CN116306377B - Method and system for rapidly predicting consequences of leakage accident of hydrogen station - Google Patents
Method and system for rapidly predicting consequences of leakage accident of hydrogen station Download PDFInfo
- Publication number
- CN116306377B CN116306377B CN202310349824.9A CN202310349824A CN116306377B CN 116306377 B CN116306377 B CN 116306377B CN 202310349824 A CN202310349824 A CN 202310349824A CN 116306377 B CN116306377 B CN 116306377B
- Authority
- CN
- China
- Prior art keywords
- hydrogen
- station
- cloud
- training
- concentration
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 239000001257 hydrogen Substances 0.000 title claims abstract description 208
- 229910052739 hydrogen Inorganic materials 0.000 title claims abstract description 208
- UFHFLCQGNIYNRP-UHFFFAOYSA-N Hydrogen Chemical compound [H][H] UFHFLCQGNIYNRP-UHFFFAOYSA-N 0.000 title claims abstract description 207
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000012549 training Methods 0.000 claims abstract description 74
- 238000004088 simulation Methods 0.000 claims abstract description 15
- 230000015654 memory Effects 0.000 claims abstract description 5
- 238000013528 artificial neural network Methods 0.000 claims description 22
- 238000009792 diffusion process Methods 0.000 claims description 12
- 238000012544 monitoring process Methods 0.000 claims description 12
- 239000007789 gas Substances 0.000 claims description 11
- 230000007787 long-term memory Effects 0.000 claims description 8
- 238000013135 deep learning Methods 0.000 claims description 6
- 230000006403 short-term memory Effects 0.000 claims description 6
- 238000012856 packing Methods 0.000 claims description 3
- 150000002431 hydrogen Chemical class 0.000 claims description 2
- 230000007423 decrease Effects 0.000 claims 1
- 230000006870 function Effects 0.000 description 8
- 238000004364 calculation method Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 239000003795 chemical substances by application Substances 0.000 description 5
- 238000004806 packaging method and process Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 238000004880 explosion Methods 0.000 description 2
- 239000007788 liquid Substances 0.000 description 2
- 210000005036 nerve Anatomy 0.000 description 2
- 230000000717 retained effect Effects 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 239000012080 ambient air Substances 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000000446 fuel Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000009965 odorless effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/28—Design optimisation, verification or simulation using fluid dynamics, e.g. using Navier-Stokes equations or computational fluid dynamics [CFD]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/08—Fluids
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
- Y02E60/30—Hydrogen technology
- Y02E60/32—Hydrogen storage
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Computer Hardware Design (AREA)
- Mathematical Optimization (AREA)
- Health & Medical Sciences (AREA)
- Mathematical Analysis (AREA)
- Fluid Mechanics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Algebra (AREA)
- Pure & Applied Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a quick predicting method and a system for the consequences of a leakage accident of a hydrogen station, which relate to the technical field of safety management of the leakage accident of the hydrogen station and have the technical scheme that: obtaining the space-time distribution characteristics of hydrogen leakage results of the hydrogen adding station by a numerical simulation method to form a training data set; training a result prediction proxy model based on a long-period memory network based on the hydrogen leakage accident result numerical simulation result of the hydrogen adding station; and taking hydrogen concentration data on site of the hydrogen adding station as input, calling a result prediction proxy model for completing training, and predicting the result of the hydrogen leakage accident. The method can realize the real-time prediction of the leakage accident result of the hydrogen station within the range of relatively acceptable accuracy; and key data support is provided for emergency measure schemes and safety distance judgment of site operators and emergency rescue personnel of the hydrogen station.
Description
Technical Field
The invention relates to the technical field of emergency safety management of a hydrogen addition station, in particular to a method and a system for rapidly predicting the consequences of a leakage accident of the hydrogen addition station.
Background
Hydrogen energy applications such as hydrogen fuel cell automobiles and hydrogen stations are rapidly expanding in industrial scale. The potential leakage risk of a large amount of high-pressure hydrogen gas is the most important factor currently restricting the development of the hydrogen energy industry.
Hydrogen is colorless and odorless as the smallest molecular weight gas and is highly flammable over a wide range of concentrations. Unexpected leakage of hydrogen can form a flammable mixture with ambient air in a short period of time, which can lead to serious fire and explosion accidents once an ignition source is encountered. The rapid development of the hydrogen industry is accompanied by such a serious fire explosion risk that cannot be ignored. To address these risks, targeted risk early warning and emergency measures are required, which in turn require rapid prediction of potential accident consequences.
In the related accident result prediction technology, a theoretical formula and a numerical simulation method are mainly adopted. The theoretical formula method is a liquid hydrogen leakage diffusion range prediction method which is combined with a theoretical formula describing a hydrogen leakage diffusion process according to a large number of experiments and accident statistics data analysis and summarized, and directly calculates a hydrogen distribution range through leakage source conditions so as to further predict accident risk grades and accident consequences, such as a liquid hydrogen leakage diffusion range prediction method proposed by CN 113128755A. However, although such methods can obtain risk values or impact ranges after leakage in a short time, the evolution of the consequences of an accident under complex conditions cannot be predicted. Another method of numerical simulation can predict accident consequences more accurately, but the calculation cost is relatively higher, and the prediction result can only aim at single scene accidents, and usually, a few hours or even days are required to perform accurate numerical calculation to determine accident consequences. Therefore, the current commonly used empirical formulas and numerical simulation methods hardly realize real-time prediction of accident results, so researchers have proposed to realize rapid prediction of accident results through a deep learning algorithm. CN114429589a proposes training a concentration distribution prediction model based on the concentration and image data after hydrogen leakage, so as to realize rapid high-precision prediction of hydrogen concentration distribution. However, in the hydrogen leakage accident of the hydrogen adding station, complicated structures such as a hydrogen adding machine, a hydrogen storage tank and the like in the hydrogen adding station have significant influence on the hydrogen concentration distribution. Therefore, it is necessary to propose and design a method and a system for quickly predicting the consequences of the hydrogen leakage accident of the hydrogen addition station, and the quick prediction of the consequences of the hydrogen leakage accident of the hydrogen addition station is realized on the premise of acceptable precision.
Disclosure of Invention
The invention aims to provide a method and a system for quickly predicting the consequences of a leakage accident of a hydrogen station, so as to solve the problems in the background technology.
In order to achieve the above purpose, the present invention provides the following technical solutions:
in a first aspect, a method for rapidly predicting the consequences of a hydrogen station leakage accident is provided, comprising the steps of:
s1, obtaining a training data set, obtaining hydrogen cloud concentration distribution conditions after hydrogen leakage of a hydrogen station by using a CFD numerical simulation method, and packing and storing sequences of hydrogen cloud concentration cloud patterns at a plurality of heights along with time as training samples;
s2, training and storing a proxy model, training a deep learning proxy model based on a long-term and short-term memory network by using the obtained training data set, so that the proxy model can relatively accurately predict the hydrogen cloud concentration distribution change condition after hydrogen leakage accidents of the hydrogen station occur;
s3, predicting hydrogen leakage results based on the stored proxy model, reading hydrogen concentration sensor data of the hydrogen station site, importing the current hydrogen concentration distribution situation of the hydrogen station into a deep learning proxy model which is trained, predicting hydrogen cloud distribution situation at the next moment by using the proxy model, and visually displaying the prediction results.
Further, the training data set obtaining process specifically includes:
s101, building a virtual three-dimensional model of the hydrogen adding station, and building a virtual three-dimensional model of a specific hydrogen adding station for predicting accident results by using the method, so as to ensure that the geometric dimensions of key components such as a hydrogen adding machine, a ceiling, a station house, a hydrogen storage tank, a compressor, a hydrogen conveying pipeline and the like are consistent with those of an actual hydrogen adding station;
s102, carrying out accident result virtual simulation, and calculating a plurality of typical hydrogen leakage accident scenes of the hydrogen addition station by using CFD numerical simulation software to obtain the concentration distribution condition of hydrogen cloud after hydrogen leakage of the hydrogen addition station occurs;
and S103, storing and structuring simulation result data as training samples, and packing and storing sequences of the hydrogen cloud concentration cloud patterns at a plurality of heights, which are obtained through numerical simulation, along with time as the training samples.
Further, the training and storing agent model process specifically includes:
s201, importing training sample data and preprocessing, importing sequence images of the hydrogen cloud concentration cloud pictures stored in a packaged mode along with time change into a nerve network model, normalizing the training sample data before training starts, and adopting the following calculation formula:
wherein,to normalize the pixel value of single pixel point of hydrogen concentration cloud image, x pix Pixel value, x of single pixel point of original hydrogen concentration cloud image obtained by CFD calculation min Is the minimum pixel value, x in a single Zhang Qing concentration cloud picture max The maximum pixel value in the single Zhang Qing concentration cloud picture;
s202, building a long-term memory network model, wherein in the long-term memory network, forgetting information needs to be determined first. For a neural network element, the "forget gate" can be expressed by the following formula:
f t ＝σ(W f ·[h t-1 ,x t ]+b f ) (2)
wherein h is t-1 And x t For input of the current neural network element, W f Weight, b f To be biased, f t The output of the forgetting gate is a number between 0 and 1. The output "1" indicates that "this information is completely retained" and "0" indicates that "this information is completely forgotten";
after determining the forgetting information, the information to be saved is further determined. For a neural network element, the "input gate" includes a sigmoid layer and a tanh layer, which can be expressed by the following formula:
wherein i is t For the output of the sigmoid layer,for output of tanh layer, W i 、W c Weight, b i 、b c For biasing, the sigmoid layer determines information that the current network element needs to be updated, and the tanh layer generates new candidate data and integrates the new candidate data into the neural network element in a subsequent process;
after determining the forgetting information and saving the information, the current neural network element state is further updated, which can be expressed by the following formula:
wherein C is t C is the output state of the current neural network unit t-1 To the output state of the last neural network element, the output of the last element is multiplied by f t Discarding the information needing to be forgotten, and adding the state information newly input by the current unit to obtain the output state information of the current unit;
finally, the output information of the current neural network unit needs to be determined, and for a neural network unit, the output gate comprises a sigmoid layer and a tanh layer, the output gate can be expressed by the following formula:
wherein o is t For output of sigmoid layer, h t For the output of the current unit, W o Weight, b o For biasing, firstly determining an output part of the current network element state by a sigmoid layer, and then normalizing the output value to a range of-1 to 1 through a tanh layer to further obtain the output state value of the current network element.
S203, training a result prediction proxy model, training the proxy model by using training super parameters, and considering that training is completed when the loss value is acceptable, wherein the specific process is as follows:
selecting proper iteration times, learning rate and loss function expression, and developing proxy model training;
the iteration times and the learning rate are related to the complexity of the model based on a specific geometric scene of the hydrogen station, and an optimal set value is obtained through a plurality of training attempts;
the loss function is an operation function for measuring the degree of difference between a model generated output value and a true value in the training process of the proxy model, is a non-negative real value function, comprises two parts of training loss and physical loss, and can be expressed by the following formula:
loss＝loss tr +loss phy (6)
wherein loss is tr Loss of training phy The mean square error value is selected for the physical loss, which can be further expressed as:
wherein n is tr To train the number of samples, y i In order to train the true value of the sample,generating a value for the model;
physical loss is defined as a loss term based on physical constraints of the hydrogen free diffusion process, in which the concentration at the hydrogen cloud boundary satisfies the continuity equation, i.e., the concentration changes continuously and exhibits a positive gradient drop from the leak point to the gas cloud boundary, can be expressed by the following formula:
wherein delta is i,t N is the concentration difference between the boundary of the hydrogen cloud and the inner side of the boundary edge For the number of concentration sample points at the boundary of the hydrogen cloud, Y i,edge-1 Sample point concentration for hydrogen cloud boundary internal measurementDegree prediction value, Y i,edge A sample point concentration predicted value at a hydrogen cloud boundary; the neural network will automatically iterate to find loss in the training process phy And (3) the prediction accuracy at the boundary of the gas cloud can be improved.
S204, storing the trained proxy model, and packaging and storing the trained proxy model.
Further, the hydrogen leakage result prediction process based on the stored proxy model specifically comprises the following steps:
s301, loading a result prediction proxy model, and loading a hydrogen leakage result prediction proxy model after training;
s302, reading and identifying current site monitoring parameters, reading current hydrogen concentration distribution data in a hydrogen adding station through a hydrogen adding station site hydrogen concentration sensor, and identifying accident scene information such as leakage sources, leakage flow and the like of current hydrogen leakage;
s303, predicting accident results, inputting the hydrogen concentration distribution condition and accident scene information in the current hydrogen station to a hydrogen leakage result prediction proxy model, outputting the hydrogen leakage accident result prediction results, and visually displaying the prediction results.
In a second aspect, a system for rapid prediction of the consequences of a hydrogen station leak event is provided, comprising:
and the training sample generation module is used for generating sample data for proxy model training to form a sample set of the space-time distribution characteristics of typical leakage accident consequences of the hydrogen station.
And the prediction model training module is used for training a hydrogen station leakage result prediction proxy model based on the long-term and short-term memory network through the generated training sample data, and storing and packaging the trained model.
The on-site information monitoring module is used for identifying and judging the current operating state of the hydrogen adding station according to the monitoring data of the hydrogen concentration sensor arranged on the site of the hydrogen adding station.
And the result prediction presentation module is used for calling the result prediction proxy model which has completed training, predicting the leakage accident result by using the field monitoring data in the current state, and visually displaying the distribution condition of the combustible gas.
Compared with the prior art, the invention has the following beneficial effects:
1. according to the quick prediction method for the leakage accident result of the hydrogen station, disclosed by the invention, the real-time prediction of the leakage accident result of the hydrogen station is realized within a relatively acceptable accuracy range through a deep learning technology, so that the time required for accident prediction is greatly shortened;
2. the rapid prediction result of the hydrogen leakage accident result of the hydrogen station, particularly the hydrogen cloud concentration space-time distribution characteristic after the hydrogen leakage occurs, can provide key data support for emergency measure schemes and safety distance judgment for site operators and emergency rescue personnel of the hydrogen station.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and together with the description, serve to explain the principles of the application.
FIG. 1 is a flow chart showing the steps of a method for rapidly predicting the consequences of a hydrogen station leak event in accordance with a first embodiment of the present invention;
FIG. 2 is a diagram illustrating a network structure of a long-short-term memory network according to a first embodiment of the present invention;
FIG. 3 is a diagram illustrating the definition of loss terms based on physical constraints according to a first embodiment of the present invention;
FIG. 4 is a virtual three-dimensional model of a hydrogen addition station in accordance with a first embodiment of the present invention;
FIG. 5 is a graph showing an example of the results of numerical simulation of hydrogen leakage diffusion concentration distribution in a hydrogen addition station according to the first embodiment of the present invention;
FIG. 6 is a time-varying sequence of hydrogen cloud concentration clouds at a certain altitude in accordance with one embodiment of the present invention;
FIG. 7 is a graph showing the predicted hydrogen concentration distribution timing of the predictive proxy model in accordance with the first embodiment of the invention;
FIG. 8 is a schematic diagram of a fast predicting system for the consequences of a hydrogen station leak in accordance with a second embodiment of the present invention.
Detailed Description
The present invention will be further described in detail by way of specific embodiments with reference to the accompanying drawings, which are only for the purpose of illustrating the invention and are not to be construed as limiting the scope of the invention.
Example 1: fig. 1 is a flow chart of steps of a method for quickly predicting the consequences of a leakage accident of a hydrogen station, which is provided by the invention, and as shown in fig. 1, specifically includes the following steps:
in step S1, a training data set is obtained. The specific process is as follows:
s101, constructing a virtual three-dimensional model of the hydrogen adding station. Fig. 2 is a schematic diagram of a virtual three-dimensional model of a hydrogen addition station according to an embodiment of the present invention. As shown in fig. 2, a virtual three-dimensional model of a specific hydrogen adding station for predicting accident results by using the method is built, so that the geometric dimensions of key components such as a hydrogen adding machine, a ceiling, a station house, a hydrogen storage tank, a compressor, a hydrogen conveying pipeline and the like are ensured to be consistent with those of an actual hydrogen adding station;
s102, carrying out accident result virtual simulation. FIG. 3 is a graph showing an example of the results of numerical simulation of hydrogen leakage diffusion concentration distribution in a hydrogen addition station according to an embodiment of the present invention. As shown in fig. 3, calculating a plurality of typical hydrogen leakage accident scenes of the hydrogen addition station by using CFD numerical simulation software to obtain the concentration distribution condition of hydrogen cloud after hydrogen leakage of the hydrogen addition station;
s103, saving and structuring simulation result data as training samples. Fig. 4 is a sequence example diagram of a hydrogen cloud concentration cloud graph at a certain height over time according to an embodiment of the present invention. As shown in fig. 4, the time-varying sequence of the hydrogen cloud concentration cloud patterns at several heights obtained by numerical simulation is packed and stored as training samples.
In step S2, the proxy model is trained and saved, and the specific process is as follows:
s201, training sample data is imported and preprocessed. The sequence images of the hydrogen cloud concentration cloud images which are stored in a packaged mode and change along with time are imported into a nerve network model, training sample data are normalized before training is started, and a calculation formula is as follows:
wherein,to normalize the pixel value of single pixel point of hydrogen concentration cloud image, x pix Pixel value, x of single pixel point of original hydrogen concentration cloud image obtained by CFD calculation min Is the minimum pixel value, x in a single Zhang Qing concentration cloud picture max The maximum pixel value in the single Zhang Qing concentration cloud picture;
s202, building a long-term and short-term memory network model. As shown in fig. 5, the long-short-term memory network structure includes a "forget gate", "input gate", "output gate", and a status updating unit. In long and short term memory networks, first it is necessary to determine forgetting information. For a neural network element, the "forget gate" can be expressed by the following formula:
f t ＝σ(W f ·[h t-1 ,x t ]+b f ) (10)
wherein h is t-1 And x t For input of the current neural network element, W f Weight, b f To be biased, f t The output of the forgetting gate is a number between 0 and 1. The output "1" indicates that "this information is completely retained" and "0" indicates that "this information is completely forgotten";
after determining the forgetting information, the information to be saved is further determined. For a neural network element, the "input gate" includes a sigmoid layer and a tanh layer, which can be expressed by the following formula:
wherein i is t For the output of the sigmoid layer,for output of tanh layer, W i 、W c Weight, b i 、b c For bias, sigmoid layer decides the current network listThe information that the element needs to be updated, the tanh layer generates new candidate data and is integrated into the neural network unit in the subsequent flow;
after determining the forgetting information and saving the information, the current neural network element state is further updated, which can be expressed by the following formula:
wherein C is t C is the output state of the current neural network unit t-1 To the output state of the last neural network element, the output of the last element is multiplied by f t Discarding the information needing to be forgotten, and adding the state information newly input by the current unit to obtain the output state information of the current unit;
finally, the output information of the current neural network unit needs to be determined, and for a neural network unit, the output gate comprises a sigmoid layer and a tanh layer, the output gate can be expressed by the following formula:
wherein o is t For output of sigmoid layer, h t For the output of the current unit, W o Weight, b o For biasing, firstly determining an output part of the current network element state by a sigmoid layer, and then normalizing the output value to a range of-1 to 1 through a tanh layer to further obtain the output state value of the current network element.
S203, training a result prediction agent model. Selecting proper training super parameters for carrying out agent model training, and considering that training is completed when the loss value is acceptable, wherein the specific process is as follows:
selecting proper iteration times, learning rate and loss function expression, and developing proxy model training;
the iteration times and the learning rate are related to the complexity of the model based on a specific geometric scene of the hydrogen station, and an optimal set value is obtained through a plurality of training attempts;
the loss function is an operation function for measuring the degree of difference between a model generated output value and a true value in the training process of the proxy model, is a non-negative real value function, comprises two parts of training loss and physical loss, and can be expressed by the following formula:
loss＝loss tr +loss phy (14)
wherein loss is tr Loss of training phy The mean square error value is selected for the physical loss, which can be further expressed as:
wherein n is tr To train the number of samples, y i In order to train the true value of the sample,generating a value for the model;
as shown in fig. 6, the physical loss is defined as a loss term based on physical constraints of the hydrogen free diffusion process, and the concentration at the boundary of the hydrogen cloud should satisfy the continuity equation during hydrogen leak diffusion in the hydrogen station, that is, the concentration changes continuously and exhibits a positive gradient from the leak point to the boundary of the gas cloud, and can be expressed by the following formula:
wherein delta is i,t N is the concentration difference measured between the boundary of the hydrogen cloud and the inside of the boundary edge For the number of concentration sample points at the boundary of the hydrogen cloud, Y i,edge-1 For the predicted value of the concentration of the sample point measured in the hydrogen cloud boundary, Y i,edge A sample point concentration predicted value at a hydrogen cloud boundary; the neural network will automatically iterate to find loss in the training process phy And (3) the prediction accuracy at the boundary of the gas cloud can be improved.
S204, saving the trained agent model. And packaging and storing the trained proxy model.
Further, the hydrogen leakage result prediction process based on the stored proxy model specifically comprises the following steps:
s301, loading a result prediction agent model. Loading a hydrogen leakage result prediction proxy model after training;
s302, reading and identifying current field monitoring parameters. Reading the current hydrogen concentration distribution data in the hydrogen adding station by a hydrogen adding station site hydrogen concentration sensor, and identifying accident scene information such as leakage source, leakage flow and the like of current hydrogen leakage;
s303, predicting accident consequences. Fig. 7 is an exemplary graph of a predicted hydrogen concentration distribution timing result predicted by a prediction proxy model according to an embodiment of the present invention. And inputting the hydrogen concentration distribution condition and accident scene information in the current hydrogen adding station to the hydrogen leakage accident result prediction proxy model, outputting a hydrogen leakage accident result prediction result, as shown in fig. 7, and visually presenting the prediction result.
Example 2: fig. 8 is a schematic diagram of a fast predicting system for the consequences of a leakage accident of a hydrogen station according to the present invention, as shown in fig. 8, the system mainly includes:
training sample generation module 1, prediction model training module 2, on-site information monitoring module 3, and result prediction presentation module 4.
The training sample generation module 1 is used for generating sample data for proxy model training to form a sample set of the time-space distribution characteristics of typical leakage accident consequences of the hydrogen station;
the prediction model training module 2 is used for training a hydrogen station leakage result prediction proxy model based on a long-term and short-term memory network through the generated training sample data, and storing and packaging the trained model;
the on-site information monitoring module 3 is used for identifying and judging the current operating state of the hydrogen adding station according to the monitoring data of the hydrogen concentration sensor arranged on the site of the hydrogen adding station;
and the result prediction presentation module 4 is used for calling the result prediction proxy model which has completed training, predicting the leakage accident result by using the field monitoring data in the current state, and visually displaying the distribution condition of the combustible gas.
In summary, the invention provides a method and a system for rapidly predicting the consequences of hydrogen leakage accidents of a hydrogen addition station, which are characterized in that the space-time distribution characteristics of the hydrogen leakage consequences of the hydrogen addition station are obtained through a numerical simulation method to form a training data set; training a result prediction proxy model based on a long-period memory network based on the hydrogen leakage accident result numerical simulation result of the hydrogen adding station; and taking hydrogen concentration data on site of the hydrogen adding station as input, calling a result prediction proxy model for completing training, and predicting the result of the hydrogen leakage accident. The rapid prediction method provided by the invention can realize the real-time prediction of the leakage accident result of the hydrogen station within the range of relatively acceptable accuracy; and key data support is provided for emergency measure schemes and safety distance judgment of site operators and emergency rescue personnel of the hydrogen station.
Other embodiments of the present application will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of the application following, in general, the principles of the application and including such departures from the present disclosure as come within known or customary practice within the art to which the application pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the application being indicated by the following claims.
It is to be understood that the present application is not limited to the precise arrangements and instrumentalities shown in the drawings, which have been described above, and that various modifications and changes may be effected without departing from the scope thereof. The scope of the application is limited only by the appended claims.
Claims (2)
1. A method for rapidly predicting the consequences of a hydrogen station leak event, comprising:
s1, obtaining a training data set, constructing a virtual three-dimensional model of a hydrogen adding station, ensuring that the geometric dimensions of key components including a hydrogen adding machine, a ceiling, a station house, a hydrogen storage tank, a compressor and a hydrogen conveying pipeline are consistent with those of an actual hydrogen adding station, obtaining the concentration distribution condition of hydrogen cloud after hydrogen leakage of the hydrogen adding station by using a CFD numerical simulation method, and packing and storing sequences of the hydrogen cloud concentration cloud patterns at a plurality of heights along with time as training samples;
s2, training and saving a proxy model, and training a deep learning proxy model based on a long-term and short-term memory network and taking physical loss term into consideration by using the obtained training data set, wherein the physical loss is a loss term based on physical constraint of a hydrogen free diffusion process, and the concentration at a hydrogen cloud boundary satisfies a continuity equation in the hydrogen leakage diffusion process of a hydrogen station, namely the concentration change is continuous and positive gradient decline is shown from a leakage point to the gas cloud boundary, wherein the physical loss is expressed asWherein delta is i,t ＝Y i,edge-1 -Y i,edge ，Δ i,t N is the concentration difference between the boundary of the hydrogen cloud and the inner side of the boundary tr To train the number of samples, n edge For the number of concentration sample points at the boundary of the hydrogen cloud, Y i,edge-1 For the predicted value of the concentration of the sample point measured in the hydrogen cloud boundary, Y i,edge For the predicted value of the concentration of the sample point at the boundary of the hydrogen cloud, the neural network automatically and iteratively searches loss in the training process phy The prediction precision at the boundary of the gas cloud can be improved by the minimum value of the gas cloud;
s3, predicting hydrogen leakage results based on the stored proxy model, reading hydrogen concentration sensor data of the hydrogen station site, importing the current hydrogen concentration distribution situation of the hydrogen station into a deep learning proxy model which is trained, predicting hydrogen cloud distribution situation at the next moment by using the proxy model, and visually displaying the prediction results.
2. A system for rapid prediction of the consequences of a hydrogen station leak event, comprising:
the training sample generation module is used for generating sample data for proxy model training, constructing a virtual three-dimensional model of the hydrogen adding station, ensuring that the geometric dimensions of key components comprising the hydrogen adding machine, the ceiling, the station house, the hydrogen storage tank, the compressor and the hydrogen conveying pipeline are consistent with those of an actual hydrogen adding station, and forming a sample set based on the time-space distribution characteristics of typical leakage accident results of the hydrogen adding station;
the prediction model training module is used for training a hydrogen station leakage result prediction proxy model based on a long-short-term memory network and taking physical loss terms into consideration through generated training sample data, wherein the physical loss is a loss term based on physical constraint of a hydrogen free diffusion process, and the concentration at a hydrogen cloud boundary satisfies a continuity equation in the hydrogen station hydrogen leakage diffusion process, namely the concentration change is continuous and positive gradient drop is shown from a leakage point to a gas cloud boundary, wherein the concentration is expressed asWherein delta is i,t ＝Y i,edge-1 -Y i,edge ，Δ i,t N is the concentration difference between the boundary of the hydrogen cloud and the inner side of the boundary tr To train the number of samples, n edge For the number of concentration sample points at the boundary of the hydrogen cloud, Y i,edge-1 For the predicted value of the concentration of the sample point measured in the hydrogen cloud boundary, Y i,edge For the predicted value of the concentration of the sample point at the boundary of the hydrogen cloud, the neural network automatically and iteratively searches loss in the training process phy The prediction precision at the boundary of the air cloud can be improved, and the trained model is stored and packaged;
the on-site information monitoring module is used for identifying and judging the current operating state of the hydrogen adding station according to the monitoring data of the hydrogen concentration sensor arranged on the site of the hydrogen adding station;
and the result prediction presentation module is used for calling the result prediction proxy model which has completed training, predicting the leakage accident result by using the field monitoring data in the current state, and visually displaying the hydrogen distribution condition.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310349824.9A CN116306377B (en) | 2023-04-04 | 2023-04-04 | Method and system for rapidly predicting consequences of leakage accident of hydrogen station |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310349824.9A CN116306377B (en) | 2023-04-04 | 2023-04-04 | Method and system for rapidly predicting consequences of leakage accident of hydrogen station |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116306377A CN116306377A (en) | 2023-06-23 |
| CN116306377B true CN116306377B (en) | 2024-04-05 |
Family
ID=86820183
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310349824.9A Active CN116306377B (en) | 2023-04-04 | 2023-04-04 | Method and system for rapidly predicting consequences of leakage accident of hydrogen station |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116306377B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118211489A (en) * | 2024-04-19 | 2024-06-18 | 中国石油大学(华东) | Case library-based intelligent rapid prediction method for consequences of leakage accidents of hydrogen station |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111144055A (en) * | 2019-12-27 | 2020-05-12 | 苏州大学 | Method, device and medium for determining toxic heavy gas leakage concentration distribution in urban environment |
| CN111365624A (en) * | 2020-03-20 | 2020-07-03 | 淮阴工学院 | Intelligent terminal and method for detecting leakage of brine transportation pipeline |
| WO2022052068A1 (en) * | 2020-09-11 | 2022-03-17 | 西门子（中国）有限公司 | Target available model-based environment prediction method and apparatus, program, and electronic device |
| CN114429589A (en) * | 2022-04-07 | 2022-05-03 | 北京理工大学 | Hydrogen leakage concentration distribution prediction method and system |
| WO2022174508A1 (en) * | 2021-02-20 | 2022-08-25 | 华南理工大学 | Oil and gas major infrastructure multi-disaster type event coupling three-dimensional simulation system |
| CN115496003A (en) * | 2022-06-08 | 2022-12-20 | 华南理工大学 | Overpressure injury assessment method for leakage explosion accident of hydrogenation station |
| CN115600643A (en) * | 2022-10-17 | 2023-01-13 | 中国科学技术大学（Cn） | Method and system for rapidly predicting toxic gas |
-
2023
- 2023-04-04 CN CN202310349824.9A patent/CN116306377B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111144055A (en) * | 2019-12-27 | 2020-05-12 | 苏州大学 | Method, device and medium for determining toxic heavy gas leakage concentration distribution in urban environment |
| CN111365624A (en) * | 2020-03-20 | 2020-07-03 | 淮阴工学院 | Intelligent terminal and method for detecting leakage of brine transportation pipeline |
| WO2022052068A1 (en) * | 2020-09-11 | 2022-03-17 | 西门子（中国）有限公司 | Target available model-based environment prediction method and apparatus, program, and electronic device |
| WO2022174508A1 (en) * | 2021-02-20 | 2022-08-25 | 华南理工大学 | Oil and gas major infrastructure multi-disaster type event coupling three-dimensional simulation system |
| CN114429589A (en) * | 2022-04-07 | 2022-05-03 | 北京理工大学 | Hydrogen leakage concentration distribution prediction method and system |
| CN115496003A (en) * | 2022-06-08 | 2022-12-20 | 华南理工大学 | Overpressure injury assessment method for leakage explosion accident of hydrogenation station |
| CN115600643A (en) * | 2022-10-17 | 2023-01-13 | 中国科学技术大学（Cn） | Method and system for rapidly predicting toxic gas |
Non-Patent Citations (3)
| Title |
|---|
| CFD simulation and experimental study of a hydrogen leak in a semi-closed space with the purpose of risk mitigation;Malakhov, A.A.等;International Journal of Hydrogen Energy;20201231;第45卷(第15期);9231-9240 * |
| LSTM在煤矿瓦斯预测预警系统中的应用与设计;李伟山;王琳;卫晨;;西安科技大学学报(第06期);全文 * |
| 基于概率神经网络的液压管路泄漏故障程度识别;王立文;刘强;霍金鉴;姜兴禹;胡建伟;唐杰;;机床与液压(第04期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116306377A (en) | 2023-06-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116306377B (en) | Method and system for rapidly predicting consequences of leakage accident of hydrogen station | |
| Nabiyan et al. | Mechanics‐based model updating for identification and virtual sensing of an offshore wind turbine using sparse measurements | |
| CN103914622A (en) | Quick chemical leakage predicating and warning emergency response decision-making method | |
| CN117387559B (en) | Concrete bridge monitoring system and method based on digital twinning | |
| CN113944888B (en) | Gas pipeline leakage detection method, device, equipment and storage medium | |
| CN113837451B (en) | Method, device, equipment and storage medium for constructing digital twin body of oil and gas pipeline | |
| KR20170124124A (en) | Gas leak concentration prediction method generated in the confined spaces | |
| CN111583067A (en) | Urban underground large space construction safety early warning and emergency decision-making method and system | |
| He et al. | Prediction model for the evolution of hydrogen concentration under leakage in hydrogen refueling station using deep neural networks | |
| KR20210078197A (en) | Safety management system for hydrogen supply station using virtual sensor | |
| CN105717257B (en) | A kind of gas source wireless location method | |
| KR20170040908A (en) | Probabilistic gas explosion scenario calculation system and probabilistic gas explosion scenario calculation method using the same | |
| He et al. | Hybrid neural network-based surrogate model for fast prediction of hydrogen leak consequences in hydrogen refueling station | |
| Song et al. | A Novel Outlier Detection Method of Long‐Term Dam Monitoring Data Based on SSA‐NAR | |
| Hu et al. | Prediction and interpretability of accidental explosion loads from hydrogen-air mixtures using CFD and artificial neural network method | |
| Li et al. | Sensor Fault Diagnosis Method of Bridge Monitoring System Based on FS-LSTM | |
| CN115879378A (en) | Training method and device for expansion force prediction model of battery cell | |
| CN115688464A (en) | Visualization method and system for hydrogen leakage safety boundary | |
| CN118114575B (en) | Combustible gas leakage diffusion range prediction simulation method and system | |
| CN114491877A (en) | Method, device, system, terminal and medium for determining pipeline leakage influence area | |
| CN117851879B (en) | Method and device for studying, judging and predicting disaster evolution of leakage explosion accident of hydrogen station | |
| CN103455689B (en) | Based on the interval model modification method of Taylor series expansion | |
| CN115326295B (en) | Hydrogen leakage detection method, device, equipment and storage medium | |
| CN113657018A (en) | Method and device for predicting leakage flow in real time after gas storage tank leaks | |
| CN118408646B (en) | Method and system for early warning of abnormal temperature of power cable connector |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |