as: (the maximum pressure value within the collection period - the minimum pressure value within the collection period) / the average value of the pressure values within the collection period.

[0034] When collecting the temperature value of multi-dimensional operating parameters, the collection period is exemplified as 5 minutes, and the temperature volatility is calculated as: , and the unit of the temperature volatility is °C / min.

[0035] When collecting the flow rate value of multi-dimensional operating parameters, the collection period is exemplified as 30 seconds, and the flow rate volatility = . The number of collections can be appropriately increased during pressure regulation, sudden changes in the external temperature, and adjustment of the valve opening.

[0036] Through the above parameter combination analysis strategy of the belief rule base, the situations of the three groups of key parameter combinations of pressure and temperature, pressure and flow rate, and temperature and flow rate are investigated. Through the evidence theory, multi-dimensional operating parameters are fused, and the combined confidence degrees for various types of faults are calculated. Finally, the fault types are determined through the combined confidence degrees.

[0037] In some embodiments, the belief rule base is constructed based on the equipment mechanism of the natural gas station and the fault event case base, and further includes: First, the operating parameters with a pressure volatility greater than the first threshold, a temperature volatility greater than the second threshold, and a flow rate volatility greater than the third threshold in the fault event operating data are respectively labeled as abnormal pressure fluctuation, abnormal temperature fluctuation, and abnormal flow rate fluctuation, where the fault event operating data is extracted from the fault event case base of the natural gas station control system.

[0038] Then, according to the equipment mechanism of the natural gas station and the proportion of abnormal fluctuations of each operating parameter, expert scoring is performed on abnormal pressure fluctuation, abnormal temperature fluctuation, and abnormal flow rate fluctuation to determine the weight coefficients corresponding to each operating parameter. Exemplarily, since the proportion of abnormal pressure fluctuation is the largest, its weight coefficient is relatively high. The weight coefficient of pressure fluctuation is 0.6, the weight coefficient of temperature fluctuation is 0.3, and the weight coefficient of flow rate fluctuation is 0.1.

[0039] Finally, based on each weight coefficient, the evidential reasoning algorithm is used to determine the initial rule confidence degrees of each fault. Exemplarily, the value range of the initial rule confidence degree of the pressure regulating valve jamming is 0.85±0.05, the value range of the initial rule confidence degree of the heat exchanger fouling is 0.78±0.07, and the value range of the initial rule confidence degree of the pipeline low-temperature blockage is 0.92±0.03.

[0040] It should be noted that using the evidential reasoning algorithm to determine the initial rule confidence degrees of each fault based on each weight coefficient further includes: Respectively determine the confidence degrees of abnormal pressure fluctuations, abnormal temperature fluctuations, and abnormal flow fluctuations corresponding to each fault, and establish a fault evidence matrix.

[0041] According to the fault evidence matrix and the weight coefficients, the original rule confidence degrees of each fault are calculated by using the weighted summation method; the original rule confidence degrees are adjusted based on the equipment operation status and expert experience to obtain the value ranges of the initial rule confidence degrees of each fault. Exemplarily, considering equipment operation status such as equipment aging factors and service life, the original rule confidence degrees are fluctuated by ±5% to obtain the value ranges of the initial rule confidence degrees, and the initial rule confidence degrees are within the value ranges of the initial rule confidence degrees.

[0042] Among them, respectively determining the confidence degrees of abnormal pressure fluctuations, abnormal temperature fluctuations, and abnormal flow fluctuations corresponding to each fault further includes: First, extract the relevant data of at least 500 fault events with clear accident causes that occurred recently from the natural gas station control system, and extract the multi-dimensional operation parameters up to 1 hour before the fault occurrence from the relevant data. Then, calculate the corresponding operation parameter volatility based on the multi-dimensional operation parameters, and conduct percentile statistical analysis on the parameter volatility to calculate the 50th percentile of the operation parameter volatility corresponding to each fault. The 50th percentile represents the typical reference state when the fault occurs. Exemplarily, the 50th percentile of the pressure volatility corresponding to the pressure regulating valve jamming fault is 4.3%, indicating that the pressure volatility does not exceed 4.3% when 50% of the pressure regulating valve jamming faults occur. Finally, use the Sigmoid function to calculate each fluctuation abnormal confidence degree as , where x represents the actual operation parameter volatility, x0 takes the P50 value corresponding to the fault, and y is the sensitivity coefficient set according to the characteristics of each operation parameter.

[0043] In some embodiments, calculating the combined confidence degree based on the confidence rule base and the reconstruction error in S400 further includes: S410 When the volatility matches the antecedent of the corresponding rule, calculate the feature matching degree based on the initial rule confidence, volatility, and the threshold value corresponding to the volatility. The volatility includes pressure volatility, temperature volatility, or flow volatility, and the threshold values corresponding to the volatility include the first threshold, the second threshold, or the third threshold respectively. The feature matching degree μ i = initial rule confidence × min(volatility / corresponding threshold, 1).

[0044] S420 Quantify the reconstruction error as the basic probability assignment function m(A) = 1 - exp(-λE), where E represents the reconstruction error and λ represents the preset fault sensitivity coefficient, which takes different values according to different faults of the natural gas station control system. Exemplarily, for the stuck pressure regulating valve fault, λ = 0.8; for the low-temperature blockage fault of the pipeline, λ = 1.5; for the heat exchanger structure fault, λ = 1.2; S430 Combine the basic probability assignment function m(A) and the feature matching degree μ i through weighted fusion to obtain the corrected probability assignment value m'(A) = α·m(A) + (1 - α)·μ i , and calculate the combined confidence according to the Dempster combination rule:

[0045] where α represents the experience coefficient, A and B represent any two of the three rule antecedents, and C represents any one of the three types of faults. Exemplarily, when A is abnormal pressure fluctuation and B is abnormal temperature fluctuation, C is the stuck pressure regulating valve; when A is abnormal pressure fluctuation and B is abnormal flow fluctuation, C is the low-temperature blockage of the pipeline; when A is abnormal temperature fluctuation and B is abnormal flow fluctuation, C is the fouling of the heat exchanger. The remaining combinations of A and B are classified into the conflict factor K. K is the conflict factor, .

[0046] In some embodiments, the interactive report further includes a warning level, countermeasures, and the timing of implementing the countermeasures. Therefore, after S400 determines the fault type when the combined confidence exceeds the confidence threshold, the method further includes: S500 Match countermeasures according to the fault type. For the stuck pressure regulating valve fault, match the measures of preferentially checking the wear of the valve core and performing lubrication measures; for the fouling fault of the heat exchanger, match the measures of starting the chemical cleaning program and checking the gap between the heat exchange fins; for the low-temperature blockage fault of the pipeline, match the measures of increasing the power of the tracing system.

[0047] When the combined confidence level is greater than the confidence threshold and less than the first confidence threshold, increase the frequency of collecting the operating parameters corresponding to the combined confidence level; when the combined confidence level is greater than or equal to the first confidence threshold and less than the second confidence threshold, trigger a yellow warning and implement countermeasures during the next low-load period; when the combined confidence level is greater than or equal to the second confidence threshold, trigger a red warning and implement countermeasures. The interactive report is pushed in real time through the HMI interface of the station control system and simultaneously uploaded to the cloud operation and maintenance management platform to generate an electronic work order.

[0048] In this embodiment, a fault warning classification mechanism is set up. By setting a three-level response strategy of increasing the frequency of collecting operation parameters, associating the yellow warning with maintenance during the low-load period, and associating the red warning with immediately implementing countermeasures, the precise scheduling of operation and maintenance resources is achieved.

[0049] In some embodiments, the fault detection method further includes the following steps: When the combined confidence level is in the range of [0.7β, β], input the feature vector and the equally spaced sampling sequence of multi-dimensional operating parameters within at least the last 12 hours into the LSTM prediction model, where β represents the confidence threshold.

[0050] Perform multi-step prediction through the trained LSTM prediction model, and output the pressure prediction value, temperature prediction value, and flow prediction value at 15-minute intervals within at least the next 3 hours.

[0051] If the pressure prediction value exceeds the pressure safety threshold, trigger the pre-maintenance alarm of the pressure regulating valve; if the temperature prediction value exceeds the temperature safety threshold, trigger the heat exchanger inspection alarm; if the flow prediction value exceeds the flow safety threshold, trigger the pipeline blockage inspection alarm, where each safety threshold is the pressure, temperature, and flow during the fault-free event of the natural gas station.

[0052] Among them, the Long Short-Term Memory (LSTM) is a special recurrent neural network. It controls the flow of information through the cell state and three gating mechanisms (input gate, forget gate, output gate), can capture complex patterns in time series data, and is capable of learning long-term dependence information. The training process of the LSTM prediction model provided by the embodiments of this application includes: Interval sampling is performed on the multi-dimensional operating parameters of the fault-free event case base for 30 consecutive days extracted from the natural gas station control system at 15-minute intervals to construct a training sample set; preprocessing is performed on the training sample set, including removing outliers, performing mean filtering through a sliding window, and standardization; configuring the network structure, and using the Huber loss function for model training. The network structure includes an input gate layer, a forget gate layer, and an output gate layer. The input gate layer is used to receive 48 time steps of 12-hour historical window data, and each time step includes 3 feature dimensions of pressure, temperature, and flow. The first forget gate layer includes 64 neurons, the second forget gate layer includes 64 neurons, and the output gate layer includes 3 dense units, and linear activation functions are used respectively for the prediction of pressure, temperature, and flow. The Huber loss function is used for model training, and the Adam optimizer is used during the training process. Through the LSTM prediction model, it is possible to predict the pressure prediction value, temperature prediction value, and flow prediction value at 15-minute intervals within at least the next 3 hours, providing a basis for preventing accidents.

[0053] Embodiment 2 This embodiment provides a fault detection device for a natural gas station control system, including: A data acquisition module for real-time acquisition of the multi-dimensional operating parameters of natural gas in the natural gas station control system, and the multi-dimensional operating parameters include pressure values, temperature values, and flow rates; A preprocessing module for preprocessing the multi-dimensional operating parameters to generate feature vectors, where the preprocessing sequentially includes outlier removal, sliding window mean filtering, and standardization; An analysis and processing module for inputting the feature vectors into a reversible neural network model and outputting a reconstruction error. The reversible neural network includes a network architecture composed of reversible transformation modules. The affine coupling layer and convolutional layer of the reversible transformation module are used to learn the multi-dimensional operating parameters of the natural gas station control system under normal operating conditions, and the reconstruction error is used to characterize the multi-dimensional deviation degree between the multi-dimensional operating parameters and the operating parameters under normal operating conditions learned by the reversible neural network; A report generation module is configured to calculate a combined confidence based on a belief rule base and a reconstruction error when the reconstruction error is greater than a reconstruction error threshold, and generate an interactive report including a fault type when the combined confidence exceeds a confidence threshold. The belief rule base is constructed based on the mechanism of natural gas station equipment and a fault event case base. The antecedents of the rules in the belief rule base include that the pressure volatility is greater than a first threshold, the temperature volatility is greater than a second threshold, and the flow rate volatility is greater than a third threshold. The consequents of the rules are associated with three types of faults: pressure regulating valve jamming, heat exchanger fouling, and low-temperature blockage of pipelines. The corresponding relationship between the antecedents of the rules and the faults is as follows: the combination of abnormal pressure fluctuation and abnormal temperature fluctuation corresponds to pressure regulating valve jamming, the combination of abnormal pressure fluctuation and abnormal flow rate fluctuation corresponds to low-temperature blockage of pipelines, and the combination of abnormal temperature fluctuation and abnormal flow rate fluctuation corresponds to heat exchanger fouling. The belief rule base also includes the initial rule confidence levels corresponding to the three types of faults, and the confidence threshold is taken as the initial rule confidence level corresponding to the fault.

[0054] Embodiment 3 This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, it executes the steps of the natural gas station control system fault detection method described in the above embodiment.

[0055] The embodiments of the present application provide a natural gas station control system fault detection method, device, and readable storage medium. By collecting multi-dimensional operation parameters such as pressure, temperature, flow rate, and valve opening in real time, and inputting them into a reversible neural network model after preprocessing, a reconstruction error that can characterize the multi-dimensional deviation degree between the current operation state and the normal working condition is output. When the reconstruction error exceeds the reconstruction error threshold, it is determined that there may be an abnormality in the current operation state, and then the belief rule base is activated for in-depth analysis. Further, the belief rule base adopts a parameter combination analysis strategy to perform multi-parameter joint analysis of pressure-temperature, pressure-flow rate, and temperature-flow rate to calculate the combined confidence. When the combined confidence exceeds the confidence threshold, an interactive report including the fault type is generated. The present application realizes accurate fault diagnosis of the real-time operation state through a double judgment mechanism of a reversible neural network and a belief rule base.

[0056] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, a segment of a program, or a part of code, and the above-mentioned module, segment of a program, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order from that marked in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram or flowchart, as well as the combination of blocks in the block diagram or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0057] It should be noted that although several modules or units of the device for action execution are mentioned in the above detailed description, such a division is not mandatory. In fact, according to the embodiments of the present invention, the features and functions of the two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0058] From the description of the above embodiments, those skilled in the art can easily understand that the example embodiments described herein can be implemented by software, or can be implemented by a combination of software and necessary hardware. Therefore, the technical solutions according to the embodiments of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the embodiments of the present invention.

[0059] In several embodiments provided in the present application, it should be understood that the disclosed systems, modules, and methods can be implemented in other ways. For example, the module embodiments described above are merely illustrative. For example, the division of units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be an indirect coupling or communication connection through some interfaces, modules, or units, and can be in an electrical, mechanical, or other form.

[0060] The above embodiments are only used to illustrate the technical solutions of the present application, rather than limiting it. The present application is not limited to the exact structures described above and illustrated in the drawings, and it cannot be determined that the specific implementation of the present application is only limited to these descriptions. For those of ordinary skill in the technical field to which the present application belongs, various changes and deformations made without departing from the concept of the present application shall be regarded as falling within the protection scope of the present application.

Claims

1. A fault detection method for a natural gas station control system, characterized in that, The method includes the following steps: Collect the multi-dimensional operating parameters of natural gas in the natural gas station control system in real time, where the multi-dimensional operating parameters include pressure value, temperature value, and flow rate; Preprocess the multi-dimensional operating parameters to generate feature vectors, where the preprocessing sequentially includes outlier removal, moving window mean filtering, and normalization; Input the feature vectors into a reversible neural network model to output a reconstruction error, where the reversible neural network includes a network architecture composed of reversible transformation modules, and the affine coupling layer and convolutional layer of the reversible transformation module are used to learn the multi-dimensional operating parameters of the natural gas station control system under normal operating conditions, and the reconstruction error is used to characterize the multi-dimensional deviation degree between the multi-dimensional operating parameters and the operating parameters under normal operating conditions learned by the reversible neural network; When the reconstruction error is greater than the reconstruction error threshold, calculate the combined confidence based on the confidence rule base and the reconstruction error. When the combined confidence exceeds the confidence threshold, generate an interactive report including the fault type. The confidence rule base is constructed based on the equipment mechanism of the natural gas station and the fault event case base. The antecedents of the rules in the confidence rule base include that the pressure volatility is greater than the first threshold, the temperature volatility is greater than the second threshold, and the flow rate volatility is greater than the third threshold. The consequents of the rules are associated with three types of faults: pressure regulating valve jamming, heat exchanger fouling, and pipeline low-temperature blockage. The corresponding relationship between the antecedents of the rules and the faults is as follows: the combination of abnormal pressure fluctuation and abnormal temperature fluctuation corresponds to pressure regulating valve jamming, the combination of abnormal pressure fluctuation and abnormal flow rate fluctuation corresponds to pipeline low-temperature blockage, and the combination of abnormal temperature fluctuation and abnormal flow rate fluctuation corresponds to heat exchanger fouling. The confidence rule base also includes the initial rule confidence levels corresponding to the three types of faults, and the confidence threshold takes the initial rule confidence level corresponding to the fault.

2. The fault detection method for the natural gas station control system according to claim 1, wherein The interactive report also includes the warning level, countermeasures, and the timing of implementing the countermeasures. After determining the fault type when the combined confidence exceeds the confidence threshold, the method further includes: Match the countermeasures according to the fault type. For the pressure regulating valve jamming fault, match the measures of preferentially checking the wear condition of the valve core and performing lubrication measures. For the heat exchanger fouling fault, match the measures of starting the chemical cleaning procedure and checking the gap between the heat exchange fins. For the pipeline low-temperature blockage fault, match the measure of increasing the power of the tracing system; When the combined confidence is greater than the confidence threshold and less than the first confidence threshold, increase the frequency of collecting the corresponding operating parameters; When the combined confidence is greater than or equal to the first confidence threshold and less than the second confidence threshold, trigger a yellow warning and execute the corresponding countermeasures during the next load valley period; When the combined confidence is greater than or equal to the second confidence threshold, trigger a red warning and execute the corresponding countermeasures.

3. The method for detecting faults in the natural gas station control system according to claim 1, characterized in that, The confidence rule base is constructed based on the equipment mechanism of the natural gas station and the fault event case base, and further includes: Mark the operating parameters in the fault event operation data with pressure volatility greater than the first threshold, temperature volatility greater than the second threshold, and flow rate volatility greater than the third threshold as abnormal pressure fluctuations, abnormal temperature fluctuations, and abnormal flow rate fluctuations respectively, where the fault event operation data is extracted from the fault event case library of the natural gas station control system; Based on the equipment mechanism of the natural gas station and the proportion of abnormal fluctuations of each operating parameter, conduct expert scoring on abnormal pressure fluctuations, abnormal temperature fluctuations, and abnormal flow rate fluctuations to determine the weight coefficients corresponding to each operating parameter; Based on each of the weight coefficients, use the evidential reasoning algorithm to determine the initial rule confidence of each fault.

4. The method for detecting faults in the natural gas station control system according to claim 3, wherein, The step of using the evidential reasoning algorithm to determine the initial rule confidence of each fault based on each of the weight coefficients further includes: Determine the confidence of abnormal pressure fluctuations, abnormal temperature fluctuations, and abnormal flow rate fluctuations corresponding to each fault respectively, and establish a fault evidence matrix; According to the fault evidence matrix and the weight coefficients, calculate the original rule confidence of each fault by means of weighted summation; Adjust the original rule confidence based on the equipment operating state and expert experience to obtain the value range of the initial rule confidence of each fault.

5. The method for detecting faults in the natural gas station control system according to claim 1, characterized in that, The step of calculating the joint confidence based on the confidence rule base and the reconstruction error further includes: When the volatility matches the antecedent of the rule, calculate the feature matching degree based on the initial rule confidence, the volatility, and the threshold value corresponding to the volatility, where the volatility includes pressure volatility, temperature volatility, or flow rate volatility, and the threshold values corresponding to the volatility include the first threshold value, the second threshold value, or the third threshold value, respectively, and the feature matching degree μ i = initial rule confidence × min(volatility / corresponding threshold value, 1); Quantize the reconstruction error into a basic probability assignment function \(m(A) = 1 - exp(-\lambda E)\), where \(E\) represents the reconstruction error, and \(\lambda\) represents a preset fault sensitivity coefficient, with different values set according to different faults of the natural gas station control system; The basic probability assignment function m(A) is weighted and fused with the feature matching degree μ i to obtain a corrected probability assignment value m'(A)=α·m(A)+(1-α)·μ i , and the combined confidence is calculated according to the Dempster combination rule: ， Among them, α represents the empirical coefficient, A and B represent any two of the three rule antecedents, C represents any one of the three types of faults, and K is the conflict factor. .

6. The method for detecting faults in the natural gas station control system according to claim 5, characterized in that, The \(A\) and \(B\) represent any two of the three rule antecedents, and \(C\) represents any one of the three types of faults. It further includes: When \(A\) is abnormal pressure fluctuation and \(B\) is abnormal temperature fluctuation, \(C\) is stuck pressure regulating valve; when \(A\) is abnormal pressure fluctuation and \(B\) is abnormal flow rate fluctuation, \(C\) is low-temperature blockage of the pipeline; when \(A\) is abnormal temperature fluctuation and \(B\) is abnormal flow rate fluctuation, \(C\) is fouling of the heat exchanger, and the remaining combinations of \(A\) and \(B\) are classified into the conflict factor \(K\).

7. The method for detecting faults in the natural gas station control system according to claim 1, wherein The method further includes: If the joint confidence is in the range of \([0.7\beta, \beta]\), input the feature vector and the equally spaced sampling sequence of the multi-dimensional operating parameters within at least the last 12 hours into the LSTM prediction model, where \(\beta\) represents the confidence threshold; Perform multi-step prediction through the trained LSTM prediction model, and output the pressure prediction value, temperature prediction value, and flow rate prediction value at 15-minute intervals for at least the next 3 hours; If the pressure prediction value exceeds the pressure safety threshold, trigger a pre-maintenance alarm for the pressure regulating valve; if the temperature prediction value exceeds the temperature safety threshold, trigger an inspection alarm for the heat exchanger; if the flow rate prediction value exceeds the flow rate safety threshold, trigger an inspection alarm for pipeline blockage, where each of the safety thresholds is the pressure, temperature, and flow rate when there is no fault event in the natural gas station.

8. The method for detecting faults in the natural gas station control system according to claim 7, characterized in that, The training process of the LSTM prediction model includes: Perform interval sampling on the multi-dimensional operating parameters in the fault-free event case library of the natural gas station control system continuously for 30 days at 15-minute intervals to construct a training sample set; Preprocess the training sample set, including removing outliers, performing mean filtering through a sliding window, and standardization; Configure the network structure and use the Huber loss function for model training. The network structure includes an input gate layer, a forget gate layer, and an output gate layer. The input gate layer is used to receive 48 time steps of 12-hour historical window data, and each time step includes 3 feature dimensions of pressure, temperature, and flow. The first forget gate layer contains 64 neurons, the second forget gate layer contains 64 neurons, and the output gate layer contains 3 dense units corresponding to the predictions of pressure, temperature, and flow, respectively. Use the Huber loss function for model training.

9. A fault detection device for a natural gas station control system, characterized in that, The device includes: A data acquisition module for real-time acquisition of multi-dimensional operating parameters of natural gas in the natural gas station control system. The multi-dimensional operating parameters include pressure values, temperature values, and flow rates; A preprocessing module for preprocessing the multi-dimensional operating parameters to generate feature vectors. The preprocessing includes outlier removal, sliding window mean filtering, and standardization in sequence; An analysis and processing module for inputting the feature vectors into a reversible neural network model and outputting a reconstruction error. The reversible neural network includes a network architecture composed of reversible transformation modules. The affine coupling layer and convolutional layer of the reversible transformation module are used to learn the multi-dimensional operating parameters of the natural gas station control system under normal conditions. The reconstruction error is used to characterize the multi-dimensional deviation degree between the multi-dimensional operating parameters and the operating parameters under normal conditions learned by the reversible neural network; A report generation module for calculating a combined confidence based on a confidence rule base and the reconstruction error when the reconstruction error is greater than a reconstruction error threshold. When the combined confidence exceeds a confidence threshold, an interactive report including the fault type is generated. The confidence rule base is constructed based on the equipment mechanism of the natural gas station and a fault event case library. The antecedents of the rules in the confidence rule base include that the pressure volatility is greater than a first threshold, the temperature volatility is greater than a second threshold, and the flow rate volatility is greater than a third threshold. The consequents of the rules are associated with three types of faults: pressure regulating valve jamming, heat exchanger fouling, and low-temperature blockage of the pipeline. The corresponding relationship between the antecedents of the rules and the faults is as follows: the combination of abnormal pressure fluctuation and abnormal temperature fluctuation corresponds to pressure regulating valve jamming, the combination of abnormal pressure fluctuation and abnormal flow rate fluctuation corresponds to low-temperature blockage of the pipeline, and the combination of abnormal temperature fluctuation and abnormal flow rate fluctuation corresponds to heat exchanger fouling. The confidence rule base also includes the initial rule confidence levels corresponding to the three types of faults, and the confidence threshold takes the initial rule confidence level corresponding to the corresponding fault.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by a processor, it executes the steps of the method for detecting faults in the natural gas station control system according to any one of claims 1 to 8.

Images