Gas pipeline operation monitoring system based on multi-source data detection

Abstract

The invention relates to the technical field of gas pipeline operation monitoring, in particular to a gas pipeline operation monitoring system based on multi-source data detection, which comprises the following steps: constructing an operation monitoring agent and integrating multi-source data (environment interference parameters, transmission data quality indexes and historical monitoring decision indexes) to generate state vectors; the system can accurately sense a complex operation environment, introduces a comprehensive reward function including sensitive area consequences, maintenance cost optimization and environmental friendliness, performs hierarchical reward calculation on a sensitive area, introduces data value density evaluation on the maintenance cost, and integrates full-chain energy consumption and carbon emission management and control on environmental friendliness. The system can quantify comprehensive benefits of a monitoring strategy from multiple dimensions, continuous optimization of action combination is achieved, and based on feedback of a reward function, an intelligent agent can dynamically adjust the weight of a sensor, a fusion algorithm and a working mode.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of gas pipeline operation monitoring, more particularly, it relates to a gas pipeline operation monitoring system based on multi-source data detection. BACKGROUND

[0002] The existing gas pipeline operation monitoring system has obvious short boards in robustness and adaptability under complex working conditions. The traditional system mostly adopts a static data fusion strategy and a fixed sensor configuration, which is difficult to dynamically respond to diversified environmental interference. When facing complex interference such as urban background noise, industrial electromagnetic radiation and extreme weather, the monitoring accuracy will suddenly decrease due to signal distortion, data conflict or sensor performance degradation, especially in high-sensitivity areas, which is prone to false negatives and false positives, and it is difficult to meet the rigid demand of pipeline network safe operation. At the same time, the traditional system lacks dynamic evaluation of data quality, and when encountering transmission interruption, data loss and other abnormalities, it lacks an adaptive strategy adjustment mechanism, which further aggravates the fluctuation of monitoring reliability.

[0003] In addition, the existing technology has significant limitations in target optimization: most systems only focus on a single safety indicator, ignoring the coordinated control of maintenance cost and environmental impact. Some solutions over-configure resources in pursuit of monitoring accuracy, resulting in high operation and maintenance costs; others simplify the monitoring process to reduce costs, sacrificing safety redundancy; and the consideration of environmental factors such as energy consumption and carbon emissions in the monitoring process is even more lacking. This single and fragmented optimization mode is difficult to adapt to the dynamic balance demand of modern gas pipeline networks between safety protection, economic operation and green development, restricting the comprehensive performance improvement of the monitoring system.

[0004] Based on the above, the present application provides a gas pipeline operation monitoring system based on multi-source data detection. SUMMARY

[0005] In view of the deficiencies in the prior art, the purpose of the present application is to provide a gas pipeline operation monitoring system based on multi-source data detection.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical solutions: The gas pipeline operation monitoring system based on multi-source data detection comprises An operation monitoring intelligent agent construction module is used to construct an intelligent agent for gas pipeline operation monitoring. A parameter index state vector generation module is used to periodically collect various environmental interference parameters, various transmission data quality indicators and various historical monitoring decision indicators that affect gas pipeline monitoring, and further determine the state vector of the current operation monitoring environment. An operation monitoring execution module is used to import the state vector of the current operation monitoring environment into the intelligent agent, and the intelligent agent outputs an action combination, and the action combination is executed. The running monitoring optimization module is run, the action combination is executed, the gas pipeline running monitoring reward function after the execution of the action combination is calculated, and the action combination of the gas pipeline running monitoring is continuously optimized based on the gas pipeline running monitoring reward function.

[0007] Further, the construction process of the intelligent agent: S1: define the state space, action space and interference influence matrix; S2: define the policy network structure.

[0008] Further, the determination step of the state vector of the current running monitoring environment: splice various environmental disturbance parameters, various transmission data quality indicators and various historical monitoring decision indicators into a state vector S=[s1, s2,..., s n ], wherein each parameter ∈ [0, 1], and n is the total number of parameters and indicators.

[0009] Further, the gas pipeline running monitoring reward function: is the sensitive area consequence reward value, is the maintenance cost optimization reward value, is the environmental friendliness reward value; a1 is the first reward coefficient, a2 is the second reward coefficient, and a3 is the third reward coefficient; a1+a2+a3=1.

[0010] Further, the sensitive area consequence reward value is obtained as follows: the gas pipeline containing the extreme sensitive area, the medium sensitive area and the low sensitive area is obtained, the single-area reward of the extreme sensitive area, the single-area reward of the medium sensitive area and the single-area reward of the low sensitive area are obtained, and the sensitive area consequence reward value is calculated based on the single-area reward of the extreme sensitive area, the single-area reward of the medium sensitive area and the single-area reward of the low sensitive area.

[0011] Further, the single-area reward of the extreme sensitive area is calculated based on the extreme sensitive area accuracy, the extreme sensitive area false alarm rate and the response delay coefficient.

[0012] Further, the single-area reward of the medium sensitive area is calculated based on the medium sensitive area accuracy, the medium sensitive area false alarm rate and the maintenance cost correction term.

[0013] Further, the single-area reward of the low sensitive area is calculated based on the low sensitive area accuracy, the low sensitive area false alarm rate and the energy consumption exceeding coefficient.

[0014] Further, the maintenance cost optimization reward value is calculated based on the standard calibration times, the actual calibration times and the data value density.

[0015] ​Further, the environmental friendliness reward value is calculated based on the energy consumption reduction rate in the trough period, the area of the region in which the gas concentration in the extremely sensitive area exceeds the standard, and the system carbon emission intensity.

[0016] Compared with the prior art, the present application has the following beneficial effects: The system of the present application can accurately perceive the complex operating environment by constructing an operating monitoring intelligent agent and integrating multi-source data (environmental interference parameters, transmission data quality indicators, and historical monitoring decision indicators) to generate a state vector. The action combination output by the intelligent agent can dynamically adapt to different interference scenarios (such as physical, electromagnetic, and meteorological interference). Through sensor weight adjustment, fusion algorithm switching, and working mode optimization, the system can effectively resist the influence of environmental interference on monitoring accuracy, solve the problem of performance degradation of traditional systems in strong interference or data anomaly scenarios, and significantly improve the robustness and adaptability of gas pipeline operation monitoring. At the same time, the system introduces a comprehensive reward function including sensitive area consequences, maintenance cost optimization, and environmental friendliness. Through hierarchical reward calculation for sensitive areas, data value density evaluation for maintenance costs, and integration of energy consumption and carbon emission control throughout the chain for environmental friendliness, the system can quantify the comprehensive benefits of monitoring strategies from multiple dimensions, continuously optimize the action combination, and based on the feedback of the reward function, the intelligent agent can dynamically adjust the sensor weight, fusion algorithm, and working mode. In a complex interference environment, the system can maintain monitoring robustness while reducing invalid energy consumption and redundant operations, reducing the total life cycle operation and maintenance cost, and through the use of renewable energy and carbon emission control, it can help green and low-carbon operation. BRIEF DESCRIPTION OF DRAWINGS

[0017] Fig. 1 The principle block diagram of the gas pipeline operation monitoring system based on multi-source data detection; Fig. 2 The construction principle diagram of the interference influence matrix. DETAILED DESCRIPTION

[0018] Reference Figs. 1-2 The gas pipeline operation monitoring system based on multi-source data detection includes an operating monitoring intelligent agent construction module, a parameter indicator state vector generation module, an operating monitoring execution module, and an operating monitoring optimization module. The running monitoring intelligent agent construction module is configured to construct an intelligent agent for gas pipeline running monitoring. The construction process of the intelligent agent includes the following steps: S1, defining a state space, an action space, and an interference influence matrix; S2, defining a policy network structure (the policy network structure is the brain of the intelligent agent, and is responsible for mapping a state vector into a specific action (sensor weight, fusion algorithm, and working mode). The policy network structure adopts a shared feature layer and a branch output layer architecture: the shared feature layer is a two-layer fully connected network (FC) that maps a multi-dimensional state vector into multi-dimensional abstract features (extracting the core law of the state, such as strong wind = unreliable sound sensor); the continuous action branch is configured to output a multi-dimensional original weight through a two-layer FC for sensor weight adjustment, and a final weight vector is obtained after Softmax and constraint processing (ensuring that the sum is 1 and a single value is less than or equal to 0.5); the discrete action branch is configured to output a probability distribution (for example, the probability of algorithm 1 is 0.7, and the probability of algorithm 2 is 0.2) through an independent FC layer for fusion algorithm switching (four options) and working mode switching (for example, three sampling frequencies), and a discrete action is generated through roulette sampling); and S3, defining a value network (the value network is configured to evaluate the expected long-term return (i.e., a Q value) of performing an action A in a current state S, and provide a feedback signal for optimization of the policy network. The network structure: inputting a state S and an action A, and outputting a Q value (scalar, a larger value indicates a better action) of the action through a three-layer FC: the first layer: concatenating S and A into a multi-dimensional vector; the middle layer: mapping into multi-dimensional abstract features; and the output layer: outputting a Q value (for example, Q = 8.5, indicating that the expected long-term return of the action is 8.5 points). The core role: telling the policy network whether the action is good or not through the Q value, and the policy network adjusts parameters according to the high or low of the Q value (for example, actions with high Q values are more adopted). The state space is defined based on various environmental interference parameters, various transmission data quality indicators, and various historical monitoring decision indicators. In the embodiment, the environmental interference parameters are divided into three categories, namely, physical interference parameters, electromagnetic interference parameters, and meteorological interference parameters. The physical interference parameters include urban background noise (sound pressure level SPL, unit dB), third-party construction vibration (vibration acceleration, unit m / s²), and ground subsidence (displacement mm). The electromagnetic interference parameters include industrial equipment electromagnetic radiation intensity (dBμV / m), signal transmission channel signal-to-noise ratio SNR (dB), and same-frequency interference signal intensity. The meteorological interference parameters include wind grade (wind speed m / s), rainfall (mm / h), and temperature and humidity (℃ / %RH), especially for open-air pipelines and overhead sections. In the detailed description, the transmission data quality indicators are divided into six categories, namely, sound sensor signal-to-noise ratio (SNR_sound, 0-1, the higher the more reliable), electromagnetic flowmeter measurement error (Err_flow, 0-1, 1 when the error is greater than 20%), soil gas sensor data integrity (Comp_gas, 0-1, 1 when the loss rate is less than 5%), infrared imager contrast (Cont_ir, 0-1, reduced in rainy and snowy weather), distributed optical fiber signal stability (Stab_fiber, 0-1, reduced in vibration interference), and wireless transmission packet loss rate (Loss_rate, 0-1, 1 when the packet loss rate is greater than 20%); The historical monitoring decision indicators are divided into three categories, namely, decision accuracy (Accuracy, 0-1, correct decision number / total decision number), average response delay (Delay, 0-1, 1 when greater than 5s), and false alarm rate (False_alarm, 0-1, false alarm number / total alarm number). The historical monitoring decision indicators usually obtain the decision accuracy, average response delay, and false alarm rate of the last 10 historical monitoring decisions. Design the action space for gas pipeline operation monitoring; The action space is divided into action 1, action 2, and action 3. Action 1: Sensor weight dynamic allocation (continuous action, core adjustment means); for all sensors involved in various environmental interference parameters, allocate fusion weights w i -w I , i=1, 2, …, I, I is the total number of sensors, i represents the corresponding sensor, and satisfies: constraint condition: ∑w i =1 (the sum of weights is 1), and wᵢ∈[0,0.5] (the weight of a single sensor does not exceed 50%, to avoid excessive dependence); adjustment granularity: ±0.1 can be adjusted based on the current weight (e.g., the sound sensor current w=0.4 can be adjusted to 0.3 or 0.5) each time.

[0019] Action 2: Fusion algorithm switching (discrete action, to deal with data conflicts / losses); four fusion algorithms are built-in, and are dynamically switched according to data quality: A1: weighted Bayesian fusion; A2: D-S evidence theory; A3: graph neural network (GNN) interpolation fusion; A4: attention mechanism Transformer fusion.

[0020] Action 3: Sensor operating mode switching (discrete + continuous action, optimize energy consumption and anti-interference); For sensors susceptible to interference, design switchable operating modes: Sound sensor: sampling frequency (1 kHz / 5 kHz / 10 kHz, reduce noise input to 1 kHz when strong interference); Infrared imager: mode switching (normal mode / rainproof mode / night enhancement mode, automatically switch to rainproof mode in rainy and snowy weather); UAV-mounted sensors: patrol frequency (5 min / 10 min / 30 min, high-risk areas encrypted to 5 min); Electromagnetic flowmeter: calibration frequency (1 h / 4 h / 24 h, strong electromagnetic interference encrypted to 1 h).

[0021] Construct an interference influence matrix based on various environmental interference parameters; Construction process of interference influence matrix: Determine the interference mechanism of physical interference parameters; Urban background noise (SPL, dB): interference object: sound sensor (the core goal is to leak sound, background noise will mask the leak sound, causing the signal-to-noise ratio to drop). No impact: electromagnetic flowmeter, soil gas sensor (not dependent on sound waves).

[0022] Third-party construction vibration (vibration acceleration, m / s²): interference object: distributed optical fiber strain sensor (the optical fiber may mistakenly consider construction vibration as pipe deformation itself), sound sensor (construction machinery noise masks leak sound). Effective signal: vibration sensor (the core goal is to detect construction vibration, so construction vibration is an effective input, not interference).

[0023] Ground subsidence (displacement mm): interference object: distributed optical fiber strain sensor (subsidence may cause additional strain on the pipe, which is confused with corrosion / damage-induced strain). No impact: sound sensor, gas sensor (not related to subsidence).

[0024] Determine the interference mechanism of electromagnetic interference parameters; Industrial equipment electromagnetic radiation (dBμV / m): interference object: electromagnetic flowmeter (based on electromagnetic induction principle, strong electromagnetic radiation will distort the magnetic field, causing flow measurement error), wireless transmission unit (such as sensor data wireless communication affected by electromagnetic interference, resulting in packet loss). No impact: soil gas sensor (electrochemical principle), infrared imager (optical principle).

[0025] Signal transmission channel signal-to-noise ratio SNR (dB): interference object: all sensors relying on wireless transmission (such as AI vision camera, UAV-mounted sensors, low SNR will cause data transmission distortion). No impact: wired transmission sensors (such as distributed optical fiber, fixed installation electromagnetic flowmeter).

[0026] Co-channel interference signal strength: interference object: sensors using wireless co-channel communication (e.g., multiple vibration sensors using the same frequency will interfere with each other, causing signal confusion). No impact: wired sensors, wireless sensors of different frequencies.

[0027] Interference mechanism for determining meteorological interference parameters; Wind level (wind speed m / s): interference object: sound sensor (wind noise covers leakage sound), infrared imager (strong wind causes air turbulence, distorts infrared thermal field distribution), sensors carried by unmanned aerial vehicles (wind causes unmanned aerial vehicles to shake, image / data acquisition is blurred). No impact: soil gas sensor (buried, not affected by wind), electromagnetic flowmeter (measured in pipeline, unrelated to wind).

[0028] Rainfall (mm / h): interference object: infrared imager (rainwater attached to the lens causes imaging blur), AI vision camera (rainwater on the lens causes image distortion), soil gas sensor (heavy rain may dilute the concentration of fuel gas in the soil, causing detection sensitivity to decrease). No impact: distributed optical fiber (buried and sealed), electromagnetic flowmeter (in the pipeline).

[0029] Temperature and humidity (℃ / %RH): interference object: soil gas sensor (high humidity may affect the service life of the electrochemical probe, causing drift), sound sensor (temperature changes may change the speed of sound wave propagation, affecting positioning accuracy). Weak impact: infrared imager (when the ambient temperature is close to the temperature of the leakage area, the contrast decreases).

[0030] Quantify interference weight, use interference coefficient (0-1) to quantify the degree of influence: 0: no interference (this environmental parameter does not affect the detection accuracy of the sensor); 0.1-0.3: weak interference (affecting detection accuracy <10%); 0.4-0.6: moderate interference (affecting 10%-30%); 0.7-1.0: strong interference (affecting >30%).

[0031] Assignment basis: laboratory data (e.g., under an electromagnetic radiation field intensity of 100 dBμV / m, the electromagnetic flowmeter error reaches 40% → interference coefficient 0.8); industry standards (e.g., requirements for sensor anti-interference performance in GB50494-2009 "Technical Code for Urban Gas"); expert experience (combined with gas company operation and maintenance data, such as a 60% increase in false alarm rate of sound sensors in strong wind weather → interference coefficient 0.8).

[0032] Based on the above analysis, an interference influence matrix of interference parameters-sensors is formed.

[0033] Parameter index state vector generation module, regularly collect various environmental interference parameters affecting gas pipeline monitoring, various transmission data quality indicators, and various historical monitoring decision indicators to further determine the state vector of the current running monitoring environment; The determination step of the state vector of the current running monitoring environment: splice various environmental interference parameters, various transmission data quality indicators, and various historical monitoring decision indicators into a state vector S=[s1, s2,..., s n ], wherein each parameter ∈ [0, 1], and n is the total number of parameters and indicators. For example, a state vector of strong wind (wind force s=0.8) + high industrial equipment electromagnetic radiation (s=0.9) + low signal-to-noise ratio of the sound sensor (s=0.2) + high false alarm rate (s=0.7) intuitively reflects the current strong interference + low-quality data + monitoring strategy optimization scenario; The running monitoring execution module imports the state vector of the current running monitoring environment into the agent, and the agent outputs an action combination. The action combination is executed (the specific process of action execution: sensor weight adjustment: real-time update the fusion weight of each sensor through the fusion engine API (for example, adjust the sound sensor weight from 0.4 to 0.1, and adjust the soil gas sensor weight from 0.3 to 0.5); fusion algorithm switching: trigger the algorithm scheduling module, interrupt the current algorithm (such as weighted Bayesian), load the target algorithm (such as D-S evidence theory), and reuse the collected sensor data to recalculate; work mode switching: send control instructions to the sensor (such as switching the infrared imager to “rainproof mode” — start lens heating and defogging, and reduce exposure time)); The running monitoring optimization module, after the execution of the action combination, calculates the gas pipeline running monitoring reward function after the execution of the action combination, and continuously optimizes the action combination of the gas pipeline running monitoring based on the gas pipeline running monitoring reward function; Gas pipeline running monitoring reward function: ; is the consequence reward value of the sensitive area, is the maintenance cost optimization reward value, is the environmental friendliness reward value; a1 is the first reward coefficient, a2 is the second reward coefficient, and a3 is the third reward coefficient; a1+a2+a3=1, in the urban core area network, a1 is 0.6-0.7, a2 is 0.2-0.25, and a3 is 0.1-0.2, the core logic is that the population is dense, the high consequence area is concentrated, the safety priority is the highest, the environment is second (avoiding leakage pollution of residential areas), and the cost is moderately considered; in the ecological sensitive area network, a1 is 0.5-0.6, a2 is 0.2, and a3 is 0.2-0.3, the core logic is that it is adjacent to water sources / natural reserves, the environmental friendliness and safety are equally important, and the cost weight is reduced (allowing moderate investment to protect the ecology), in the remote suburban network, a1 is 0.5, a2 is 0.3-0.4, and a3 is 0.2-0.3, the core logic is that the population is sparse, the high consequence area is less, the cost control priority is raised, but the safety is still the basis (avoiding major accidents); Consequence reward value of sensitive area The acquisition process is as follows: Identify the extremely sensitive, moderately sensitive, and low-sensitivity areas within the gas pipeline network. (The classification of these areas must be based on the severity of potential consequences of a leak, taking into account regional population density, environmental sensitivity, intensity of third-party activities, and the pipeline's inherent risk characteristics. Quantitative indicators are developed according to industry standards (such as China's "Technical Guidelines for Identification and Risk Assessment of High-Consequence Areas in Gas Pipelines" and the US ASME B31.8S). The core logic of this classification is: the higher the potential risk of personal injury, environmental damage, and economic loss after a leak, the higher the area's sensitivity level.) Obtain the single-area reward R1 for extremely sensitive areas, R2 for moderately sensitive areas, and R3 for low-sensitivity areas. Calculate the consequence reward value for sensitive areas The values ​​for r1 can be 0.5, r2 can be 0.3, and r3 can be 0.2. Statistical data shows that the consequences (personnel casualties + economic losses) of a single leak in an extremely sensitive area are approximately 1.5-2 times that of a moderately sensitive area, and approximately 1.5 times that of a low-sensitivity area. The coefficients 0.5:0.3:0.2 (approximately a ratio of 5:3:2) precisely match this "difference in the magnitude of consequences"—0.5 is 1.67 times that of 0.3, and 0.3 is 1.5 times that of 0.2, consistent with the actual risk ratio.

[0034] Single-zone reward for extremely sensitive areas ; The accuracy rate in the extremely sensitive area is (number of correct identifications / total number of decisions, ∈ [0,1], where the number of correct identifications is the total number of decisions made by the system in the extremely sensitive area that are completely consistent with the actual situation, and the total number of decisions is the total number of all independent decisions made by the system in the extremely sensitive area). The false alarm rate for extremely sensitive areas is (actual number of unidentified leaks / total number of leaks, ∈ [0,1]. The number of unidentified leaks is the number of times a gas leak actually occurred in the extremely sensitive area (confirmed by manual review or equipment verification), but the monitoring system did not issue any warning (including "judged as normal" or "no decision output"). The total number of leaks is the total number of confirmed actual leaks in the extremely sensitive area. The response delay coefficient is calculated as (actual response time / standard time 30 seconds, ∈ [0,2], with a maximum of 2 for any time exceeding 60 seconds; actual response time refers to the entire chain of time from the system's first detection of a risk signal to the commencement of emergency intervention measures). Single-area reward for moderately sensitive areas Acc2 represents the accuracy rate in moderately sensitive areas, Miss2 represents the false negative rate in moderately sensitive areas, and Cost2 represents the maintenance cost adjustment item (if the monthly maintenance cost (labor + equipment) > 10% of the budget, deduct 2-5 points (balance safety and cost, avoid over-investment)). Single-zone reward for low-sensitivity areas ; Acc3 is the accuracy rate in the low-sensitivity area, Miss3 is the false negative rate in the low-sensitivity area, and Energy3 is the energy consumption overrun coefficient (actual energy consumption / standard energy consumption, ∈[0, 1.5], where the actual energy consumption is the total energy actually consumed by the monitoring system within a specific period (such as 1 day or 1 month), and the standard energy consumption is the reasonable upper limit of energy consumption preset according to the characteristics of the monitoring area, equipment configuration, and operation strategy); L1, L2, and L3 are the grade coefficients of the extremely sensitive area, moderately sensitive area, and low-sensitivity area respectively, with L1 = 2.0, L2 = 1.5, and L3 = 1.2; The design logic of L1 = 2.0: The core goal of the extremely sensitive area is to "prevent fatal false negatives", and it is necessary to amplify the rewards and punishments with a high coefficient to force the system to tilt resources towards this area; The design logic of L2 = 1.5: The transition weight of "balancing safety and efficiency", and the design logic of L1 = 2.0: The lowest weight of "taking basic safety as the bottom line"; Maintenance cost optimization reward value The acquisition process is as follows: ; is the standard calibration frequency: According to industry specifications (such as the "Safety Technical Regulations for the Operation, Maintenance and Emergency Repair of Urban Gas Facilities" CJJ51), the standard calibration frequency of aging sensors is "once a month" (i.e., once every 30 days, and the quarterly standard frequency = 3 times); is the actual calibration frequency: Recorded through the "Sensor Operation and Maintenance Work Order System", and the actual calibration frequency of aging sensors within a quarter is statistically calculated (for example, if an aging sensor is only calibrated once within a quarter, the savings = 3 - 1 = 2); is the data value density (number of effective features / acquisition energy consumption, normalized to [0, 1], such as the highest value corresponding to 1 and the lowest value corresponding to 0. The number of effective features is the key features extracted from sensor data that can be directly used to judge whether the pipeline is abnormal (such as leakage, corrosion, third-party damage), and the acquisition energy consumption is all the energy consumed to obtain the above "effective features", covering the entire chain of data acquisition, transmission, and preprocessing); Environmental friendliness reward value The acquisition process is as follows: ; is the energy consumption reduction rate during the low valley period, ; is the benchmark energy consumption: the standard energy consumption during the low valley period (such as the energy consumption when the sensor samples at 1Hz and the equipment is in normal standby), which is determined by the average energy consumption during the low valley period in the previous 3 months (data source: historical logs of the sensor energy consumption monitoring module); is the actual energy consumption: the actual energy consumption during the current low valley period (such as the energy consumption when the sensor automatically reduces to 0.1Hz sampling and non-essential equipment goes into sleep mode), which is recorded by the real-time energy consumption monitoring system (data source: smart meter, equipment standby power sensor); = Area of the region where the gas concentration exceeds the standard in the extremely sensitive area (unit: m²), calculated by the area measurement tool of the GIS system (more than 500 m² is counted as 500 m² to avoid extreme values) , valued ∈ [0, 0.1] (more than 0.1 is counted as 0.1 to avoid extreme values), is the system carbon emission intensity, is the total system carbon emission: all carbon emissions generated by the system operation; A is the monitoring area: the monitoring area corresponding to the total length of the gas pipeline covered by the system (such as a range of 50 m on both sides of the pipeline, total area = pipeline length x 100 m) (data source: pipeline coverage measurement of GIS system); is the industry average: the latest monitoring system carbon emission intensity average of the gas industry (such as 0.05 kg CO2 / m² / year), data source: “Low-carbon Development Report of Gas Industry” published by the industry association.

[0035] The above system generates a state vector by constructing a running monitoring agent and integrating multi-source data (environmental interference parameters, transmission data quality indicators, historical monitoring decision indicators), so that the system can accurately perceive the complex operating environment. The action combination output by the agent can dynamically adapt to different interference scenarios (such as physical, electromagnetic, and meteorological interference), effectively resist the influence of environmental interference on monitoring accuracy through sensor weight adjustment, fusion algorithm switching, and working mode optimization, solve the performance degradation problem of traditional systems in strong interference or data anomaly scenarios, and significantly improve the robustness and adaptability of gas pipeline operation monitoring. At the same time, the system introduces a comprehensive reward function including sensitive area consequences, maintenance cost optimization, and environmental friendliness, implements hierarchical reward calculation on sensitive areas, introduces data value density evaluation on maintenance cost, and integrates whole-chain energy consumption and carbon emission control on environmental friendliness, so that the system can quantify the comprehensive benefits of monitoring strategies from multiple dimensions, realize the continuous optimization of action combination, based on the feedback of the reward function, the agent can dynamically adjust the sensor weight, fusion algorithm and working mode, maintain the monitoring robustness in complex interference environment, reduce invalid energy consumption and redundant operation, reduce the whole life cycle operation and maintenance cost, and through the use of renewable energy and carbon emission control, help green and low-carbon operation.

[0036] The above formulas are dimensionless to calculate their numerical values, and the preset parameters in the formulas are set by those skilled in the art according to actual conditions.

[0037] The above-described embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented in software, the above-described embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are wholly or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center through a wired or wireless (such as infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. containing one or more available medium collections. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state disk.

[0038] It should be understood that the size of the sequence number of each process described above in various embodiments of the present application does not mean the order of execution, and the execution order of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0039] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be realized in electronic hardware, or a combination of computer software and electronic hardware. Whether the functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0040] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0041] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other manners. For example, the division of the above-described device embodiments is only a logical function division, and there can be another division manner for actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different units, or the among different units, can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0042] If the functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, and the computer software product is stored in a storage medium, and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, and various other media that can store program codes.

[0043] The above describes only the specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A gas pipeline operation monitoring system based on multi-source data detection, characterized in that, include The operation monitoring intelligent agent construction module is used to build intelligent agents for monitoring the operation of gas pipelines; The parameter index state vector generation module periodically collects various environmental interference parameters, various transmission data quality indicators, and various historical monitoring decision indicators that affect gas pipeline monitoring, and further determines the state vector of the current operating monitoring environment. The operation monitoring and execution module imports the state vector of the current operation monitoring environment into the agent, and the agent outputs action combinations and executes the action combinations. The operation monitoring and optimization module calculates the gas pipeline operation monitoring reward function after the action combination is executed, and continuously optimizes the action combination for gas pipeline operation monitoring based on the gas pipeline operation monitoring reward function.

2. The gas pipeline operation monitoring system based on multi-source data detection according to claim 1, characterized in that, The process of building an intelligent agent: S1: Define the state space, action space, and disturbance effect matrix; S2: Define the policy network structure.

3. The gas pipeline operation monitoring system based on multi-source data detection according to claim 1, characterized in that, The steps for determining the state vector of the current operating monitoring environment are as follows: Various environmental interference parameters, various transmission data quality indicators, and various historical monitoring decision indicators are concatenated into a state vector S=[s1,s2,...,s...]. n ], where each parameter ∈ [0,1], and n is the total number of parameters and indicators.

4. The gas pipeline operation monitoring system based on multi-source data detection according to claim 1, characterized in that, Gas pipeline operation monitoring reward function: ; Rewards for consequences in sensitive areas Optimize reward values ​​to reduce maintenance costs. The environmental friendliness reward value is represented by a1, a2, and a3; a1 + a2 + a3 = 1.

5. The gas pipeline operation monitoring system based on multi-source data detection according to claim 4, characterized in that, The process for obtaining the consequence reward value of sensitive areas is as follows: Identify the extremely sensitive areas, moderately sensitive areas, and lowly sensitive areas in the gas pipeline; obtain the single-area reward 1 for the extremely sensitive area, the single-area reward for the moderately sensitive area, and the single-area reward for the lowly sensitive area; and calculate the consequence reward value of sensitive areas based on the single-area rewards for the extremely sensitive area, the moderately sensitive area, and the lowly sensitive area.

6. The gas pipeline operation monitoring system based on multi-source data detection according to claim 5, characterized in that, The single-area reward for extremely sensitive areas is calculated based on the accuracy, false negative rate, and response delay coefficient of extremely sensitive areas.

7. The gas pipeline operation monitoring system based on multi-source data detection according to claim 5, characterized in that, The single-area reward for medium-sensitive areas is calculated based on the accuracy rate, the false negative rate, and the maintenance cost correction item for medium-sensitive areas.

8. The gas pipeline operation monitoring system based on multi-source data detection according to claim 5, characterized in that, The single-zone reward for low-sensitivity areas is calculated based on the accuracy rate of low-sensitivity areas, the false alarm rate of low-sensitivity areas, and the energy consumption exceeding the standard coefficient.

9. The gas pipeline operation monitoring system based on multi-source data detection according to claim 4, characterized in that, The maintenance cost optimization reward value is calculated based on the standard calibration count, the actual calibration count, and the data value density.

10. The gas pipeline operation monitoring system based on multi-source data detection according to claim 4, characterized in that, The environmental friendliness bonus value is calculated based on the energy consumption reduction rate during off-peak hours, the area of ​​gas concentration exceeding the standard in extremely sensitive areas, and the system's carbon emission intensity.

Images