CN120296364B

CN120296364B - Intelligent analysis and diagnosis system for caliber matching of gas metering appliance

Info

Publication number: CN120296364B
Application number: CN202510779290.2A
Authority: CN
Inventors: 丰波; 诸葛鹏; 孙勐; 杨光; 李燕艳; 陈宸; 齐玉祥
Original assignee: Aude Group Co ltd
Current assignee: Aude Group Co ltd
Priority date: 2025-06-12
Filing date: 2025-06-12
Publication date: 2025-09-23
Anticipated expiration: 2045-06-12
Also published as: CN120296364A

Abstract

The invention provides an intelligent analysis and diagnosis system for caliber matching of a gas metering appliance, which relates to the technical field of data processing and comprises an extraction module, a correction module and a processing module, wherein the extraction module is used for selecting a main pipe junction point P1, a user access point P2 and a pressure regulating station output point P3 as three dynamic topological datum points based on pipeline pressure gradient data, flow phase data and user gas time sequence fluctuation data, constructing dynamic topological units according to the three points, dividing the units into areas, extracting pressure-flow coupling characteristic values of subareas, carrying out weighted correction on user gas demand mutation probability characteristic vectors through the pressure-flow coupling characteristic values, generating corrected mutation probability characteristic vectors, and the processing module is used for tensor splicing the corrected mutation probability characteristic vectors with pipeline equivalent inner diameter dynamic characteristic vectors and inputting the corrected mutation probability characteristic vectors into a full-connection diagnosis network to generate caliber matching deviation index. The intelligent operation efficiency of gas supply can be improved.

Description

Intelligent analysis and diagnosis system for caliber matching of gas metering appliance

Technical Field

The invention relates to the technical field of data processing, in particular to an intelligent analysis and diagnosis system for caliber matching of a gas metering appliance.

Background

In a large commercial complex, the complex covers various gas use scenes such as markets, hotels, restaurants and the like, and the gas use requirement is complex and has large fluctuation. During initial construction, the caliber of the gas metering device is configured according to conventional estimation and experience. Along with the promotion of operation, the number of catering merchants in a market is increased, business hours are different, so that the gas supply is insufficient in the gas consumption peak period, and the gas equipment of partial merchants cannot normally run at full load, thereby influencing the normal operation of the merchants.

In the low peak period, the excessive caliber of the metering device reduces the metering precision, so that the gas metering is inaccurate, and disputes are often generated between gas companies and merchants due to the problem of gas settlement. In addition, because the long-term unreasonable gas supply also causes frequent faults of partial gas equipment due to unstable pressure, the equipment maintenance cost is increased, and therefore, the defect that the caliber matching mode of some traditional gas metering appliances is in response to complex gas use scenes is highlighted.

Disclosure of Invention

The invention aims to solve the technical problem of providing an intelligent analysis and diagnosis system for caliber matching of a gas metering appliance, which can improve the intelligent operation efficiency of a gas supply system.

In order to solve the technical problems, the technical scheme of the invention is as follows:

In a first aspect, an intelligent analysis and diagnosis system for aperture matching of a gas meter, comprising:

the acquisition module is used for acquiring pipeline pressure gradient data, flow phase data and user gas time sequence fluctuation data in real time through pressure sensors deployed at three-dimensional space coordinate positions of a main node of the gas pipe network, a user access point and a pressure regulating station;

The extraction module is used for selecting a main pipe junction point P1, a user access point P2 and a voltage regulating station output point P3 as three dynamic topology datum points based on the pipeline pressure gradient data, the flow phase data and the user gas time sequence fluctuation data;

the correction module is used for carrying out weighted correction on the mutation probability feature vector of the user gas demand through the pressure-flow coupling feature value to generate a corrected mutation probability feature vector;

The processing module is used for tensor splicing of the corrected mutation probability feature vector and the pipeline equivalent inner diameter dynamic feature vector, inputting the tensor spliced mutation probability feature vector and the pipeline equivalent inner diameter dynamic feature vector into the fully-connected diagnosis network to generate an caliber matching deviation index, and outputting a safety redundant caliber replacement scheme by the linkage time sequence prediction network when the deviation exceeds a set threshold value.

In a second aspect, a method for intelligent analysis and diagnosis of gas meter aperture matching, the method comprising:

Step 1, acquiring pipeline pressure gradient data, flow phase data and user gas time sequence fluctuation data in real time through pressure sensors deployed at three-dimensional space coordinate positions of a main node of a gas pipe network, a user access point and a pressure regulating station;

Step 2, selecting a main pipe junction point P1, a user access point P2 and a voltage regulating station output point P3 as three vertexes based on pipeline pressure gradient data, flow phase data and user gas time sequence fluctuation data, and forming a dynamic triangle according to the three vertexes;

step 3, carrying out weighted correction on the mutation probability feature vector of the user gas demand through the pressure-flow coupling feature value so as to obtain a corrected mutation probability feature vector;

and step 4, tensor splicing is carried out on the corrected mutation probability feature vector and the equivalent inner diameter dynamic feature vector of the pipeline, the tensor spliced mutation probability feature vector and the equivalent inner diameter dynamic feature vector of the pipeline are input into a fully-connected diagnosis network to generate an caliber matching deviation degree index, and when the deviation degree exceeds a threshold value, the linkage time sequence prediction network outputs a safety redundant caliber replacement scheme.

In a third aspect, a computing device includes:

One or more processors;

And storage means for storing one or more computer programs that, when executed by the one or more processors, cause the one or more processors to implement the system.

In a fourth aspect, a computer readable storage medium has a computer program stored therein, which when executed by a processor, implements the system.

The above scheme of the invention at least comprises the following beneficial effects.

The acquisition module is used for acquiring pipeline pressure gradient data, flow phase data and user gas time sequence fluctuation data in real time by deploying pressure sensors at a main node of a gas pipe network, a user access point and a pressure regulating station, and compared with the traditional mode relying on manual or static monitoring, the system can comprehensively capture dynamic information in the pipe network operation process, and the caliber matching deviation caused by data loss or hysteresis is avoided.

The extraction module creatively selects three dynamic topological reference points to construct a dynamic topological unit based on the acquired multidimensional data, and utilizes a delaunay subdivision algorithm to divide areas so as to extract pressure-flow coupling characteristic values of the subareas, the method can effectively integrate the spatial topological structure and the hydrodynamic characteristics of the pipe network, converts the complex pipe network operation state into quantifiable characteristic parameters, and greatly improves the analysis capability of the pipe network operation rule.

The correction module carries out weighted correction on the mutation probability feature vector of the gas demand of the user by utilizing the pressure-flow coupling feature value, the influence of the pipe network operation state on the gas consumption behavior of the user is fully considered, the dynamic correction mechanism can accurately capture the sudden change of the gas demand of the user, and compared with the traditional fixed parameter prediction model, the dynamic correction mechanism can be more fit with the actual gas consumption scene, and the caliber matching error caused by the demand pre-estimated deviation is avoided.

The processing module performs tensor splicing on the corrected mutation probability feature vector and the equivalent inner diameter dynamic feature vector of the pipeline, generates an caliber matching deviation index through the fully-connected diagnosis network, and outputs a safe and redundant caliber replacement scheme through the linkage time sequence prediction network when the deviation exceeds a threshold value.

Drawings

FIG. 1 is a schematic diagram of an intelligent analysis and diagnosis system for aperture matching of a gas metering appliance, which is provided by an embodiment of the invention.

FIG. 2 is a schematic flow chart of an intelligent analysis and diagnosis method for aperture matching of a gas metering appliance according to an embodiment of the invention.

Detailed Description

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

As shown in fig. 1, an embodiment of the present invention proposes an intelligent analysis and diagnosis system for caliber matching of a gas meter, including:

In the embodiment of the invention, the acquisition module can acquire pipeline pressure gradient, flow phase and user gas time sequence fluctuation data in real time by arranging the pressure sensor at the key position of the gas pipe network, and compared with the traditional manual monitoring or intermittent data acquisition mode, the system can realize omnibearing and dynamic monitoring on the running state of the pipe network, and effectively avoid caliber matching errors caused by data lag or deletion.

The extraction module constructs a topological unit by using a dynamic topological datum point, divides the area by using a Delaunay subdivision algorithm, accurately extracts the pressure-flow coupling characteristic value, fully considers the dynamic association of the pipe network space topology and the fluid mechanics, can deeply excavate the pipe network operation rule, and converts the complex pipe network working condition into the characteristic parameter which can be quantized and analyzed. The correction module carries out weighted correction on the mutation probability feature vector of the gas demand of the user by utilizing the pressure-flow coupling feature value, the influence of the pipe network operation state on the gas consumption behavior of the user is fully considered, the mechanism can dynamically adjust the gas demand prediction in real time, the sudden change of the gas demand of the user is accurately captured, and compared with the traditional fixed parameter prediction model, the mechanism can be more fit with the actual gas consumption scene, and the caliber matching error caused by the demand prediction deviation is effectively avoided. The processing module realizes the automatic generation of caliber matching deviation index through tensor splicing and full-connection diagnosis network, and the linkage time sequence prediction network outputs a safe and redundant caliber replacement scheme when the deviation exceeds a threshold value, the process is completely based on an intelligent algorithm and a model, the full-flow automation is realized from data processing to decision output, the diagnosis efficiency is improved, the gas conveying pressure loss is effectively reduced, the metering precision is improved, the equipment abrasion and the energy waste are reduced, and the safe and stable operation of a gas supply system is ensured.

In a preferred embodiment of the present invention, pressure gradient data, flow phase data and user gas time sequence fluctuation data of a pipeline are collected in real time by pressure sensors deployed at three-dimensional space coordinate positions of a trunk node of a gas pipe network, a user access point and a pressure regulating station, and the method includes:

establishing a pipe network topology mapping relation based on three-dimensional space coordinates of each sensor;

Synchronously acquiring real-time pressure values among trunk nodes along the trend of the pipeline, calculating the pressure difference correlation space distance among adjacent nodes, and generating pressure gradient data;

Monitoring the output of a flow sensor of a user access point, extracting the phase offset and the amplitude characteristic of a flow waveform in a period, and generating flow phase data;

Recording an instantaneous fluctuation signal of the gas consumption at the downstream of the pressure regulating station, and calculating the fluctuation frequency and the absolute difference value of wave crest and wave trough in unit time to form gas consumption time sequence fluctuation data of a user.

In the embodiment of the present invention, the above steps, when applied specifically, may be implemented specifically by the following steps, for example:

Three-dimensional space coordinates of pressure sensors deployed at a trunk node, a user access point and a pressure regulating station of a gas pipe network are obtained, each sensor has corresponding X, Y, Z coordinate values, the coordinates reflect the space positions of the sensors in an actual pipe network, and then the sensors with the connection relationship are associated according to the actual laying condition and the connection relationship of the gas pipe network. For example, if two trunk nodes are connected through a pipeline, the sensors corresponding to the two nodes are marked as a connection state in space, and for the user access point and the voltage regulating station, the connection relation between the user access point and the voltage regulating station and the trunk node sensors is determined according to the position of the actual access pipe network, and finally the topology mapping relation of the whole gas pipe network is constructed, just like drawing a pipe network map which marks the connection condition of each node in detail.

By establishing the pipe network topology mapping relation, the space structure and node connection condition of the gas pipe network can be clearly presented, and when data calculation and analysis such as pressure, flow and the like are carried out, the transmission path and the mutual influence of the data can be more accurately judged by combining the topology relation, so that the accuracy and the reliability of data processing are improved, and the overall operation condition of the pipe network can be comprehensively mastered.

And enabling the pressure sensors distributed at the trunk nodes to simultaneously start to acquire data, and acquiring the real-time pressure value of each trunk node at the same time. Then, for every two adjacent trunk nodes, subtracting the node pressure with the smaller pressure value from the node pressure with the larger pressure value to obtain the pressure difference between the two adjacent nodes. And measuring the actual distance between two adjacent nodes in the three-dimensional space, and correlating the pressure difference with the corresponding space distance, for example, the pressure difference between the node A and the node B is 5 units, and the space distance is 10 meters, then taking the combination of the pressure difference and the space distance as a group of data, repeating the process, calculating the pressure difference and the space distance between all adjacent trunk nodes, and integrating the data to generate the pressure gradient data reflecting the pressure change trend of the pipe network.

The pressure gradient data can intuitively reflect the pressure change condition of the fuel gas flowing in the pipeline, and by analyzing the pressure gradient data, whether a pressure abnormal region exists in the pipeline or not can be found in time, for example, the problem that the pipeline is leaked or blocked due to pressure suddenly drops can be solved, so that the quick positioning of a fault point is facilitated for a fuel gas enterprise, corresponding maintenance measures are adopted, and the safety and stability of fuel gas transportation are ensured.

Continuously monitoring flow data output by a flow sensor of a user access point, forming a continuously-changed flow waveform along with the time, selecting a fixed time period, observing the change condition of the flow waveform in the period, comparing the standard flow waveform under normal conditions with the phase offset, judging whether the current flow waveform deviates leftwards or rightwards on a time axis, measuring the time length of the deviation, finding the maximum value and the minimum value of the flow waveform in the period according to the amplitude characteristic, calculating the difference value between the maximum value and the minimum value, wherein the difference value is the change range of the flow amplitude, recording two groups of data of the phase offset and the amplitude characteristic, repeating the operation for each user access point for a plurality of times, obtaining data in a plurality of periods, and summarizing the data to generate flow phase data.

The flow phase data can accurately reflect the change rule and the characteristic of the gas consumption of the user, the phase offset can reflect the change of the gas consumption time of the user, such as whether to advance or delay the peak of the gas consumption, and the amplitude characteristic reflects the fluctuation amplitude of the gas consumption of the user. By analyzing the data, the gas enterprise can better know the gas consumption habit of the user, reasonably allocate the gas resources, improve the efficiency and stability of gas supply, and simultaneously, be helpful for predicting the future gas consumption demand of the user.

And recording gas consumption data downstream of the pressure regulating station in real time, wherein the data can generate instantaneous fluctuation with time to form a series of fluctuation signals. A fixed time unit, such as 1 minute, is selected, and the number of times the air consumption fluctuates in the 1 minute is counted, wherein the number of times is the fluctuation frequency in unit time. Then, the maximum value (peak) and the minimum value (trough) of the air consumption in this time unit are found, and the absolute difference between them is calculated. For example, within 1 minute, the maximum value of the air consumption is 20 cubic meters, the minimum value is 10 cubic meters, and then the absolute peak-trough difference is 10 cubic meters. Recording the fluctuation frequency and the absolute difference of wave crest and wave trough in each time unit, continuously counting a plurality of time units, and integrating the data together to form the time sequence fluctuation data of the user gas, wherein the time sequence fluctuation data can clearly show the fluctuation condition of the user gas on the time sequence.

The user gas consumption time sequence fluctuation data can intuitively present the instantaneous change condition of the user gas consumption at the downstream of the pressure regulating station, the fluctuation frequency reflects the frequent degree of the gas consumption change, and the absolute difference value of the wave crest and the wave trough reflects the intensity of the gas consumption change. By analyzing the data, the gas enterprise can timely find abnormal fluctuation of the gas consumption of the user, make countermeasures in advance, ensure the stability and safety of gas supply, and simultaneously is beneficial to optimizing the operation parameters of the pressure regulating station and improving the gas conveying efficiency.

In a preferred embodiment of the present invention, based on the pipeline pressure gradient data, the flow phase data and the user gas time sequence fluctuation data, the main pipe junction point P1, the user access point P2 and the voltage regulating station output point P3 are selected as three dynamic topology reference points, including:

According to the topological mapping relation of the pipe network, selecting a main pipe intersection point with the largest pressure change rate in the pressure gradient data as P1, and adopting a three-dimensional space coordinate as an initial reference coordinate;

Calculating the space influence weight of the user access point P2 by taking the initial reference coordinate of the main pipe intersection point P1 as a reference and combining the phase offset and the amplitude characteristic in the flow phase data, and carrying out offset compensation on the original three-dimensional coordinate of the user access point P2 according to the space influence weight to obtain the weighted real-time monitoring coordinate of the user access point P2;

based on the weighted real-time monitoring coordinates of the user access point P2, performing time sequence displacement compensation on the original coordinates of the output point P3 of the voltage regulating station through the fluctuation frequency and the peak-trough difference value in the air time sequence fluctuation data of the user so as to obtain the real-time regulation coordinates of the output point P3 of the voltage regulating station;

The method comprises the steps of carrying out space synchronization on an initial reference coordinate of a main pipe intersection point P1, a real-time monitoring coordinate of a user access point P2 and a real-time regulation coordinate of a voltage regulating station output point P3, converting the initial reference coordinate, the real-time monitoring coordinate and the real-time regulation coordinate into a two-dimensional plane through dimension reduction mapping, and connecting three points P1, P2 and P3 in the two-dimensional plane to form a dynamic topology unit.

And then, checking the performance of each main pipe junction in the pressure gradient data, and focusing on the change condition of the pressure along with the space distance, namely the pressure change rate. For example, comparing the relation between the pressure difference and the distance between adjacent nodes of different junction points, the large pressure change rate means that the pressure change is obvious in a shorter distance, finding out the junction point of the main pipe with the largest pressure change rate, determining the junction point as P1, recording the three-dimensional space coordinate of the junction point as an initial reference coordinate of the subsequent operation, and selecting a key reference anchor point in the pipe network.

The main pipe junction point with the largest pressure change rate is selected as P1, because the main pipe junction point is often positioned at a position with more severe pressure fluctuation in a pipe network, and the main pipe junction point can be a key node or a potential problem area for gas transportation. The method is used as an initial reference coordinate, a very representative reference point can be provided for subsequent analysis, and the method is helpful for rapidly positioning abnormal pressure changes in a pipe network.

The method comprises the steps of taking three-dimensional coordinates of P1 as a reference, analyzing phase offset and amplitude characteristics in flow phase data of each user access point P2, wherein the phase offset reflects the difference between gas consumption time of a user and the conventional condition, the amplitude characteristics reflect fluctuation of gas consumption, comprehensively considering the two factors, such as the user access points with early gas consumption time and large fluctuation, giving higher space influence weight to a pipe network, otherwise, giving low influence weight, determining the space influence weight of each P2 point, and adjusting the original three-dimensional coordinates according to the weight. If the weight is large, the coordinate is offset in space by a relatively large extent according to a certain rule, and if the weight is small, the offset is small, and finally the weighted real-time monitoring coordinate of the user access point P2 is obtained, so that the coordinate can reflect the influence degree of the point on the actual operation of the network.

For example, the user access point P2 spatial impact weight determination process and the value range are as follows:

setting evaluation indexes and grading:

And (5) evaluating the phase offset, namely dividing the phase offset of the user gas into five levels. The offset time is regarded as 'minimum offset' within +/-15 minutes, 15-30 minutes as 'small offset', 30-60 minutes as 'medium offset', 60-120 minutes as 'large offset', and more than 120 minutes as 'maximum offset' based on the conventional gas time. The longer the offset time, the greater the difference in user gas usage time from the conventional case, the more significant the impact on the pipeline network scheduling and pressure balance may be.

Amplitude characteristic evaluation, the amplitude characteristic of the air consumption is also classified into five grades. The fluctuation amplitude (the difference between the wave crest and the wave trough) of the air consumption is 'minimum fluctuation' within 5% of rated air consumption, 5% -15% is 'small fluctuation', 15% -30% is 'medium fluctuation', 30% -50% is 'large fluctuation', and more than 50% is 'large fluctuation'. The larger the amplitude is, the more severe the change of the air consumption of the user is, and the larger the impact on the pipe network is.

Quantization scoring and weight calculation:

Each rank is assigned a respective score, with "very small offset or wave" corresponding to 1 score, "small offset/wave" corresponding to 2 scores, "medium offset/wave" corresponding to 3 scores, "large offset/wave" corresponding to 4 scores, and "very large offset or wave" corresponding to 5 scores. And respectively acquiring grade scores corresponding to the phase offset and the amplitude characteristics of each user access point P2, and then adding the two scores to obtain the comprehensive influence score of the point. For example, a user access point P2 may have a "large offset" (4 points) in phase, a "medium fluctuation" (3 points) in amplitude, and a combined impact score of 7 points.

And carrying out normalization processing on the comprehensive influence scores of all the user access points P2, and mapping the comprehensive influence scores to the value range of [0,1] so as to obtain the space influence weight of each P2 point. Assuming that the sum of the comprehensive influence scores of all the user access points P2 is S, and the comprehensive influence score of a certain P2 point is S, the spatial influence weight w=s/S of the point.

Weight value range:

The weight is close to 0, when the phase offset of the user access point P2 is 'minimum offset', and the amplitude characteristic is 'minimum fluctuation', the spatial influence weight is close to 0, the air consumption time of the user is stable, the air consumption fluctuation is small, and the influence on the operation of the pipeline network is negligible.

The weight is 0< 0.2, and is in this range if the phase offset and amplitude characteristics are in the "small offset", "small ripple" or lower level combination. The air consumption behavior of the user is less influenced on the pipe network, and the air consumption behavior is a conventional and stable air consumption condition.

The weight of 0.2< 0.5 is in this interval when "medium offset" and "medium ripple" and below-level combinations occur, or "large offset or ripple" and "small or small offset or ripple" are combined. The method shows that the gas consumption behavior of the user has a certain influence on the pipe network, and proper attention is required in pipe network analysis and scheduling.

Weight 0.5< weight 0.8 or less, if a combination of "large offset" and "medium/large fluctuation" or a combination of "large offset/fluctuation" and "small offset or fluctuation" occurs, the weight is in the range, which means that the user has a large influence on the pipe network by using gas behavior, and the local pressure fluctuation or the flow distribution change of the pipe network can be caused.

The weight is close to 1, and when the phase offset of the user access point P2 is "maximum offset" and the amplitude characteristic is "maximum fluctuation", the weight is close to 1. Such user gas behavior has a great influence on the operation of the pipe network, and may need to perform targeted adjustment or enhanced monitoring on the pipe network.

According to the invention, the space influence weight is calculated by combining the flow phase data, the coordinates are compensated, the influence of the gas consumption behavior of the user on the pipe network is fully considered, and the obtained real-time monitoring coordinates of the user access point P2 are not the pure geographic position coordinates, but are integrated with the gas consumption characteristics of the user, so that the actual effect of the user access point in the operation of the pipe network can be reflected more accurately, the influence of the gas consumption of the user on the pipe network can be more accurately mastered by a gas enterprise, and the gas distribution and scheduling strategies are optimized.

The number of fluctuations per unit time (e.g., 1 hour) is divided into five levels, the number of fluctuations is 5 times or less, the number of fluctuations is "extremely low frequency", the number of fluctuations is 5-15 times "low frequency", the number of fluctuations is 15-30 times "medium frequency", the number of fluctuations is 30-50 times "high frequency", and the number of fluctuations exceeds 50 times "extremely high frequency". The higher the frequency, the more frequently the user gas changes, and the more frequently the voltage regulating station needs to adjust the output state.

And grading the peak-trough difference, and dividing the proportion of the peak-trough difference to the rated output into five grades by taking the rated output of the voltage regulating station as a reference. The difference ratio is less than 5% and is 'tiny fluctuation', 5% -15% is 'small fluctuation', 15% -30% is 'medium fluctuation', 30% -50% is 'big fluctuation', more than 50% is 'maximum fluctuation', the larger the difference value is, the more severe the air consumption change of a user is, and the larger the impact on the output stability of the voltage regulating station is.

Evaluation of the degree of comprehensive influence:

an evaluation matrix is constructed, and five levels of fluctuation frequency and five levels of peak-to-valley difference are combined to form a 5×5 evaluation matrix. For example, when the ripple frequency is "high frequency" and the peak-to-trough difference is "large ripple", one particular combination in the matrix is corresponded.

And determining the influence level, and dividing the degree of influence of the user on the output state of the voltage regulating station into five levels according to different combination conditions. For example, "high frequency + large fluctuations", "extremely high frequency + medium fluctuations" and combinations thereof correspond to "extremely large influence" levels, combinations of "high frequency + medium fluctuations", "medium frequency + large fluctuations" and the like correspond to "large influence" levels, and so on until "extremely low frequency + extremely small fluctuations" correspond to "extremely small influence" levels.

And (3) determining a compensation direction:

Forward compensation (time axis backward), when the user uses air to bring the wave crest forward, the fluctuation frequency increases, etc., which means that the voltage regulating station needs to respond to the changes earlier, at this time, forward displacement compensation is carried out on the original coordinate of P3 on the time axis, and the coordinate point is moved backward.

And negative compensation (time axis forward), if the user delays with the air wave peak and the fluctuation frequency is reduced, the response of the voltage regulating station can be properly delayed, and at the moment, the original coordinate of P3 is subjected to negative displacement compensation on the time axis, and the coordinate point is moved forward.

And (3) calculating compensation amplitude:

The basic compensation unit is set, and a basic compensation time unit is set, for example, 5 minutes. This unit represents the amount of time shift performed on the P3 coordinate at the minimum influence level (minimal influence).

And (3) adjusting the grading multiple, and respectively setting corresponding compensation multiple according to five grades of the comprehensive influence degree. The "minimal effect" corresponds to 1-fold basis units, the "small effect" corresponds to 2-fold, the "medium effect" corresponds to 4-fold, the "large effect" corresponds to 8-fold, and the "large effect" corresponds to 16-fold. For example, when the integrated effect is "great effect", the compensation amplitude is 16×5=80 minutes.

And determining final compensation amplitude, and carrying out fine adjustment by combining specific numerical values of the peak-trough difference values on the basis of the grading multiple. If the difference is in the upper half of the level interval, the upper limit value of the level compensation multiple is taken, and if the difference is in the lower half, the lower limit value is taken. For example, under the "great influence" scale, the difference is compensated for 80 minutes in the upper half and for 70 minutes in the lower half.

Coordinate adjustment and real-time regulation coordinate generation:

And (3) time dimension displacement, namely, the original coordinate of P3 is displaced in the time dimension according to the determined compensation direction and amplitude. For example, if the compensation direction is positive and the amplitude is 60 minutes, the time coordinate value of P3 is increased by 60 minutes, and if it is negative, the corresponding time value is decreased.

And (3) synchronously adjusting the space coordinates, taking the influence of the user on the space state of the voltage regulating station by using the pneumatic fluctuation into consideration when the time dimension is displaced, and carrying out tiny adjustment on the P3 coordinates in the space dimension if the fluctuation causes the output pressure or flow of the voltage regulating station to change obviously, wherein the adjustment amplitude is in direct proportion to the fluctuation influence degree.

And generating a real-time regulation coordinate, and obtaining the real-time regulation coordinate of the output point P3 of the voltage regulating station after the adjustment of time and space dimensions, wherein the coordinate reflects the time correlation between the output state of the voltage regulating station and the fluctuation of the user gas and also reflects the influence of the fluctuation on the space state of the voltage regulating station.

According to the invention, the time sequence displacement compensation is carried out on the coordinates of the output point P3 of the voltage regulating station by utilizing the time sequence fluctuation data of the user gas, and the dynamic change of the user gas is closely related to the regulation and control of the voltage regulating station, so that the real-time regulation and control coordinates of the P3 can embody the response of the voltage regulating station according to the user gas fluctuation, thereby being beneficial to a gas enterprise to better know the running state of the voltage regulating station, ensuring the stability and safety of gas supply and realizing the fine regulation and control of a pipe network.

The dynamic topology unit construction process specifically comprises the following steps:

space unified calibration:

The coordinates are aligned, and a unified space origin is determined, for example, the geometric center of the coverage area of the whole gas pipe network is selected as the origin. Then, for the initial reference coordinates of P1, the real-time monitoring coordinates of P2, and the real-time regulation coordinates of P3, their offsets with respect to this origin are calculated, respectively. For example, the original coordinates of P1 are (X1, Y1, Z1), the distance that it needs to translate in three directions of X, Y, Z is calculated with reference to the new origin, the coordinates are adjusted to the coordinate system with reference to the new origin, and the translation operation is performed on the coordinates of P2 and P3 by the same method, so that the coordinates of the three points are all based on the same origin, and preliminary alignment is achieved.

The dimensions are uniform, and whether the dimensions of the three points in the X, Y, Z axis direction are uniform is checked. Because different areas in an actual pipe network may have measurement units or accuracy differences, a uniform scale is required. For example, if the Z-axis coordinate measurement unit of P1 is found to be meters and the Z-axis coordinate measurement units of P2 and P3 are found to be centimeters, the Z-axis coordinate values of P2 and P3 are divided by 100 and converted into meters, so that the dimensions of the three points in the three axial directions are ensured to be the same, and the spatial calibration is completed so that the three points are positioned in the same reference system.

Dimension reduction mapping to a two-dimensional plane:

The projection direction is determined, and an appropriate projection direction is selected, and it is common to select a direction perpendicular to the main laying plane of the pipe network as the projection direction, for example, if the gas pipe network is mainly laid on the underground plane, a direction perpendicular to the main laying plane may be selected as the projection direction.

And (3) projection operation, namely projecting three-dimensional coordinates of the three points P1, P2 and P3 after calibration to a two-dimensional plane along a selected projection direction, imagining that a beam of parallel light irradiates the three points from the projection direction, wherein shadows of the three points on the two-dimensional plane are coordinate positions after projection, respectively calculating corresponding X and Y coordinate values of the three points on the two-dimensional plane, discarding Z coordinate values, and obtaining new coordinates of the three points on the two-dimensional plane.

And (3) coordinate adjustment and optimization, wherein the projected two-dimensional coordinates are checked and adjusted, so that the relative position relationship among points is ensured to be in accordance with the spatial relationship in an actual pipe network. If the positions of some points after projection are found to have unreasonable overlapping or offset, the coordinate values are finely adjusted, so that the layout of the coordinate values on a two-dimensional plane can more accurately reflect the relative positions of three points in an actual pipe network for subsequent analysis.

Building a dynamic topology unit:

And connecting lines, namely sequentially connecting three points P1, P2 and P3 on a two-dimensional plane by using a line segment tool, firstly drawing a line segment from the point P1 to the point P2, then drawing a line segment from the point P2 to the point P3, and finally drawing a line segment from the point P3 to the point P1 to form a closed triangle region.

The information integration and labeling are carried out by labeling the data information related to three points, such as the pressure change rate of the P1 point, the space influence weight of the P2 point, the time sequence displacement compensation basis of the P3 point and the like, in the corresponding points or triangle areas, and annotating can be added to illustrate the relationship between the data and the points and the meaning of the data and the points on the whole dynamic topological unit, so that the triangle area is not only a geometric figure, but also a visual analysis unit integrating the important information of the key nodes of the pipe network, and the spatial relationship and the mutual influence among the nodes can be intuitively presented.

The invention integrates and simplifies the complex spatial relationship and operation data of key nodes in the pipe network by constructing the dynamic topology unit. Through the unit, the association among the main pipe junction, the user access point and the output point of the voltage regulating station can be intuitively analyzed, and the operation characteristics of the local area of the pipe network can be rapidly mastered. The method is beneficial to analyzing the gas pipe network from the angles of integral and local combination, provides a clear analysis framework for the subsequent operations of carrying out region division, extracting pressure-flow coupling characteristic values and the like based on the Delaudiences subdivision algorithm, and improves the analysis efficiency and accuracy of the pipe network running state.

In a preferred embodiment of the present invention, a dynamic topology unit is constructed according to three points, the unit is divided into regions based on a delaunay subdivision algorithm, and a pressure-flow coupling characteristic value of a sub-region is extracted, including:

The method comprises the steps of executing a delaunay subdivision algorithm on a dynamic topological unit to generate a minimum angle division structure comprising at least three subregions, constructing an initial convex hull boundary by taking real-time two-dimensional plane coordinates of three points P1, P2 and P3 in the dynamic topological unit as subdivision input point sets, dynamically inserting auxiliary vertexes Q1 at the midpoints of connecting lines P2 and P3 based on real-time variation amplitude of air time sequence fluctuation data used by a user when the fluctuation amplitude exceeds a set threshold value, inserting auxiliary vertexes Q2 at the midpoints of connecting lines P1 and P2 according to phase offset of flow phase data, adding Q1 and Q2 into the subdivision point sets and updating the convex hulls to obtain updated subdivision point sets, executing diagonal exchange detection on quadrilation formed by any four adjacent points by traversing the updated subdivision point sets, updating topological connection relation if the minimum internal angle is increased after exchange, iteratively optimizing until all quadrilation meets the delaunay round criterion to generate a final triangulated structure, defining each minimum topological unit in the final triangulated structure as a subregion coordinate set and an adjacency relation;

Extracting pipeline pressure gradient data associated with vertex coordinates of each divided subregion, and synchronously acquiring flow phase data amplitude characteristics in the same spatial domain;

Calculating the arithmetic mean value of the product of the pressure gradient data and the flow phase amplitude characteristic, and outputting a basic space coupling factor of the subarea;

Based on the basic space coupling factor, positioning gas utilization time sequence fluctuation data of user access points governed by the subarea, calculating a statistical variance value of the governed user fluctuation data in a continuous time window, and performing linear normalization processing of a [0,1] interval on the variance value to generate a time sequence correction coefficient;

Adding the basic spatial coupling factor and the time sequence correction coefficient to generate an intermediate coupling characteristic;

and multiplying the intermediate coupling characteristic by a space weight coefficient of the subarea in the pipe network topology, and finally outputting a pressure-flow coupling characteristic value.

The relative position relation of the three points is judged based on two-dimensional coordinates of the three points P1, P2 and P3, and as only three points are collinear or can form a triangle, in the actual pipe network analysis, the three points are not collinear, so that the three points are directly connected to form a triangle, and the triangle is the smallest convex polygon which can surround the three points, namely the initial convex hull.

Marking boundary directions, namely marking three sides of the triangle in turn according to the clockwise or anticlockwise directions, and determining the connection sequence of each vertex.

Dynamically inserting auxiliary vertices:

and detecting fluctuation conditions, monitoring the variation amplitude of the user gas time sequence fluctuation data in real time, and comparing the variation amplitude with a preset threshold value. For example, the threshold is set to 1.5 times the average fluctuation width, and when the actual fluctuation width exceeds the threshold, the unstable gas consumption condition of the area is indicated.

And inserting a vertex Q1, and calculating the midpoint coordinate of the connecting line of P2 and P3 when the fluctuation amplitude exceeds a threshold value. The specific method is that the average value of the coordinate values of P2 and P3 on the X axis and the Y axis is taken respectively, and the obtained new coordinate point is Q1. Q1 is inserted into the point set, which contains four points P1, P2, Q1, P3.

And detecting phase offset, analyzing the phase offset of the flow phase data, and judging whether the flow phase data reaches a certain degree or not. For example, when the phase shift amount exceeds 20% of the normal fluctuation range, it is considered that finer division of the region is required.

And inserting a vertex Q2, and if the phase offset reaches a set degree, calculating the midpoint coordinate of the connecting line of P1 and P2, wherein the method is similar to that of calculating Q1, and the obtained new coordinate point is Q2. Q2 is also inserted into the point set, forming a new split point set containing five points P1, P2, P3, Q1, Q2.

Updating convex hull boundaries:

and (3) redefining peripheral points, and after adding Q1 and Q2, redefining the position relation of the five points to find out the points positioned at the outermost periphery, wherein the points form a new convex hull boundary. Starting from any one point, each point is checked in turn along one direction (clockwise, for example) to judge whether other points are outside the connecting line of the point and the adjacent points, if so, the convex hull boundary is adjusted until all peripheral points are determined.

And connecting the new boundaries, and sequentially connecting the determined peripheral points to form a new convex polygon, wherein the convex polygon is the minimum convex hull boundary containing five points. For example, if the new peripheral points are P1, Q2, P2, Q1, and P3 in this order, these five points are connected in sequence to form a pentagonal convex hull boundary.

Diagonal exchange detection and optimization:

traversing the quadrangle, and finding out any quadrangle formed by four adjacent points in the updated split point set, wherein for example, there may be quadrangles formed by P1, Q2, P2 and Q1, quadrangles formed by Q2, P2, Q1 and P3, and the like.

The diagonal exchange effect is detected, considering two diagonals for each quadrilateral, respectively. For example, for the quadrangle P1-Q2-P2-Q1, two diagonals are P1-P2 and Q2-Q1 respectively, the minimum interior angle of the new triangle formed after exchanging the diagonals is calculated, and the sizes of the minimum interior angles before and after exchanging are compared.

Updating the topological connection, if the minimum internal angle is increased after diagonal lines are exchanged, the new connection mode is more in line with the requirement of Delay internal triangularization, the original diagonal lines are replaced by the new diagonal lines, and the quadrilateral topological connection relation is updated. For example, if the smallest interior angle increases after exchanging P1-P2 and Q2-Q1, the quadrilateral P1-Q2-P2-Q1 is split into two triangles, P1-Q2-Q1 and Q2-P2-Q1.

Iterative optimization, repeating the diagonal exchange detection process, and checking and optimizing all possible quadrilaterals until all quadrilaterals meet the delaunay circle criterion, namely that no other points are contained in the circumscribed circle of each triangle. At this time, the split structure reaches an optimal state, and a final triangulated structure is generated.

Defining sub-regions and adjacency relations:

Dividing triangle subareas, defining each triangle in the final triangularization structure as a subarea, and recording three vertex coordinates of each subarea. For example, one sub-region may be constituted by vertices P1, Q2, Q1, another sub-region may be constituted by vertices Q2, P2, Q1, and so on.

Establishing an adjacency list, checking the shared edge condition of each sub-area and other sub-areas, recording the numbers of adjacent sub-areas, for example, if the sub-area A and the sub-area B share one edge, recording the adjacency of A and B in the adjacency list, and forming a complete adjacency list in this way, and describing the space connection relation between the sub-areas.

And extracting and calculating the sub-region pressure-flow coupling characteristic value:

And searching the pressure value, and for each sub-area, checking the coordinates of three vertexes of the sub-area in turn. In the original pipeline pressure gradient data acquired and stored in advance, pressure values corresponding to the three vertex coordinates are found. For example, if the three vertex coordinates of the sub-region are (X1, Y1), (X2, Y2), and (X3, Y3), the pressure gradient data is searched for the pressure values recorded at the three coordinate points, and the pressure values are extracted in a one-to-one correspondence.

And (3) obtaining flow phase amplitude, namely finding flow phase data in the same spatial range, namely a pipe network area covered by the subareas, and extracting the amplitude characteristic of the flow waveform, namely the difference value between the maximum value and the minimum value of the flow waveform, from the data. For example, the flow waveform in a certain subarea has a maximum value of 100 cubic meters per hour and a minimum value of 20 cubic meters per hour, and then the flow phase amplitude characteristic of the subarea is 80 cubic meters per hour.

Calculating a basic spatial coupling factor:

and multiplying the data, namely multiplying the pressure gradient data in the extracted subarea by the flow phase amplitude characteristics according to the corresponding relation. For example, if the pressure values corresponding to the three vertexes of the subarea are P1, P2, and P3, and the flow phase amplitude is a, then p1× A, P2 × A, P3 ×3×a is calculated respectively, and three sets of product data are obtained.

Arithmetic mean value is calculated, the obtained three groups of product data are added and divided by the number of data (here 3), and arithmetic mean value is calculated, and the mean value is the basic space coupling factor of the subarea and reflects the association degree of pressure and flow in the subarea in space dimension.

Generating a time sequence correction coefficient:

and locating fluctuation data, finding out the managed user access points according to the range of the subareas, screening out the fluctuation data belonging to the subareas from the gas time sequence fluctuation data of the user access points, and completely extracting the gas time sequence fluctuation data of the 3 user access points if 3 user access points are covered by the subarea A.

And calculating a variance value, and calculating a statistical variance value in a continuous time window for the extracted fluctuation data. The variance value is calculated to measure the degree of dispersion of the fluctuation data, and the larger the variance value is, the more intense the user's fluctuation with air is. For example, by calculating the air consumption data in a period of time, the variance value of the air fluctuation data for users in the subarea is 25.

And (3) carrying out linear normalization, carrying out linear transformation on the calculated variance value, and mapping the variance value into the interval of [0,1 ]. The specific operation is that the relative size of the variance value in all the subarea variance values is scaled according to a certain proportion. For example, if the variance value in all the sub-regions is 100 at the maximum and 0 at the minimum, and the current sub-region variance value is 25, the obtained value is in the [0,1] interval after calculation and conversion, and the value is the time sequence correction coefficient.

Generating an intermediate coupling feature:

and adding the data, and adding the basic space coupling factor obtained by the previous calculation with the time sequence correction coefficient to obtain a new value, wherein the value is an intermediate coupling characteristic value, and in this way, the pressure-flow correlation characteristic of the space dimension and the aerodynamic characteristic of the user of the time dimension are preliminarily fused.

Calculating a final characteristic value:

And (3) distributing a space weight coefficient, and distributing a space weight coefficient for each sub-region according to the importance degree of the sub-region in the whole pipe network topological structure, wherein the judgment of the importance degree can refer to whether the sub-region is close to a main pipe or not, whether the sub-region contains key user access points and other factors. For example, a sub-region near a voltage regulator station where the users are dense may be assigned a higher weight factor of 0.8, while a sub-region at the end of the network where the users are less may be assigned a lower weight factor of 0.2.

The final value is calculated, the intermediate coupling characteristic value is multiplied by the space weight coefficient distributed to the subarea, the obtained result is the final pressure-flow coupling characteristic value, the characteristic value comprehensively considers the space characteristic of the subarea, the time characteristic of gas consumption of a user and the importance degree of the subarea in a pipe network, and the coupling relation between the pressure and the flow in the subarea can be comprehensively reflected.

The pipeline pressure gradient data, the flow phase amplitude characteristic and the user gas time sequence fluctuation data are combined, the operation state of the pipeline is comprehensively evaluated from two dimensions of space and time, the limitation of the previous single dimension analysis is changed, the influence of gas flow characteristics and user gas behaviors on the pipeline is more comprehensively reflected, the basic space coupling factor reflects the spatial association of the pressure and the flow in the subarea, the time sequence correction coefficient reflects the fluctuation characteristics of the user gas in time, the unique operation characteristics of each subarea can be accurately described by combining the pressure gradient data and the flow phase amplitude characteristic with the gas time sequence fluctuation data, and the importance of key areas in the pipeline is highlighted by combining the space weight coefficient, so that the analysis result is more fit with the actual pipeline operation condition, and the detection of weak links and potential problem areas in the pipeline is facilitated.

The auxiliary vertex is dynamically inserted into the middle point of the connecting line between the P2 and the P3, between the P1 and the P2, the subdivision structure can be adaptively adjusted according to the gas time sequence fluctuation and the flow phase deviation condition of a user, the pressure and flow change sensitive area in the pipe network can be captured more finely, the analysis precision is improved, the triangulated structure generated by adopting the Delay subdivision algorithm has good geometric property (minimum internal angle maximization), the occurrence of an elongated triangle is avoided, the subdivision division is more uniform and reasonable, the subsequent feature extraction and analysis are facilitated, the operation state of the pipe network is comprehensively estimated from the two dimensions of space and time by combining the pressure gradient data, the flow phase amplitude feature and the gas time sequence fluctuation data, the influence of the gas flow characteristic and the gas behavior of the user on the pipe network is reflected more comprehensively, and the influence of key nodes (such as the vicinity of a pressure regulating station and a user dense area) in the pipe network can be highlighted by distributing the space weight coefficient for the subregion.

In a preferred embodiment of the present invention, the weighting correction is performed on the mutation probability feature vector of the user air demand by the pressure-flow coupling feature value, so as to generate a corrected mutation probability feature vector, which includes:

inputting the pressure-flow coupling characteristic values of all the subareas into a matrix converter, sorting according to the space weight coefficients of the subareas in the pipe network topology, and scaling the characteristic values by taking the pipe lengths governed by all the subareas as normalization factors to generate a diagonal correction coefficient matrix with the same dimension as the mutation probability characteristic vector;

carrying out Hadamard product operation on the correction coefficient matrix and the mutation probability feature vector according to a preset mutation probability feature vector required by the user to obtain a weighted feature vector;

and performing L2 norm normalization processing on the weighted feature vector to obtain a corrected mutation probability feature vector.

and sequencing the pressure-flow coupling characteristic values obtained by calculation of each subarea according to the size of the spatial weight coefficient of the subarea in the pipe network topology. For example, if the spatial weight coefficient of the subarea A is 0.8, the subarea B is 0.6, and the subarea C is 0.4, the characteristic value is ordered to be A, B, C.

Normalizing and scaling, namely scaling the sequenced characteristic values by taking the length of the pipeline governed by each subarea as a normalization factor. Dividing the characteristic value of each sub-area by the length of the pipeline governed by the sub-area to obtain a scaled characteristic value. For example, the characteristic value of the subarea a is 10, the length of the administered pipeline is 5 km, and the scaled value is 10+.5=2.

Generating a diagonal matrix, sequentially arranging the scaled eigenvalues on diagonal lines of the matrix, and filling other positions with 0 to form a diagonal matrix, wherein the dimension of the diagonal matrix is the same as that of the mutation probability eigenvector of the air demand of the user.

The Hadamard product operation specifically comprises:

preparing an original vector, and acquiring a preset user gas consumption demand mutation probability feature vector, wherein the vector comprises a plurality of dimensions, and each dimension represents different types of gas consumption demand mutation probabilities.

And multiplying the elements by each other, and carrying out Hadamard product operation on the diagonal correction coefficient matrix and the mutation probability feature vector, namely multiplying the elements at corresponding positions. For example, a first diagonal element of the diagonal matrix is multiplied by a first element of the eigenvector, a second diagonal element is multiplied by a second element, and so on, to obtain a new vector, which is the weighted eigenvector, with each element modified by the pressure-flow coupling eigenvalue.

L2 norm normalization:

The method comprises the steps of calculating the modular length of a vector, calculating the L2 norm of the weighted feature vector, namely the square root of the sum of squares of all elements of the vector, dividing each element of the weighted feature vector by the L2 norm of the element to obtain a corrected mutation probability feature vector, and carrying out normalization processing to obtain the modular length of the vector which is 1, so that the subsequent analysis and comparison are convenient, and meanwhile, the relative proportional relation among all elements of the vector is maintained.

In the embodiment of the invention, the influence of the actual running state of the pipe network on the gas demand of the user is fully considered by integrating the pressure-flow coupling characteristic value into the correction of the mutation probability characteristic vector, the change of the pipe network pressure and flow is often a precursor of the change of the user demand, and the data-driven correction method can more accurately capture the mutation trend of the gas demand of the user and improve the accuracy of a prediction model. The application of the spatial weight coefficient enables the pressure-flow characteristics of key areas (such as close to a voltage regulating station and a user dense area) in the pipe network to have greater influence on the mutation probability. The method is favorable for highlighting potential risks of the areas, so that gas enterprises can monitor and regulate more pertinently, the possible large fluctuation of gas demand is prevented in advance, the dimension influence caused by the difference of the lengths of pipelines of different subareas is eliminated by normalization processing, and the characteristic values of the subareas are comparable. Meanwhile, the L2 norm normalization ensures that the corrected feature vector is in a reasonable range, is convenient for subsequent fusion analysis with other models or indexes, and improves the overall stability and reliability of the system.

In a preferred embodiment of the present invention, tensor stitching is performed on the corrected mutation probability feature vector and the equivalent inner diameter dynamic feature vector of the pipeline, and the tensor stitching is input into a fully-connected diagnostic network to generate an caliber matching deviation index, including:

Expanding the corrected mutation probability feature vector and the real-time updated pipeline equivalent inner diameter dynamic feature vector into a three-dimensional tensor along a feature dimension, wherein the first dimension is a time step, the second dimension is a space partition, and the third dimension is a feature channel;

inputting a three-dimensional tensor into a three-layer fully connected diagnostic network, wherein:

The first layer, the linear transformation layer reduces the dimension of the input characteristic to 1/2 of the original dimension, and activates the function through the ReLU;

the second layer, the linear transformation layer further reduces to 1/4 of the original dimension, and activates the function through Sigmoid;

the third layer, the single neuron linear layer outputs the original value of the deviation degree;

and carrying out dynamic threshold calibration on the original value output by the third layer, and calculating the Z-score standardized offset of the current value relative to the reference distribution through a preset reference distribution curve fitted by the data of the historical normal working conditions.

And acquiring a corrected mutation probability feature vector reflecting the possibility of mutation of the user gas demand and a real-time updated pipeline equivalent inner diameter dynamic feature vector reflecting the condition that the inner diameter of the pipeline changes along with the running state of the pipe network.

The two feature vectors are expanded into a three-dimensional tensor along the feature dimension. The method comprises the steps of firstly determining a time step as a first dimension, for example, taking 1 hour as a time step, recording data at different moments, taking a second dimension as a space partition, corresponding to the previously divided pipe network subareas, each subarea having corresponding data record, taking a third dimension as a characteristic channel, respectively storing mutation probability characteristics and equivalent inner diameter dynamic characteristics of a pipeline, and filling data into positions of a three-dimensional tensor sequentially according to the sequence of the time step, the space partition and the characteristic channel to form a complete three-dimensional data structure.

Full connection diagnostic network operation:

And the first layer of operation inputs the constructed three-dimensional tensor into the first layer of the fully-connected diagnosis network, wherein the layer is a linear transformation layer, and the layer processes the input characteristic dimension and reduces the input characteristic dimension to 1/2 of the original dimension. After the processing is completed, the function can directly change all the characteristic values of negative numbers into 0 through the ReLU activation function, and positive characteristic values are reserved, so that more valuable characteristics are screened, and the nonlinear expression capacity of the network is enhanced.

And the second layer of operation, the data processed by the first layer enters the second layer, which is also a linear transformation layer, and the layer further reduces the characteristic dimension to 1/4 of the original dimension, so that deeper compression and characteristic extraction are carried out on the data. The function is then activated by Sigmoid, which maps the feature values between 0 and 1, converting the features into a probabilistic form.

And the third layer of operation, wherein the data output by the second layer enter a third layer of single neuron linear layer, the layer carries out final processing on the input data and outputs an original deviation value, and the original deviation value reflects the primary evaluation result of the caliber matching condition of the gas metering device in the current pipe network state.

Dynamic threshold calibration:

And screening all records marked as normal working conditions from the historical data, removing the data in abnormal states such as pipe network leakage, equipment failure, extreme gas consumption peak and the like, cleaning the screened data, removing missing values and obvious abnormal values, and ensuring the integrity and reliability of the data.

And calculating statistical characteristics, namely calculating statistics of caliber matching related indexes in normal working condition data, wherein the statistics comprise a mean value (reflecting the average level in a normal state), a standard deviation (reflecting the discrete degree of the data), a fractional number (such as 25% fractional and 75% fractional, used for describing the range of data distribution) and the like.

Normal distribution fitting calculation process:

The method comprises the steps of extracting an original data set (such as an original value of a historical deviation degree) of caliber matching related indexes from historical normal working condition data, ensuring that the data only comprise records of stable pipe network operation and no abnormal events, and sequencing the data set from small to large in value so as to facilitate the follow-up observation of data distribution trend. For example, 1000 pieces of deviation data under normal working conditions are arranged from small to large, and the concentration trend and the discrete range of the deviation data are observed.

Drawing a histogram, dividing the ordered data into a plurality of intervals (for example, taking 0.1 as a group distance), counting the data frequency in each interval, and drawing the histogram. And preliminarily judging whether the data show bell-shaped distribution characteristics of high middle and low two sides through the histogram, namely whether the data accord with the visual characteristics of normal distribution.

Calculating normal distribution key parameters:

and calculating a mean value mu, and dividing all the data by the total data to obtain the mean value. The mean value represents the center position of the normal distribution, for example, the mean value of 1000 pieces of data is 1.2, and the mean level of the deviation index under the normal working condition is 1.2.

And then squaring the variance to obtain the standard deviation, wherein the standard deviation reflects the discrete degree of the data, for example, the standard deviation is 0.3, and the deviation index generally fluctuates within the range of +/-0.3 of the mean under normal working conditions.

Generating a reference distribution curve:

determining the shape of a curve, drawing a bell-shaped curve which takes the mean as the center and takes the standard deviation as the width based on the calculated mean and standard deviation, wherein the highest point of the curve corresponds to the mean position, the curve gradually descends to two sides and theoretically extends to plus or minus infinity, but in practice, about 95% of data fall within the range of the mean plus or minus 2 times of the standard deviation, and about 99.7% of data fall within the range of the mean plus or minus 3 times of the standard deviation.

And marking key positions, namely marking key positions such as a mean value mu, a mean value + -1 time standard deviation (mu + -sigma), a mean value + -2 time standard deviation (mu + -2 sigma) and the like on the curve, and determining the distribution interval of data under normal working conditions. For example, if the mean is 1.2 and the standard deviation is 0.3, the curve contains about 68% data between 0.9 (μ - σ) and 1.5 (μ+σ) and about 95% data between 0.6 (μ -2σ) and 1.8 (μ+2σ).

The invention superimposes a probability density function curve of normal distribution on the histogram, and observes whether the curve basically coincides with the shape of the histogram. For example, if the peak position of the histogram is consistent with the peak position of the curve and the frequency attenuation trend of the two sides is consistent with the curve, the fitting effect is good, and whether the fitted normal distribution can reasonably describe the data distribution characteristics is verified through visual observation or a simple statistical method (such as whether the actual duty ratio of the calculated data in the range of mu+/-sigma and mu+/-2 sigma is close to the theoretical duty ratio of 68% and 95%). And converting the index distribution under normal working conditions into two key parameters of mean value and standard deviation by fitting normal distribution, and clearly defining the numerical range (such as mu+/-2 sigma) of a normal state. The operator can quickly determine whether the current data falls within the normal interval, for example, when the deviation index is 2.0 at a certain time and μ=1.2 and σ=0.3, 2.0> μ+2σx1.8 can be immediately identified as an abnormality.

Unifying abnormal judgment standards, and enabling indexes with different dimensions and different value ranges to be uniformly measured by using the standard deviation multiple of the distance mean value by using the standardized characteristic (such as Z-score) of normal distribution. For example, both the pressure difference index (unit kPa) and the flow fluctuation index (unit m ³/h) can be converted into comparable abnormal grades by the Z-score, thereby avoiding confusion of judgment caused by index unit difference.

The reference distribution curve is not fixed, the mean value and the standard deviation can be regularly recalculated according to the newly collected normal working condition data, and the curve form is updated, so that the system can adapt to long-term changes (such as user structure adjustment and equipment aging) of pipe network operation, and the timeliness of the abnormal judgment standard is continuously maintained. For example, the increase of the air consumption of the user in the winter heating period may cause the overall upward shift of the mean value and standard deviation of the deviation index, and the erroneous judgment of the normal fluctuation as abnormal can be avoided after re-fitting.

Based on the "3σ principle" of normal distribution (i.e., 99.7% of the data fall within μ±3σ), a case where the Z-score absolute value is greater than 3 can be defined as "extremely high risk", and emergency pre-warning is automatically triggered. The early warning logic based on the statistical rule does not need manual experience intervention, reduces subjective judgment errors, and simultaneously ensures quick response to extreme abnormal states.

And (3) visually verifying, namely comparing the fitted distribution curve with a histogram or a probability density chart of actual data, checking whether the fitting effect accords with the expectation, and ensuring that the curve can accurately reflect the distribution rule of the index under the normal working condition.

Calculate normalized offset (Z-score):

and (3) positioning the original value position, substituting the original value of the deviation degree output by the fully-connected network into the mathematical expression of the reference distribution curve, and determining the relative position of the original value in the reference distribution. For example, if the reference distribution is a normal distribution and the original value is located on the right side of the mean value, the direction indicating that the current state deviates from the normal level is "positive".

And calculating the deviation degree, namely calculating the difference value (x-mu) between the original value x and the mean value according to the mean value mu and the standard deviation sigma of the reference distribution, and dividing the difference value by the standard deviation to obtain the standardized offset (Z-score). This value represents the standard deviation multiple of the original value from the mean, e.g. z=1 represents 1 standard deviation above the mean and z= -1 represents 1 standard deviation below the mean.

Mapping the standard deviation amount to a specific deviation degree index according to the service requirement. For example, the corresponding relation between the absolute value range of the Z-score and the deviation degree grade is set to be normal, the absolute value of the Z is less than or equal to 1, the absolute value of the Z is less than or equal to 2, and the absolute value of the Z is more than 2, so that operators can intuitively understand the current pipe network state.

The method realizes data fusion of time, space and different feature dimensions by constructing the corrected mutation probability feature vector and the pipeline equivalent inner diameter dynamic feature vector into the three-dimensional tensor, can comprehensively capture dynamic information of user demand change and pipeline state change in the pipe network operation process, and gradually performs feature extraction and dimension reduction on input data through multi-layer linear transformation and activation function processing by the fully connected diagnosis network. The method has the advantages of effectively reducing data dimension, reducing calculation complexity, retaining key characteristics, highlighting factors which have important influence on aperture matching, improving analysis precision and efficiency, carrying out dynamic threshold calibration on the basis of a reference distribution curve fitted by historical normal working condition data, and being capable of adaptively adjusting the evaluation standard of deviation indexes according to the actual condition of pipe network operation. Compared with a fixed threshold mode, the method can be more suitable for the change of the running state of the pipe network, accurately identify abnormal conditions, avoid misjudgment and missed judgment, and improve the reliability of the caliber matching deviation index.

As shown in fig. 2, the embodiment of the invention further provides an intelligent analysis and diagnosis method for caliber matching of a gas metering appliance, which comprises the following steps:

Embodiments of the invention also provide a computing device comprising a processor, a memory storing a computer program which, when executed by the processor, performs a method as described above. All the implementation manners in the method embodiment are applicable to the embodiment, and the same technical effect can be achieved.

Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform a method as described above. All the implementation manners in the method embodiment are applicable to the embodiment, and the same technical effect can be achieved.

While the foregoing is directed to the preferred embodiments of the present invention, it will be appreciated by those skilled in the art that various modifications and adaptations can be made without departing from the principles of the present invention, and such modifications and adaptations are intended to be comprehended within the scope of the present invention.

Claims

1. An intelligent analysis and diagnosis system for gas metering instrument caliber matching, comprising:

The acquisition module is used to collect pipeline pressure gradient data, flow phase data, and user gas consumption time series fluctuation data in real time through pressure sensors deployed at the three-dimensional spatial coordinate positions of the gas pipeline network trunk nodes, user access points, and pressure regulating stations;

The extraction module is used to select the main pipeline intersection point P1, the user access point P2, and the pressure regulating station output point P3 as three dynamic topology reference points based on the pipeline pressure gradient data, the flow phase data, and the user gas consumption time series fluctuation data; construct a dynamic topology unit based on the three points, divide the unit into regions based on the Delaunay subdivision algorithm, and extract the pressure-flow coupling characteristic value of the subregion, including: executing the Delaunay subdivision algorithm on the dynamic topology unit to generate a minimum angle division structure containing at least three subregions, including: using the real-time two-dimensional plane coordinates of the three points P1, P2, and P3 in the dynamic topology unit as the subdivision input point set, and constructing an initial convex Envelope boundary, based on the real-time change amplitude of the user's gas consumption time series fluctuation data, when the fluctuation amplitude exceeds the set threshold, an auxiliary vertex Q1 is dynamically inserted at the midpoint of the line connecting P2 and P3. According to the phase offset of the flow phase data, an auxiliary vertex Q2 is inserted at the midpoint of the line connecting P1 and P2. Q1 and Q2 are added to the segmentation point set and the convex hull is updated to obtain the updated segmentation point set. The updated segmentation point set is traversed and a diagonal exchange test is performed on the quadrilateral formed by any four adjacent points. If the minimum internal angle increases after the exchange, the topological connection relationship is updated. The iterative optimization is carried out until all quadrilaterals meet the Delaunay empty circle criterion to generate the final triangulated structure.

Each minimum topological unit in the final triangulated structure is defined as a sub-region to obtain a sub-region vertex coordinate set and an adjacency table; for each divided sub-region, the pipeline pressure gradient data associated with the vertex coordinates of the sub-region is extracted, and the flow phase data amplitude characteristics in the same spatial domain are simultaneously obtained; the arithmetic mean of the product of the pressure gradient data and the flow phase amplitude characteristics is calculated, and the basic spatial coupling factor of the sub-region is output; based on the basic spatial coupling factor, the gas consumption time series fluctuation data of the user access point under the jurisdiction of the sub-region is located, the statistical variance value of the user fluctuation data under the jurisdiction in the continuous time window is calculated, and the variance value is linearly normalized in the interval [0, 1] to generate a time series correction coefficient; the basic spatial coupling factor and the time series correction coefficient are added to generate an intermediate coupling feature; the intermediate coupling feature is multiplied by the spatial weight coefficient of the sub-region in the pipeline network topology, and finally the pressure-flow coupling feature value is output;

The correction module is used to perform weighted correction on the user gas demand mutation probability characteristic vector using the pressure-flow coupling characteristic value to generate a corrected mutation probability characteristic vector, including:

The pressure-flow coupling eigenvalues of each sub-region are input into the matrix converter, sorted by the spatial weight coefficients of the sub-regions in the pipe network topology, and scaled by the length of the pipes under each sub-region as a normalization factor to generate a diagonal correction coefficient matrix with the same dimension as the mutation probability eigenvector.

According to the preset user gas demand mutation probability eigenvector, the correction coefficient matrix and the mutation probability eigenvector are subjected to a Hadamard product operation to obtain a weighted eigenvector;

Perform L2 norm normalization on the weighted feature vector to obtain the corrected mutation probability feature vector;

The processing module is used to perform tensor splicing on the modified mutation probability feature vector and the pipeline equivalent inner diameter dynamic feature vector, and input the result into a fully connected diagnostic network to generate a caliber matching deviation index, including:

According to the corrected mutation probability feature vector and the real-time updated dynamic feature vector of the pipeline equivalent inner diameter, it is expanded into a three-dimensional tensor along the feature dimension, where the first dimension is the time step, the second dimension is the spatial partition, and the third dimension is the feature channel. The three-dimensional tensor is input into a three-layer fully connected diagnostic network, where: in the first layer, the linear transformation layer reduces the input feature dimension to 1/2 of the original dimension and passes the ReLU activation function; in the second layer, the linear transformation layer further reduces it to 1/4 of the original dimension and passes the Sigmoid activation function; in the third layer, the single neuron linear layer outputs the original value of the deviation, and the original value output by the third layer is dynamically threshold calibrated. The Z-score standardized offset of the current value relative to the benchmark distribution is calculated through the benchmark distribution curve fitted with the preset historical normal operating data; when the deviation exceeds the set threshold, the linkage time series prediction network outputs a safe redundant caliber replacement plan.

2. The intelligent analysis and diagnosis system for gas metering instrument caliber matching according to claim 1 is characterized by real-time collection of pipeline pressure gradient data, flow phase data, and user gas usage time series fluctuation data through pressure sensors deployed at three-dimensional spatial coordinate locations of gas pipeline network trunk nodes, user access points, and pressure regulating stations, including:

Establish the pipe network topology mapping relationship based on the three-dimensional spatial coordinates of each sensor;

Synchronously collect real-time pressure values along the pipeline between trunk nodes, calculate the pressure difference and associated spatial distance between adjacent nodes, and generate pressure gradient data;

Monitor the flow sensor output at the user access point, extract the phase offset and amplitude characteristics of the flow waveform within the cycle, and generate flow phase data;

The instantaneous fluctuation signal of gas consumption downstream of the pressure regulating station is recorded, and the fluctuation frequency and the absolute difference between the peak and the trough in unit time are calculated to form the time series fluctuation data of gas consumption of users.

3. The intelligent analysis and diagnosis system for gas metering instrument caliber matching according to claim 2 is characterized in that, based on the pipeline pressure gradient data, flow phase data, and user gas consumption time series fluctuation data, the main pipeline intersection point P1, the user access point P2, and the pressure regulating station output point P3 are selected as three dynamic topology reference points, including:

According to the pipe network topology mapping relationship, the main pipe intersection point with the largest pressure change rate in the pressure gradient data is selected as P1, and its three-dimensional space coordinates are used as the initial reference coordinates;

Using the initial reference coordinates of the main pipeline intersection point P1 as a reference, combined with the phase offset and amplitude characteristics in the flow phase data, the spatial influence weight of the user access point P2 is calculated. The original three-dimensional coordinates of the user access point P2 are offset compensated based on the spatial influence weight to obtain the weighted real-time monitoring coordinates of the user access point P2.

Based on the weighted real-time monitoring coordinates of the user access point P2, the original coordinates of the pressure regulating station output point P3 are compensated for time series displacement using the fluctuation frequency and peak-to-trough difference in the user's gas consumption time series fluctuation data to obtain the real-time control coordinates of the pressure regulating station output point P3.

The initial reference coordinates of the main pipeline intersection point P1, the real-time monitoring coordinates of the user access point P2, and the real-time control coordinates of the pressure regulating station output point P3 are spatially synchronized and converted to a two-dimensional plane through dimensionality reduction mapping; the three points P1, P2, and P3 in the two-dimensional plane are connected to form a dynamic topological unit.

4. A method for intelligent analysis and diagnosis of gas metering instrument caliber matching, characterized in that the method is applied to the system according to any one of claims 1 to 3, and the method comprises:

Step 1: Use pressure sensors deployed at the three-dimensional coordinate locations of the gas pipeline network trunk nodes, user access points, and pressure regulating stations to collect pipeline pressure gradient data, flow phase data, and user gas consumption time series fluctuation data in real time;

Step 2: Based on pipeline pressure gradient data, flow phase data, and user gas consumption time series fluctuation data, the main pipeline intersection point P1, user access point P2, and pressure regulating station output point P3 are selected as three vertices, and a dynamic triangle is formed based on the three vertices. The dynamic triangle is meshed using the Delaunay triangulation algorithm, and the pressure-flow coupling eigenvalue of the mesh is extracted.

Step 3: Perform weighted correction on the user gas demand mutation probability characteristic vector using the pressure-flow coupling characteristic value to obtain a corrected mutation probability characteristic vector;

In step 4, the corrected mutation probability feature vector is tensor-concatenated with the pipeline equivalent inner diameter dynamic feature vector and input into the fully connected diagnostic network to generate a caliber matching deviation index. When the deviation exceeds the threshold, the linked timing prediction network outputs a safe and redundant caliber replacement plan.

5. A computing device, comprising:

one or more processors;

A storage device for storing one or more computer programs, wherein when the one or more computer programs are executed by the one or more processors, the one or more processors implement the method according to claim 4.

6. A computer-readable storage medium, characterized in that a computer program is stored in the computer-readable storage medium, so that when the computer program is executed by a processor, the method according to claim 4 is implemented.