Natural gas data management method and system

Abstract

The invention relates to the technical field of data processing and management, and particularly discloses a natural gas data management method and system. The method comprises the steps that original data in an SCADA system, sensor equipment and a historical database are collected, data cleaning and semantic annotation are carried out, and a structured semantic data set is generated; key features are extracted, and an ontology network and entity relation graph is constructed; joint modeling is carried out based on a multi-task deep neural network model, and natural gas volume prediction and path scheduling are achieved; inputting real-time state data to generate a scheduling vector, and adjusting and generating a scheduling instruction in combination with a system boundary condition; scheduling execution feedback is collected, risk events are recognized, and then model self-adaptive optimization is achieved. The method has the advantages of being complete in structure closed loop, clear in semantic expression, timely in risk early warning and sustainable in model evolution.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing and management, and in particular to a natural gas data management method and system. Background Art

[0002] With the continuous expansion of urban gas scale and the advancement of natural gas clean energy strategy, the natural gas transmission and distribution system has gradually shown the characteristics of multi-source heterogeneous equipment access, complex data structure, dynamic changes in operating status and rapid propagation of safety risks. In the whole process of natural gas transmission and distribution, a large number of operating parameters and business data from SCADA systems, edge sensors, component analysis equipment and historical databases are involved, such as pressure, flow, temperature, gas quality components and other indicators, and are distributed among multiple nodes, pipe sections and end users. There are problems such as large structural differences, high timeliness requirements, complex business semantics, and weak correlation between these data, which are difficult to be directly used for intelligent analysis and scheduling optimization.

[0003] In the existing technology, natural gas data management is mainly based on single-point collection and offline processing, lacking semantic modeling capabilities and task coordination mechanisms for the entire system, and unable to support real-time perception and dynamic response to operating conditions. At the same time, the prediction and scheduling model is usually designed in isolation and cannot form a closed-loop optimization path with risk warning or historical strategies, resulting in low scheduling efficiency, delayed risk response, and lagging safety control. Therefore, there is an urgent need for a data management method and system that supports multi-source data fusion, has semantic expression capabilities, and integrates prediction scheduling and risk warning functions to realize intelligent and adaptive operation control of natural gas transmission and distribution systems. Summary of the invention

[0004] The present invention provides a natural gas data management method and system to solve the problem of how to build an integrated intelligent management model with prediction, scheduling, early warning and adaptive optimization capabilities based on multi-source monitoring data, operating status parameters and semantic ontology relationships in the natural gas transmission and distribution system, so as to realize dynamic perception, intelligent scheduling and risk closed-loop control of the entire natural gas process.

[0005] In order to solve the above technical problems, the present invention provides a natural gas data management method, the method comprising: Obtain raw data from SCADA systems, sensor devices, and historical databases, perform missing value filling, field standardization, and semantic labeling to generate structured semantic data sets; Extracting the flow nodes, pipe segment attributes and gas quality index information in the structured semantic data set, building a natural gas system ontology network in combination with preset ontology rules, and generating an entity relationship graph and a feature vector set; Training a multi-task deep neural network model based on the graph and feature set to obtain a joint optimization model; The natural gas state data is input into the model to generate a gas volume prediction sequence and a preliminary scheduling vector, and the path constraint is adjusted according to the system boundary conditions to generate a task execution vector; Execute the vector to generate distribution instructions and collect pressure deviation and flow fluctuation, extract risk factors and combine with the knowledge rule base to identify abnormal events and generate risk warning parameters; The execution results, error sequences and risk indicators are fed back to the model to complete parameter updates and weight corrections, and output a new round of optimized model parameters.

[0006] Furthermore, the field standardization includes time alignment and unit format unification of SCADA data, sensor data and historical data.

[0007] Furthermore, the semantic labeling includes adding node labels and functional categories to each sampling field for ontology network construction.

[0008] Furthermore, the constructed ontology network forms a graph structure through triple extraction, where nodes contain natural gas equipment entities and edges represent operational dependencies.

[0009] Further, wherein Calculated based on the softmax ranking loss function, it is used to optimize the scheduling path score differences.

[0010] Furthermore, the system boundary conditions include the maximum allowable flow rate of the node, the pressure difference limit and the network connectivity rules.

[0011] Furthermore, the task execution vector is converted into a device layer scheduling instruction through a mapping relationship, including a target pressure, a valve opening state and a compression ratio setting.

[0012] Furthermore, the pressure offset is the pressure difference between adjacent time steps, and the flow fluctuation rate is the flow change ratio per unit time.

[0013] Furthermore, the risk knowledge rule base includes a plurality of abnormal type mapping rules, and generates risk levels according to fluctuation feature matching.

[0014] Furthermore, a natural gas data management system is provided. The system is used to implement the above method, and the system comprises: Data collection and preprocessing module, which is used to collect raw data from SCADA systems, sensor monitoring equipment and historical databases, and perform missing value filling, field standardization and semantic labeling processing; Semantic modeling and feature extraction module, used to extract natural gas feature information from structured semantic data sets, build ontology networks, and generate entity relationship graphs and feature vector sets; A multi-task modeling module, used to train a multi-task deep neural network model based on the graph and feature set to generate a joint optimization model; The scheduling generation and instruction issuing module is used to adjust the scheduling vector according to the prediction results and boundary constraints output by the model and generate equipment control instructions; The status monitoring module is used to collect the operating status data during the distribution execution process and generate the pressure deviation and flow fluctuation sequence; The anomaly identification and feedback module is used to identify the level of abnormal events based on the risk rule base, and feed back the risk results to the model to complete parameter updates, forming a closed-loop optimization mechanism; The strategy feedback and model updating module is used to feed back the task execution results, prediction error sequence and risk identification to the joint optimization model, update the strategy parameters and correct the training weights of the model, and generate an updated model parameter set.

[0015] The following are its main beneficial effects: (1) Realize unified integration and structural semantic expression of heterogeneous data: By standardizing fields, filling missing values, and annotating semantic labels for operating data in SCADA systems, sensor devices, and historical databases, a structured semantic data set is generated. This solves the problem of inconsistent formats and unclear semantics of multi-source data in the natural gas industry, and improves subsequent modeling efficiency and system adaptability.

[0016] (2) Construct a multi-task joint optimization model to improve the coordinated performance of prediction and scheduling: This invention extracts feature information based on the graph structure, trains a deep neural network model with traffic prediction and path scheduling functions, and introduces a joint optimization mechanism of scheduling score and prediction error. Compared with the traditional single-objective model, it has higher overall performance in prediction accuracy and scheduling rationality.

[0017] (3) Establish a model adaptive update mechanism to enhance the system robustness and dynamic adjustment capabilities: The system collects operating status data during the scheduling execution process, identifies abnormal fluctuations and makes event level judgments based on the risk rule library. The abnormal characteristics and scheduling deviations are fed back to the model for parameter correction and strategy fine-tuning, thus building an intelligent scheduling system with sustainable iterative optimization and enhancing the system's ability to adapt to complex operating conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A flow chart of a natural gas data management method provided in an embodiment of the present application; Figure 2 This is a structural block diagram of a natural gas data management system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0020] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0021] Example 1: Reference Figure 1 , is a flow chart of a natural gas data management method provided by an embodiment of the present invention, and the flow chart may at least include steps S100-S700: S100, obtaining raw data from a natural gas SCADA system, sensor monitoring equipment, and a historical database, performing missing value filling, field standardization, and semantic labeling on the raw data to generate a structured semantic data set; S200, extracting natural gas flow nodes, pipe segment attributes and gas quality index information from the structured semantic data set, building an ontology network of the natural gas system in combination with preset ontology rules, and generating an entity relationship graph and a feature vector set; S300, based on the entity relationship graph and the feature vector set, training a multi-task deep neural network model to obtain a joint optimization model with natural gas flow prediction and path scheduling capabilities; S400, inputting the natural gas state data to be processed into the joint optimization model, generating a gas volume prediction sequence and a preliminary scheduling vector, and adjusting the path constraints according to the system boundary conditions to generate a task execution vector; S500, based on the task execution vector and the real-time operation monitoring data, the natural gas transmission and distribution process is command-scheduled and executed, and the pressure offset and flow fluctuation sequence in the scheduling process are collected; S600, extracting features from the pressure offset and flow fluctuation sequence, identifying anomalies and determining event levels in combination with a risk knowledge rule base, and generating dynamic risk identification and warning parameters; S700, feeding back the task execution results, prediction error sequence and risk identification to the joint optimization model, updating the strategy parameters and correcting the training weights, and generating model parameters for a new round of iterative optimization; Step S100 at least includes steps S110-S130: S110, obtaining operating parameter data in the natural gas SCADA system, field sampling data in the distributed sensor monitoring equipment, and structured data in the historical business database, and performing field format unification processing on the data to obtain a standardized data set; Specifically, the data input of the natural gas data management system includes three key sources: natural gas pipeline network operation parameter data in the SCADA system , high-frequency gas physical detection data collected by on-site sensors , and structured operation record data in the historical business management system ,in represents the sampling time, Indicates the SCADA system acquisition source, Indicates the monitoring device node. Indicates the history database index.

[0022] The above three types of heterogeneous data are input into the data management engine. , and Perform field standard mapping. Use a unified field template ,in Indicates uniform fields (such as pressure ,temperature , flow rate , component concentration etc.), align the fields of each data source to obtain a standardized data set .

[0023] The field normalization function can be expressed as:

[0024] in, Indicates field format conversion and standard mapping operations. To unify the field set, output Represents the standardized data sample set.

[0025] S120, performing missing value repair processing and outlier elimination operations on the standardized data set to obtain a cleaned data set after integrity verification.

[0026] In the Based on this, the possible missing data values ​​and observed abnormal values ​​are jointly processed to construct a complete data set ; First, the missing fields are filled with sliding window mean and time series interpolation repair. Indicates Fields at time The value of , using the following repair function:

[0027] in, is the sliding window length, which can be 5 or 10, depending on the sampling frequency of historical data. j is a historical time step index variable under a sliding window, which means the jth moment before the current time point t. The value range of j is from 1 to W, that is, j=1,2,…,W.

[0028] Secondly, The value of is eliminated by the 3σ rule, and its mean is , the standard deviation is , if it satisfies:

[0029] It is determined as an outlier and removed, and finally forms a clean data set after integrity verification:

[0030] in, Represents the combined missing repair and anomaly removal operation.

[0031] S130, performing entity annotation and semantic label generation for each data record in the cleaned data set, constructing a semantic annotation mapping table, and obtaining a structured semantic data set.

[0032] Furthermore, Each data record in Perform entity annotation and semantic tag generation processing, extract structured semantic units, and map out information such as device entities, traffic channels, geographic locations, and transmission behaviors. Build semantic tag sets , each Corresponding to a semantic label (such as "inlet pressure node", "critical gas pipeline section", "fluctuation point", "risk source", etc.).

[0033] Define the entity annotation function as follows:

[0034] in, is the semantic mapping function, Represents the ontology model (preset ontology structure), Represents a rule set, Indicates at time For data logging Generated semantic labels.

[0035] Finally, a semantic annotation mapping table is established:

[0036] And output a structured semantic dataset:

[0037] Said It will be directly used as the input data set for the S210 feature extraction and ontology construction module. In particular, in S210, natural gas flow nodes, pipeline segment attributes and gas quality indicators will be extracted from this data set for feature vector construction and ontology network generation to ensure the contextual traceability and structural consistency of the semantic data.

[0038] Step by step instructions: Connection between S110 and S120: In step S110, the fields of multi-source raw data from SCADA systems, sensor devices, and historical databases are standardized to form a unified data set. The standardized data set is used as the input of S120 to further perform missing value repair and outlier removal operations.

[0039] Description of the connection between S120 and S130: The cleaned data set after integrity verification obtained in step S120 , providing S130 with a stable data foundation without missing or abnormal data. S130 uses this as input to generate semantic labels and entity annotations, and outputs a structured semantic dataset. .

[0040] Description of the connection between S130 and subsequent S200: The structured semantic dataset constructed in S130 , which includes the semantically annotated natural gas operation data and its corresponding entity label information, serving as the basic input data source for feature extraction and ontology relationship construction in S200.

[0041] Step S200 at least includes steps S210-S230: S210, extracting flow node features, pipe segment attribute features and gas quality index features related to natural gas flow, pressure, temperature, etc. from the structured semantic data set to generate an initial set of feature vectors; Specifically, the structured semantic data set received from S130 output ,in Indicates The cleaned data records, is its corresponding semantic label, Indicates the sampling time. Semantic label screening and field analysis are performed on the data set to extract the core indicators of natural gas operation, including but not limited to: Natural gas flow characteristics: , unit: m³ / h; Gas transmission pressure characteristics: , unit: MPa; Pipeline temperature characteristics: , unit: °C; Gas component concentration characteristics: ,in Indicates the component category, such as methane, ethane, etc.

[0042] On this basis, construct the initial set of feature vectors , which is defined as:

[0043] in Indicates the total number of component types. It will serve as the basic input for S220 ontology classification and semantic relationship modeling.

[0044] S220, based on the preset natural gas field ontology rules, semantically analyze and conceptually classify the entity attributes and relationship dimensions in the initial set of feature vectors to construct a natural gas entity hierarchical structure model; Furthermore, the ontology rule base built based on the expert knowledge in the natural gas field ,right The various attributes are classified and hierarchically mapped: Represents a category collection, such as "trunk section", "voltage regulating node", "user terminal", etc.; Represents a collection of attributes, such as "maximum allowable pressure", "normal operating temperature range"; Represents a set of relationships, such as "connection", "transmission", "dependency" and other semantic connection relationships.

[0045] Mapping functions according to the ontology Perform attribute resolution operations:

[0046] in, Indicates that at the sampling time The constructed natural gas domain entity set, each entity contains attribute information and semantic category labels. Combined with the semantic relationship set , build an entity hierarchy model :

[0047] The model describes the upstream and downstream dependencies, distribution logic and functional classification between entities.

[0048] S230, performing triple extraction and graph structure encoding on the natural gas entity hierarchical structure model to generate an entity relationship graph and a trainable feature vector set of the natural gas system; In the completed On this basis, we further extract triples of entity attribute relationships to form a standard knowledge graph structure. Define the extraction function :

[0049] in, and Represents two semantic entities, Represent the semantic relations connecting them in triples (e.g., “connected downstream,” “controlled by,” “belongs to”); That is, the natural gas system at time Entity relationship diagram.

[0050] Perform graph neural network (GNN) input vector construction on the graph to generate a set of trainable feature vectors , used for subsequent model training modules:

[0051] in is a graph structure vector encoding function that combines the graph structure adjacency matrix and node features to generate an embedded representation vector. It will be used as input in the next module (S300) to train the multi-task optimization model.

[0052] Step by step instructions: Connection between S210 and S220: Initial set of feature vectors constructed in S210 , as the input data for S220 ontology classification and semantic mapping, it ensures that the data field semantics are interpretable.

[0053] Description of the connection between S220 and S230: Natural gas entity hierarchical structure model constructed in S220 As the structural basis for performing triple extraction and graph encoding in S230, used to generate graph structure Vector collection with .

[0054] Step S300 at least includes steps S310-S330: S310, based on the entity relationship graph and the feature vector set, construct a deep neural network input layer and a feature embedding structure, and embed the semantic vector into the input layer; Specifically, the entity relationship graph output from step S230 is received The corresponding graph structure encoding vector set ,in Indicates The entity node at time The semantic feature embedding vector of .

[0055] The embedding propagation function is constructed based on the Graph Neural Network (GNN) structure, and the following information aggregation mechanism is used to fuse the input layer vectors:

[0056] in, Indicates The embedding representation vector of each node is Representation and Node The set of connected neighbors, represents the attention weight of the neighboring nodes, represents the learnable transformation weight matrix, Represents a non-linear activation function (such as ReLU).

[0057] Further, the As the joint model input layer Components and construct a unified input tensor for the multi-task network structure:

[0058] S320, based on the embedded structure and the scheduling objective function, jointly modeling the gas volume prediction task and the path scheduling task, training a multi-task deep neural network, and obtaining a pre-trained joint optimization model; Based on the input tensor , construct a dual-branch multi-task deep neural network structure to perform two subtasks, namely gas volume prediction and path scheduling. Assume: Indicates Nodes in the future The predicted gas volume; Representation Node To Node Scheduling priority score.

[0059] Define the joint loss function for:

[0060] in, is the gas volume prediction error loss term:

[0061] For the scoring sorting error loss in the path scheduling task (such as pair wise ranking):

[0062] in, Sum all training sample pairs (i, j), where i represents the task number and j represents the number of the target path in the current task; It means that the predicted score values ​​of all candidate paths k of the i-th task are normalized by softmax to form the normalized term of the probability distribution; The joint optimization model predicts the score of the jth path of the i-th task at time step t.

[0063] By minimizing the above joint loss function , perform back propagation and weight update, iteratively train the multi-task model, and generate a set of pre-trained parameters:

[0064] in, , are the network weights of the prediction branch and the scheduling branch respectively; , is the multi-task loss balance weight parameter, and its value is , , obtained by tuning the historical validation set.

[0065] S330, fine-tuning and training the joint optimization model, and solidifying the model structure and parameter weights, to obtain an optimization model parameter set with natural gas flow prediction and path scheduling functions; Furthermore, based on the actual operation feedback data set Regarding the Perform fine-tuning training, including scheduling execution deviation, prediction error and system feedback indicators. Introduce scheduling deviation function and the prediction residuals :

[0066] according to and The statistical distribution of is used to adjust the weighting strategy of the loss function, and the model is trained iteratively to finally solidify the set of optimized parameters:

[0067] in, It represents the fine-tuned optimization model parameters, including the final network structure, number of layers, weight matrix and scheduling scoring mechanism, which will serve as the basis for the S400 module to generate scheduling task vectors.

[0068] Step by step instructions: Connection between S310 and S320: Embedding representation constructed in S310 , as the input data structure of the multi-task neural network in S320, it supports simultaneous gas volume forecasting and scheduling modeling.

[0069] Connection between S320 and S330: Pre-training parameters generated by S320 training As the initial model for further fine-tuning and optimization of S330, it is used to generate the refined final optimization parameter set in combination with subsequent feedback data .

[0070] Step S400 at least includes steps S410-S430: S410, inputting the natural gas state data to be processed into the joint optimization model to generate a natural gas volume prediction sequence and a preliminary path scheduling vector; Specifically, get the pending time Real-time natural gas status data , Including pressure distribution Traffic distribution , Temperature distribution and site load requests .

[0071] The Input the joint optimization model trained in step S300 , respectively generate the gas volume prediction value sequence at the future time and the path scheduling vector :

[0072] in, Represents the future gas volume forecast sequence for each flow node; Represents a slave node To Node Scheduling priority scoring matrix.

[0073] S420, performing system boundary condition detection and constraint rule matching on the preliminary path scheduling vector, adjusting the path plan, and generating a task execution vector that satisfies the constraints; Understandably, The system-level boundary condition constraint matching is performed by scoring each scheduling path in the system. Assume that the natural gas system constraint condition set is , including but not limited to: Maximum allowed traffic flow constraint for a node ; Constraint on maximum pressure difference of pipe segment ; Pipeline network connectivity and energy consumption constraints

[0074] According to the scheduling score vector With the above constraints Constructing a scheduling feasibility function , reorder and constrain the scoring vector to generate a task execution vector that meets the conditions :

[0075] in, Each scheduling relationship in Indicates at time By Node Delivering natural gas to nodes executable path that satisfies all physical and business constraints.

[0076] S430, performing node scheduling mapping and device instruction conversion on the task execution vector to generate an executable transmission and distribution scheduling instruction set; Further, the task execution vector The scheduling path relationship in the natural gas scheduling control system is mapped to the device layer node to form a node device mapping table ,in Indicates Controllable valves, compressors or pressure regulating station equipment under the node.

[0077] in accordance with right Perform device-level translation to generate transmission and distribution scheduling instruction sets :

[0078] in, For the device mapping function, the output Indicates at time The control instructions that can be executed include opening and closing valve operations, setting pressure differences, adjusting compression ratios, etc. The instruction content is identified and issued by the SCADA control interface.

[0079] Step by step instructions: Description of the connection between S410 and S420: S410 based on the optimization model Predicted and , as the input vector for S420 to perform path feasibility judgment and scheduling constraint matching, and to generate a .

[0080] Description of the connection between S420 and S430: Task execution vector output by S420 As the basic input for executing instruction mapping in S430, through the device mapping relationship Cooperate and transform into a scheduling instruction set with device-level execution capability .

[0081] Step S500 at least includes steps S510-S530: S510, sending the transmission and distribution scheduling instruction set to the natural gas transmission and distribution system execution module to trigger the on-site equipment to perform corresponding transmission and distribution operations; Specifically, receiving the transmission and distribution scheduling instruction set generated by step S430 , Indicates Control instructions, including device operating parameters , such as valve opening and closing status, compressor speed setting, pressure regulation target value, etc.

[0082] Will Send to the edge execution module of the natural gas transmission and distribution system , which accesses the field control interface SCADA / PLC layer and executes specific actions. , the execution process is as follows:

[0083] in, Indicates actual on-site response actions, including gas flow path switching, pressure setting update, etc.

[0084] S520, during the scheduling execution process, collecting the operating status parameters of each flow node and pressure monitoring point to generate an operating status data sequence; After the execution module is activated, the distributed monitoring device is started to sample the natural gas operation status parameters at key nodes. The collected indicators include: Pressure monitoring data: ;Flow monitoring data: ; Temperature and other optional environmental parameters: .

[0085] The collection cycle is recorded as , construct a continuous running status data sequence:

[0086] in, represents the number of time steps in the observation window. It will serve as input for subsequent fluctuation detection and anomaly identification.

[0087] S530, sampling and calculating the pressure offset and flow rate fluctuation rate in the operation status data sequence to generate a state fluctuation sequence; Further, The continuous state data in the system is differentially calculated and normalized to extract the pressure offset and flow fluctuation factors and construct the state fluctuation sequence. .

[0088] First, calculate the Nodes in adjacent time steps and Pressure offset under With flow rate change :

[0089] in, It is a trace positive number to avoid the denominator being zero; and Node In time Pressure and flow values ​​at the moment.

[0090] The above indicators are spliced ​​together according to the time axis to form a state fluctuation sequence:

[0091] Said It serves as the input variable basis for the subsequent S600 risk identification module to quantify potential operational anomalies in the natural gas system.

[0092] Step by step instructions: Connection between S510 and S520: Control instructions issued by S510 Directly trigger the on-site equipment to perform scheduling operations, and the resulting flow changes and pressure changes are the operating status data sequence collected in S520 The sampling source.

[0093] Description of the connection between S520 and S530: The operating status sequence collected in S520 Provide data support for differential analysis and volatility calculation in S530, and finally generate state volatility sequence As input to subsequent modules.

[0094] Step S600 at least includes steps S610-S630: S610, performing feature extraction and variability analysis on the state fluctuation sequence to extract features of possible abnormal risk factors; Specifically, the state fluctuation sequence output from S530 is received:

[0095] in, Representation Node At the moment The pressure offset, Representation Node Construct a risk factor feature vector for each node:

[0096] in, For Node The mean pressure offset within the window, is its standard deviation, and is the mean and standard deviation of traffic volatility, which is used to characterize abnormal fluctuation trends.

[0097] Furthermore, in order to enhance the ability to identify mutation risk factors, the coefficient of variation was introduced Characterizing the intensity of fluctuations:

[0098] in is a trace constant used to prevent the denominator from being zero; when Significantly higher than the node's historical stability threshold When Derived from historical data distribution, such as experience value .

[0099] S620, matching the risk factor characteristics with a preset risk knowledge rule base, identifying the abnormal type and determining the event level; Based on the risk factor characteristics extracted above Coefficient of variation , enter the abnormal identification process. Call the preset natural gas system risk knowledge rule base:

[0100] Each rule Indicates the matching conditions of a type of risk scenario (such as "sudden rise in high pressure", "continuous fluctuation", "dramatic drop in traffic", etc.), in the following format:

[0101] in, is the rule triggering threshold, is the risk type label. and In the rule, identify each node at time Under the Risk Type tab .

[0102] Furthermore, according to the frequency of events, risk intensity and system impact, the event level function is set , output risk level :

[0103] in, Representation Node The importance weight of the pipeline network can be defined by domain experts or calculated by topological location (such as the main pipeline segment , end user ), It is a risk level scoring function, and the output level classification is such as "low", "medium", "high", "severe", etc.

[0104] S630, performing multi-dimensional risk identification coding on the determination result to generate dynamic identification and early warning parameters of risk events; Furthermore, the risk type With grade results Encoded into multi-dimensional dynamic risk markers , including the following components:

[0105] Said It is a structured dynamic risk event object that supports multi-source system identification, storage and visualization.

[0106] Furthermore, the risk event identification of all nodes is integrated to build the moment System-level risk warning parameter set:

[0107] Said It will be used as one of the inputs of the subsequent strategy feedback module (S700) to trigger model updating and parameter correction.

[0108] Step by step instructions: Connection between S610 and S620: Risk factor vector extracted from S610 and variation index , as the core input for rule matching and event identification in S620, driving risk types With level judgment.

[0109] Connection between S620 and S630: The event type and level results output by S620 are used as the basis for dynamic coding in S630. Unified description of risk events and integration into system-level warning parameters , for use by the next module.

[0110] Step S700 at least includes steps S710-S730: S710, using the scheduling execution result, the prediction error sequence and the risk identification as feedback input to generate a joint feedback data set; Specifically, obtain the following feedback input: Actual operating status data from S530 ; Prediction output from S410 , used to calculate the forecast error series; risk event identification set from S630 , containing dynamic risk levels and impact factors.

[0111] Based on the above data, construct the error indicator function:

[0112] in, represents the gas volume prediction error, represents the path scheduling deviation, Score the constraint-adjusted real scheduling path (from S420 ).

[0113] Combining error sequences and risk events, construct a joint feedback dataset:

[0114] Said Serves as correction input for online training of the joint model.

[0115] S720, performing gradient correction and strategy fine-tuning on parameter weights in the joint optimization model based on the joint feedback data set; Furthermore, the correction loss function is defined:

[0116] in: A weighting parameter for the feedback loss, e.g. , , , determined by tuning the historical validation set; It represents a function that maps risk level and volatility factor to penalty value, for example, high risk events are given larger update amplitude; The third item introduces the impact of risk identification on gradient adjustment, making the model more sensitive to risk event areas.

[0117] Based on the above loss function, the back propagation algorithm is used to optimize the joint optimization model. The parameter set in Perform gradient update:

[0118] in, is the learning rate (e.g. ), represents the gradient with respect to the parameter.

[0119] S730, iteratively training the adjusted joint optimization model to generate an updated model parameter set for the next round of reasoning; based on The obtained weight parameters are used as the initial model structure, and the data in the subsequent observation window is iteratively fine-tuned to build a dynamically evolving version of the optimization model. .

[0120] Suppose the new round of input sequence is , perform the incremental training process:

[0121] in, Indicates an incremental training function, with the current model and its new data samples as input and the new round of model parameters after fine-tuning as output. , replace the original optimization model , used for the next round of reasoning and prediction tasks in step S410.

[0122] Step connection instructions: Connection instructions between S710 and S720: The joint feedback dataset generated in S710 Provides prediction error and risk event coding for the correction loss function constructed in S720 Provides complete input to achieve gradient update.

[0123] Connection between S720 and S730: Parameters updated by S720 As the basic model for iterative fine-tuning in S730, the incremental training process is performed and solidified into new inference model parameters , realizing parameter adaptive closed loop.

[0124] The key innovative features of the present invention include: (1) Semantic structure-driven multi-source natural gas data integration method.

[0125] (2) Design of joint optimization model for both prediction and scheduling.

[0126] (3) Risk identification and model adaptive closed-loop mechanism based on operational feedback. This method is applicable to data management and intelligent control scenarios of various natural gas transmission and distribution networks, and has good promotion value and application prospects.

[0127] The following are its main beneficial effects: (1) Realize the unified integration and structural semantic expression of heterogeneous data: By standardizing the fields, filling in missing values, and annotating semantic labels for the operating data in the SCADA system, sensor equipment, and historical database, a structured semantic data set is generated. This solves the problem of inconsistent formats and unclear semantics of multi-source data in the natural gas industry, and improves the subsequent modeling efficiency and system adaptability.

[0128] (2) Construct a multi-task joint optimization model to improve the coordinated performance of prediction and scheduling: This invention extracts feature information based on the graph structure, trains a deep neural network model with traffic prediction and path scheduling functions, and introduces a joint optimization mechanism of scheduling score and prediction error. Compared with the traditional single-objective model, it has higher overall performance in prediction accuracy and scheduling rationality.

[0129] (3) Establish a model adaptive update mechanism to enhance the system robustness and dynamic adjustment capabilities: The system collects operating status data during the scheduling execution process, identifies abnormal fluctuations and makes event level judgments based on the risk rule library. The abnormal characteristics and scheduling deviations are fed back to the model for parameter correction and strategy fine-tuning, thus building an intelligent scheduling system with sustainable iterative optimization and enhancing the system's ability to adapt to complex operating conditions.

[0130] Embodiment 2: Figure 2 FIG. 2 shows a structural block diagram of a natural gas data management system according to an embodiment of the present invention. Figure 2 As shown, the system may include: The data acquisition and preprocessing module 10 is used to obtain raw data related to the operation of the transmission and distribution system from multi-source natural gas data inlets, including pressure, flow and temperature parameters in the SCADA system, gas component concentration, valve status and node operation information collected by edge sensor devices, and structured business data in the historical database. This module generates a semantically unified and complete structured data set through field standardization, missing value filling and outlier removal, providing a high-quality data source for downstream semantic modeling and analysis.

[0131] The semantic modeling and ontology construction module 20 is used to perform semantic enhancement and ontology abstraction on structured data sets based on the knowledge system of the natural gas industry. This module constructs an entity label system and a semantic label mapping relationship, extracts key features such as pipe segment attributes, node features, gas quality indicators, etc., forms a semantic graph and ontology network structure of the natural gas transmission and distribution system, and provides a high-dimensional and interpretable structural representation for subsequent task scheduling and model training.

[0132] The multi-task prediction and scheduling optimization module 30 is used to construct a multi-task neural network model based on the constructed semantic graph and feature vector, and jointly model the natural gas flow trend and the transmission and distribution path. This module integrates the gas volume prediction loss function and the path scheduling scoring function, obtains the pre-trained model parameters with linkage optimization capabilities through historical data training, and realizes the flow prediction of the target node and the generation of the path scheduling vector.

[0133] The boundary constraint and task generation module 40 is used to perform system physical constraint and business rule verification on the predicted scheduling vector, matching boundary conditions including maximum pressure limit, node load capacity and pipe network connectivity. The module adjusts the scheduling plan according to the constraint detection result, generates a task execution vector that meets the constraint conditions, and maps the task to a specific node device to form a scheduling execution instruction set.

[0134] The dispatch execution and status monitoring module 50 is used to send dispatch instructions to the natural gas transmission and distribution field execution system, realize the operation of valves, compressors, pressure regulators and other equipment through the SCADA control port, and collect the pressure, flow and temperature parameters of each node in real time during the execution process. This module records the operation status data through the sliding window mechanism and constructs a time-series status monitoring sequence for subsequent volatility assessment and abnormality identification.

[0135] The risk identification and early warning module 60 is used to extract features and perform variation analysis on the state fluctuations in the monitoring data, and identify the abnormal type and determine the risk level in combination with the preset risk rule library. This module intelligently matches scenarios such as high-frequency fluctuations, high-voltage mutations, or sharp drops in traffic, forms a multi-dimensional risk event identification structure, generates risk early warning parameters, and provides a quantitative basis for strategy feedback.

[0136] The strategy feedback and model update module 70 is used to use the feedback information such as scheduling errors, risk events and prediction deviations as joint inputs to fine-tune the strategy and adaptively update the parameters of the multi-task optimization model. This module constructs an online feedback loss function to implement incremental training of the neural network model, generates updated optimization model parameters, and uses them to replace the original model, continuously improving the prediction accuracy and scheduling robustness of the system under different working conditions.

[0137] The natural gas data management method and system provided by the present invention can realize efficient fusion of multi-source heterogeneous data, semantic structure modeling, flow trend prediction, scheduling task generation, risk perception warning and adaptive closed-loop optimization in the natural gas transmission and distribution process, and has the following significant advantages: Support semantic modeling and graph construction of natural gas full-process business data, and improve data interpretability and task reasoning capabilities; Establish a multi-task deep learning model to achieve coordinated optimization of gas volume forecasting and route scheduling, and improve the efficiency of forecasting and scheduling decisions; Based on the feedback-driven mechanism, the closed-loop control logic of scheduling instructions - execution monitoring - abnormal identification - strategy update is realized to enhance the intelligence and adaptability of the system; The modular architecture has good scalability and deployment flexibility, and is suitable for various transmission and distribution pipeline scenarios in the natural gas industry.

[0138] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A natural gas data management method, characterized in that: The method comprises: Obtain raw data from SCADA systems, sensor devices, and historical databases, perform missing value filling, field standardization, and semantic labeling to generate structured semantic data sets; Extracting the flow nodes, pipe segment attributes and gas quality index information in the structured semantic data set, building a natural gas system ontology network in combination with preset ontology rules, and generating an entity relationship graph and a feature vector set; Based on the graph and feature set, a multi-task deep neural network model is trained to obtain a joint optimization model; the model training takes the joint loss function as the optimization target: ； in, is the joint loss function, is the gas volume prediction error loss term, is the score ranking error loss in the path scheduling task, and Balancing weight parameters for multi-task losses; The natural gas state data is input into the model to generate a gas volume prediction sequence and a preliminary scheduling vector, and the path constraint is adjusted according to the system boundary conditions to generate a task execution vector; Execute the vector to generate distribution instructions and collect pressure deviation and flow fluctuation, extract risk factors and combine with the knowledge rule base to identify abnormal events and generate risk warning parameters; The execution results, error sequences and risk indicators are fed back to the model to complete parameter updates and weight corrections, and output a new round of optimized model parameters.

2. The method according to claim 1, characterized in that The field standardization includes time alignment and unit format unification of SCADA data, sensor data and historical data.

3. The method according to claim 1, characterized in that The semantic labeling includes adding node labels and functional categories to each sampling field for ontology network construction.

4. The method according to claim 1, characterized in that: The ontology network constructed forms a graph structure through triple extraction. The nodes contain natural gas equipment entities and the edges represent operation dependencies.

5. The method according to claim 1, characterized in that: It is stated Calculated based on the softmax ranking loss function, it is used to optimize the scheduling path score differences.

6. The method according to claim 1, characterized in that The system boundary conditions include the maximum allowable node flow, pressure difference limit and network connectivity rules.

7. The method according to claim 1, characterized in that The task execution vector is converted into a device-level scheduling instruction through a mapping relationship, including a target pressure, a valve opening state, and a compression ratio setting.

8. The method according to claim 1, characterized in that The pressure offset is the pressure difference between adjacent time steps, and the flow fluctuation rate is the ratio of flow change per unit time.

9. The method according to claim 1, characterized in that: The risk knowledge rule base includes a plurality of abnormal type mapping rules, and generates risk levels according to fluctuation feature matching.

10. A natural gas data management system, characterized in that: The system is used to implement the method described in any one of claims 1 to 9, and the system includes: Data collection and preprocessing module, which collects raw data from SCADA systems, sensor monitoring equipment and historical databases, and performs missing value filling, field standardization and semantic labeling processing; The semantic modeling and feature extraction module extracts natural gas feature information from structured semantic data sets, builds an ontology network, and generates entity relationship graphs and feature vector sets; A multi-task modeling module, which trains a multi-task deep neural network model based on the graph and feature set to generate a joint optimization model; The scheduling generation and instruction issuing module adjusts the scheduling vector according to the prediction results and boundary constraints output by the model, and generates equipment control instructions; The status monitoring module collects the operating status data during the distribution execution process and generates pressure deviation and flow fluctuation sequences; The anomaly identification and feedback module identifies the level of abnormal events based on the risk rule base, and feeds back the risk results to the model to complete parameter updates, forming a closed-loop optimization mechanism; The strategy feedback and model update module feeds back the task execution results, prediction error sequence and risk identification to the joint optimization model, updates the strategy parameters and corrects the training weights of the model, and generates an updated model parameter set.

Images