Tracing positioning and concentration deduction method for hydrogen leakage noise based on hydrogen-related scene
By deploying sensors and decibel meters in hydrogen leak scenarios, and combining deep learning models and accident case libraries, the problems of slow response and high cost in hydrogen leak location in existing technologies have been solved. This enables rapid and accurate location of hydrogen leak sources and concentration prediction, supporting rescue and evacuation efforts.
Patent Information
- Application Number
- CN202511669812.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-14
- Publication Date
- 2026-02-13
AI Technical Summary
Existing hydrogen leak detection and location technologies suffer from slow response, complex deployment, and high costs, making it difficult to achieve rapid and flexible location of hydrogen leak sources and concentration estimation.
By deploying multiple hydrogen concentration sensors and decibel meters in hydrogen-related scenarios, and combining deep learning models of convolutional neural networks and long short-term memory networks, we analyze hydrogen leakage noise and concentration data, calculate leakage orifice diameter and flow rate using dimensionless formulas, and conduct source tracing and concentration extrapolation using a hydrogen leakage diffusion accident case library.
It enables rapid and accurate location and concentration estimation of hydrogen leak sources, adapts to various complex scenarios, provides guidance for personnel rescue and evacuation, and has good engineering application value.
Smart Images
Figure CN121525568A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of hydrogen energy safety, and in particular to a method for tracing and positioning hydrogen leakage noise and deducing concentration based on a hydrogen-related scene. BACKGROUND
[0002] Hydrogen, as a clean energy, is widely used in industries, energy and transportation, etc. However, hydrogen is flammable, explosive and has a fast diffusion speed, and once it leaks, it can easily cause safety accidents.
[0003] Quick and accurate deduction of the location of the leakage can help rescue personnel to infer the development of the accident and reduce the risk of hydrogen leakage explosion. The existing public hydrogen leakage detection and positioning technology mainly arranges hydrogen sensors in the area where hydrogen leakage may occur to monitor the hydrogen concentration, and locates the leakage source through the hydrogen concentration. Although the application of hydrogen sensors can provide basic monitoring function to a certain extent, such monitoring method has problems such as response lag, complex point arrangement and high cost. In comparison, the leakage gas flow generates characteristic noise through the orifice, and the sound can be captured at a long distance. Therefore, the monitoring path based on noise can realize quick tracing of the leakage source location under the condition of fewer sensors and more flexible arrangement.
[0004] Therefore, a method for quick positioning and concentration deduction of hydrogen leakage noise in a hydrogen-related scene is developed, which can realize positioning of the leakage source location by monitoring with a few noise sensors, and reflect the hydrogen leakage diffusion distribution in a timely manner by retrieving the accident case library, to assist personnel in rescue and evacuation. SUMMARY
[0005] The purpose of the present application is to overcome the defects in the prior art and provide a method for tracing and positioning hydrogen leakage noise and deducing concentration based on a hydrogen-related scene.
[0006] The present application is realized by the following technical steps:
[0007] S1, analyze the characteristics of hydrogen leakage accidents in a hydrogen-related scene, determine the possible hydrogen leakage positions and working conditions, carry out hydrogen leakage experiments on representative hydrogen-related scenes, verify the numerical simulation model through multi-point noise and concentration data collection, and evaluate the accuracy and reliability of the model; after verifying the reliability of the model, use the verified simulation model to study potential hydrogen leakage accidents in the hydrogen-related scene;
[0008] S2, arrange a plurality of potential hydrogen concentration sensors and decibel meters in the hydrogen-related scene, and obtain the concentration data and leakage noise at the positions of the hydrogen concentration sensors and decibel meters after hydrogen leakage;
[0009] S3, based on different leakage scenarios, the collected data are feature extracted and labeled, and a hydrogen leakage noise data set and a hydrogen leakage diffusion accident typical case library are established, containing working condition labels and accident consequence data;
[0010] S4, based on the deep learning model combining convolutional neural network and long short-term memory network, time-frequency feature analysis is performed on the multi-point noise signal to realize the source location of the leakage source;
[0011] S5, by obtaining the noise of the monitoring point, the noise of the leakage origin and the distance from the noise monitoring point to the leakage origin, a dimensionless formula is fitted for the calculation of the leakage aperture;
[0012] wherein, is the average noise (dB) at the monitoring point, is the noise at the origin at t seconds (dB), is the leakage aperture; is the distance from the monitoring point to the leakage point;
[0013] In field application, according to the aperture and the leakage pressure, the leakage flow rate is calculated by using the corresponding flow rate formula, which satisfies the following function:
[0014]
[0015] wherein d is the equivalent diameter of the leakage port and P is the leakage pressure.
[0016] S6, input the leakage aperture and the flow rate size into the deep learning model and call the hydrogen leakage diffusion accident case library to obtain the hydrogen leakage accident consequence for estimating the leakage time on site;
[0017] S7, the field application stage is as follows,
[0018] S71, the leakage source is located by inputting the long-distance noise signal of the leakage point collected by the decibel instrument into the model, the leakage pressure size of the leakage source is determined according to the layout structure characteristics of the hydrogen-related scene, and the distance from the monitoring point to the leakage point and the noise of the leakage point are obtained based on the position of the leakage source by using the range finder and the movable remote control robot.
[0019] S72, according to the monitoring point noise, the leakage source noise and the distance parameter from the monitoring point to the leakage origin obtained by S71, the leakage aperture size is calculated by the dimensionless formula; based on the leakage aperture and the leakage pressure, the leakage flow rate size is further obtained by the flow rate calculation formula;
[0020] S73, input the leakage pressure and flow rate and based on the hydrogen leakage diffusion accident case library, the hydrogen leakage accident consequence is predicted to estimate the leakage time, which assists personnel rescue and evacuation.
[0021] In the step S1, the hydrogen leakage test is carried out on the hydrogen leakage involved hydrogen scene platform, the numerical simulation model is a full-size leakage-diffusion model built by using CFD simulation software according to the actual scene size structure characteristics, and the potential leakage scene is the area, equipment and other places where hydrogen leakage may occur in the hydrogen involved scene.
[0022] In the step S2, in the test and numerical simulation process, 5 decibel instruments are deployed, 1 of which is located at the leakage point, and 4 of which are arranged at different directions and distances from the leakage point for monitoring the noise at a distance, wherein the noise data at the origin point is used for dimensionless formula calculation, the noise at the 4 monitoring points is used for input of the machine learning model and leakage position tracing, and all the monitoring points of the decibel instruments are at a vertical distance of 1.5 m from the ground, and the hydrogen concentration sensors are arranged at key positions according to the equipment layout structure of the hydrogen involved scene.
[0023] In the step S3, the label includes a leakage aperture and a leakage flow rate. The hydrogen involved scene includes a hydrogen production plant, a hydrogen refueling station and other areas and equipment prone to hydrogen leakage, and a hydrogen fuel cell test workshop and plant.
[0024] In the step S4, the convolutional neural network is a deep learning model for extracting features of discrete noise data and concentration data, and the long short-term memory network is a recurrent neural network for capturing long and short-term dependencies of noise and concentration data.
[0025] In the step S5, the dimensionless formula is obtained by fitting the acoustic feature extraction of the noise loudness in the hydrogen leakage stage.
[0026] In the step S6, the label is an index of the hydrogen leakage diffusion accident case library, including a leakage aperture and a leakage flow rate, and the hydrogen concentration distribution consequence data can be obtained by inputting the label.
[0027] In the step S71, in actual application, the noise data is monitored on site, and is used as the data input of the aforementioned noise tracing model to locate the leakage source, determine the position information of the leakage, and determine the hydrogen leakage pressure according to the structure characteristics of the hydrogen leakage accident area in the hydrogen involved scene. The noise at the monitoring point is collected by using a decibel instrument, the distance from the monitoring point to the leakage origin point is measured by using a range finder, and the noise at the leakage point is monitored by using a movable robot to obtain the three parameters. When the noise is measured on site, the monitoring points are all at a height of 1.5 m from the ground.
[0028] In the step S72, the monitoring point noise, leakage point noise and distance parameters obtained in the foregoing are input into the dimensionless formula to obtain the size of the leakage aperture, and the size of the leakage flow rate is obtained by using a flow rate calculation formula.
[0029] In the step S73, the leakage aperture and the flow rate are input, and based on a hydrogen leakage diffusion accident case library, the hydrogen leakage accident consequence is determined, compared with the hydrogen concentration data obtained by the field monitoring, the leakage time is estimated, and the subsequent leakage diffusion situation is predicted.
[0030] The beneficial technical effects of the present application are:
[0031] The present application provides a hydrogen leakage noise source positioning and concentration deduction method based on a hydrogen-related scene,
[0032] The simulation model is verified by experiments to ensure reliable data sources and high prediction accuracy; the noise sensor is flexible and suitable for various complex scenes; the physical interpretability is realized through the dimensionless acoustic relationship and the flow equation to improve the model universality; the concentration deduction model combines deep learning and case retrieval to realize fast prediction; it provides guidance for personnel rescue and evacuation, and has good engineering application value. BRIEF DESCRIPTION OF DRAWINGS
[0033] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the specific embodiments of the present application are described in detail below, which are implemented on the premise of the solutions described in the present application, and give detailed implementation manners and specific operation processes, but the protection scope of the present application is not limited to the following examples.
[0034] Figure 1 A flow chart of a hydrogen leakage noise source positioning and concentration deduction method based on a hydrogen-related scene;
[0035] Figure 2 A three-dimensional simulation model of the present application;
[0036] Figure 3 A noise monitoring point deployment map
[0037] Figure 4 Fitting results of the decibel number at the dimensionless monitoring point and the leakage aperture and the dimensionless monitoring distance and the leakage aperture;
[0038] Figure 5 A hydrogen fuel cell vehicle leakage diffusion concentration distribution result map of the present application DETAILED DESCRIPTION
[0039] The present application provides a hydrogen leakage noise source positioning and concentration deduction method based on a hydrogen-related scene, taking a hydrogen fuel cell vehicle leakage scene as an example, and realizing hydrogen leakage noise source positioning and concentration deduction according to the following implementation technical method, the specific process is as shown in Figure 1 The specific steps are as follows:
[0040] Step S1, hydrogen leakage test of hydrogen fuel cell vehicle is carried out for verification of simulation model, potential hydrogen leakage accident scene test research is carried out, specifically:
[0041] The hydrogen fuel cell passenger car is selected as the test vehicle, the appearance structure and technical parameters of the fuel cell vehicle at the time of leaving the factory are maintained, during the test, the vehicle is kept stationary, except for the necessary test equipment and daily operation parts of the vehicle, the lighting devices and auxiliary devices on the vehicle are turned off; external hydrogen supply is used during the test, mainly including hydrogen storage bottle, pressure reducing valve, flow meter and customized leakage pipeline and leakage port; the leakage pressure is adjusted by the pressure reducing valve, and the leakage flow is monitored by the flow meter.
[0042] CFD simulation software is used to build a virtual three-dimensional model of the hydrogen fuel cell vehicle to which the method is applied, to ensure that the three-dimensional model is consistent with the actual hydrogen fuel cell vehicle size and equipment material properties, etc.
[0043] Step S2, decibel meter and concentration sensor are deployed for data collection, according to Figure 3 As shown in the figure, the decibel meters are arranged in a diamond shape, with a 3-meter distance between S1-S3, and a 5-meter distance between S1-S2, S2-S3, S3-S4 and S4-S1. When monitoring data, 5 decibel meters are used, one for monitoring the noise at the leakage origin, and the other 4 are arranged as shown. All decibel meters are at a vertical distance of 1.5m from the ground. The noise at the origin is used for dimensionless formula calculation, and the noise at the three monitoring points is used for input of the machine learning model and leakage location tracing; hydrogen concentration sensors are deployed at key positions according to actual leakage conditions.
[0044] Step S3, simulate different hydrogen leakage conditions to obtain noise and concentration data after hydrogen leakage, only low-pressure leakage is done during the test, and the authenticity of the simulation data is verified to ensure that it meets the real scene, so as to realize the research on subsequent conditions through simulation; based on the simulation results, a hydrogen leakage diffusion accident case library is constructed, containing labels and consequence data, the labels are the index of the case library, including leakage aperture and leakage flow, and by inputting the labels, the hydrogen leakage diffusion distribution result can be quickly obtained.
[0045] Step S4, the collected data is preprocessed, the missing values and abnormal values in the data are filled and removed to ensure the integrity of the data, and the data set for model training is constituted;
[0046] When training the convolutional neural network and long short-term memory network model, the sample data is randomly divided into training set and test set, and the proportion of division is usually 80% and 20%.
[0047] A hydrogen leak noise tracing model architecture was built based on convolutional neural networks (CNNs) and long short-term memory (LSTM) networks. The CNN is a deep learning model capable of extracting data features, comprising convolutional layers, pooling layers, fully connected layers, and an output layer. The convolutional layer is the core of the CNN, responsible for extracting features from the input data. The pooling layer reduces the number of parameters and computational cost, while making the feature detector more invariant. The fully connected layer maps the learned high-level features to the final output. The final output layer typically uses the Softmax activation function, which maps the input to a probability distribution for multi-class classification problems.
[0048] Long Short-Term Memory (LSTM) networks are time-recurrent neural networks used to capture long-term dependencies in noisy time-series data. They consist of an input layer, hidden layers, and an output layer. The input layer primarily receives external data and passes it to the hidden layers. The hidden layers process the input data and update their information by combining the current input with the state of the previous time step, thus capturing long-term dependencies in the time series. Over time, the hidden layers gradually accumulate and update important historical information. Finally, the output layer receives the final state from the hidden layers and transforms it into the actual output.
[0049] Step S5: Based on the measured noise at the monitoring point, the noise at the leakage point, and the monitoring distance, a dimensionless formula is proposed, specifically as follows: Parameters affecting the sound level at the monitoring point include the sound level (dB) at the leak point and the distance between the monitoring point and the leak point. ), Leakage diameter ( Atmospheric pressure (p0), ambient density (ρ0), and ambient wind speed (V).
[0050] Based on dimensional analysis, select , and As independent variables, equation (1) can be rewritten as: By multiplying by the reciprocals of the first and second terms of equation (2), we can obtain the expression for the dimensionless distance and leakage orifice diameter, as shown in equation (3): Based on formula (3), it is assumed that the decibel level at the dimensionless monitoring point / the decibel level at the leak point has the following relationship with the dimensionless monitoring distance / the leak orifice diameter: The dimensionless monitoring point decibel number / leak decibel number and the dimensionless monitoring distance / leak hole diameter are fitted when the leak hole diameter is 0.5 mm, and the fitting curve is shown in formula (5). The curve is verified by using data of leak hole diameters of 1 mm, 1.5 mm and 1.8 mm. It is verified that formula (5) has a good prediction effect on the experimental results.
[0051] Step S6, input the label to call the hydrogen leakage diffusion accident case library, and obtain the consequences of the hydrogen leakage accident.
[0052] Step S7, the specific steps are as follows: Step S71, in actual application, four decibel meters are used to arrange monitoring points according to S1-S4 to monitor long-distance noise data, which are used as data input of the noise tracing model built before, to locate the leakage source, determine the position information of the leakage, and determine the hydrogen leakage pressure according to the structure characteristics of the hydrogen leakage accident area in the hydrogen-related scene. The parameter measurement uses a decibel meter to detect the noise at the monitoring point, a range finder to measure the distance from the arranged decibel meter to the leakage origin, and a movable robot to approach the leakage point to monitor the noise, to obtain the above three parameters. The noise is measured at a height of 1.5 m from the ground.
[0053] Step S72, the obtained monitoring point noise, leakage point noise and distance parameters are put into the dimensionless formula to obtain the size of the leakage hole diameter, and then the leakage flow rate is obtained through the flow calculation formula; Step S73, input the leakage flow rate and the leakage hole diameter label to call the hydrogen leakage diffusion accident case library to quickly obtain the consequences of the hydrogen leakage accident, compare with the measured hydrogen concentration data, estimate the leakage time on site, further predict the future hydrogen diffusion trend, and assist personnel rescue and evacuation.
[0054] The above examples are only used to illustrate the technical solutions of the present application, but not to limit them. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents, and these modifications or replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present application.
Claims
1. A method for tracing the source and determining the concentration of hydrogen leakage noise in hydrogen-related scenarios, characterized in that... Includes the following steps: S1. Analyze the characteristics of hydrogen leakage accidents in hydrogen-related scenarios, determine the possible locations and conditions of hydrogen leakage, conduct hydrogen leakage tests on representative hydrogen-related scenarios, verify the numerical simulation model through multi-point noise and concentration data acquisition, and evaluate the accuracy and reliability of the model; after verifying the reliability of the model, use the verified simulation model to study potential hydrogen leakage accidents in hydrogen-related scenarios. S2. Deploy multiple potential hydrogen concentration sensors and decibel meters in hydrogen-related scenarios to obtain concentration and noise data at the locations of the hydrogen concentration sensors and decibel meters after a hydrogen leak. S3. Based on different leakage scenarios, feature extraction and labeling of the collected data are performed, and a hydrogen leakage noise dataset and a typical case library of hydrogen leakage diffusion accidents are established, including operating condition labels and accident consequence data. S4. Based on a deep learning model combining convolutional neural networks and long short-term memory networks, time-frequency feature analysis is performed on multi-point noise signals to achieve source tracing and localization of leakage sources. S5. By obtaining the noise at the monitoring point, the noise at the leakage origin, and the distance from the noise monitoring point to the leakage origin, a dimensionless formula is derived for calculating the leakage orifice diameter. in, The average noise level (dB) at the monitoring point. The noise level at the origin is t seconds (dB). The leakage orifice diameter; The distance between the monitoring point and the leak point; The leakage flow rate is calculated using the corresponding flow rate formula based on the orifice diameter and leakage pressure. The formula satisfies the following function: Where d is the equivalent diameter of the leak outlet, and P is the leakage pressure; S6. Input the leakage orifice diameter and flow rate into the deep learning model and call the hydrogen leakage diffusion accident case library to obtain the consequences of the hydrogen leakage accident for on-site analysis and to estimate the leakage time. S7. The specific details of the field application phase are as follows. S71. Collect long-distance noise signals at the leak point using a decibel meter and input them into the model to locate the leak source. Determine the leakage pressure of the leak source based on the layout structure of the hydrogen-related scenario. Based on the location of the leak source, use a rangefinder and a mobile remote-controlled robot to obtain the distance from the monitoring point to the leak point and the noise at the leak point. S72. Based on the noise at the monitoring point, the noise at the leakage point, and the distance from the monitoring point to the leakage point obtained in S71, calculate the leakage orifice size using a dimensionless formula; based on the leakage orifice size and leakage pressure, further derive the leakage flow rate using a flow rate calculation formula. S73. Input the operating condition label and, based on the hydrogen leakage and diffusion accident case library, obtain the hydrogen leakage concentration distribution and estimate the on-site leakage time, further predict the future hydrogen concentration distribution, and assist personnel rescue and evacuation.
2. The method for tracing the source and determining the concentration of hydrogen leakage noise in hydrogen-related scenarios according to claim 1, characterized in that, In S1: The analysis of the characteristics of hydrogen leakage accidents in hydrogen-related scenarios includes analyzing the operation of equipment in hydrogen-related scenarios, environmental factors, and possible leakage locations; the hydrogen leakage test is conducted on a hydrogen leakage scenario platform that has been built, and the numerical simulation model is a full-size hydrogen leakage model built using CFD simulation software based on the actual scenario size and structural characteristics. Potential leakage scenarios are areas, equipment, and other places in the hydrogen-related scenario where hydrogen leakage may occur.
3. The method for tracing the source and determining the concentration of hydrogen leakage noise in hydrogen-related scenarios according to claim 1, characterized in that, In S2: Considering the layout characteristics of equipment in hydrogen-related scenarios, one decibel meter is deployed at the location where hydrogen leakage may occur to monitor noise at the leak point, and four decibel meters are deployed around the leak point to monitor the long-distance noise characteristics. The noise data at the origin is used for dimensionless formula calculation, and the noise data at the four monitoring points are used as input for machine learning models and for tracing the leak location. All decibel meters are placed 1.5m above the ground. The hydrogen concentration sensor's placement scheme is determined according to the actual situation of the hydrogen-related scenario. For hydrogen-related scenarios with fixed sensors, data is collected based on the original sensor locations. The sampling accuracy of both the decibel meter and the hydrogen concentration sensor is no greater than 1 second.
4. The method for tracing the source and determining the concentration of hydrogen leakage noise in hydrogen-related scenarios according to claim 1, characterized in that, In S3: Hydrogen-related scenarios include areas and equipment prone to hydrogen leakage, such as hydrogen production plants and hydrogen refueling stations, as well as fuel cell vehicle testing workshops and factories; operating condition labels include leakage orifice diameter and leakage flow rate.
5. The method for tracing the source and determining the concentration of hydrogen leakage noise in hydrogen-related scenarios according to claim 1, characterized in that, In S4: the convolutional neural network is a deep learning model used to extract features from discrete data; the long short-term memory network is a recurrent neural network used to capture the long-short-term dependencies of data.
6. The method for tracing the source and determining the concentration of hydrogen leakage noise in hydrogen-related scenarios according to claim 1, characterized in that, In S5: The dimensionless formula includes leakage orifice diameter, distance, and noise decibel terms, which are obtained by extracting acoustic features of the noise loudness during the hydrogen leakage stage.
7. The method for tracing the source and determining the concentration of hydrogen leakage noise in hydrogen-related scenarios according to claim 1, characterized in that, In S6: By inputting tags, the system can directly access the hydrogen leak diffusion consequence case library, obtain hydrogen leak accident consequence data, compare it with the concentration results monitored on site, estimate the leak time, further predict the future hydrogen concentration distribution, and assist in personnel rescue and evacuation.
8. The method for tracing the source and determining the concentration of hydrogen leakage noise in hydrogen-related scenarios according to claim 1, characterized in that, In S71: In practical applications, noise data is monitored on-site and used as data input for the aforementioned noise source tracing model to locate the leak source and determine its location. Based on the structural characteristics of the area where the hydrogen leak accident occurred in a hydrogen-related scenario, the magnitude of the hydrogen leak pressure is determined. A decibel meter is used to collect noise at the monitoring point, and a rangefinder is used to measure the distance from the monitoring point to the leak origin. A mobile robot is used to monitor the noise at the leak point to obtain the above three parameters. During on-site noise measurement, the monitoring points are all at a height of 1.5m above the ground.
9. The method for tracing the source and determining the concentration of hydrogen leakage noise in hydrogen-related scenarios according to claim 1, characterized in that, In S72: The noise at the monitoring point, the noise at the leak point, and the distance parameters obtained above are applied to the dimensionless formula to obtain the size of the leak orifice, and then the leakage flow rate is obtained through the flow rate calculation formula.
10. The method for tracing the source and determining the concentration of hydrogen leakage noise in hydrogen-related scenarios according to claim 1, characterized in that, In S73: By retrieving the hydrogen leak and diffusion accident case database, the consequences of hydrogen leak accidents are clarified, and the hydrogen concentration data obtained from on-site monitoring is compared to estimate the time of the leak and predict the subsequent leakage and diffusion, thus assisting in personnel rescue and evacuation.