CN107798467A - Water pollution burst accident based on deep learning quickly meet an urgent need assess and decision-making technique - Google Patents

Water pollution burst accident based on deep learning quickly meet an urgent need assess and decision-making technique Download PDF

Info

Publication number
CN107798467A
CN107798467A CN201710939980.5A CN201710939980A CN107798467A CN 107798467 A CN107798467 A CN 107798467A CN 201710939980 A CN201710939980 A CN 201710939980A CN 107798467 A CN107798467 A CN 107798467A
Authority
CN
China
Prior art keywords
decision
accident
assessment
deep learning
water
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.)
Granted
Application number
CN201710939980.5A
Other languages
Chinese (zh)
Other versions
CN107798467B (en
Inventor
卢滨
陈义中
常文婷
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hangzhou Environmental Protection Science Research Institute
Original Assignee
Hangzhou Environmental Protection Science Research Institute
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Hangzhou Environmental Protection Science Research Institute filed Critical Hangzhou Environmental Protection Science Research Institute
Priority to CN201710939980.5A priority Critical patent/CN107798467B/en
Publication of CN107798467A publication Critical patent/CN107798467A/en
Application granted granted Critical
Publication of CN107798467B publication Critical patent/CN107798467B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0637—Strategic management or analysis, e.g. setting a goal or target of an organisation; Planning actions based on goals; Analysis or evaluation of effectiveness of goals
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/26—Government or public services
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16Z—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS, NOT OTHERWISE PROVIDED FOR
    • G16Z99/00—Subject matter not provided for in other main groups of this subclass
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2219/00—Indexing scheme relating to application aspects of data processing equipment or methods
    • G06F2219/10—Environmental application, e.g. waste reduction, pollution control, compliance with environmental legislation

Abstract

The present invention relates to field of environment protection.Purpose is to provide a kind of water pollution burst accident based on deep learning quickly emergent assessment and decision-making technique, quick to form emergent assessment result and be used for the technical support of Emergency decision on the basis of the neural network model established by deep learning.Technical scheme is：A kind of water pollution burst accident based on deep learning quickly meet an urgent need assess and decision-making technique, comprise the following steps：(1) upstream hydrologic condition is determined；(2) downstream hydrologic condition is determined；(3) high-precision river underwater topography data are obtained by surveying and drawing；(4) ecological environmental protection object when selecting target location (i.e. sensitive target) the most typical as emergent assessment and decision-making, while as the object output of deep learning；(5) typical flow fields database is established；(6) determine that Riverine area accident easily sends out frequent point position；(7) typical case's burst water pollution accident case library is established；(8) the emergent assessment of burst accident and decision system are established.

Description

Water pollution burst accident based on deep learning quickly meet an urgent need assess and decision-making technique
Technical field
The present invention relates to field of environment protection, specifically a kind of quick meet an urgent need for being used to tackle water pollution burst accident is commented Estimate and decision-making technique.
Background technology
For a long time, worsening water pollution burst accident is domestic and international most cities water head site safety all the time Important threat.These water pollution burst accidents both include the conventional pollution accidents such as trade effluent, sanitary wastewater, face source, also include The sudden water pollution event such as ship, the chemicals of harbour and Oil spills, industrial accident discharge, the attack of terrorism.Statistics It has been shown that, pop-up threat frequency is existing caused by oil spilling, toxic chemical spills etc. all over the world in recent years, and environmental pollution is tight Weight, Ecological Loss are huge.
Generally eco-environmental impact is carried out as water pollution burst accident using hydrodynamic force mathematical modeling both at home and abroad at present A kind of effective means with Emergency decision is assessed, the decision-making of science is carried out based on forecast assessment result, takes corresponding meet an urgent need to arrange Apply.Although this method can provide the assessment result of scientific quantitative analysis, because condition limits in reality, while by water The influence of numerous uncertain factors such as literary flow field, pollutant kind and leakage rate, place where the accident occurred point, substantially can not possibly be in pole Reliable Water Environment Mathematical Model is established in short time rapidly, is particularly running into wider deeper big-and-middle-sized river or pollution Thing needs to use threedimensional model to carry out simulation and forecast, and threedimensional model when belonging to the complex situations such as the material of insoluble or semi-soluble property More it is difficult to the foundation and computing that emergent model is completed in the short time.Once burst accident occurs, if can not be in a short time There is provided the forecast assessment result of reliable quantification for decision-making to judge, just can not quickly formulate the emergency measure of science to subtract Influence of the small burst accident to ecological environment and production and living.
The content of the invention
The purpose of the present invention is to overcome the shortcomings of above-mentioned background technology, there is provided a kind of water pollution based on deep learning is dashed forward Hair accident quickly meet an urgent need assess and decision-making technique, can within the very short time according to the scene of the accident obtain limited information, It is quick to form emergent assessment result and for the skill of Emergency decision on the basis of the neural network model established by deep learning Art supports, and helps the emergency measure of contingency management department rapid development science.
Technical scheme provided by the invention is：A kind of water pollution burst accident based on deep learning is quickly met an urgent need and assessed And decision-making technique, comprise the following steps：
1) upstream hydrologic condition is determined.It is boundary by affiliated administrative division for specific river, analyzes the stream of upland water Measure feature, choose typical hydrology condition data, including the historical data such as upstream flowrate, water level or measured data.If any artificial structure Thing is built, then needs to analyze the regularity of distribution of its letdown flow, therefrom summary and induction and sets different flow by particular flow rate interval As upstream hydrology boundary condition；The flow of setting must cover history maximum and minimum, and setting upstream hydrologic condition is big In equal to 10 groups.
2) downstream hydrologic condition is determined.If specific river is tidal river, downstream hydrology border is tide reversing current (downstream flow field is according to tide dynamic change), for downstream hydrologic condition due to tidal effect, the situation of change in flow field is more complicated, Including 12 kinds of tide classifications：Big tidal bulge is had a rest, big tidal bulge is anxious, big ebb is had a rest, big ebb is anxious, middle tidal bulge is had a rest, middle tidal bulge is anxious, middle tide Fall to have a rest, middle ebb is anxious, small tidal bulge is had a rest, small tidal bulge is anxious, small ebb is had a rest, small ebb is anxious.Tide classification in order distinguish assignment 1~ 12, as downstream hydrologic condition.If downstream boundary flow field is not tide reversing current, but steady flow, then using the level of tail water or Flow is as downstream hydrologic condition.
3) high-precision river underwater topography data are obtained by surveying and drawing.Commented to tackle all kinds of the emergent of complicated pollutant Estimate, detailed river underwater topography data need to be obtained, in order to establish the three-dimensional flow field in river.Longitudinal length is less than during mapping 500m, laterally it is less than 100m, needs to encrypt horizontal measuring point if running into river bend or there are island.
4) Riverine area potable water source district, the longitude and latitude position of water environment ecology sensitive target or scope, choosing are combed comprehensively Ecological environmental protection object when fixed number group target location (i.e. sensitive target) the most typical is as emergent assessment and decision-making, Object output as deep learning simultaneously.
5) according to the characteristic contamination such as the storage of river periphery, the main hazard chemicals of production and transportation, oil product, choose Common attribute pollutant simultaneously inquires about physicochemical property data, including density, solubility, degradation rate, viscosity, surface tension, boiling point And the parameter such as evaporation constant.Characteristic contamination is more than or equal to 5, and characteristic contamination character type should include suspended state, glue State, the class of solubilised state 3.
6) typical flow fields database is established
According to the upstream and downstream boundary condition determined in 1), with reference to the terrain data in step 3), numerical simulation meter is carried out Calculate, the combination for different hydrological conditions establishes typical three-dimensional flow field database (three by the calculating of three-dimensional flow field model respectively Dimension flow field model equation is that the formula of hydrodynamics process described in model represents.Three-dimensional flow field database is model in difference The result of calculation inquiry storehouse that the three dimensional flow simulation result data being calculated under hydrologic condition is formed)；Three-dimensional flow field Model carries out secondary development, the three-dimensional hydrodynamic model of foundation, vertical to be divided into 5 layers, grid using ECOMSED source codes pattern Ultimate resolution 50-60m.The substantially continuous equation of three-dimensional flow field model is as follows：
In formula：G is coordinate transform tensor；(U, V) is the vertical mean flow rate in (ξ, η) direction；D is the depth of water, and ζ is water level, D+ ζ=H are total depth of water；Q is source, remittance item, is expressed as：
In formula：qinAnd qoutThe respectively source of unit volume and remittance, Pr are precipitation item, and Ev is evaporation item；
(2) boundary condition；On bottom, the boundary condition of the equation of momentum is：
In formula：τbξAnd τbηRespectively bottom shearing stress τbComponent on ξ and η directions；
(3) surface boundary condition is：
In formula：For surface wind-stress, θ is wind-stress and the angle in η directions；
(4) open boundaryconditions；It is used as the driving of pattern by SEA LEVEL VARIATION, water level ζ expression formula is：
In formula：A is remaining water level, is calculated by great Qu；fiFor the intersection point factor of each partial tide, (V0+u)iFor the day of partial tide Literary phase angle, it can be tried to achieve by geographical position and specific year, month, day；ωiFor the angular frequency of partial tide, giAnd HiFor the place of tidal wave Delay angle and amplitude；mtdFor partial tide number.
7) determine that Riverine area accident easily sends out frequent point position
The topography and geomorphology of Riverine area, traffic route, bridge, Industrial Concentrated Area etc. are analyzed, it is determined that accident along the line The position for easily sending out occurred frequently, with accident point position and starting point (typically using river administration intersection, important artificial works or source as Starting point) distance represent；The accident point position of determination is more than or equal to 10.
8) simulation of typical case's burst water pollution accident scene calculates；Position occurs according to 7) the middle accident determined, chooses 5) In characteristic contamination and set leakage rate (general span is in 0.2~50t) respectively, with reference to the flow field database in 6), The scene case of formation is more than or equal to 1500, carries out numerical simulation calculation with environmental careers model one by one；Pollutant is obtained to enter Enter the transfer after water body and diffusion model, including volatilize, hydrolyze, the following mathematical modeling of extended surface, the process such as evaporation (is changed The environmental model such as product or oil spilling)；
(1) prediction of rate of volatilization：
In formula：C is the concentration of chemical substance in dissolving phase；KvFor rate of volatilization constant；Kv' it is unit time mixed water body Rate of volatilization constant；Z is the interacting depth of water body；P is on the water body studied, point of chemical substance in an atmosphere Pressure；KHFor henry 's law constant；
(2) process of hydrolysis：RH=Kh[c]={ KA[H+]+KN+KB[OH-]}[c]
In formula：KA、KB、KNThe respectively second order reaction hydrolytic rate constant of acidic catalyst, base catalysis and neutral process； KhFor the pseudo first order reaction hydrolytic rate constant under a certain pH value；
(3) extended surface rate of change calculates：
In formula：AtkFor oil film surface area (m2)；K1 is spreading rate coefficient (1/s)；VmFor sea surface oil slick volume (m3)；t For the time (s)；For single oil particles surface area rate of change；
(4) evaporation rate calculates：Fv=ln [l+B (TG/T)θexp(A-BTo/T)][T/BTG]
In formula：T0To correct the Initial Boiling Point (K) of distillation curve；TGTo correct the slope of distillation curve；T is environment temperature (K)；A, B are dimensionless constant；T is the time (s)；θ is evaporation direction.
9) above-mentioned scene case result of calculation deposit database is formed into typical case's burst water pollution accident case library, calculated As a result pollutant due in, superthreshold duration and the peak-peak concentration of each sensitive target are included.Randomly select wherein The training dataset of 1350 groups (90%) as deep learning, remaining 150 groups of case (10%) are used as test data set, are used for The test checking of deep learning neutral net.
10) limited scene is confined to solve the accident case built in advance, and the structure of full basin three-dimensional flow field is fast Spend slowly can not the physical presence such as Fast simulation burst accident the problem of, the present invention is in foregoing hydrodynamics and environmental careers mould On the basis of type, using the method for deep learning, the emergent assessment of burst water pollution accident is established based on deep neural network and determined Plan model；Specific steps include：
(1) data prediction；9) the middle 1350 groups of training datasets prepared are normalized, to the every of data The value of one dimension is readjusted, and transfer function is as follows：
X=(x-min)/(max-min)
(2) training environment；The present invention uses Tensorflow deep learning frameworks, and depth is realized with Python Neural network model；
(3) deep neural network model uses Wide And Deep models, and hidden layer is no less than 3 layers, and activation primitive makes Linear unit is corrected with ReLU, the majorized function of each layer weight gradient updating algorithm uses Adagrad algorithms；
(4) 1350 groups of training datas after being readjusted in (1) are combined, deep neural network model is trained, are led to Cross training and determine the weighted value of each node, deviation value parameter in model, establish deep neural network model；
(5) degree of accuracy calculating is carried out to model using 150 groups of other test datas, set when test the result meets The fixed degree of accuracy, model learning process is completed, the model of final output is as the emergent assessment of pop-up threat and decision model Type.
11) the emergent assessment of burst accident and decision system are established；System is by input block, arithmetic element and output unit Composition；
Burst accident occurs input block position, pollutant leakage rate, the density of pollutant, solubility, degraded speed Rate, viscosity, surface tension and evaporation constant, and (data can be from the scene of the accident or net for the data such as upstream and downstream flow or tidal level Network inquiry obtain) be normalized after, be input to arithmetic element；
Arithmetic element carries out computing with decision model by emergent assess of the pop-up threat 10) established to input data, Result is quickly calculated in 15min；
Output unit carries out renormalization conversion to operation result, when exporting the peak concentration of sensitive downstream target, reaching Quarter and duration, provided strong support for emergency disposal.
Emergent assess of the burst accident is realized with decision system by software.
12) apply
Emergent assess of pop-up threat has two-way decision support function with decision system, can be carried according to emergent expert The effect of the scheme rapid evaluation emergency measure gone out；Several emergency measures can also be preset to be assessed, then will be assessed Conclusion is supplied to emergent expert or the decision-making of emergency command portion.Can also be reached according to pollution in emergent assess the moment and it is lasting when Between, personnel to live emergency monitoring, layouting and monitoring the time carries out science distribution and guidance.
The beneficial effects of the invention are as follows：
Practicality of the present invention is very strong, solves and is continually faced with pop-up threat, can not fast and effectively establish number The problem that model carries out quantitative analysis is learned, breaches the limitation that can only rely on micro-judgment in the past.The present invention both overcomes specialty The problem of difficult is modeled under environmental hydrology model emergency condition, solving deep learning multilayer neural network again needs a large amount of numbers According to the problem being trained.In actual applications, emergent determine can also be established even if the river that can not obtain real-time traffic data Plan system.The best advantage is that the reaction time is extremely short, due to the various conditions of contamination accident taken into full account, lead to Cross condition setting and cover various burst scenes, therefore system adapts to the actual emergency requirement of burst accident, can be according to existing The needs in emergency command portion, provide rapidly assessment result, have quantitative conclusion to the influence degree of contamination accident, can be with The implementation result of emergency measure is prejudged, at utmost reduces the influence degree of contamination accident.The present invention can also instruct Time distribution is layouted and monitored to field monitoring, effectively improves the monitoring efficiency under emergency rating, reduces manpower and materials input.
Brief description of the drawings
Fig. 1 is the job step flow chart of the present invention.
Fig. 2 is deep neural network structural representation.
Embodiment
It is further to the present invention with specific implementation method below in conjunction with the accompanying drawings in order to be more clearly understood that the present invention Detailed description, but the invention is not limited in following examples.
Water pollution burst accident provided by the invention based on deep learning quickly meet an urgent need assess and decision-making technique, be one The emergent forecasting system that kind professional hydrology, water quality model are combined with deep learning neutral net, can within the very short time root The limited information obtained according to the scene of the accident, on the basis of the neural network model established by deep learning, quick formed is met an urgent need Assessment result and the technical support for being used for Emergency decision, can help contingency management department quickly to formulate the emergent of science and arrange Apply.
The present invention is implemented in the Qiantang River (Hangzhou Section), and embodiment illustrates by taking the Qiantang River as an example.To protect the Qiantang River Drinking water source along the line, the Qiantang River (Hangzhou Section) is set to research object, the Qiantang River is analyzed using historical data and actual measurement mode The traffic characteristic of (Hangzhou Section) upland water, choose typical hydrologic condition, including the historical data such as upstream flowrate, water level or reality Survey data.Because the Qiantang River (Hangzhou Section) has Fuchunjiang River dam, the regularity of distribution of dam letdown flow need to be analyzed, and press specific stream Amount interval sets different flows as flow boundary condition.The flow covering history maximum and minimum of setting, setting stream It is 10 groups to measure number, and the Qiantang River (Hangzhou Section) flow takes equally distributed 10 groups of stream from 40 cubes of meter per seconds to 5000 cubes of meter per seconds Measure data.The Qiantang River (Hangzhou Section) downstream is tidal river, and downstream hydrologic condition is used as input using tide classification, totally 12 class.
The Qiantang River (Hangzhou Section) underwater topography data are obtained by surveying and drawing, the longitudinally spaced of data is less than 500 meters, laterally Interval is less than 50m.By reconnaissance trip and data check, it (is typically to drink to obtain sensitive target along the Qiantang River (Hangzhou Section) Water head site, and fish backflow ground etc. water environment Ecological Target) position, select one group of target location conduct the most typical Ecological environmental protection object when emergent assessment and decision-making, while be also the object output of deep learning.The Qiantang River (Hangzhou Section) The centralized potable water source district intake in 5, downstream is chosen as object output.
Polluted according to the feature such as the storage of the Qiantang River (Hangzhou Section) periphery, the main hazard chemicals of production and transportation, oil product Thing, choose common attribute pollutant and inquire about physicochemical property data, including density, solubility, degradation rate, viscosity, surface The parameters such as power, boiling point and evaporation constant.Characteristic contamination is more than or equal to 5, and characteristic contamination character type should at least include outstanding Floading condition, gluey state, the class of solubilised state 3.Embodiment choose pollutant include tetrachloroethanes, toluene, 1,2,4- trichloro-benzenes, phenol, Diesel oil totally 5.
Establish typical flow fields storehouse.According to upstream and downstream boundary condition, with reference to underwater topography data, numerical simulation calculation is carried out, Typical three-dimensional flow field database is established in combination for different hydrological conditions.
Determine that accident easily sends out frequent point position along the Qiantang River (Hangzhou Section).Topography and geomorphology, traffic road to Riverine area Road, bridge, Industrial Concentrated Area etc. are analyzed, it is determined that the position that accident is easily sent out occurred frequently along the line, with starting point (typically with river Administrative intersection, important artificial works or source are starting point) distance represent.The accident mould that the Qiantang River (Hangzhou Section) determines Intend point position totally 10, predominantly riverine road accident easily sends out section or crossing-fiver bridge, and input number is used as using the distance with starting point According to.
The simulation of typical case's burst water pollution accident scene calculates.Position occurs for the accident drafted with reference to the Qiantang River (Hangzhou Section) At the generation moment put and set, selected characteristic pollutant simultaneously sets leakage rate (span is in 0.2~50t) respectively, with three-dimensional Based on the database of flow field, the scene case of formation 1500, numerical simulation calculation is carried out with environmental careers model one by one.Will Above-mentioned scene case result of calculation deposit database forms typical case's burst water pollution accident case library (being shown in Table 1), result of calculation bag Including each sensitive target, (the present embodiment is the centralized potable water source district intake in 5, the Qiantang River (Hangzhou Section) downstream, specific visible Table 1) pollutant due in, superthreshold duration and peak-peak concentration.Randomly select wherein 1350 groups and be used as depth The training dataset of study, remaining 150 cases are as test data set, for follow-up deep learning deep neural network Test exercise.
The typical case's burst water pollution accident case library of table 1
(in view of the case library data are excessively huge, to save space, table 1 only provides 4 cases and illustrated)
1350 groups of training datasets are normalized, the value of each dimension of data is readjusted, Transfer function is as follows：
X=(x-min)/(max-min)
Using Tensorflow deep learning frameworks, deep neural network model is realized with Python；Depth god Wide And Deep models are used through network model；The hidden layer of embodiment sets 5 layers, hidden layer number of unit totally 50；Swash Function living uses Adagrad algorithms using ReLU amendment linear units, the majorized function of each layer weight gradient updating algorithm；Instruction Fit functions can be used by practicing model, and evaluation and test uses evaluate functions.Above-mentioned function can pass through Tensorflow depth Learning framework directly invokes, and is realized by Python Programming with Pascal Language.
Using 1350 groups of training datas, deep neural network model is trained, respectively tied by training in certain model The weighted value of point, the parameter such as deviation, establish deep neural network model；
Degree of accuracy calculating is carried out to model using test data, the degree of accuracy that setting is met when test the result (is implemented 99.5%) example takes, complete model learning process, and the deep learning model of final output is dirty as the Qiantang River (Hangzhou Section) burst Contaminate accident emergency assessment and decision model；
The emergent foundation assessed with decision system of the Qiantang River (Hangzhou Section) burst accident.System is by input block, computing list Member and output unit composition, completion (being programmed using Python) is write by common software technical staff.Input block is to prominent Position, pollutant leakage rate, the density of pollutant, solubility, degradation rate, viscosity, surface tension and the steaming that hair accident occurs Constant is sent out, and (data can be obtained the data such as upstream and downstream flow or tidal level from the scene of the accident or network inquiry, and embodiment is built Stood the physicochemical property data storehouse of 500 kinds of common contaminants) be normalized after, be input to arithmetic element.Arithmetic element That is the emergent assessment of pop-up threat and decision model, carry out computing to input data, result are quickly calculated in 15min. Output unit carries out renormalization conversion to operation result, exports contaminant peak concentrations, the pollutant of sensitive downstream target Due in and superthreshold duration, provided strong support for emergency disposal；
The use of Emergency decision.The emergent assessment of the Qiantang River (Hangzhou Section) pop-up threat and decision system have two-way Decision support function.The effect for the scheme rapid evaluation emergency measure that can be proposed according to emergent expert；It can also preset several Kind emergency measure is assessed, and assessment result then is supplied into emergent expert or the decision-making of emergency command portion.Further, it is also possible to Moment and duration are reached according to pollution in emergent assessment, personnel to live emergency monitoring, layouting and monitor the time is carried out Science is distributed and instructed.
Assuming that certain year, in such a month, and on such a day some time divided on the Qiantang River (Hangzhou Section) expressway the acetaldehyde groove tank car rollover thing that happens suddenly Therefore the accident information obtained rapidly is：It is 3.1km that position, which occurs, for accident, and acetaldehyde leakage rate is 12t；Inquiry data obtains acetaldehyde Information be：Density 0.783g/cm3, solubility 406000ppm, degradation rate 0.1131, viscosity 0.3cP, surface tension 23.90dyn/cm and evaporation constant 0.0421；And upstream hydrology flow 700m3/ s and downstream hydrologic condition (tide classification 9)；Data fully enter the emergent assessment system of pop-up threat that the present invention establishes, can be fast by the computing within 15min Speed obtains contaminant peak concentrations, pollutant due in and the superthreshold of 5, downstream sensitive target (important water factory's intake) Duration；By taking the sensitive target output result nearest apart from accident as an example：Acetaldehyde reaches downstream Tonglu water factory intake Moment is 1.5h after accident occurs, and the peak concentration of intake acetaldehyde reaches 795ppb, and the duration of acetaldehyde superthreshold is 3.5h.Assessment result may be directly applied to accident emergency disposal and decision-making.
The general principle and principal character of the present invention has been shown and described above.It should be understood by those skilled in the art that The present invention is not limited to the above embodiments, merely illustrating the principles of the invention described in above-described embodiment and specification, Without departing from the spirit and scope, various changes and modifications of the present invention are possible, and these changes and improvements all fall Enter in scope of the claimed invention, claimed scope of the invention is by appended claims and its equivalent circle It is fixed.

Claims (8)

  1. Assessment and decision-making technique 1. a kind of water pollution burst accident based on deep learning is quickly met an urgent need, comprise the following steps：
    1) upstream hydrologic condition is determined；It is boundary by affiliated administrative division for specific river, the flow for analyzing upland water is special Sign, chooses typical hydrology condition data, therefrom summary and induction and sets different flow as upper water by particular flow rate interval Literary boundary condition；
    2) downstream hydrologic condition is determined；
    If specific river is tidal river, downstream hydrology border is tide reversing current, and 12 kinds of tide classifications are distinguished in order Assignment 1~12, as downstream hydrologic condition；
    If downstream boundary flow field is steady flow, downstream hydrologic condition is used as using the level of tail water or flow；
    3) high-precision river underwater topography data are obtained by surveying and drawing, establishes the three-dimensional flow field in river；
    4) Riverine area potable water source district, the longitude and latitude position of water environment ecology sensitive target or scope are combed comprehensively, select number Assessed and ecological environmental protection object during decision-making as emergent group target location the most typical；
    5) according to the characteristic contamination such as the storage of river periphery, the main hazard chemicals of production and transportation, oil product, common spy is chosen Sign pollutant simultaneously inquires about physicochemical property data, including density, solubility, degradation rate, viscosity, surface tension, boiling point and evaporation Constant parameter；
    6) typical flow fields database is established
    According to the upstream and downstream boundary condition determined in 1), with reference to the terrain data in (2), numerical simulation calculation is carried out, for not Typical three-dimensional flow field database is established in combination with hydrologic condition；
    7) determine that Riverine area accident easily sends out frequent point position, represented with the distance of accident point position and starting point；
    8) simulation of typical case's burst water pollution accident scene calculates；According to the accident determined in 7) occur position and 7) in determine Moment occurs for accident, and the characteristic contamination in choosing 4) simultaneously sets leakage rate respectively, with reference to the flow field database in 5), formation Scene case is more than or equal to 1500, carries out numerical simulation calculation with environmental careers model one by one；Obtain pollutant and enter water body Transfer afterwards and the mathematical modeling of diffusion；
    9) above-mentioned scene case result of calculation deposit database is formed into typical case's burst water pollution accident case library, result of calculation bag Include pollutant due in, superthreshold duration and the peak-peak concentration of each sensitive target；Randomly select in scene case 90% training dataset as deep learning, remaining 10% case is as test data set, for follow-up deep learning Deep neural network；
    10) method for using deep learning, the emergent assessment of burst water pollution accident and decision model are established based on deep neural network Type；
    11) the emergent assessment of burst accident and decision system are established；System is made up of input block, arithmetic element and output unit；
    It is burst accident occurs input block position, pollutant leakage rate, the density of pollutant, solubility, degradation rate, viscous Degree, surface tension and evaporation constant, and after the data such as upstream and downstream flow or tidal level are normalized, it is input to computing list Member；
    Arithmetic element carries out computing with decision model by emergent assess of the pop-up threat 10) established to input data, Result is quickly calculated in 15min；
    Output unit carries out renormalization conversion to operation result, export the peak concentration of sensitive downstream target, due in and Duration；
    12) apply
    The emergent assessment of pop-up threat and decision system, the scheme rapid evaluation emergency measure that can be proposed according to emergent expert Effect；Or preset several emergency measures and assessed, assessment result is then supplied to emergent expert or emergency command Portion's decision-making；Also moment and duration can be reached according to pollution in emergent assessment, personnel to live emergency monitoring, layout and supervise Survey time progress science distribution and guidance.
  2. Assessment and decision-making technique 2. the water pollution burst accident according to claim 1 based on deep learning is quickly met an urgent need, It is characterized in that：Typical hydrology condition data in the step 1), including upstream flowrate, downstream flow, water level or tidal level etc. Historical data or measured data；If any artificial works, then need to analyze the regularity of distribution of its letdown flow, therefrom summary and induction is simultaneously Different flows is set as flow boundary condition by particular flow rate interval；The flow of setting must cover history maximum and pole Small value, setting flow number are more than or equal to 10 groups.
  3. Assessment and decision-making technique 3. the water pollution burst accident according to claim 2 based on deep learning is quickly met an urgent need, It is characterized in that：12 kinds of tide classifications in the step 2) are：Big tidal bulge is had a rest, big tidal bulge is anxious, big ebb is had a rest, big ebb is anxious, Middle tidal bulge is had a rest, middle tidal bulge is anxious, middle ebb is had a rest, middle ebb is anxious, small tidal bulge is had a rest, small tidal bulge is anxious, small ebb is had a rest and small ebb is anxious.
  4. Assessment and decision-making technique 4. the water pollution burst accident according to claim 3 based on deep learning is quickly met an urgent need, It is characterized in that：In the step 3), to tackle the emergent assessment of all kinds of complicated pollutants, detailed river need to be obtained under water Graphic data, in order to establish the three-dimensional flow field in river；Longitudinal length is less than 500m during mapping, is laterally less than 100m, such as runs into river Stream bending has island then to need to encrypt horizontal measuring point.
  5. Assessment and decision-making technique 5. the water pollution burst accident according to claim 4 based on deep learning is quickly met an urgent need, It is characterized in that：Characteristic contamination is more than or equal to 5 in the step 5), and characteristic contamination character type includes suspended state, glue State, the class of solubilised state 3.
  6. Assessment and decision-making technique 6. the water pollution burst accident according to claim 5 based on deep learning is quickly met an urgent need, It is characterized in that：Three-dimensional flow field database in the step 6), secondary development is carried out using ECOMSED source codes pattern, established Three-dimensional hydrodynamic model, vertical to be divided into 5 layers, grid ultimate resolution 50-60m；Three-dimensional flow field model equation is as follows：
    (1)
    In formula：G is coordinate transform tensor；(U, V) is the vertical mean flow rate in (ξ, η) direction；D is the depth of water, and ζ is water level, d+ ζ= H is total depth of water；Q is source, remittance item, is expressed as：
    In formula：qinAnd qoutThe respectively Yuan Hehui of unit volume, Pr are precipitation item, and Ev is evaporation item；
    (2) boundary condition.On bottom, the boundary condition of the equation of momentum is：
    In formula：τbξAnd τbηRespectively bottom shearing stress τbComponent on ξ and η directions；
    (3) surface boundary condition is：
    In formula：For surface wind-stress, θ is wind-stress and the angle in η directions；
    (4) open boundaryconditions.It is used as the driving of pattern by SEA LEVEL VARIATION, water level ζ expression formula is：
    In formula：A is remaining water level, is calculated by great Qu；fiFor the intersection point factor of each partial tide, (V0+u)iFor the astronomical phase of partial tide Angle, it can be tried to achieve by geographical position and specific year, month, day；ωiFor the angular frequency of partial tide, giAnd HiFor tidal wave local delay angle and Amplitude；mtdFor partial tide number.
  7. Assessment and decision-making technique 7. the water pollution burst accident according to claim 6 based on deep learning is quickly met an urgent need, It is characterized in that：Pollutant is into the transfer after water body and diffusion model in the step 8), including volatilizees, hydrolyzes, surface expansion The following mathematical modeling of the processes such as exhibition, evaporation；
    (1) prediction of rate of volatilization：
    In formula：C is the concentration of chemical substance in dissolving phase；KvFor rate of volatilization constant；Kv' waving for unit time mixed water body Send out speed constant；Z is the interacting depth of water body；P is on the water body studied, the partial pressure of chemical substance in an atmosphere；KH For henry 's law constant；
    (2) process of hydrolysis：RH=Kh[c]={ KA[H+]+KN+KB[OH-]}[c]
    In formula：KA、KB、KNThe respectively second order reaction hydrolytic rate constant of acidic catalyst, base catalysis and neutral process；KhFor Pseudo first order reaction hydrolytic rate constant under a certain pH value；
    (3) extended surface rate of change calculates：
    In formula：AtkFor oil film surface area (m2)；K1 is spreading rate coefficient (1/s)；VmFor sea surface oil slick volume (m3)；T is the time (s)；For single oil particles surface area rate of change；
    (4) evaporation rate calculates：Fv=ln [l+B (TG/T)θexp(A-BT0/T)][T/BTG]
    In formula：T0To correct the Initial Boiling Point (K) of distillation curve；TGTo correct the slope of distillation curve；T is environment temperature (K)； A, B are dimensionless constant；T is the time (s)；θ is evaporation direction.
  8. Assessment and decision-making technique 8. the water pollution burst accident according to claim 7 based on deep learning is quickly met an urgent need, It is characterized in that：The specific steps of the emergent assessment of burst water pollution accident and decision model are established in the step 10) to be included：
    (1) data prediction；9) the middle training dataset prepared is normalized, to the value of each dimension of data Readjusted, transfer function is as follows：
    X=(x-min)/(max-min)
    (2) training environment；Using Tensorflow deep learning frameworks, deep neural network mould is realized with Python Type；
    (3) deep neural network model uses Wide And Deep models, and hidden layer is no less than 3 layers, and activation primitive uses ReLU Linear unit is corrected, the majorized function of each layer weight gradient updating algorithm uses Adagrad algorithms；
    (4) training data after being readjusted in (1) is combined, deep neural network model is trained, is determined each in model The weighted value of node, deviation value parameter, establish deep neural network model；
    (5) degree of accuracy calculating is carried out to model using other test data, when test the result meets the degree of accuracy of setting, Model learning process is completed, the model of final output is as the emergent assessment of pop-up threat and decision model.
CN201710939980.5A 2017-10-11 2017-10-11 Water pollution sudden accident rapid emergency assessment and decision-making method based on deep learning Active CN107798467B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710939980.5A CN107798467B (en) 2017-10-11 2017-10-11 Water pollution sudden accident rapid emergency assessment and decision-making method based on deep learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710939980.5A CN107798467B (en) 2017-10-11 2017-10-11 Water pollution sudden accident rapid emergency assessment and decision-making method based on deep learning

Publications (2)

Publication Number Publication Date
CN107798467A true CN107798467A (en) 2018-03-13
CN107798467B CN107798467B (en) 2020-08-21

Family

ID=61532946

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710939980.5A Active CN107798467B (en) 2017-10-11 2017-10-11 Water pollution sudden accident rapid emergency assessment and decision-making method based on deep learning

Country Status (1)

Country Link
CN (1) CN107798467B (en)

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108537587A (en) * 2018-04-03 2018-09-14 广州优视网络科技有限公司 It is lost in user's method for early warning, device, computer readable storage medium and server
CN108564376A (en) * 2018-04-20 2018-09-21 阿里巴巴集团控股有限公司 Risk control method, device, server and readable storage medium storing program for executing
CN108647251A (en) * 2018-04-20 2018-10-12 昆明理工大学 The recommendation sort method of conjunctive model is recycled based on wide depth door
CN108960525A (en) * 2018-07-20 2018-12-07 郑州轻工业学院 Pollution monitoring method and device based on mobile computer and neural network
CN110866367A (en) * 2019-08-13 2020-03-06 河海大学 Real-time tracking method for conventional sudden water pollution clusters under complex river network water flow condition
CN110991697A (en) * 2019-11-05 2020-04-10 天津大学 Marine oil spill rapid prediction system based on tidal periodicity and self-similarity
CN111639456A (en) * 2020-06-17 2020-09-08 太原理工大学 Full space-time water area pollutant emergency situation sensing system based on unmanned aerial vehicle
CN112288200A (en) * 2019-07-25 2021-01-29 中国石油天然气股份有限公司 Method and device for determining inland river oil spill emergency disposal scheme
CN113128105A (en) * 2021-03-31 2021-07-16 北京工业大学 Method and device for monitoring sudden watershed water pollution accident
CN113824800A (en) * 2021-11-23 2021-12-21 武汉超云科技有限公司 Big data analysis method and device based on hybrid energy data

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102221389A (en) * 2011-04-11 2011-10-19 国家海洋信息中心 Method for predicting tide-bound water level by combining statistical model and power model
CN102663222A (en) * 2011-12-27 2012-09-12 中国科学院生态环境研究中心 Calculating method for dynamic pollution field of open water body emergent pollution accident
CN106228007A (en) * 2016-07-19 2016-12-14 武汉大学 Accident polluter retroactive method
CN106339586A (en) * 2016-08-23 2017-01-18 环境保护部华南环境科学研究所 Tracing method and prediction method of mustard gas pollution
CN106779418A (en) * 2016-12-20 2017-05-31 河海大学常州校区 Water contamination accident Intelligent Decision-making Method based on neutral net and evidence theory

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102221389A (en) * 2011-04-11 2011-10-19 国家海洋信息中心 Method for predicting tide-bound water level by combining statistical model and power model
CN102663222A (en) * 2011-12-27 2012-09-12 中国科学院生态环境研究中心 Calculating method for dynamic pollution field of open water body emergent pollution accident
CN106228007A (en) * 2016-07-19 2016-12-14 武汉大学 Accident polluter retroactive method
CN106339586A (en) * 2016-08-23 2017-01-18 环境保护部华南环境科学研究所 Tracing method and prediction method of mustard gas pollution
CN106779418A (en) * 2016-12-20 2017-05-31 河海大学常州校区 Water contamination accident Intelligent Decision-making Method based on neutral net and evidence theory

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
ZHANG JIBIAO 等: "pH Numerical Simulation of the Desulfuration Acidic Effluent Discharge of Power Plant and Its Impact Prediction on Phytoplankton in Seawater", 《METEOROLOGICAL AND ENVIRONMENTAL RESEARCH》 *

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108537587A (en) * 2018-04-03 2018-09-14 广州优视网络科技有限公司 It is lost in user's method for early warning, device, computer readable storage medium and server
CN108564376A (en) * 2018-04-20 2018-09-21 阿里巴巴集团控股有限公司 Risk control method, device, server and readable storage medium storing program for executing
CN108647251A (en) * 2018-04-20 2018-10-12 昆明理工大学 The recommendation sort method of conjunctive model is recycled based on wide depth door
CN108647251B (en) * 2018-04-20 2021-06-18 昆明理工大学 Recommendation sorting method based on wide-depth gate cycle combination model
CN108960525B (en) * 2018-07-20 2021-11-16 郑州轻工业学院 Pollution monitoring method and device based on mobile computer and neural network
CN108960525A (en) * 2018-07-20 2018-12-07 郑州轻工业学院 Pollution monitoring method and device based on mobile computer and neural network
CN112288200A (en) * 2019-07-25 2021-01-29 中国石油天然气股份有限公司 Method and device for determining inland river oil spill emergency disposal scheme
CN110866367A (en) * 2019-08-13 2020-03-06 河海大学 Real-time tracking method for conventional sudden water pollution clusters under complex river network water flow condition
CN110991697A (en) * 2019-11-05 2020-04-10 天津大学 Marine oil spill rapid prediction system based on tidal periodicity and self-similarity
CN110991697B (en) * 2019-11-05 2023-10-17 天津大学 Rapid prediction system for offshore oil spill based on tide periodicity and self-similarity
CN111639456A (en) * 2020-06-17 2020-09-08 太原理工大学 Full space-time water area pollutant emergency situation sensing system based on unmanned aerial vehicle
CN113128105A (en) * 2021-03-31 2021-07-16 北京工业大学 Method and device for monitoring sudden watershed water pollution accident
CN113824800A (en) * 2021-11-23 2021-12-21 武汉超云科技有限公司 Big data analysis method and device based on hybrid energy data
CN113824800B (en) * 2021-11-23 2022-02-11 武汉超云科技有限公司 Big data analysis method and device based on hybrid energy data

Also Published As

Publication number Publication date
CN107798467B (en) 2020-08-21

Similar Documents

Publication Publication Date Title
CN107798467A (en) Water pollution burst accident based on deep learning quickly meet an urgent need assess and decision-making technique
CN107563139A (en) A kind of point source sudden water pollution source accident source percentage contribution computational methods
Dargahi et al. Combined 3D hydrodynamic and watershed modelling of Lake Tana, Ethiopia
Rai et al. Planning and evaluation of irrigation projects: methods and implementation
Martin et al. Development of a GIS-based spill management information system
CN107609742A (en) A kind of water environment early warning comprehensive management platform
CN107944608B (en) Sea surface drift and oil spill drift diffusion forecasting method based on satellite remote sensing
CN104898183A (en) Modeling evaluation method for urban heavy rain inundation
Wu et al. Integrated hydrologic and hydrodynamic modeling to assess water exchange in a data-scarce reservoir
Xu et al. A numerical study on land-based pollutant transport in Singapore coastal waters with a coupled hydrologic-hydrodynamic model
Li et al. Water accessibility assessment of freshwater wetlands in the yellow river delta national nature reserve, China
Amiri et al. Selecting wells for an optimal design of groundwater monitoring network based on monitoring priority map: A Kish Island case study
Wang et al. Assessing the total maximum allocated load of jurisdiction petroleum pollutants in the Bohai Sea
KR20060092420A (en) System and method for coastal water quality management
CN105844073A (en) Source/sink landscape recognition method and apparatus of non-point source pollution
Conrads et al. Simulation of salinity intrusion along the Georgia and South Carolina coasts using climate-change scenarios
Akhralov et al. Geoinformation technologies and methods of mathematical modeling in hydrogeological research
Wang Water Resources and Hydraulics
Nasrulin et al. Environmental monitoring in construction for hydro-technical structures and land reclamation facilities of Uzbekistan
Mogheir et al. Comparative identification of wellhead protection areas for municipal supply wells in Gaza
Duarte et al. Pollutant dispersion modelling for Portuguese river water uses protection linked to tracer dye experimental data
McGreevy et al. Development of a digital model of ground-water flow in deeply weathered crystalline rock, Chester County, Pennsylvania
Di et al. Analysis of the use of the method of hydroecological monitoring in order to improve the ecological condition of the hydrotechnical constructions of Uzbekistan
Singh et al. Site selection for ground water recharge using treated municipal wastewaters-a case study of Varanasi (India)
Akhralov et al. Mathematical modeling in hydrogeological research

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