A neural network monitoring method and system for a global gas pipeline network structure

Abstract

The application discloses a neural network monitoring method and system for a global gas pipe network structure, which comprises the following steps: inputting a plurality of local pipe network structure corresponding key node sequences, pressure parameter values, the number of monitoring nodes in the local pipe network structure and the global pipe network structure as input information into a neural network model, and obtaining the output of the neural network model as a monitoring result. The application improves the selection efficiency of the monitoring nodes in the global gas pipe network structure under the condition that the data accumulation amount is large based on the mutual information method of pressure interval division, so that it is possible to accurately analyze the neural network model with the input vector composed of the key nodes, and the monitoring accuracy is further improved.

Description

Technical Field

[0001] This invention relates to the field of urban gas pipeline network monitoring technology, specifically to a neural network monitoring method and system for the global gas pipeline network structure. Background Technology

[0002] Urban gas pipeline systems are highly complex multivariable control systems, characterized by large areas, numerous influencing factors, strong internal correlations, long time lags, and severe nonlinearity. Current technologies for analyzing gas pipeline network data often focus on individual analytical objectives and rely on traditional simulation software. While such software is effective for analyses based on independent objectives, it struggles to produce effective results in large-scale data environments.

[0003] In a big data analytics environment, large-scale, global gas pipeline network data analysis is often required, such as for urban gas management agencies or even city-wide gas data analysis. However, we have already accumulated a massive amount of data across a wide geographical area. my country's big data technology has reached a relatively mature stage; and gas has been used for many years. With the continuous integration, recording, and processing of data, a large amount of various types of gas data has been generated. To fully utilize this data, a large amount of data may be generated, leading many analyses to discard some data to improve efficiency or make big data analysis possible. However, such practices are not only unhelpful to subsequent analyses but may also lead to errors.

[0004] If all information is crammed into advanced analytical tools, such as neural network models, the resulting input vectors will inevitably contain too many components, leading to overly complex model structures unsuitable for routine analysis. Therefore, how to effectively analyze global gas pipeline network structure data across a broad geographical and data scope is a problem to be solved. Thus, designing a practical neural network monitoring method and system for global gas pipeline network structures is essential. Summary of the Invention

[0005] The purpose of this invention is to provide a neural network monitoring method and system for global gas pipeline network structures to solve the problems mentioned in the background art.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a neural network monitoring method and system for global gas pipeline network structure, including step S1: reading the local gas data file corresponding to the local gas pipeline network structure;

[0007] Step S2: reading pressure parameter values from the local gas data file and selecting a key node sequence in the local gas pipeline network structure based on the pressure parameter; the key node sequence contains one or more monitoring nodes arranged in order of number;

[0008] Step S21: obtaining historical pressure parameters of the monitoring nodes in the local gas pipeline network;

[0009] Step S22: obtaining the change range of the historical pressure parameters and dividing the change range into X intervals;

[0010] Step S23: dividing all historical pressure parameters of the monitoring nodes into X different pressure intervals;

[0011] Step S24: calculating the mutual information value MI(P i , P j ) between any two monitoring nodes;

[0012]

[0013]

[0014] wherein P i , P j represent two monitoring nodes; x is the interval number, x = 1 ~ X; p x (P i , P j ) represents the proportion of the monitoring nodes P i and P j located in interval x at the same time; p x (P i ) represents the proportion of the monitoring node P i located in interval x; p x (P j ) represents the proportion of the monitoring node P j located in interval x; W x is the weight coefficient of interval x; N x is the number of pressure parameters falling into interval x; NL is the number of all pressure parameters;

[0015] Step S25: arranging all mutual information values from large to small and finding the inflection point of the mutual information value;

[0016] Step S26: calculating the average mutual information value

[0017] Step S27: intercepting the monitoring nodes with average mutual information values greater than the inflection point of the mutual information value and arranging their numbers in order from large to small to form the key node sequence (KP1, KP2...);

[0018] Step S3: input the key node sequence corresponding to the plurality of local pipe network structures, the pressure parameter values thereof, and the number of monitoring nodes in the local pipe network structures and the global pipe network structure as input information into the neural network model, and obtain the output of the neural network model as the monitoring result.

[0019] Further, the preprocessing is to fill in the missing data in the local gas data file and delete the contradictory data.

[0020] Further, the numbering is based on the position of the monitoring node.

[0021] Further, the numbering is based on the flow size of the monitoring node.

[0022] Further, the numbering is random numbering.

[0023] A neural network monitoring system for a global gas pipe network structure, the system being used to implement the neural network monitoring method for the global gas pipe network structure.

[0024] Further, one or more cloud servers are arranged in the system.

[0025] A processor for running a program, wherein the program, when running, executes the neural network monitoring method for the global gas pipe network structure.

[0026] A computer-readable storage medium comprising a program, which, when run on a computer, causes the computer to execute the neural network monitoring method for the global gas pipe network structure.

[0027] An execution device comprising a processor, the processor and a memory being coupled, and the memory storing program instructions, which, when executed by the processor, implement the neural network monitoring method for the global gas pipe network structure.

[0028] Based on the same inventive concept, the present application also provides a processor for running a program, wherein the program, when running, executes the neural network monitoring method for the global gas pipe network structure.

[0029] Compared with the prior art, the present application has the following beneficial effects:

[0030] (1) The present application improves the key node selection efficiency under large data conditions based on the mutual information method of pressure interval division, and also makes it possible to combine and analyze the local pipe network structure and the global pipe network structure;

[0031] (2) By taking the key node sequence based on the number as the input component, that is, each input component is composed of an integer component value, on the one hand, the calculation degree is reduced, and the number itself has brought the topological information, and on the basis of not losing the global or local pipe network structure information, the calculation complexity of the subsequent neural network model is greatly reduced;

[0032] (3) The span value of the pressure parameter and the proportion value of the number of monitoring nodes are substituted into the neural network model, and the sensitivity of the specific local pipe network structure monitoring is improved. BRIEF DESCRIPTION OF DRAWINGS

[0033] The accompanying drawings are included to provide a further understanding of the application, and constitute a part of the specification, illustrate the application, and are used to explain the application, and do not constitute a limitation on the application. In the drawings:

[0034] Figure 1 is a schematic diagram of the neural network monitoring method for the global gas pipe network structure of the application;

[0035] Figure 2 is a schematic diagram of the neural network model based on the span value of the pressure parameter of the application; DETAILED DESCRIPTION

[0036] The technical solutions in the embodiments of the application will be described in detail below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0037] Please refer to Figures 1-2 The application provides a technical solution: a neural network monitoring method and system for a global gas pipe network structure, comprising the following steps:

[0038] Step S1: reading the local gas data file corresponding to the local gas pipe network structure; specifically: the gas pipe network structure is divided into a plurality of local gas pipe network structures according to its covered area relationship, the monitoring nodes in the covered area obtain monitoring data, and save the monitoring data to the local gas data file; read the local gas data file corresponding to each local gas pipe network structure covered area; respectively, the local gas data file is preprocessed;

[0039] The local gas pipe network structure is the target object analyzed by common simulation software, when the global gas pipe network structure needs to be analyzed, the local gas pipe network structure files cannot be simply and independently analyzed, because the pipe networks are connected with each other, independent analysis will inevitably lead to the loss of analysis factors, the application carries out overall analysis on the local gas pipe network structure data based on the neural network model analysis of the key node sequence, and the stability of the analysis is improved;

[0040] Preferably, the preprocessing is to fill the missing data in the local gas data file and delete the contradictory data.

[0041] Step S2: reading the pressure parameter value from the local gas data file, and selecting the key node sequence in the local gas pipe network structure based on the pressure parameter; the key node sequence comprises one or more monitoring nodes arranged in sequence according to the number;

[0042] For example, for a local gas pipe network structure comprising 10 monitoring nodes, 3 key nodes <P2, P4, P7> are finally selected, so that the sequence 2, 4 and 7 is the sequence of the monitoring nodes, and for the key node sequence <P4, P2, P7>, the key node sequence is changed;

[0043] Preferably, the number is based on the position of the monitoring node;

[0044] Preferably, the number is based on the flow size of the monitoring node;

[0045] Preferably, the number is a random number;

[0046] The key node sequence in the local gas pipe network structure is selected based on the pressure parameter, and specifically comprises the following steps:

[0047] Step S21: obtaining the historical pressure parameters of the monitoring nodes in the local gas pipe network;

[0048] Step S22: obtaining the change range of the historical pressure parameters, and dividing the change range into X intervals; and the X intervals have the identification in the sense of pressure parameters; that is, the change range of the interval cannot be too small, so that the interval loses the identification;

[0049] Step S23: dividing all the historical pressure parameters of the monitoring nodes into X different pressure intervals;

[0050] Since the same monitoring node in the historical data can correspond to multiple pressure parameters at different times, the multiple pressure parameters of the monitoring node will be respectively divided into X different pressure intervals; in this way, the information about the time dimension will not be lost;

[0051] Step S24: Calculate mutual information value MI(P i , P j ) between any two monitoring nodes;

[0052]

[0053]

[0054] Wherein: P i , P j represent two monitoring nodes; x is interval number, x = 1 ~ X; p x (P i , P j ) represents the proportion of monitoring nodes P i and P j located in interval x at the same time; p x (P i ) represents the proportion of monitoring nodes P i located in interval x; p x (P j ) represents the proportion of monitoring nodes P j located in interval x; W x is the weight coefficient of interval x; N x is the number of pressure parameters falling into interval x; NL is the number of all pressure parameters; the mutual information method analysis method is used for analyzing the classification problem of one data, and the application is used for analyzing the correlation between pressure data by interval analysis, so that the degree of mutual correlation between data values between different monitoring nodes is measured without losing the time information of pressure data, which can be measured by mutual information; the greater MI(P i , P j ), the greater the interaction between the two monitoring nodes, and vice versa;

[0055] Step S25: arrange all mutual information values from large to small and find the inflection point of mutual information value; wherein the inflection point of mutual information value is the numerical point where the mutual information value changes obviously;

[0056] Alternatively, the inflection point of mutual information value is the numerical point at a specific position; for example, the numerical point at the third position from large to small;

[0057] Step S26: calculate the average mutual information value of each monitoring node

[0058] Step S27: intercept the monitoring nodes with average mutual information value greater than the inflection point of mutual information value, and arrange the numbers in descending order to form a key node sequence (KP1, KP2...);

[0059] Step S3: input the key node sequence and the pressure parameter value corresponding to the plurality of local pipe network structures as input information into the neural network model, and obtain the output of the neural network model as the monitoring result;

[0060] Preferably, the neural network model is an RBF, LTSM or BP neural network model.

[0061] The step S3 comprises the following steps:

[0062] Step S31: length-padded the key node sequence corresponding to the mi-th local pipe network structure to form an input component I mi mi，1 mi，nj mi，n ; wherein: mi=1~m; nj=1~n; that is, all the key node sequences need to be length-padded to n; m is the number of local pipe network structures, that is, the number of input components;

[0063] Preferably, n is the maximum length of the key node sequence in all local pipe network structures.

[0064] Step S32: calculate and set the weight value ω mi of each input component; after weighting each input component, fully connect to each neuron node of the hidden layer; wherein:

[0065] mi ; nj=1~n:

[0066]

[0067] PRS mi，nj is the nj-th input component in the mi-th local pipe network structure; wherein: PRS mi is the pressure monitoring value of the nj-th key monitoring node in the mi-th local pipe network structure; that is, the specific monitoring value is taken as the input information of the neural network model by means of weighting, but since the pressure monitoring value itself is not taken as the input vector, the dimension of the setting information is reduced and the influence of the weight information itself on the calculation amount brought by the full connection is not exponential;

[0068] As can be seen, since the key node sequence itself may change, especially when an anomaly occurs, the weight value may change once every time the key node sequence changes. Then in the subsequent model training process, the weight value can be changed once after each iteration, that is, after learning a sample, or the weight value can be modified uniformly after learning a key node sequence, thereby improving the training efficiency. ​​​​

[0069] Step S33: Calculate and set the connection weights ν of the hidden layer and the output layer. mi The hidden layer has m neurons, and the output is produced by fully connecting the hidden layer neurons to the m output layer neurons; where: ν mi =(ν mi，nj ), nj = 1 ~ n;

[0070]

[0071] PN mi，nj This represents the ratio of the number of monitoring nodes in the local pipeline network structure mi to the number of monitoring nodes in the global pipeline network structure. For some large local pipeline network structures, the types of monitoring nodes are relatively simple, and the number of key nodes selected is also relatively small. The sensitivity of monitoring these types of local pipeline network structures can be adjusted by outputting weights.

[0072] It can be seen that since the number of monitoring nodes in the local pipeline network structure is relatively stable, the connection weight between the hidden layer and the output layer is a relatively stable value.

[0073] Step S34: Input the input components sequentially into the neural network model to obtain the neural network model output Y = (y mi As a monitoring result; here y mi It is unary information, and Y is the mi-ary output component;

[0074] Preferred: y mi = 0 or 1;

[0075] This invention uses a sequence of key nodes based on their numbers as input components, meaning each input component is an integer value. This reduces computational complexity, and the numbering itself incorporates topological information. Furthermore, by substituting the pressure parameter span value and the proportion of monitoring nodes into the neural network model, the sensitivity of monitoring specific local pipeline structures is improved.

[0076] As attached Figure 2 As shown, s and g are the activation functions of the hidden layer and the output layer, respectively; of course, different activation functions can be set for the neurons in each row of the hidden layer and the output layer, and the activation functions are nonlinear functions; and the hidden layer can also be set to a multi-layer structure;

[0077] Preferably, the radial basis function in the neural network model is a Gaussian kernel function;

[0078] Preferably, the historical monitoring data of the key nodes in the key node sequence are used as samples to train the neural network model; the samples include input components corresponding to each key node sequence and output components corresponding to monitoring results after artificial feedback, and each tuple in the output components corresponds to a monitoring result of a local pipe network structure;

[0079] Alternatively, the output components as a whole correspond to a monitoring result respectively as a binary value;

[0080] Preferably, each neuron has multiple inputs, multiple outputs, and an internal threshold value;

[0081] A computer program (also known as a program, software, software application, script, or code) can be written in any form of programming language, including compiled or interpreted languages, declarative or procedural languages, and it can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, object, or other unit suitable for use in a computing environment. A computer program may, but need not, correspond to a file in a file system. A program can be stored in a portion of a file that holds other programs or data (e.g., one or more scripts stored in a markup language document), in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files that store one or more modules, sub programs, or portions of code). A computer program can be deployed to be executed on one computer or on multiple computers that are located at one site or distributed across multiple sites and are interconnected by a communication network.

[0082] Those skilled in the art will appreciate that embodiments of the present application can be supplied as a method, a system, or a computer program product. Thus, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage media, etc.) having computer usable program code embodied in the medium.

[0083] The present application is described in reference to the flowcharts and / or block diagrams of the methods, apparatus (systems) and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and combinations of flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, a special purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions, which are executed via the processor of the computer or other programmable data processing apparatus, generate a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 Each flow or multiple flows and / or blocks Figure 1means for performing the function specified by the block or blocks.

[0084] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a Figure 1 one or more processes and / or blocks Figure 1 means for performing the function specified by the block or blocks.

[0085] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the process Figure 1 one or more processes and / or blocks Figure 1 means for performing the function specified by the block or blocks.

[0086] It should be noted that, in the present document, the terms such as first and second, etc., are used only to distinguish one entity or operation from another, and do not necessarily require or imply these entities or operations to be in any such actual relationship or order. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover a non-exclusive inclusion, such that a process, method, article or apparatus that comprises a list of elements does not include only those elements recited, but can also include other elements not expressly listed or inherent to such process, method, article or apparatus.

[0087] Finally, it should be noted that the above-mentioned only constitutes the preferred embodiments of the present application, and is not intended to limit the present application. Although the present application has been described in detail with reference to the foregoing embodiments, modifications or equivalent replacements to the technical solutions recorded in the foregoing embodiments can still be made by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A neural network monitoring method for a global gas pipeline network structure, comprising: Step S1: reading a local gas data file corresponding to a local gas pipeline network structure; Step S2: reading pressure parameter values from the local gas data file and selecting a key node sequence in the local gas pipeline network structure based on the pressure parameters; the key node sequence includes one or more monitoring nodes arranged in order of numbering; Step S21: obtaining historical pressure parameters of the monitoring nodes in the local gas pipeline network; Step S22: obtaining a change range of the historical pressure parameters and dividing the change range into X intervals; Step S23: dividing all the historical pressure parameters of the monitoring nodes into X different pressure intervals; Step S24: Calculate the mutual information value MI(P i , P j ) between any two monitoring nodes; ； wherein: P i , P j represent two monitoring nodes; x is the interval number, x = 1 ~ X; p x (P i , P j ) represents the proportion of monitoring nodes P i and P j located in interval x at the same time; p x (P i ) represents the proportion of monitoring nodes P i located in interval x; p x (P j ) represents the proportion of monitoring nodes P j located in interval x; W x is the weight coefficient of interval x; N x is the number of pressure parameters falling into interval x; NL is the total number of pressure parameters; Step S25: arranging all the mutual information values from large to small and finding a mutual information value inflection point; Step S26: Calculate the average mutual information value of each monitoring node ; Step S27: intercepting monitoring nodes with average mutual information values greater than the mutual information value inflection point and arranging their numbers in order from large to small to form a key node sequence (KP1, KP2,...); Step S3: inputting the key node sequences corresponding to multiple local pipeline network structures, the pressure parameter values, the number of monitoring nodes in the local pipeline network structure and the global pipeline network structure as input information into a neural network model, and obtaining the output of the neural network model as a monitoring result; specifically: Step S31: length complement of the key node sequence corresponding to the mi-th local pipe network structure to form the input component I mi = (KP mi，1 ,...KP mi，nj ...KP mi，n ); wherein: mi = 1 ~ m; nj = 1 ~ n; that is, all the key node sequences need to be complemented to n length; m is the number of local pipe network structures; Step S32: calculate and set the weight value ω corresponding to each input component mi ; after weighting each input component, fully connected to each neuron node of the hidden layer; wherein: mi = (ω mi，nj ); nj = 1 ~ n: ； PRS mi，nj Input component for the nthj element in the local pipe network structure mi; Step S33: Calculate and set the connection weight v of the hidden layer and the output layer mi ; the number of hidden layer neuron nodes is m, and the output after full connection between the hidden layer neuron nodes and the m output layer neuron nodes of the output layer; wherein: mi = (v mi，nj ), n j = 1 ~ n; ； PN mi，nj is the ratio of the number of monitoring nodes in the local pipe network structure mi to the number of monitoring nodes in the global pipe network structure.

2. The neural network monitoring method for the global gas pipeline network structure according to claim 1, characterized in that: The preprocessing is to fill in the missing data in the local gas data file and delete the contradictory data.

3. The neural network monitoring method for the global gas pipeline network structure according to claim 1, characterized in that: The numbering is based on the position of the monitoring node.

4. The neural network monitoring method for the global gas pipeline network structure according to claim 1, characterized in that: The numbering is based on the flow size of the monitoring node.

5. The neural network monitoring method for the global gas pipeline network structure according to claim 1, characterized in that: The numbering is random numbering.

6. A neural network monitoring system for a global gas network structure, characterized by: The system is used to implement the neural network monitoring method for the global gas pipeline network structure according to any one of claims 1-5.

7. The neural network monitoring system for a global gas network structure according to claim 6, characterized in that: One or more cloud servers are provided in the system.

Images