Unattended gas station monitoring system and method

Abstract

The invention relates to the technical field of gas field station monitoring, in particular to an unattended gas field station monitoring system and method, and the system comprises a data sensing module which is used for receiving real-time data streams from various sensors deployed in a gas field station, the sensor at least comprises a laser methane remote sensing instrument, a vibration sensor, a pressure transmitter and a high-definition camera; the module is configured to trigger a primary alarm when any sensor data exceeds its dynamic threshold; the causal diagnosis and situation deduction fusion analysis module is in communication connection with the data sensing module, and the intelligent decision-making and scheduling module is in communication connection with the fusion analysis module. According to the system, the crossing from passive alarm to active cognition, decision making and disposal is realized, the safety management and control level and the emergency response efficiency of the unmanned station are greatly improved, and intelligent unattended operation is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of gas station monitoring, and in particular to an unattended gas station monitoring system and method. BACKGROUND

[0002] City gas pipeline network is a key infrastructure for maintaining city energy security and normal social operation. Among them, the gas gate station, storage and distribution station and other stations distributed in the suburbs or remote areas of the city bear the core functions of gas reception, storage, pressure regulation, and transmission and distribution. In order to reduce operating costs and improve management efficiency, realizing "unattended" or "few people" of the station has become an explicit development trend in the industry. However, this is accompanied by unprecedentedly higher requirements for the safety monitoring system of the station.

[0003] At present, in the field of safety monitoring of unattended gas stations, the mainstream technical solution still has significant limitations. First, the existing monitoring system mostly relies on discrete, isolated sensor networks for monitoring, such as independently setting up gas concentration detectors, pressure sensors, and video monitoring. These systems usually only have simple threshold alarm functions, triggering an alarm when the detected data exceeds the pre-set fixed value. This way has a high false alarm rate, because transient concentration fluctuations (such as vehicle exhaust interference), sensor drift or environmental factors (such as strong winds, temperature changes) can easily trigger invalid alarms, causing maintenance personnel to be overwhelmed, and may have a "Wolf Coming" paralysis effect on real danger.

[0004] Secondly, and more importantly, the existing system is generally "heavy monitoring, light analysis, heavy alarm, light decision". When an alarm occurs, the system can only provide an isolated, superficial information point (such as "A area methane concentration exceeds the standard"), but cannot answer the core decision-making questions such as "where is the source of the leak?", "how will the risk develop?", and "what should be done now?" Maintenance personnel still need to rely on personal experience to manually analyze and judge through telephone communication and review of data from multiple independent subsystems, which not only responds slowly, but also in complex multi-device, multi-parameter coupled fault scenarios, it is easy to miss the best disposal opportunity due to incomplete information and judgment errors.

[0005] In addition, the existing technology has limited automation, and lacks an integrated intelligent management and control closed loop that can integrate multi-source information, conduct deep causal diagnosis and risk prediction, and drive automated equipment and dispatch human resources accordingly. Although a large number of automated equipment (such as emergency shut-off valves, ventilation systems) are deployed in the station, their linkage logic is often simple and fixed, and cannot adapt to complex and changing real risk situations, nor can they achieve forward-looking and precise disposal based on risk prediction.

[0006] Therefore, there is an urgent need in the art for a new type of monitoring system and method that can deeply integrate perception, analysis, decision-making, and execution capabilities to address the high false alarm rate, the lack of root cause diagnosis, the insufficient risk predictability, and the low efficiency of emergency response in the prior art, and to truly provide a solid technical guarantee for the safe, reliable, and efficient operation of unattended gas stations. SUMMARY

[0007] Therefore, the purpose of the present application is to provide an unattended gas station monitoring system and method to address the high false alarm rate, the lack of root cause diagnosis, the insufficient risk predictability, and the low efficiency of emergency response in the prior art.

[0008] To achieve the above purpose, the present application provides an unattended gas station monitoring system, comprising: A data perception module for receiving real-time data streams from various sensors deployed at the gas station, including at least laser methane remote sensors, vibration sensors, pressure transmitters, and high-definition cameras; the module is configured to trigger subsequent calculations when any sensor data exceeds its dynamic threshold; a causal diagnosis and situation deduction fusion analysis module in communication with the data perception module, comprising: A causal diagnosis unit configured to use a causal graph model based on a Bayesian network as input to perform root cause reasoning on the real-time data streams and output root cause hypotheses with confidence levels; A situation deduction unit configured to use the real-time data streams and the root cause hypotheses as initial conditions to call an embedded physical model library for forward simulation to predict risk evolution paths in a specific future time period; A risk portrait synthesis unit configured to fuse the root cause hypotheses and the risk evolution paths to generate a structured risk portrait object; an intelligent decision-making and dispatching module in communication with the fusion analysis module, comprising: A task decomposer configured to analyze the risk portrait object and generate at least one autonomous control instruction and at least one manual disposal work order based on a pre-defined emergency response rule base; A dynamic dispatcher configured to access a responsibility database, assign appropriate responsibility persons to the manual disposal work order based on a multi-objective optimization algorithm, and generate a push message; A device controller configured to execute the autonomous control instructions through an industrial control network to drive the actuators in the station to act.

[0009] Optionally, the Bayesian network includes device state nodes, environmental parameter nodes, and abnormal event nodes, as well as edges representing their causal relationships; The causal relationships specifically include: Causal edges formed by device physical connection relationships; a causal edge constituted by a device failure mode; the environment-induced causal edge; the root cause inference on the received real-time data stream outputs a root cause hypothesis with a confidence level, including: inputting the real-time data stream as observation evidence into the Bayesian network, updating the posterior probability of each node by using the conditional probability table of the nodes in the network and the Bayesian inference algorithm, and determining the abnormal event node with the highest posterior probability as the current root cause output and outputting its probability as the confidence level.

[0010] Optionally, the causal diagnosis unit is further configured to perform dynamic confidence adjustment of environmental reliability awareness: in the causal graph model, a sub-node for evaluating the reliability of sensor readings is set for the environmental parameter node; when the value of the environmental parameter node is wind speed or precipitation and exceeds a preset reliability influence threshold, the causal diagnosis unit automatically adjusts the weight of the causal edge based on the inference of a single gas concentration sensor reading.

[0011] Optionally, the causal diagnosis unit is further configured to perform spatiotemporal collaborative analysis of leakage plume based on a sensor network: a plurality of gas concentration sensors connected by the data awareness module constitute a sensor network; when the causal diagnosis unit receives non-continuous concentration fluctuation signals from different positions in the sensor network, a spatiotemporal collaborative analysis process is started; the spatiotemporal collaborative analysis process includes: obtaining real-time wind direction and wind speed data, analyzing the time sequence and spatial distribution of the concentration fluctuation signals on different sensors; if the analysis shows that the occurrence sequence of the concentration fluctuation signals matches the gas plume drift path constituted by the real-time wind direction in space and time, the causal diagnosis unit will determine that there is a real leakage source, and according to the drift path, the possible position of the leakage source is located.

[0012] Optionally, the situation deduction unit is specifically configured as follows: the physical model library at least includes a Gaussian plume diffusion model and a device remaining life prediction model; the process of calling the built-in physical model library for forward simulation includes: taking the detected leakage source position and intensity as the input of the Gaussian plume diffusion model, and taking the real-time wind speed and wind direction as the model parameters, calculating and outputting the diffusion range, path and concentration time curve of the gas cloud in the key area within 5-30 minutes in the future.

[0013] Optionally, the risk evolution path specifically includes: the time and space boundary of the gas concentration exceeding the lower limit of explosion, and the predicted time of functional failure of the key device.

[0014] Optionally, the dynamic scheduler is specifically configured, wherein the field stored in the responsibility database at least includes: personnel ID, name, skill qualification list, real-time GPS / UWB position coordinates, current number of work orders and priority; the optimization target of the multi-objective optimization algorithm is to minimize the total response time, and the constraint conditions include: the assigned responsibility person has the skill qualification required for processing the root cause; the straight-line distance between the current position of the assigned responsibility person and the entrance of the station needs to be within a preset threshold; the current total workload of the assigned responsibility person needs to be lower than a preset threshold; and the suitable responsibility person is: performing preliminary screening from the database according to the root cause, and then sorting the candidates by using a weighted scoring method, and finally assigning the work order to the one with the highest score.

[0015] Optionally, the device controller is further configured to: if the risk portrait indicates that there is a gas leakage and it is spreading to a dangerous area, automatically send a “maximum power start” instruction to the ventilation system control cabinet.

[0016] Optionally, the device controller is further configured to: if the root cause is diagnosed as a specific device failure, automatically send a sequence control instruction to the underlying control execution platform, the instruction sequence including: closing the standby device power supply, starting the standby device, and then shutting down the faulty device; if the deduction shows that there is an immediate fire risk of leakage and the video analysis confirms that there is no personnel on site, automatically send a “trigger inert gas fire extinguishing” instruction to the fire extinguishing system.

[0017] Based on the same invention, the application further provides an unattended gas station monitoring method, which comprises the following steps: a data acquisition step: receiving real-time data streams through various sensors deployed in the gas station, the sensors at least including a laser methane remote sensor, a vibration sensor, a pressure transmitter and a high-definition camera; when any sensor data exceeds its dynamic threshold, triggering subsequent calculation; a fusion analysis step, which comprises: a cause and effect diagnosis step: based on a cause and effect graph model constructed by a Bayesian network, taking the real-time data stream as input, performing root cause reasoning, and outputting a root cause hypothesis with a confidence level; a situation deduction step: taking the real-time data stream and the root cause hypothesis as initial conditions, calling a built-in physical model library to perform forward simulation, and predicting the risk evolution path in a specific time period in the future; a risk portrait synthesis step: fusing the root cause hypothesis and the risk evolution path to generate a structured risk portrait object; a decision and scheduling step, which comprises: a task decomposition step: analyzing the risk portrait object, and generating at least one autonomous control instruction and at least one artificial disposal work order according to a pre-defined emergency response rule library; Dynamic scheduling step: access the responsibility person database, assign an appropriate responsibility person to the artificial handling order based on multi-objective optimization algorithm, and generate a push message; Control execution step: execute the autonomous control instruction through the industrial control network to drive the actuator in the field station to act.

[0018] The system conducts all-round, multi-parameter state monitoring on the field station through the data perception module, triggers subsequent calculation when any sensor data is abnormal, and the causal diagnosis and situation deduction fusion analysis module is started immediately. The causal diagnosis unit inside the module uses Bayesian network to infer the root cause and confidence of the fault from complex data, while the situation deduction unit predicts the future development trend of the risk based on the physical model, and the risk portrait synthesis unit integrates the diagnosis and prediction results into a unified and structured cognitive object (risk portrait). Finally, the intelligent decision and scheduling module automatically decomposes the macro response strategy into executable device control instructions and artificial handling orders according to the portrait, and executes them through the device controller and dynamic scheduler respectively, thus forming a closed loop from perception to disposal.

[0019] The system realizes the leap from "passive alarm" to "active cognition, decision and disposal". It fuses dispersed perception information into unified situation understanding, and drives automated systems and human resources to carry out precise and efficient collaborative response, greatly improving the safety control level and emergency response efficiency of unmanned field stations, and realizing intelligent "unattended operation". BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the present application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only a part of the present application, and other drawings can also be obtained by those skilled in the art without creative labor based on these drawings.

[0021] Figure 1 The system block diagram of the embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application will be further described in detail below in combination with specific embodiments.

[0023] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0024] like Figure 1 As shown, an unattended gas station monitoring system and method include: A data sensing module is used to receive real-time data streams from various sensors deployed at the gas station, including at least a laser methane telemetry instrument, a vibration sensor, a pressure transmitter, and a high-definition camera. The module is configured to trigger subsequent calculations when any sensor data exceeds its dynamic threshold. A causal diagnosis and situational analysis fusion module is communicatively connected to the data sensing module and includes: The causal diagnosis unit is configured as a causal graph model based on a Bayesian network. It takes the real-time data stream as input, performs root cause reasoning, and outputs a root cause hypothesis with confidence. The situation simulation unit is configured to use the real-time data stream and the root cause hypothesis as initial conditions, call the built-in physical model library to perform forward simulation, and predict the risk evolution path within a specific future time period. A risk profile synthesis unit is configured to fuse the root cause hypothesis and the risk evolution path to generate a structured risk profile object; an intelligent decision-making and scheduling module is communicatively connected to the fusion analysis module, and includes: The task decomposer is configured to parse the risk profile object and generate at least one autonomous control instruction and at least one manual handling work order based on a predefined emergency response rule base. A dynamic scheduler is configured to access a database of responsible persons, assign appropriate responsible persons to the manual handling work orders based on a multi-objective optimization algorithm, and generate push messages; The equipment controller is configured to execute the autonomous control commands through an industrial control network to drive the actuators within the site.

[0025] The system conducts all-round and multi-parameter state monitoring on the station through a data perception module, triggers subsequent calculation when any sensor data is abnormal, and a causal diagnosis and situation deduction fusion analysis module is started immediately. A causal diagnosis unit in the module uses a Bayesian network to infer the root cause and confidence of failure from complex data, while a situation deduction unit predicts the future development trend of risk based on a physical model, and a risk portrait synthesis unit integrates the diagnosis and prediction results into a unified and structured cognitive object (risk portrait). Finally, an intelligent decision and scheduling module automatically decomposes a macro response strategy into executable device control instructions and manual handling work orders according to the portrait, and executes them through a device controller and a dynamic scheduler respectively, thereby forming a closed loop from perception to disposal.

[0026] The system realizes the leap from “passive alarm” to “active cognition, decision and disposal”. It fuses dispersed perception information into unified situation understanding, and drives automated systems and human resources to carry out precise and efficient collaborative response, greatly improving the safety control level and emergency response efficiency of unmanned stations, and realizing intelligent “unattended operation”.

[0027] In some embodiments, the Bayesian network includes device state nodes, environment parameter nodes and abnormal event nodes, and edges representing the causal relationship thereof; The causal relationship specifically includes: causal edges constituted by device physical connection relationships; causal edges constituted by device failure modes; causal edges constituted by environmental inducements; performing root cause reasoning on the received real-time data stream, and outputting a root cause hypothesis with a confidence level includes: inputting the real-time data stream as observation evidence into the Bayesian network, updating the posterior probability of each node by using a Bayesian inference algorithm through the conditional probability table of the nodes in the network; determining the abnormal event node with the highest posterior probability as the current root cause output, and outputting its probability as the confidence level.

[0028] The nodes in the model represent the states of various entities (devices, environments, events) in the station, and the edges represent the known causal logic (such as physical connection, failure mode, environmental induction) between them. When the sensor detects abnormal data (observation evidence), the system inputs it into the Bayesian network. The network updates the posterior probability of all nodes in the network according to the preset conditional probability table through the Bayesian inference algorithm. Finally, the abnormal event node at the top of the causal chain that best explains all observation evidence will be determined as the root cause.

[0029] This method enables the system to have "deep" fault reasoning ability like an expert, accurately locate the root cause of the problem through the surface phenomenon, and avoid "treating the symptoms" instead of the root cause. The output with confidence also provides a basis for subsequent decision-making, improving the scientificity and reliability of decision-making.

[0030] In some embodiments, the causal diagnosis unit is further configured to perform environment reliability-aware dynamic confidence adjustment: in the causal graph model, a sub-node for evaluating the reliability of sensor readings is set for the environmental parameter node; when the value of the environmental parameter node is wind speed or precipitation exceeds a preset reliability influence threshold, the causal diagnosis unit automatically reduces the weight of the causal edge based on a single gas concentration sensor reading for reasoning.

[0031] This scheme is an optimization of the causal diagnosis process. The system sets a sub-node for evaluating the reliability of data in the causal graph for environmental parameters such as wind speed and precipitation. When these environmental parameters exceed the threshold (such as extremely high wind speed), it means that the readings of some sensors (such as gas concentration sensors) are greatly affected by the environment and their reliability is reduced. At this time, the system will automatically reduce the weight of the causal edge that relies only on a single, unreliable sensor reading for reasoning, thereby avoiding misdiagnosis caused by environmental interference.

[0032] This significantly improves the robustness and reliability of the system in adverse weather conditions. It enables the system to "know" the limitations of its perception ability and dynamically adjust its "judgment criteria", effectively filtering environmental noise and significantly reducing false positives, ensuring the accuracy of the diagnosis conclusion in a high-interference environment.

[0033] In some embodiments, the causal diagnosis unit is further configured to perform sensor network-based leak plume spatiotemporal collaborative analysis: a plurality of gas concentration sensors connected by the data perception module form a sensor network; when the causal diagnosis unit receives non-continuous concentration fluctuation signals from different locations in the sensor network, it starts the spatiotemporal collaborative analysis process; the spatiotemporal collaborative analysis process includes: obtaining real-time wind direction and wind speed data, analyzing the time sequence and spatial distribution of the concentration fluctuation signals on different sensors; if the analysis shows that the order of appearance of the concentration fluctuation signals matches the gas plume drift path formed by the real-time wind direction in space and time, the causal diagnosis unit will determine that there is a real leak source, and according to the drift path, it will locate the possible position of the leak source.

[0034] For the problem of micro and intermittent leakage being difficult to capture, the system uses multiple gas sensors deployed in different locations to form a monitoring network. When non-continuous and weak concentration fluctuations are detected in multiple sensors, the system will start the analysis process, combine real-time wind direction and speed data, and check whether these fluctuations meet a "gas plume" model that moves with the wind in terms of time and spatial sequence. If the pattern matches, even if each sensor reading is weak and short-lived, the system can be sure that there is a real leak, and can estimate the approximate location of the leak source according to the plume path.

[0035] The problem of insufficient sensitivity and easy to miss detection of traditional single sensor in detecting weak and intermittent leakage is solved. By the strategy of "space for information", the weak signals of multiple points are associated into a strong evidence, realizing the early detection and accurate positioning of hidden leakage, and preventing trouble from happening.

[0036] In some embodiments, the situation deduction unit is specifically configured to: the physical model library at least includes a Gaussian plume diffusion model and a device remaining life prediction model; the process of calling the built-in physical model library for forward simulation includes: taking the detected leakage source position and intensity as the input of the Gaussian plume diffusion model, taking the real-time wind speed and direction as the model parameters, calculating and outputting the diffusion range, path and concentration time curve of the gas cloud in the key area within 5-30 minutes in the future.

[0037] The situation deduction unit makes predictions based on solid physical laws. For leakage risk, the Gaussian plume diffusion model is called with the leakage source position and intensity provided by the diagnosis unit as input, combined with real-time weather conditions, to simulate the movement trajectory, diffusion range and concentration change of the gas cloud in the key position within a certain period of time. For device failure, the device remaining life prediction model is called to predict the time point of functional failure based on the current device state.

[0038] The safety management is extended from "now" to "near future", realizing a revolutionary change from static monitoring to dynamic prediction. It enables the system to "predict" risks, providing a critical time window for personnel evacuation, emergency resource prepositioning and other decision-making, and changing passive response to proactive prevention.

[0039] In some embodiments, the risk evolution path specifically includes: the time and space boundary of the gas concentration exceeding the lower explosive limit, and the predicted time of functional failure of the key device.

[0040] The specific content of the situation deduction output is defined, i.e. the risk evolution path. It quantitatively describes specific risk events that may occur in the future, such as "the gas will reach the explosion lower limit at the entrance of the power distribution room in 8 minutes" or "the compressor is expected to be shut down due to bearing damage in 2 hours". These time and space boundaries provide a direct basis for formulating accurate emergency response plans.

[0041] The risk information becomes specific and operable. Decision makers no longer face vague "high risk" warnings, but clear timelines and roadmaps, so that they can formulate the best response strategy that is time-saving and appropriate, greatly improving the accuracy and effectiveness of emergency response.

[0042] In some embodiments, the dynamic scheduler is specifically configured: the field stored in the responsibility person database includes at least: personnel ID, name, skill qualification list, real-time GPS / UWB position coordinates, current total work order quantity and priority; the optimization target of the multi-objective optimization algorithm is to minimize the total response time, and the constraint conditions include: the assigned responsibility person has the skill qualification required to handle the root cause; the straight-line distance between the current position of the assigned responsibility person and the entrance of the station needs to be within the preset threshold; the current total workload of the assigned responsibility person needs to be lower than the preset threshold; the assigned responsibility person is: according to the root cause, the database is preliminarily screened, and then the weighting scoring method is used to sort the candidates, and finally the work order is assigned to the one with the highest score.

[0043] When assigning tasks, the dynamic scheduler does not simply poll or designate, but performs an efficient "optimal matching". It accesses the responsibility person database containing personnel skills, real-time location, and workload, with the goal of "minimizing total response time", under the condition of meeting "skill matching, closest distance, lightest load" and other constraints, the current most suitable responsibility person for handling the work order is calculated through optimization algorithm (such as weighted scoring method).

[0044] It realizes the optimal and most efficient use of limited emergency resources (especially high-level maintenance personnel). This ensures that the most suitable person can be sent to the most critical site at the fastest speed when multiple events occur, avoiding uneven task allocation and response delay, and maximizing repair efficiency.

[0045] In some embodiments, the device controller is further configured to: if the risk portrait indicates that there is a gas leak and it is spreading to a dangerous area, automatically send a "maximum power start" instruction to the ventilation system control cabinet.

[0046] When the risk profile clearly indicates that there is a leak and gas is spreading towards a dangerous area such as the power distribution room, the system does not need to wait for manual confirmation. The equipment controller will immediately and automatically send an instruction to the control cabinet of the ventilation system, starting it to maximum power, to dilute and disperse the leaked gas and prevent its accumulation.

[0047] A "second-level" automatic response to major safety risks is achieved. This immediate physical intervention can effectively block the development of the accident chain and prevent the leakage of gas into dangerous areas to cause combustion and explosion, which is the most critical and direct automated defense line to ensure the safety of the station.

[0048] In some embodiments, the equipment controller is further configured to: if the root cause is diagnosed as a specific equipment failure, automatically send a sequence control instruction to the underlying control execution platform, the instruction sequence including: closing the standby equipment power supply, starting the standby equipment, and then shutting down the faulty equipment; if the deduction indicates that there is an immediate fire risk and the video analysis confirms that there is no personnel on site, automatically send a "trigger inert gas fire extinguishing" instruction to the fire extinguishing system.

[0049] This scheme defines an automatic operation sequence under specific failure. When a key device (such as a compressor) is diagnosed as faulty, the system automatically executes a safe "switch-off" process, ensuring that the standby equipment is online first, and then isolating the faulty equipment to ensure uninterrupted production and on-site safety. Further, when the system comprehensively judges (such as through deduction and video analysis) that there is an immediate fire risk on site and there is no possibility of personnel injury, it has the right to automatically trigger the highest level of fire protection measures.

[0050] On the one hand, through automatic equipment switching, the economic loss caused by unplanned shutdown is minimized, ensuring the stability of gas supply. On the other hand, in extremely urgent situations, the system dares to and can automatically take decisive measures, avoiding the catastrophic consequences that may be caused by the inability of personnel to arrive in time or delayed decision-making.

[0051] In order to further implement the present application, the present application further provides an unattended gas station monitoring method, the method comprising the following steps: a data acquisition step: receiving real-time data streams through a plurality of sensors deployed at the gas station, the sensors at least including a laser methane remote sensor, a vibration sensor, a pressure transmitter and a high-definition camera; when any sensor data exceeds its dynamic threshold, triggering subsequent calculation; a fusion analysis step, the step comprising: a causal diagnosis step: based on a causal graph model constructed by a Bayesian network, taking the real-time data stream as input, performing root cause reasoning, and outputting a root cause hypothesis with a confidence level; Trend deduction step: calling built-in physics model library to perform forward simulation with the real-time data stream and the root cause hypothesis as initial conditions, to predict the risk evolution path in a specific future time period; Risk portrait synthesis step: fusing the root cause hypothesis and the risk evolution path to generate a structured risk portrait object; Decision and scheduling step, which comprises: Task decomposition step: analyzing the risk portrait object and generating at least one autonomous control instruction and at least one manual disposal work order according to a predefined emergency response rule base; Dynamic scheduling step: accessing a responsibility person database, assigning an appropriate responsibility person to the manual disposal work order based on a multi-objective optimization algorithm, and generating a push message; Control execution step: executing the autonomous control instruction through an industrial control network to drive the actuator in the station to act.

[0052] The unmanned gas station monitoring method provided by the application is based on the intelligent closed-loop management and control concept of "perception-cognition-decision-execution". First, the multi-source heterogeneous sensor network is used to collect the station environment and equipment state data in real time, and the intelligent analysis process is triggered immediately when any monitored parameter exceeds the dynamically adjusted threshold. In the fusion analysis stage, the system performs two core tasks in parallel: on the one hand, based on the causal graph model constructed by the Bayesian network, the real-time data stream is input into the network as observation evidence, and the posterior probability of each abnormal event node is calculated through the conditional probability propagation mechanism, so as to determine the root cause and output the confidence evaluation; on the other hand, the Gaussian smoke plume diffusion, equipment life prediction and other physical models are called to perform forward simulation, taking the real-time data and the diagnosis result as initial conditions, to deduce the risk spatio-temporal evolution law in the next 5-30 minutes. Then, the system fuses the diagnosis conclusion and the predicted situation to generate a structured risk portrait object, which fully represents the essential characteristics, spatio-temporal attributes and development trajectory of the current risk. In the decision and scheduling stage, the task decomposer converts the risk portrait into a specific response scheme based on the preset emergency rule base, and simultaneously generates a device control instruction and a manual disposal work order; the dynamic scheduler realizes the optimal allocation of emergency resources by comprehensively considering the personnel skills, real-time location, work load and other constraint conditions through a multi-objective optimization algorithm. Finally, the system issues the autonomous decision instruction to the station actuator through the industrial control network, completing the whole-process automatic closed-loop management from risk perception to disposal response.

[0053] The method realizes the whole-process automatic closed-loop from risk perception to control execution, can still maintain an efficient and reliable safety management and control level in the unmanned scenario, greatly reduces the manual operation and maintenance cost, and provides a solid technical guarantee for the intrinsic safety of the gas station.

[0054] Those skilled in the art should understand that the above discussion of any embodiment is merely exemplary in nature and is not intended to imply that the present application, including the claims, is limited to these examples; the technical features among the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other changes of the different aspects of the present application as described above, which are not provided in details for the sake of brevity.

[0055] The present application is intended to cover all such alternatives, modifications, and variations as come within the scope of the broadest possible interpretation of the appended claims. Accordingly, any and all such alternations, modifications, equivalents, improvements and the like are intended to be encompassed by the present application.

Claims

1. An unattended gas station monitoring system, characterized in that, include: A data sensing module is used to receive real-time data streams from various sensors deployed at the gas station. This module is configured to trigger subsequent calculations when any sensor data exceeds its dynamic threshold. A causal diagnosis and situational analysis fusion module is communicatively connected to the data sensing module and includes: The causal diagnosis unit is configured as a causal graph model based on a Bayesian network. It takes the real-time data stream as input, performs root cause reasoning, and outputs a root cause hypothesis with confidence. The situation simulation unit is configured to use the real-time data stream and the root cause hypothesis as initial conditions, call the built-in physical model library to perform forward simulation, and predict the risk evolution path within a specific future time period. A risk profile synthesis unit is configured to fuse the root cause hypothesis and the risk evolution path to generate a structured risk profile object; an intelligent decision-making and scheduling module is communicatively connected to the fusion analysis module, and includes: The task decomposer is configured to parse the risk profile object and generate at least one autonomous control instruction and at least one manual handling work order based on a predefined emergency response rule base. A dynamic scheduler is configured to access a database of responsible persons, assign appropriate responsible persons to the manual handling work orders based on a multi-objective optimization algorithm, and generate push messages; The equipment controller is configured to execute the autonomous control commands through an industrial control network to drive the actuators within the site.

2. The unattended gas station monitoring system according to claim 1, characterized in that, The Bayesian network includes device status nodes, environmental parameter nodes, and abnormal event nodes, as well as edges representing their causal relationships. The causal relationship specifically includes: Causal edges formed by the physical connections between devices; Causal edges formed by equipment failure modes; The causal edges formed by environmental factors; Root cause inference is performed on the received real-time data stream, and a root cause hypothesis with confidence is output, including: using the real-time data stream as observational evidence, inputting it into the Bayesian network, updating the posterior probability of each node using the conditional probability table of the nodes in the network and the Bayesian inference algorithm; determining the abnormal event node with the highest posterior probability as the current root cause output, and outputting its probability as the confidence level.

3. The unattended gas station monitoring system according to claim 1, characterized in that, The causal diagnosis unit is also configured to perform dynamic confidence adjustment for environmental reliability perception: in the causal graph model, sub-nodes are set for the environmental parameter nodes to evaluate the reliability of sensor readings; when the value of the environmental parameter node, such as wind speed or precipitation, exceeds a preset reliability impact threshold, the causal diagnosis unit automatically reduces the weight of the causal edge based on the reading of a single gas concentration sensor.

4. The unattended gas station monitoring system according to claim 1, characterized in that, The causal diagnosis unit is also configured to perform spatiotemporal collaborative analysis of a leak plume based on a sensor network: multiple gas concentration sensors connected to the data sensing module constitute a sensor network; when the causal diagnosis unit receives discontinuous concentration fluctuation signals from different locations in the sensor network, it initiates the spatiotemporal collaborative analysis process. The spatiotemporal collaborative analysis process includes: acquiring real-time wind direction and wind speed data, and analyzing the time series and spatial distribution of the concentration fluctuation signal on different sensors; If the analysis shows that the order of occurrence of the concentration fluctuation signal matches the gas plume drift path formed by the real-time wind direction in time and space, then the causal diagnosis unit will determine that there is a real leak source, and locate the possible location of the leak source in reverse according to the drift path.

5. The unattended gas station monitoring system according to claim 1, characterized in that, The specific configuration of the situation simulation unit is as follows: the physical model library includes at least a Gaussian plume diffusion model and an equipment remaining life prediction model; the process of calling the built-in physical model library for forward simulation includes: using the detected leak source location and intensity as input to the Gaussian plume diffusion model, using real-time wind speed and wind direction as model parameters, calculating and outputting the diffusion range, path, and concentration-time curve of the gas cloud in the key area within the next 5-30 minutes.

6. The unattended gas station monitoring system according to claim 1, characterized in that, The risk evolution path specifically includes: the time and space boundaries of gas concentration exceeding the lower explosive limit, and the expected time of functional failure of key equipment.

7. The unattended gas station monitoring system according to claim 1, characterized in that, The dynamic scheduler is specifically configured as follows: the fields stored in the responsible person database include at least: personnel ID, name, skill qualification list, real-time GPS / UWB location coordinates, current number of work orders and priority; the optimization objective of the multi-objective optimization algorithm is to minimize the total response time, and its constraints include: the assigned responsible person has the skills and qualifications required to handle the root cause; the straight-line distance between the assigned responsible person's current location and the station entrance must be within a preset threshold; the assigned responsible person's current total workload must be lower than a preset threshold; the method for assigning a suitable responsible person is: to perform an initial screening from the database based on the root cause, then to sort the candidates using a weighted scoring method, and finally to assign the work order to the person with the highest score.

8. The unattended gas station monitoring system according to claim 1, characterized in that, The device controller is also configured to automatically send a "maximum power start" command to the ventilation system control cabinet if the risk profile indicates a gas leak that is spreading to a dangerous area.

9. The unattended gas station monitoring system according to claim 1, characterized in that, The device controller is also configured to: if the root cause is diagnosed as a specific device failure, automatically send a sequence of control instructions to the underlying control execution platform, the sequence of instructions including: closing the power supply to the backup device, starting the backup device, and then shutting down the faulty device; if the simulation indicates that the leak poses an immediate fire risk and video analysis confirms that there are no personnel on site, automatically send a "trigger inert gas extinguishing" instruction to the fire protection system.

10. A method for monitoring unattended gas stations, characterized in that, Applied to the system as described in any one of claims 1-9, the method includes the following steps: a data acquisition step: receiving real-time data streams through multiple sensors deployed at the gas station, the sensors including at least a laser methane telemeter, a vibration sensor, a pressure transmitter, and a high-definition camera; triggering subsequent calculations when the data from any sensor exceeds its dynamic threshold; The fusion analysis step includes: Causal diagnosis steps: Based on the causal graph model constructed by Bayesian network, the real-time data stream is used as input to perform root cause reasoning and output root cause hypotheses with confidence. Situation simulation steps: Using the real-time data stream and the root cause assumption as initial conditions, call the built-in physical model library to perform forward simulation and predict the risk evolution path within a specific future time period; Risk profile synthesis steps: Integrate the root cause hypothesis and the risk evolution path to generate a structured risk profile object; Decision-making and scheduling steps, the steps including: Task decomposition steps: parse the risk profile object, and generate at least one autonomous control instruction and at least one manual handling work order according to the predefined emergency response rule base; Dynamic scheduling steps: Access the responsible person database, assign a suitable responsible person to the manual handling work order based on a multi-objective optimization algorithm, and generate a push message; Control execution steps: The autonomous control commands are executed through the industrial control network to drive the actuators within the site to move.

Images