An artificial intelligence-based liquefied petroleum gas safety distribution supervision method

Abstract

The present application relates to liquefied petroleum gas supervision technical field, disclose a kind of based on artificial intelligence's liquefied petroleum gas safety distribution supervision method, comprising: by dynamic calculation saturated vapor pressure eliminates the temperature fluctuation interference of tank body, superimposed motion disturbance compensation correction vehicle jounce influence, and the leakage suspiciousness of tank body is analyzed and calculated, in combination with voiceprint spectrum features, infrared temperature gradient and vibration signal, analysis obtains the leakage point coordinate of tank body;By analyzing the dynamic balance relationship of tank body, verify filling compliance, in combination with leakage suspiciousness and environmental parameters, leakage risk value is calculated using model to judge leakage risk;High-risk action recognition is dynamically associated with environmental parameters, and behavior environmental risk score is obtained by analysis, and input global control system;In combination with leakage suspiciousness, behavior environmental risk score and leakage point coordinate, build phase change leakage coupling factor, synchronous control alarm threshold and decision planning, and carry out leakage diffusion simulation, carry out whole-process response.

Description

Technical Field

[0001] This invention relates to the field of liquefied petroleum gas (LPG) monitoring technology, specifically to an artificial intelligence-based method for monitoring the safe delivery of LPG. Background Technology

[0002] In existing technologies, the use of mathematical models to optimize delivery routes relies solely on historical data and does not integrate real-time dynamic information such as weather data, road conditions, and equipment status. This results in the inability to dynamically adjust route planning to avoid potential dangers. Existing technologies primarily process sensor data with single-point threshold alarms, lacking the ability to perform multi-dimensional data correlation analysis, making it difficult to identify complex patterns of early leaks or equipment failures.

[0003] Existing technologies have implemented leak alarm and automatic order placement functions, but have not established a complete emergency decision support system. When a leak is detected, the system requires manual intervention to determine whether to initiate evacuation or road closure measures, lacking real-time simulation capabilities. Existing technologies have limited remote control capabilities for delivery vehicles, only able to close emergency shut-off valves, and do not integrate autonomous driving technology to automatically guide vehicles to safe areas or coordinate with fire protection systems to optimize resource scheduling. The optimization models in existing patents are mostly trained based on historical data from specific regions. When applied to new regions or scenarios, the road network structures of different cities vary greatly, requiring the path planning algorithm to be readjusted, lacking adaptive capabilities. The black-box nature of deep learning models makes it difficult for regulatory authorities to understand the decision-making basis, increasing the compliance risks of technology implementation. Existing technologies only focus on monitoring the status of gas cylinders and do not conduct intelligent assessments of the user's gas usage environment. They cannot automatically detect user violations through visual recognition or IoT devices, resulting in safety hazards not being eliminated after the delivery process.

[0004] Therefore, there is a need to provide an artificial intelligence-based method for the safe delivery and supervision of liquefied petroleum gas. Summary of the Invention

[0005] The purpose of this invention is to provide an artificial intelligence-based method for the safe delivery and supervision of liquefied petroleum gas. To solve the aforementioned problems in the prior art, this invention achieves this through the following technical solution:

[0006] The first part, an embodiment of the present invention, provides a method for supervising the safe delivery of liquefied petroleum gas based on artificial intelligence, specifically including the following steps:

[0007] Step 1: Eliminate the temperature fluctuation interference of the tank by dynamically calculating the saturated vapor pressure, and correct the impact of vehicle bumps by superimposing motion disturbance compensation to obtain the leakage suspicion of the tank. Combine the acoustic spectrum characteristics, infrared temperature gradient and vibration signal to analyze and obtain the coordinates of the leakage point of the tank.

[0008] Step 2: Based on the coordinates of the leak point, the filling compliance is verified by analyzing the dynamic balance relationship of the tank. Combined with the suspected leak and environmental parameters, the attention mechanism ELM model is used to analyze and calculate the leak risk value to determine the leak risk.

[0009] Step 3: Based on the leakage risk value, the identification of high-risk actions is dynamically correlated with environmental parameters to obtain a behavioral environmental risk score. The behavioral environmental risk score is then input into the global control system to dynamically adjust the monitoring accuracy of the delivery end.

[0010] Step 4: Combining the leakage suspicion level, behavioral environmental risk score and leakage point coordinates, construct a phase change leakage coupling factor, simultaneously adjust the alarm threshold and decision planning, and conduct leakage diffusion simulation to carry out a full-process response.

[0011] Specifically, the method for eliminating temperature fluctuation interference in the tank is as follows:

[0012] For any tank, determine the component-specific parameters of the Antoine equation based on the liquefied petroleum gas composition, and substitute the final parameters into the average temperature of the tank to calculate the basic saturated vapor pressure.

[0013] By combining vibration data, the tank structure constant is set, and multiplied by the square root of the product of the tank structure constant and the acceleration component to obtain the motion disturbance compensation value. This value is then superimposed on the basic saturated vapor pressure to calculate the dynamic saturated vapor pressure.

[0014] Specifically, the method for obtaining the suspected leakage level is as follows:

[0015] By calculating the absolute deviation between the actual pressure and the dynamic saturated vapor pressure, dividing it by the calibration error of the pressure sensor, and then multiplying it by the ambient temperature difference correction coefficient, if the temperature difference between the tank and the ambient temperature is greater than the preset temperature difference threshold, the false pressure deviation caused by condensation is suppressed by an exponential function to obtain the leakage suspicion level. If the leakage suspicion level is greater than the preset suspicion level threshold, a leakage warning is triggered; otherwise, it is not triggered.

[0016] Specifically, the method for analyzing the coordinates of the leak point in the tank is as follows:

[0017] Based on the real-time pressure of the tank and the ideal gas equation, the theoretical frequency of liquefied petroleum gas injection is derived and calculated. The deviation ratio between the main resonant frequency of the acoustic signature and the theoretical frequency is calculated and multiplied by the acoustic signature weighting coefficient that decays over time to obtain the acoustic signature influence term.

[0018] Calculate the ratio of the peak value of the infrared temperature gradient to the highest temperature of the tank, multiply it by the complement of the acoustic fingerprint weighting coefficient with respect to 1 to obtain the temperature influence term, find the position in the three-dimensional model of the tank where the sum of the acoustic fingerprint influence term and the temperature influence term is the smallest, and mark it as the coordinate of the leak point;

[0019] Specifically, the method for verifying filling compliance is as follows:

[0020] Calculate the maximum legal filling volume based on the physical properties of liquefied petroleum gas and tank parameters;

[0021] The liquid density of liquefied petroleum gas (LPG) is calculated based on the tank temperature. Combined with the tank's rated volume, the basic volume of LPG that the tank can hold is calculated. Based on the current tank pressure, atmospheric pressure, and the adiabatic index of LPG, the influence of pressure on the filling amount is corrected to obtain the theoretical maximum filling amount.

[0022] Compare the deviation between the actual filling volume and the theoretical filling volume, and monitor the matching between the pressure change rate and the filling flow rate. If the deviation between the actual filling volume and the theoretical value exceeds the preset deviation threshold, and the pressure change rate is lower than the theoretical proportional coefficient corresponding to the flow rate, it is judged as illegal filling.

[0023] Specifically, the method for determining leakage risk is as follows:

[0024] A linear transformation is performed on each risk feature, and the risk feature is multiplied by the weight matrix and a bias is added to obtain the feature importance score. The attention weight of each risk feature is obtained by normalization through an exponential function.

[0025] The weighted risk features are input into the ELM model, and the leakage risk value is obtained by nonlinear transformation of the hidden layer and generalized inverse solution of the output layer. The leakage risk value is then used to classify the risk level.

[0026] Specifically, the method for dynamically associating high-risk actions with environmental parameters is as follows:

[0027] Detect high-risk actions of gas delivery workers and output the detection probability of each type of high-risk action in a normalized manner;

[0028] Assign weights to each type of high-risk action, and sum the products of the detection probability of each type of high-risk action and the assigned weights to obtain the action violation degree;

[0029] By combining pollutant concentrations at the user end, space volume, and ventilation conditions, the level of environmental safety is quantified. The environmental risk entropy is obtained by dividing the product of CO concentration and VOC concentration by the product of space volume and ventilation volume.

[0030] Specifically, the method for dynamically adjusting the accuracy of end-of-delivery monitoring is as follows:

[0031] The behavioral environmental risk score is obtained by multiplying the degree of violation of the action by the environmental risk entropy, multiplying it by the night enhancement coefficient, and multiplying it by the emergency response delay term. If the behavioral environmental risk score is greater than the preset behavioral environmental risk threshold, a rectification prompt is triggered on the user terminal; otherwise, it is not triggered.

[0032] Specifically, the method for synchronously controlling alarm thresholds and decision-making planning is as follows:

[0033] Based on the suspected leakage level, the impact of behavioral environmental risk on overall risk is amplified to obtain the basic term. The area of ​​the safe zone is multiplied by the safe zone weight coefficient and then divided by the population density to obtain the correction term. The square root of the correction term is taken to balance the influence of the safe zone and the population density. The basic term and the correction term are multiplied to obtain the phase change leakage coupling factor.

[0034] Based on the obtained phase change leakage coupling factor, the path replanning frequency is set; an improved Gaussian diffusion model is used to simulate the diffusion process after liquefied petroleum gas leakage, and the diffusion coefficient is corrected by the phase change leakage coupling factor.

[0035] The path planning algorithm automatically generates fire resource dispatch routes and personnel evacuation routes;

[0036] Specifically, the method for performing a full-process response is as follows:

[0037] Add filling verification nodes and end-point risk nodes;

[0038] The node data of each filling verification node and the end risk node are stored on the blockchain after being cross-verified by the alliance nodes.

[0039] Query the data of all nodes in the entire chain by tank ID or order number.

[0040] The second part, an embodiment of the present invention, provides an artificial intelligence-based liquefied petroleum gas safe delivery monitoring system, which specifically includes the following units:

[0041] Detection and positioning unit: Eliminates temperature fluctuation interference of the tank by dynamically calculating saturated vapor pressure, and corrects the impact of vehicle bumps by superimposing motion disturbance compensation to obtain the leakage suspicion of the tank. Combined with acoustic spectrum characteristics, infrared temperature gradient and vibration signal, the coordinates of the leakage point of the tank are obtained.

[0042] Verification and evaluation unit: Based on the coordinates of the leak point, the filling compliance is verified by analyzing the dynamic balance relationship of the tank. Combined with the suspected leakage and environmental parameters, the attention mechanism ELM model is used to analyze and calculate the leakage risk value to judge the leakage risk.

[0043] Risk monitoring unit: Based on leakage risk value, high-risk actions are identified and dynamically correlated with environmental parameters to obtain a behavioral environmental risk score. The behavioral environmental risk score is then input into the global control system to dynamically adjust the monitoring accuracy of the delivery end.

[0044] Control and Response Unit: Combining leakage suspicion, behavioral environment risk score and leakage point coordinates, a phase change leakage coupling factor is constructed to simultaneously control alarm thresholds and decision planning, and leakage diffusion simulation is performed to conduct a full-process response.

[0045] The beneficial effects of this invention are:

[0046] 1. By dynamically calculating the saturated vapor pressure using the Antoine equation and combining component-specific parameters with the tank's average temperature, temperature fluctuation interference is accurately eliminated. A motion disturbance compensation model is constructed based on three-dimensional acceleration data and tank structural constants to correct pressure fluctuations caused by vehicle bumps. By dynamically adjusting the weights using the pressure deviation ratio and time decay function, thermodynamic phase change, kinematics, and multimodal sensing are deeply coupled to achieve dynamic self-calibration of leak detection, solving the false alarm problem caused by environmental interference and transportation vibration in traditional methods.

[0047] 2. Based on the ideal gas equation and adiabatic index correction, and simultaneously monitoring actual filling volume deviation and pressure change rate, an environmental risk entropy is proposed. This entropy comprehensively assesses the degree of environmental hazard by considering pollutant concentration, space volume, and ventilation conditions. A nighttime enhancement coefficient and an emergency response delay term are introduced to construct a behavioral environmental risk score. Qualitative risks are transformed into calculable quantitative indicators, enabling dynamic adjustment of the accuracy of end-of-delivery monitoring. A phase change leakage coupling factor is constructed as the core of global control, achieving a paradigm shift from single-indicator monitoring to multi-dimensional risk linkage, thus improving the intelligence level of emergency response. Dynamic risk factors are embedded into a Gaussian diffusion model to achieve accurate and real-time leakage diffusion prediction, providing a scientific basis for emergency resource allocation. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 This is a flowchart of the steps of an artificial intelligence-based method for supervising the safe delivery of liquefied petroleum gas, as provided in Embodiment 1 of the present invention.

[0050] Figure 2 This is a schematic diagram of the structure of an artificial intelligence-based liquefied petroleum gas safe delivery monitoring system provided in Embodiment 2 of the present invention. Detailed Implementation

[0051] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0052] Example 1: As Figure 1 As shown in the figure, this invention provides an artificial intelligence-based method for the safe delivery and supervision of liquefied petroleum gas (LPG). It achieves precise location of LPG leaks through multi-sensor fusion, dynamically quantifies leak risks and implements graded responses using an attention-based ELM model, constructs a phase change leak coupling factor to drive global control, and leverages digital twin simulation of leak diffusion and blockchain to achieve full-process data traceability. This improves safety performance and regulatory efficiency, and addresses the shortcomings of existing solutions such as static risk assessment and insufficient scenario adaptability. It achieves closed-loop management of potential hazards at the user end and intelligent safety supervision throughout the entire process. Specifically, it includes the following steps:

[0053] Step 1: Integrate the thermodynamic phase change principle into sensor data processing, eliminate the temperature fluctuation interference of the tank by dynamically calculating the saturated vapor pressure, and compensate for the impact of vehicle bumps by superimposing motion disturbance compensation. Combine acoustic spectrum characteristics, infrared temperature gradient and vibration signal, and calculate and analyze the coordinates of the leakage point of the tank by using time-varying fusion weights.

[0054] In a specific embodiment, for any tank, temperature sensors are evenly arranged on the surface of the tank to collect the tank temperature at different locations in real time; real-time pressure data is collected by a pressure sensor installed on the top of the tank.

[0055] Motion data during transportation is captured by a three-dimensional accelerometer installed at the bottom of the tank, and ambient temperature and humidity are collected by an external temperature and humidity sensor. Real-time location and environmental correlation data are obtained by combining GPS.

[0056] The average real-time data from multiple temperature sensors is taken as the average temperature of the tank to eliminate the influence of single-point temperature deviation on pressure judgment.

[0057] Based on the Antoine equation and combined with the specific parameters of liquefied petroleum gas components, the theoretical saturated vapor pressure after temperature compensation is calculated, and a motion disturbance compensation term is added to correct the pressure fluctuation caused by vehicle bumps.

[0058] Specifically, the component-specific parameters of the Antoine equation are determined based on the composition of liquefied petroleum gas. For example, the preset parameters for propane are: A=6.809, B=803.8, C=247.0; and the preset parameters for butane are: A=6.897, B=945.6, C=249.7. The component proportions are determined in advance using gas chromatography and weighted to determine the final parameters. The final parameters are substituted into the average temperature of the tank to calculate the basic saturated vapor pressure. Finally, based on the vibration data collected by the three-dimensional accelerometer, the tank structural constant is set. The preset structural constant for steel tanks is 0.002. Multiplying this by the square root of the product of the tank structural constant and the acceleration component yields the motion disturbance compensation value, which is then added to the basic saturated vapor pressure using the formula:

[0059]

[0060] The dynamic saturated vapor pressure was calculated. ,in, These are all component-specific parameters for liquefied petroleum gas. The average temperature of the tank. For time steps, This is a motion disturbance compensation term that corrects for pressure fluctuations caused by vehicle bumps. For three-dimensional acceleration, This is a structural constant for the tank, calibrated based on the tank wall thickness;

[0061] By calculating the absolute deviation between the actual pressure and the dynamic saturated vapor pressure, dividing it by the calibration error of the pressure sensor, and then multiplying it by the ambient temperature difference correction coefficient, if the temperature difference between the tank and the ambient temperature exceeds a preset temperature difference threshold, then an exponential function is used to suppress spurious pressure deviations caused by condensation, using the formula:

[0062]

[0063] Analysis yields a high degree of suspicion of leakage. ,in, Let t be the actual pressure. For pressure sensor calibration error, All are environmental temperature difference sensitivity coefficients. The default value is 0.3. The preset value is 0.1. When the temperature difference between the tank and the ambient temperature exceeds the preset temperature difference threshold of 5°C, it suppresses false pressure deviations caused by condensation. The average temperature of the tank. The ambient temperature is used as the value range for the leakage suspicion level [0, 5]. If the leakage suspicion level is greater than the preset suspicion level threshold, a leakage warning will be triggered; otherwise, it will not be triggered.

[0064] A channel microphone array is deployed on the surface of the tank to collect the jet sound when liquefied petroleum gas leaks; an uncooled infrared thermal imager is installed to capture the temperature field on the surface of the tank; and a piezoelectric vibration sensor is installed to capture the vibration signal caused by the leak.

[0065] Fast Fourier Transform is performed on the acoustic signal collected by the microphone array. The frequency with the strongest signal intensity is found in the preset frequency band of [3kHz, 15kHz]. The frequency with the strongest signal intensity is the main resonant frequency of the leakage sound. The leakage sound frequency under different pressures is used as the basis for judging liquefied petroleum gas leakage.

[0066] Analyze the temperature field of the tank captured by the infrared thermal imager, calculate the rate of change of temperature in space, i.e., the temperature gradient. Near the leak point, there will be a significant temperature gradient peak due to the local low temperature.

[0067] The key to determining the location of a leak by fusing acoustic signatures and infrared features is to balance the reliability of the two modes by using a weighting coefficient that changes over time. In the early stages of a leak, the acoustic signature signal is clear, meaning the weighting coefficient is high. As the leak time increases, the infrared temperature field becomes more stable, and the weighting coefficient gradually decreases.

[0068] The specific logic is as follows: Based on the real-time pressure of the tank and the ideal gas equation, the theoretical frequency of liquefied petroleum gas injection is derived and calculated. The higher the pressure, the higher the theoretical frequency. The deviation ratio between the main resonance frequency of the acoustic signature and the theoretical frequency is calculated and multiplied by the acoustic signature weighting coefficient that decays over time to obtain the acoustic signature influence term.

[0069] Calculate the ratio of the peak value of the infrared temperature gradient to the highest temperature of the tank, multiply it by the complement of the acoustic fingerprint weighting coefficient with respect to 1 to obtain the temperature influence term, find the position in the three-dimensional model of the tank where the sum of the acoustic fingerprint influence term and the temperature influence term is the smallest, and mark it as the coordinate of the leak point;

[0070] Step 2: Based on the coordinates of the leak point, the filling compliance is verified by analyzing the dynamic balance relationship of the tank. Combined with the suspected leak and environmental parameters, the attention mechanism ELM model is used to analyze and calculate the leak risk value to determine the leak risk.

[0071] It should be noted that the risk characteristics include: leak point coordinates, leak suspicion level, and environmental parameters; the environmental parameters include: real-time wind speed, traffic flow, and hazardous materials restricted areas; all data are unified to the same time interval through linear interpolation to form a risk characteristic dataset for risk assessment;

[0072] In a specific embodiment, the mass change of the tank during the filling process is collected in real time by a filling scale, and the filling flow rate is collected by an electromagnetic flow meter built into the filling gun.

[0073] The tank's rated volume and warranty period are obtained, and the tank pressure is collected in real time during the filling process using pressure and temperature sensors. With tank temperature ;

[0074] Calculate the legal maximum filling volume based on the physical properties of liquefied petroleum gas and tank parameters to avoid overfilling or tank mismatch.

[0075] The specific logic is as follows: Calculate the liquid density of liquefied petroleum gas (LPG) based on the tank temperature; combine this with the tank's rated volume to calculate the basic volume of LPG the tank can hold; based on the current tank pressure, atmospheric pressure, and the LPG's adiabatic index, correct for the impact of pressure on the filling volume using the formula:

[0076]

[0077] Analysis yielded the theoretical maximum filling capacity ,in, This represents the liquid density of liquefied petroleum gas after temperature compensation. The temperature of the tank. This refers to the rated volume of the tank. Atmospheric pressure For tank pressure, For preset reference temperature, For isobaric specific heat capacity, Specific heat capacity at constant volume;

[0078] Compare the deviation between the actual filling volume and the theoretical filling volume, and monitor the matching between the pressure change rate and the filling flow rate. If the deviation between the actual filling volume and the theoretical value exceeds the preset deviation threshold, and the pressure change rate is lower than the theoretical proportional coefficient corresponding to the flow rate, it is judged as illegal filling. Otherwise, it is not illegal. The larger the flow rate, the faster the pressure rises. The theoretical proportional coefficient is preset to 0.002.

[0079] Once a violation is detected, the system automatically shuts off the filling gun, records the tank ID and adds it to the blacklist. At the same time, the verification data is uploaded to the blockchain to prevent data tampering. The verification data includes: theoretical filling volume, actual filling volume and judgment result.

[0080] An attention mechanism is used to assign dynamic weights to each risk feature, with higher weights for risk features that have a greater impact on risk. The weights are automatically optimized through model training.

[0081] The specific logic is as follows: A linear transformation is performed on each risk feature; the risk feature is multiplied by the weight matrix and a bias is added to obtain the feature importance score; then, through exponential function normalization, the attention weight of each risk feature is obtained. The sum of the weights of all risk characteristics is 1;

[0082] The weighted risk features are input into the ELM model, and the solution is obtained through nonlinear transformation of the hidden layer and generalized inverse of the output layer, using the formula:

[0083]

[0084] Analysis yields leakage risk values ,in, This is the weight matrix from the hidden layer to the output layer. Output for hidden layer The transpose of the matrix, This is the output of the j-th hidden layer node. For activation function, Let be the weight matrix from the input layer to the j-th hidden layer node. Let be the attention weight for the i-th feature. For the i-th input feature, This is the bias term from the input layer to the j-th hidden layer node;

[0085] It should be noted that the ELM model is a fast learning algorithm for single-hidden-layer feedforward neural networks. Its core feature is that it breaks through the iterative training mode of traditional neural networks and achieves efficient model training by randomly initializing the hidden layer parameters and directly solving the output layer weights.

[0086] Based on the leakage risk value, five levels are defined: if the leakage risk value is [0, 2], it is marked as no risk and delivery proceeds normally; if the leakage risk value is (2, 4], it is marked as low risk and a notification is sent; if the leakage risk value is (4, 6], it is marked as medium risk and the route is adjusted; if the leakage risk value is (6, 8], it is marked as high risk and delivery is suspended; if the leakage risk value is (8, 10], it is marked as extremely high risk and an emergency response is activated.

[0087] Step 3: Dynamically correlate high-risk actions with environmental parameters, analyze and obtain a behavioral environmental risk score, and input the behavioral environmental risk score into the global control system to dynamically adjust the monitoring accuracy of the delivery end.

[0088] In a specific embodiment, action data and images of the surrounding environment are collected during the safe delivery of liquefied petroleum gas, and pollutant concentrations are monitored in the liquefied petroleum gas delivery scenario. The volume of space is scanned by lidar, and ventilation conditions are collected by ventilation volume sensors.

[0089] The YOLOv8 image recognition model is used to detect high-risk actions of gas delivery workers. High-risk actions include: rough handling, valves not being closed tightly, dragging hoses, operating near fire sources, and not wearing protective equipment. The detection probability of each type of high-risk action is normalized and output.

[0090] A weight is assigned to each type of high-risk action, and the product of the detection probability of each high-risk action and the assigned weight is summed to obtain the action violation degree. The numerical range is [0, 1]. The higher the degree of violation of the action, the higher the risk of liquefied petroleum gas delivery operations.

[0091] By combining pollutant concentrations at the user end, space volume, and ventilation conditions, the level of environmental safety is quantified. The environmental risk entropy is obtained by dividing the product of CO concentration and VOC concentration by the product of space volume and ventilation volume. The higher the value of environmental risk entropy, the more dangerous the environment.

[0092] Multiply the violation rate by the environmental risk entropy, then multiply by the nighttime enhancement factor. Set the nighttime operation period; if the delivery time falls within the nighttime operation period, the nighttime enhancement factor is 1; otherwise, it is 0. Finally, multiply by the emergency response delay term using the formula:

[0093]

[0094] The behavioral environmental risk score at time t was obtained through analysis. ,in, The action was a violation. For environmental risk entropy, This is the nighttime enhancement factor. For the emergency response delay time, if the behavioral environment risk score is greater than the preset behavioral environment risk threshold, a rectification prompt will be triggered on the user's end; otherwise, it will not be triggered.

[0095] Step 4: Combining the leakage suspicion level, behavioral environment risk score and leakage point coordinates, construct a phase change leakage coupling factor, simultaneously adjust the alarm threshold and decision planning, and conduct leakage diffusion simulation to carry out a full-process response;

[0096] In a specific embodiment, based on the suspected leakage level, the impact of behavioral environmental risk on overall risk is amplified to obtain a basic term. The higher the behavioral environmental risk score, the larger the basic term. Combining the area of ​​the safe zone around the leakage point with real-time population density, real-time population density is obtained and integrated with video surveillance data. The denser the population, the higher the risk. The safe zone area is multiplied by the safe zone weight coefficient and then divided by the population density to obtain a correction term. The square root of the correction term is taken to balance the influence of the safe zone and population density. The basic term and the correction term are then multiplied using the formula:

[0097]

[0098] The phase change leakage coupling factor at time t was obtained through analysis. ,in, To increase the likelihood of a leak, This represents the sensitivity coefficient to behavioral environmental risks. Scoring behavioral environmental risks, For the safety zone weighting coefficient, The area of ​​the safe zone surrounding the leak point. The coordinates of the leak point, Real-time population density around the leak point. For time steps;

[0099] Set leakage alarm threshold The relationship with the phase change leakage coupling factor is as follows: The higher the phase change leakage coupling factor, the lower the leakage alarm threshold, thus improving the leakage detection sensitivity.

[0100] For example, when the phase change leakage coupling factor is 2, the leakage alarm threshold is 0.008 MPa; when the phase change leakage coupling factor is 5, the leakage alarm threshold drops to 0.008 ÷ 5 ^ 0.3 ≈ 0.005 MPa.

[0101] Based on the obtained phase change leakage coupling factor, the path replanning frequency is set using the formula:

[0102]

[0103] Analysis yields path replanning frequency ,in, This is the phase change leakage coupling factor. The higher the phase change leakage coupling factor, the more frequent the path replanning.

[0104] For example, when the leakage alarm threshold is 1, it is preset to replan once every 5 minutes; when the leakage alarm threshold is 4, it is preset to replan once every 5 ÷ ln(1+4) ≈ 2.1 minutes to avoid risk areas in time.

[0105] Based on the obtained leak point coordinates and tank pressure, a 1:1 tank-environment digital twin was constructed in Unity3D, and the diffusion parameters driven by the phase change leakage coupling factor were imported into the model.

[0106] An improved Gaussian diffusion model was used to simulate the diffusion process after liquefied petroleum gas leakage. The diffusion coefficient was corrected by the phase change leakage coupling factor. The higher the phase change leakage coupling factor, the larger the diffusion coefficient. The simulation results are consistent with the actual diffusion trend in high-risk scenarios.

[0107] Based on the simulation results, the diffusion boundaries for the next 5 minutes, 10 minutes, and 30 minutes are generated, and the evacuation radius is determined. The evacuation radius = phase change leakage coupling factor × 50 meters.

[0108] Based on the diffusion range and evacuation radius, the path planning algorithm automatically generates fire resource dispatch routes and personnel evacuation routes to avoid congested and risky areas and achieve precise emergency response.

[0109] Add filling verification nodes and end-point risk nodes. The filling verification nodes include: theoretical filling quantity, actual filling quantity and illegal judgment results. The end-point risk nodes include: storage environment risk score and user rectification record, covering key data throughout the entire process.

[0110] The node data of each filling verification node and the end-risk node are encrypted with SHA-256 and then cross-verified by three alliance nodes: the filling station, the delivery company, and the regulatory department. After verification, the data is stored on the blockchain to ensure that the node data cannot be tampered with.

[0111] By querying the tank ID or order number, data from all nodes in the supply chain can be retrieved. For example, when querying any tank, one can check whether the tank filling was compliant, whether there were any leakage warnings during transportation, and whether there were any violations, thus enabling accountability and problem tracing.

[0112] Example 2: As Figure 2As shown in the figure, an artificial intelligence-based liquefied petroleum gas (LPG) safe delivery monitoring system provided by an embodiment of the present invention specifically includes the following units:

[0113] Detection and positioning unit: Eliminates temperature fluctuation interference of the tank by dynamically calculating saturated vapor pressure, and corrects the impact of vehicle bumps by superimposing motion disturbance compensation to obtain the leakage suspicion of the tank. Combined with acoustic spectrum characteristics, infrared temperature gradient and vibration signal, the coordinates of the leakage point of the tank are obtained.

[0114] Verification and evaluation unit: Based on the coordinates of the leak point, the filling compliance is verified by analyzing the dynamic balance relationship of the tank. Combined with the suspected leakage and environmental parameters, the attention mechanism ELM model is used to analyze and calculate the leakage risk value to judge the leakage risk.

[0115] Risk monitoring unit: Based on leakage risk value, high-risk actions are identified and dynamically correlated with environmental parameters to obtain a behavioral environmental risk score. The behavioral environmental risk score is then input into the global control system to dynamically adjust the monitoring accuracy of the delivery end.

[0116] Control and Response Unit: Combining leakage suspicion, behavioral environment risk score and leakage point coordinates, a phase change leakage coupling factor is constructed to simultaneously control alarm thresholds and decision planning, and leakage diffusion simulation is performed to conduct a full-process response.

[0117] The above provides a detailed description of one embodiment of the present invention, but the content described is only a preferred embodiment of the present invention and should not be considered as limiting the scope of the present invention. The above formulas are all dimensionless numerical calculations, and the formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world situation. The preset parameters in the formulas are set by those skilled in the art based on actual conditions and historical experience, and can be adjusted according to actual conditions. The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. All equivalent changes and improvements made in accordance with the scope of the present invention should still fall within the patent coverage of the present invention.

Claims

1. An artificial intelligence-based liquefied petroleum gas safety distribution monitoring method, characterized by, The method comprises the following steps: The temperature fluctuation interference of the tank body is eliminated by dynamically calculating the saturated vapor pressure, the influence of vehicle jolt is compensated and corrected by superimposing motion disturbance, the leakage suspiciousness of the tank body is obtained, the leakage point coordinates of the tank body are analyzed by combining the voiceprint spectrum features, the infrared temperature gradient and the vibration signal; Based on the leakage point coordinates, the filling compliance is verified by analyzing the dynamic balance relationship of the tank body, the leakage risk value is analyzed and calculated by combining the leakage suspiciousness and the environmental parameters by using the attention mechanism ELM model to judge the leakage risk; Based on the leakage risk value, the high-risk action is recognized and dynamically associated with the environmental parameters, the behavior environmental risk score is obtained, and the behavior environmental risk score is input into the global regulation system to dynamically adjust the distribution end supervision accuracy; The phase change leakage coupling factor is constructed by combining the leakage suspiciousness, the behavior environmental risk score and the leakage point coordinates, the alarm threshold and the decision planning are synchronously regulated and controlled, and the leakage diffusion simulation is performed to perform the whole process response; The method for dynamically adjusting the distribution end supervision accuracy is: The action violation degree is multiplied by the environmental risk entropy, multiplied by the night enhancement coefficient, and multiplied by the emergency response delay item to obtain the behavior environmental risk score, if the behavior environmental risk score is greater than the preset behavior environmental risk threshold, the user end rectification prompt is triggered, otherwise, it is not triggered; The method for synchronously regulating and controlling the alarm threshold and the decision planning is: Based on the leakage suspiciousness, the influence of the behavior environmental risk on the overall risk is amplified to obtain a basic item, the safety area is multiplied by the safety area weight coefficient, and then divided by the population density to obtain a correction item; the correction item is square rooted to balance the influence of the safety area and the population density, and the basic item and the correction item are multiplied to obtain the phase change leakage coupling factor; Based on the obtained phase change leakage coupling factor, the path re-planning frequency is set; the improved Gaussian diffusion model is used to simulate the diffusion process after the liquefied petroleum gas leaks, and the diffusion coefficient is corrected by the phase change leakage coupling factor; The fire-fighting resource scheduling route and the personnel evacuation route are automatically generated by the path planning algorithm.

2. The method for supervising the safety distribution of liquefied petroleum gas based on artificial intelligence according to claim 1, characterized in that, The method for eliminating the temperature fluctuation interference of the tank body is: For any tank body, the component-specific parameters of the Antoine equation are determined according to the liquefied petroleum gas components, and the final parameters are substituted into the tank body average temperature to calculate the basic saturated vapor pressure; Combined with the vibration data, the tank body structure constant is set, the square root of the product of the tank body structure constant and the acceleration component is multiplied to obtain the motion disturbance compensation value, which is superimposed into the basic saturated vapor pressure to calculate the dynamic saturated vapor pressure.

3. The method for supervising the safety distribution of liquefied petroleum gas based on artificial intelligence according to claim 1, characterized in that, The method for obtaining the leakage suspiciousness is: By calculating the absolute deviation of the actual pressure and the dynamic saturated vapor pressure, dividing by the calibration error of the pressure sensor, and multiplying by the environmental temperature difference correction coefficient, when the tank body and the environmental temperature difference are large, the false pressure deviation caused by condensation is suppressed by an exponential function to obtain the leakage suspiciousness, if the leakage suspiciousness is greater than the preset suspiciousness threshold, the leakage warning is triggered, otherwise, it is not triggered.

4. The method for supervising the safety distribution of liquefied petroleum gas based on artificial intelligence according to claim 1, characterized in that, The method for analyzing the leakage point coordinates of the tank body is: According to the real-time pressure of the tank body and the ideal gas equation, the theoretical frequency of liquefied petroleum gas injection is derived and calculated, the deviation proportion of the main resonance frequency of the acoustic fingerprint and the theoretical frequency is calculated, the acoustic fingerprint weight coefficient is multiplied by the time attenuation, and the acoustic fingerprint influence term is obtained; The proportion of the infrared temperature gradient peak value and the highest temperature of the tank body is calculated, the complement of the acoustic fingerprint weight coefficient about 1 is multiplied, and the temperature influence term is obtained. In the three-dimensional model of the tank body, the position with the minimum sum of the acoustic fingerprint influence term and the temperature influence term is marked as the leakage point coordinate.

5. The method for supervising the safety distribution of liquefied petroleum gas based on artificial intelligence according to claim 1, characterized in that, The method for verifying the filling compliance is: According to the physical properties of liquefied petroleum gas and the tank body parameters, the maximum legal filling amount is calculated; According to the liquefied petroleum gas liquid density calculated by the tank body temperature, combined with the rated volume of the tank body, the basic volume of liquefied petroleum gas that the tank body can accommodate is calculated; according to the current tank body pressure, atmospheric pressure and liquefied petroleum gas adiabatic index, the influence of pressure on filling amount is corrected to obtain the theoretical maximum filling amount; Compare the deviation of the actual filling amount and the theoretical filling amount, and monitor the matching of the pressure change rate and the filling flow rate. If the deviation of the actual filling amount and the theoretical value exceeds the preset deviation threshold, and the pressure change rate is lower than the theoretical proportional coefficient corresponding to the flow rate, it is determined that the filling is illegal.

6. The method for supervising the safety distribution of liquefied petroleum gas based on artificial intelligence according to claim 1, characterized in that, The method for judging the leakage risk is: Linearly transform each risk feature, multiply the risk feature by the weight matrix, add the bias to obtain the feature importance score, and normalize through the exponential function to obtain the attention weight of each risk feature; Input the weighted risk features into the ELM model, and obtain the leakage risk value through the nonlinear transformation of the hidden layer and the generalized inverse solution of the output layer. According to the leakage risk value, the risk value is divided into grades.

7. The method for supervising the safety distribution of liquefied petroleum gas based on artificial intelligence according to claim 1, characterized in that, The method for dynamically associating high-risk actions with environmental parameters is: Detect the high-risk actions of the gas delivery worker, and normalize the detection probability of each type of high-risk action; Assign a weight to each type of high-risk action, sum the product of the detection probability of each type of high-risk action and the assigned weight to obtain the action violation degree; Quantify the environmental safety degree by combining the pollutant concentration of the user end, the space volume and the ventilation condition. The environmental risk entropy is obtained by multiplying the CO concentration and the VOC concentration, and dividing the product by the product of the space volume and the ventilation volume.

8. The method for supervising the safety distribution of liquefied petroleum gas based on artificial intelligence according to claim 1, characterized in that, The method for performing full-process response is: Add filling verification nodes and end risk nodes; The node data of each filling verification node and end risk node is stored on the chain after passing through the cross-validation of the alliance nodes; Query the full-chain node data through the tank body ID or order number.

Images