CN116151487B - Physical knowledge and data hybrid-driven prediction algorithm for predicting sea surface oil spill track - Google Patents
Physical knowledge and data hybrid-driven prediction algorithm for predicting sea surface oil spill track Download PDFInfo
- Publication number
- CN116151487B CN116151487B CN202310416873.XA CN202310416873A CN116151487B CN 116151487 B CN116151487 B CN 116151487B CN 202310416873 A CN202310416873 A CN 202310416873A CN 116151487 B CN116151487 B CN 116151487B
- Authority
- CN
- China
- Prior art keywords
- oil
- oil spill
- data
- grid
- prediction
- 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
- 239000003305 oil spill Substances 0.000 title claims abstract description 208
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 30
- 230000001413 cellular effect Effects 0.000 claims abstract description 56
- 238000004458 analytical method Methods 0.000 claims abstract description 23
- 238000012937 correction Methods 0.000 claims abstract description 18
- 238000009960 carding Methods 0.000 claims abstract description 3
- 238000000034 method Methods 0.000 claims description 75
- 238000004364 calculation method Methods 0.000 claims description 43
- 238000009792 diffusion process Methods 0.000 claims description 34
- 230000008569 process Effects 0.000 claims description 33
- 238000004088 simulation Methods 0.000 claims description 28
- 239000013535 sea water Substances 0.000 claims description 26
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 24
- 238000001704 evaporation Methods 0.000 claims description 18
- 230000008020 evaporation Effects 0.000 claims description 18
- 238000013528 artificial neural network Methods 0.000 claims description 12
- 238000004090 dissolution Methods 0.000 claims description 12
- 230000000694 effects Effects 0.000 claims description 10
- 239000013598 vector Substances 0.000 claims description 10
- 230000005540 biological transmission Effects 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 9
- 238000011160 research Methods 0.000 claims description 8
- 230000008859 change Effects 0.000 claims description 7
- 238000004945 emulsification Methods 0.000 claims description 7
- 230000009466 transformation Effects 0.000 claims description 7
- 238000009835 boiling Methods 0.000 claims description 6
- 238000007667 floating Methods 0.000 claims description 6
- 239000000126 substance Substances 0.000 claims description 6
- 238000012549 training Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 239000000839 emulsion Substances 0.000 claims description 3
- 230000003993 interaction Effects 0.000 claims description 3
- 238000005259 measurement Methods 0.000 claims description 3
- 238000002844 melting Methods 0.000 claims description 3
- 230000000704 physical effect Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 230000005570 vertical transmission Effects 0.000 claims description 3
- 230000005574 cross-species transmission Effects 0.000 claims description 2
- 238000009826 distribution Methods 0.000 claims description 2
- 230000001804 emulsifying effect Effects 0.000 claims description 2
- 230000004927 fusion Effects 0.000 claims description 2
- 238000013507 mapping Methods 0.000 claims description 2
- 238000012546 transfer Methods 0.000 claims description 2
- 239000003921 oil Substances 0.000 description 115
- 230000006399 behavior Effects 0.000 description 9
- 239000000463 material Substances 0.000 description 8
- 230000000875 corresponding effect Effects 0.000 description 6
- 238000011161 development Methods 0.000 description 6
- 230000007613 environmental effect Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 238000005553 drilling Methods 0.000 description 5
- 239000010779 crude oil Substances 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000011156 evaluation Methods 0.000 description 3
- 239000003208 petroleum Substances 0.000 description 3
- 238000007781 pre-processing Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 230000000903 blocking effect Effects 0.000 description 2
- 238000007670 refining Methods 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 241000287196 Asthenes Species 0.000 description 1
- 238000010521 absorption reaction Methods 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 238000004873 anchoring Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000004040 coloring Methods 0.000 description 1
- 238000002485 combustion reaction Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000001066 destructive effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 239000000295 fuel oil Substances 0.000 description 1
- 238000011065 in-situ storage Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 230000002085 persistent effect Effects 0.000 description 1
- 238000012950 reanalysis Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000004064 recycling Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000010845 search algorithm Methods 0.000 description 1
- 239000010802 sludge Substances 0.000 description 1
- 238000011425 standardization method Methods 0.000 description 1
- 238000005211 surface analysis Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
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
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
- G06N5/022—Knowledge engineering; Knowledge acquisition
- G06N5/025—Extracting rules from data
-
- 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- 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
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Economics (AREA)
- Software Systems (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- Mathematical Physics (AREA)
- General Business, Economics & Management (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Operations Research (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Quality & Reliability (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to the field of track prediction, and particularly relates to a prediction algorithm for predicting a physical knowledge and data hybrid drive of a sea surface oil spill track, which comprises the following steps: step one: carding and theoretical analysis are carried out on the cellular automaton theory; step two: confirming influence factors, establishing a foundation for rules, and carrying out supplementary correction on the results; step three: fusing influence factors, constructing an improved prediction model, and calculating a prediction result; step four: analyzing the used parameters, determining a prediction model, and calculating a prediction value of the prediction model; step five: summarizing and analyzing the calculated results to construct a prediction model under hybrid driving; step six: the data are subjected to dimension consistency and grid unification, and physical knowledge and the data are subjected to mixed driving; step seven: and the oil spill track prediction result is obtained through operation, an existing hybrid intelligent system is taken as a framework, the data driving is assisted to perform oil spill track prediction by using a model containing physical priori knowledge, and the prediction efficiency and accuracy are improved.
Description
Technical Field
The invention belongs to the technical field of track prediction, and particularly relates to a prediction algorithm for predicting a physical knowledge and data hybrid drive of a sea surface oil spill track.
Background
With the rapid development of offshore oil exploitation and transportation industry, the number of offshore drilling platforms is continuously increased, and the total petroleum transportation amount is rapidly increased. According to EIA data, the world oil limit reserves are 1 trillion tons, and the recoverable reserves are 3000 billions tons, wherein the marine oil reserves account for 45 percent.
However, a great deal of oil exploitation and transportation causes a series of problems, wherein the most prominent is the frequent occurrence of marine oil spill accidents. The marine oil spill accident is the phenomenon that crude oil leaks into the sea due to unexpected situations or human misoperation and other reasons in the process of petroleum exploration, development, storage and refining, and mainly comprises the oil drilling platform leakage accident and the ship collision oil spill accident. The two oil spills have the greatest effect on coastal and marine environments and can cause serious losses to surrounding fisheries and travel industries. According to statistics, leakage accidents with oil spilling quantity exceeding 1000 tons occur about 4.4 times each year on average, and the total oil spills into the ocean for about 600 ten thousand tons, so that huge economic resource loss is caused, the ocean ecological environment is more seriously polluted, partial organisms face the extinct danger, and the living environment of the ocean organisms is greatly destroyed. For example, the "deep water horizon" event of month 4 of 2010 is the most representative event in oil drilling platform leakage accidents, called the "most serious one-time" oil spill accident, a large amount of crude oil flows into the bay, and the oil spill is completely controlled only by more than 3 months from the accident, so that the economic loss is as high as billions of cents. In the aspect of ship oil spill accidents, only in the coastal ports from 1973 to 2018, the ship oil spill accidents 3336 are commonly generated, and the average of the ship oil spill accidents is 76 per year, so that the marine environment is adversely affected, and huge economic loss is caused.
According to the OPRC convention (International oil pollution control, reaction and Cooperation convention 1990), the definition of "oil" in offshore spilled oil refers to any form of petroleum, including sludge, oil residues, fuel oil, refining products, crude oil, and the like. After an accident occurs, leaked oil floats on the sea surface, so that serious pollution is caused to the marine environment, the ecological environment is destroyed, life casualties and property loss are caused, and the safety of the region is possibly even influenced. The main causes of the hazard are as follows: (1) the overflowed oil products are inflammable, and are easy to explode and fire, so that the casualties and economic losses of crew and rescue personnel are caused; (2) because the accident site is on the sea surface, spilled oil stains can drift and spread along with sea waves and wind; (3) after the oil product is covered on the water surface, the seabirds and marine products can die in a large amount in the marine environment due to anoxic choking or oil inhalation, and the marine ecological environment is greatly damaged. The original ecological environment is restored, so that huge manpower and material resources are needed, and huge economic losses are caused; (4) with the drift diffusion of the greasy dirt, coastal tourist areas and coastal ports are destroyed, the safety of the areas is endangered, and the regional and national rights and interests are involved.
Marine oil spill accidents are sudden events with a series of characteristics of sudden, urgent, highly uncertain, complex, destructive, persistent and derivative. When the overflowed oil is treated, emergency measures such as oil spilling combustion, oil spilling recovery, oil spilling blocking and the like are selected according to accident conditions. In the process, emergency materials such as blocking oil spill materials, recycling oil spill materials, dispersing oil spill materials, oil absorption materials and the like are needed. In order to reduce the loss caused by the emergency as much as possible, an effective scheduling decision must be adopted in time, and the oil spill emergency materials are scheduled in time and accurately according to the requirements of the accident point. Therefore, when the accident of sudden oil spill on the sea is dealt with, no matter the damage of the accident of oil spill to the sea environment is reduced, or the economic loss caused by the accident is reduced, the present position and the possible future position of the oil spill can be reasonably and effectively found, the emergency measures can be timely taken, and the pollution damage of the oil spill is controlled to be in the minimum range.
Currently, there are two broad classes of different approaches to predicting oil spill trajectories. Traditional oil spill track prediction is to divide the oil spill process and create corresponding physical models in stages. These models, typically working with weather forecast systems, acquire a large amount of information for a particular area and can be modeled. This type of solution works well on the overall trajectory prediction of spilled oil, but the duty cycle prediction of spilled oil is poor in a particularly small area, and it is quite difficult to generalize the same solution to a new area. In order to create more adaptive predictive models, some scholars have proposed data-driven hybrid intelligent system predictive methods aimed at generating solutions to new problems by analyzing solutions to previously addressed problems. But such methods also encounter some challenges: (1) on the whole track of predicted oil spill, particularly, a large error exists in the prediction of a grid area without any oil spill, and the result cannot be interpreted or is inconsistent physically; (2) whether the case library structure can be reasonably designed, the obvious characteristics can be correctly selected, and an effective indexing technology can be developed; (3) whether a suitable method can be found to process the retrieved set of similar cases to obtain a solution for the new case. Thus, it is an in-point herein to utilize known physical knowledge driven models to assist in building hybrid intelligent systems, solving the building challenges and challenges of hybrid intelligent systems.
Disclosure of Invention
Aiming at the problems existing in the prior art, the invention aims at: the sea surface oil spill track prediction algorithm can consider various influencing factors and has more accurate prediction results, and can accurately predict future oil spill diffusion positions according to the current state of oil spill point positions, environmental factors and the like after an actual oil spill accident occurs.
In order to achieve the above purpose, the technical scheme adopted by the invention is as follows: the prediction algorithm for predicting the physical knowledge and data mixed driving of the sea surface oil spill track comprises the following steps:
step one: carding and theoretical analysis are carried out on a common model for solving the problem of predicting the sea surface oil spill track by using a cellular automaton theory;
step two: confirming influence factors of sea surface oil spill tracks, analyzing the influence factors of the sea surface oil spill tracks on the basis of the first step, establishing a basis for rules of a cellular automaton model, and carrying out supplementary correction on results obtained by a traditional cellular automaton model;
step three: fusing the influence factors of the sea surface oil spill track calculated in the second step, constructing an improved cellular automaton physical prediction model, and calculating a prediction result driven by physical knowledge;
Step four: analyzing parameters used in the data-driven prediction process, determining a data-driven prediction model, and calculating a prediction value of the prediction model under the driving of data;
step five: summarizing and analyzing the results calculated by the physical driving model and the data driving model calculated in the second step and the third step, and constructing a sea surface oil spill track prediction model under mixed driving;
step six: on the basis of the sea surface oil spill track prediction model constructed in the step five, carrying out dimension consistency and grid unification on related data of different data sources, and carrying out hybrid driving on physical knowledge and data in a residual error fusion mode;
step seven: and step six, carrying out operation through mixed driving to obtain a sea surface oil spill track prediction result.
The above prediction algorithm for predicting the sea surface oil spill track by mixing the physical knowledge and the data, wherein the first step comprises the steps of constructing a sea surface oil spill track prediction model based on cellular automaton, and realizing a driving model of the physical knowledge, and the specific steps are as follows:
step 1-1: constructing a traditional cellular automaton model, judging whether oil overflows from each cell or not by the mass of each cell, and determining the mass of each cell (i, j, k) at the time of t+1 The oil spill quality calculation formula is:
wherein: m represents the transmission coefficients of four positive directions of east, west, south and north; md represents southeast,Bevel angle correction coefficients in four bevel angle directions of northeast, southwest and northwest,is the quality of the spilled oil in the cell (i, j, k) at time t,/i>Representing the oil spill quality of the cell (i, j, k) at time t+1; />Respectively representing the oil spill quality of the cells (i, j-1, k), (i, j+1, k), (i+1, j, k) and (i-1, j, k) at the time t;respectively representing the oil spill quality of the cells (i+1, j+1, k), (i+1, j-1, k), (i-1, j-1, k) and (i-1, j+1, k) at the time t;
step 1-2: comprehensively considering the problems of cell space, cell state, cell neighbor, cell rule and initial oil spill quality of cells, and constructing a sea surface oil spill track prediction model based on a cellular automaton.
The above prediction algorithm for predicting the physical knowledge and data hybrid driving of the sea surface oil spill track, wherein the cellular space refers to a space formed by all the ocean oil spill cells according to a certain rule, the selection of different shapes of a triangle, a quadrangle or a hexagon is also important except the dimension of the cellular space, the coupling is carried out with the data divided by gridding according to the need of a cellular automaton model, the vertical diffusion of the oil spill has a transformation influence on the oil spill, the cellular simulation of the ocean oil spill adopts a three-dimensional cellular space, and a research area is divided into N multiplied by N areas, wherein N is a natural number;
The cell state refers to two states of uncontaminated and contaminated of each cell, and is setFor the state threshold, the mass of the cell (i, j, k) at time t is denoted +.>Then at +.>In the case of a cell state being contaminated; />The lower cell state is pollution-free;
the cell neighbors refer to surrounding cells which play a role in determining the state of a cell in a cellular automaton model, the cell neighbors adopt a molar neighbor type, and the number of neighbors is 3 d -1, d is the dimension, i.e. each central cell considers four positive directions around, east, west, south, north and four oblique directions, the influence of north-east, south-east, north-west, south-west, the working distance is 1 cell distance;
the cell rule is that the track and quality of the spilled oil are influenced by the spilled oil and the influence of external factors in the diffusion process, and the cell rule is used for correcting the spilled oil quality output by a cell automaton model.
The above prediction algorithm for predicting the physical knowledge and data hybrid driving of the sea surface oil spill track, wherein the second step comprises:
step 2-1: and (3) analyzing the influence of wind and flow states, wherein a wind flow influence coefficient W represents the influence of different neighbor cells on a central cell in the wind and flow states, and the wind influence coefficient is as follows: when the wind speed is from west to east, only the neighboring cells (i-1, j, k) affect the central cells (i, j, k), and the influence coefficient W generated by the wind at t F Introducing the average value of the wind speeds at (i, j, k) and (i-1, j, k), and taking the ratio of the average value to the maximum value of the wind speed as an influence coefficient of wind, wherein the formula is as follows:
wherein->And->East wind velocity at cells (i, j, k) and (i-1, j, k) at t, respectively; FV (FV) max Is the maximum wind speed; r is R W Is the wind conversion coefficient, the influence coefficients in other directions are calculated in turn,
W s to the influence coefficient of water flow in a certain direction at t, W s Correcting the ratio of the water flow speed in the certain direction to the maximum flow speed in the river basin, wherein the formula is as follows:
wherein->The velocity of the water flow at the cell (i, j, k) in the east direction at time t; SV (SV) max For this purpose, the maximum flow rate observed in the basin, the influence coefficients in the other directions are calculated in sequence,
calculating the influence coefficient of wind and flow as W=W according to the wind influence coefficient and the water flow influence coefficient F +W S The influence coefficient specific to a certain grid is;
Step 2-2: the quality consumption influence analysis of the sea surface spilled oil in the vertical direction is carried out, the spilled oil has vertical diffusion phenomenon in the vertical direction in the process of the diffusion and drifting of the sea surface spilled oil, when the cells are all sea surface cells, wind power is the main power for the vertical transmission of the spilled oil, and the transmission quantity of the spilled oil in the vertical direction is as follows The calculation formula is as follows:
wherein->Representing the relative wind speed of the cell (i, j, k) at time t, and +.>Wherein->Representing the wind speed, WV, of a cell (i, j, k) at time t max Represents the maximum wind speed value once observed, +.>Representing the oil spill quality of the cell (i, j, k) at time t, R W Is the transmission rate in the vertical direction of the oil;
step 2-3: analysis of influence of evaporation, obtaining evaporation quality of spilled oil by evaporation modelThe corrected calculation formula is as follows: />Wherein F is the evaporation rate, and the equation is expressed as:wherein a and b are constants, T is oil temperature, T G Is the gradient of the boiling point curve, T 0 Is the initial boiling point temperature of spilled oil, K is the evaporation coefficient, S is the contact area of oil and seawater, V 0 Is the initial volume of spilled oil, t is the time;
step 2-4: analysis of diffusion influence, the sea surface spilled oil is influenced by physical properties, chemical properties and environmental factors of the spilled oil in the diffusion process, including a dissolution process and an emulsification process, wherein the dissolution process refers to the process that part of spilled oil is dissolved in sea water, and the mass M of spilled oil dissolved in sea water r The calculation formula of (2) is as follows:
wherein D is a diffusion coefficient, (-) ->For concentration gradients in seawater, the negative sign indicates that the direction of oil spill mass increase is different from the direction of oil spill diffusion;
The emulsification process refers to diffusion change of quality influence on central cells, and milkQuality of oil spillIs calculated according to the formula: />Emulsion moisture content Y W The calculation formula is as follows: />Wherein, K A Is constant, u w For wind speed>,/>The final water content is constant;
step 2-5: the analysis of the influence of dissolution loss is that a small part of water-melting substances are dissolved in the sea water in the process of floating the spilled oil in the sea, and the quality parameters that the spilled oil is dissolved in the sea water and lost are set asThe specific calculation formula is as follows:wherein->Is the diffusion coefficient>Is the contact area of spilled oil and sea water, +.>Is the gradient of spilled oil concentration in the ocean.
The above prediction algorithm for predicting the physical knowledge and data hybrid driving of the sea surface oil spill track, wherein the third step comprises:
step 3-1: based on each oil spill grid C i,j The marine element behavior information of (2) is obtained by using a traditional cellular automaton model to obtain an oil spill net under the drive of pure physical knowledgeLattice C i,j Predicted oil spill quality at next timeThe metamorphic oil spill quality is expressed as M i,j,k K represents different depths of the grid area in the sea water according to the wind current influence coefficient W F Vertical oil spill transfer>Evaporation mass of spilled oil>Amount of emulsification loss- >Amount of dissolution loss->And calculating the oil spill quality at the time t+1 according to a traditional cellular automaton model oil spill quality calculation formula>Obtained with the following formula:
step 3-2: rasterizing the prediction region, wherein a large grid of 5km×5km and a small grid of 500m×500m are arranged, so that the oil spill quality data output by the physical model is converted into oil spill percentages, namely each large grid comprises 100 small grids, wherein the oil spill quality of each small grid is provided with a threshold valueIf the oil spilling quality of the small grid at a certain momentThe small grid is considered to have oil spill, and finally, the large grid oil spill duty ratio under the physical knowledge driving model is outputI.e. the predicted outcome of the physical driving model, wherein +.>Is the number of small cells in the large cell for which oil spills are deemed to exist.
The above prediction algorithm for predicting the physical knowledge and data hybrid driving of the sea surface oil spill track is characterized in that the fourth step comprises:
step 4-1: training the change of floating oil in the data prediction grid through an RBF neural network;
step 4-2: selecting a radial base center of the RBF neural network by using a fixed center method, and selecting original data points from an input data set as a center to obtain sample distribution conditions of data samples of the same cluster group after being clustered;
Step 4-3: the Gaussian function is used as an activation function to ensure the training speed of the neural network, the RBF neural network local approximation is realized, and when the input data is far from the center of the activation function, the output value of the hidden layer is ignored;
step 4-4: each oil spill grid C in a given 0-T moment i,j (I, j=1, 2, …, n) the spilled oil has a specific intensity Y i,j,0-T Ocean element behavior information based on each oil spill grid at 0-T momentAnd the oil spilling ratio intensity Y corresponding to the next moment i,j,t+1 Training a predictive neural network NN (X, T; θ), where θ is a trainable weight +.>And a set of deviations b, where the pure data-driven error is +.>Wherein->The prediction result is driven for the data at the next moment.
The above prediction algorithm for predicting the physical knowledge and data hybrid driving of the sea surface oil spill track, wherein the fifth step comprises:
step 5-1: the prediction result of the physical driving model obtained by calculation is thatThe result of the calculation to obtain the data driving model is +.>According to the related concept of the residual error, the residual error of the calculated physical driving result compared with the data driving result is: />;
Step 5-2: after obtaining the residual error of data driving and physical driving and the result of data driving, respectively processing the two items of data according to regression theory, and finally predicting the value Wherein beta is 1 And beta 2 Is a weight for balancing the interaction between the data driven result and the residual by minimizing the lossThe optimal weight parameter beta is obtained, and the sea surface oil spill track data driving model based on residual error correction of physical priori knowledge is obtained.
The above prediction algorithm for predicting the physical knowledge and data hybrid driving of the sea surface oil spill track, wherein the sixth step adopts a minimum-maximum standardization method and a bilinear interpolation method to carry out dimension consistency and grid unification on related data from different data sources, and the specific steps are as follows:
step 6-1: minimum-maximum normalization method for linear transformation of original data and mapping of data value to [0,1 ]]Between, convert the formula into;
Step 6-2: bilinear interpolation, in two directionsTo respectively perform linear interpolation once, at Q 11 =(x 1, y 1 )、Q 12 =(x 1, y 2 )、Q 21 =(x 2, y 1 )、Q 22 =(x 2, y 2 ) When known, to obtain the value of the unknown function f at the point p= (x, y), first, a linear interpolation is performed in the x direction:
Then, linear interpolation is carried out in the y direction, and the following results are obtained:
further, bilinear interpolation results are obtained:
step 6-3: the weighted Euclidean distance is adopted to carry out similarity measurement, and a weight coefficient is introduced to represent the importance degree of each characteristic variable, so that different influences of different variables on the change of the oil spilling track are eliminated, and the similarity between a target case and a source case clustering center is achieved The method comprises the following steps:
european distanceThe calculation formula of (2) is as follows: />Wherein->Is a source case clustering center, wherein->Numbering each source case in the case library; />Is a target case; />The method is used for representing the influence degree of each characteristic variable on sea surface oil spill drift; />Is->No. H of individual Source case>A characteristic variable value, wherein->For feature vector number, ++>;/>First->A characteristic vector value; />And->Respectively the firstCharacteristic direction ofMaximum and minimum values of the quantity.
The step seven includes validity analysis, wherein a calculation example is set based on published global meteorological data, a physical driving model, a data driving model and a mixed driving model are respectively used for solving and predicting the calculation example, and validity of the model and the algorithm is verified according to comparison of the predicted track obtained by solving the real calculation example and the real spilled oil diffusion track.
The above prediction algorithm for predicting the sea surface oil spill track by using the physical knowledge and data mixed driving, wherein the effectiveness analysis adopts a Kappa coefficient method for detection, and the calculation formula is as follows:wherein->For the actual agreement of the simulation result with the actual result, < > >For the simulation of the actual agreement with the actual results, the investigation region is divided into +.>Grid space->For simulation and net number of non-spilled oil as actual result, +.>Grid number of grid which is actually non-spill grid for simulation as spill grid, +.>Grid number actually being the oil spill grid for simulation as non-oil spill grid, +.>For the simulation and the actual result are the grid number of the oil spill grid, the +.>And->The calculation formula of (2) is as follows: />,Combine->And->The specific calculation method of Kappa coefficient can be expressed as: />Wherein, the value of the Kappa coefficient indicates the consistency degree of the simulation result and the actual result.
The data-driven mixed track prediction algorithm combining physical knowledge has the beneficial effects that: the sea surface oil spill track prediction is generally developed and researched based on a physical chemistry principle, and the invention takes the existing hybrid intelligent system under data driving as a framework, and utilizes a model containing physical priori knowledge to assist in solving various challenges encountered by sea surface oil spill track prediction under a data driving method, so that the prediction efficiency and precision of the whole hybrid intelligent system are improved. The cellular automaton model established based on physical priori knowledge is innovatively utilized, the case reuse and correction stage in the data-driven hybrid intelligent system is optimized, the accuracy and robustness of system prediction are improved, and the method has important theoretical significance in expanding the research range of oil spill track prediction and improving the oil spill track prediction method; in real life, if the current position and possible future position of oil spill in the oil spill accident can be accurately and effectively found, emergency countermeasures can be arranged from the container for the ocean management department, the oil spill emergency efficiency is improved, and the peripheral economic and environmental losses are reduced; for surrounding tourists and fishery practitioners, the influence time of oil spill can be reduced, the normal production life and the local normal development can be maintained.
Drawings
FIG. 1 is a diagram showing the effect of gridding an oil spill satellite image;
FIG. 2 is a schematic diagram of a mesh coloring effect;
FIG. 3 is a schematic flow chart of a sea surface oil spill track prediction model based on cellular automata;
fig. 4 is a schematic diagram of a circulation flow based on a case-based reasoning system under hybrid drive;
FIG. 5 is a schematic diagram of bilinear interpolation;
fig. 6 is a schematic diagram comparing the simulated image with the real result.
Detailed Description
In order to enable those skilled in the art to better understand the technical scheme of the present invention, the technical scheme of the present invention will be described below with reference to the detailed description and the accompanying drawings.
According to the invention, a fire explosion accident of a coastal deepwater horizon oil drilling platform is selected as a basic material, and the effect and performance of a mixed prediction model are verified by constructing the prediction model and predicting the oil spill diffusion track of the accident after the recording time.
The invention achieves the aim through the following technical scheme:
as shown in fig. 1-6, in the present technical solution, after an oil spill accident occurs at sea, the spilled oil drifts and spreads from the leaking place to the outside rapidly under the action of wind, current, surge, wave, etc., so as to form a large-area dispersed offshore oil film and oil band, and the formed offshore oil film is generally irregular. At this time, we can perform gridding processing on the nearby area affected by the oil spill, and then determine the amount of oil spill at different time points in each grid corresponding to the oil spill satellite image. Depending on the image and the calculated amount of oil slivers, each grid will have a different number of bright spots, and the oil slivers corresponding to different times will be colored differently, as shown in fig. 1. Thereafter, the grids are colored with different intensities according to the number of floating oil spots occurring in each grid, as shown in fig. 2, the more the number of spots, the darker the color of the grid, and the more serious the oil spill pollution is suffered from the area. To this end, the color shades of each grid can be quantitatively represented as output variables in the data: the oil spill ratio in the oil spill grid.
Research of all parties proves that marine element behaviors in a complex marine system play a decisive role in the development of a sea surface oil spill track, the marine element behaviors comprise ocean current direction and strength, wave height, offshore air pressure and the like, and the behavior information of the marine elements can be obtained through marine weather stations, satellite images and the like. Thus, the sea condition of each oil spill grid can be used as an input characteristic parameter for oil spill track prediction. In summary, the ocean situation where each oil spill case is located is used as the case feature to construct the case library, and a case reasoning system or a data driver related to the oil spill track is constructed according to the frame of the case reasoning system to become a possible method for predicting the oil spill track.
According to the satellite image recognition or calculated oil spill pollution degree, each grid can be represented by different numbers of bright spots, and the oil spill bright spots corresponding to different times can be coated with different colors, wherein the former time period is represented by reddish brown, and the latter time period is represented by purple.
According to the number of floating oil bright spots in each grid, the grids are colored with different intensities, and the more the number of bright spots is, the darker the color of the grids is, and the more serious the oil spill pollution is on the area.
Considering that a pure data-driven model may be well suited for observations, it leads to poor generalization performance and situations where the predicted results are physically inconsistent or unexplained. Therefore, residual correction is carried out on the data driving result through the cellular automaton model prediction result driven by physical knowledge, and the residual correction is used as a core model of a case correction and reuse stage.
The sea surface oil spill track prediction problem under the mixed driving of physical priori knowledge and data can be summarized as follows: on the sea surface where spills may affect, a case-based reasoning system for spilled oil trajectory prediction will grid the spilled oil areas to be analyzed, each of whichThe grid of the oil spilling area is the case main bodyIn combination with specific cases, the specific steps are as follows:
step 1: the sea surface oil spill track prediction model based on cellular automata is constructed, and a driving model of physical knowledge is realized, wherein the method comprises the following specific steps:
1.1. there are many kinds of sea surface oil spill track prediction models driven by physical knowledge, but the data form of the oil spill prediction model based on cellular automata and the grid data form based on a GIS geographic satellite system are considered to have natural consistency, so that the output forms of the two models are unified. Therefore, the sea surface oil spill track prediction under the drive of physical knowledge is realized by selecting the sea surface oil spill prediction model based on the cellular automaton. FIG. 3 is a flowchart of prediction using cellular automaton model, first meshing the prediction region, calculating the oil spill quality M of each mesh, and combining M with a threshold M 0 And comparing to judge whether oil overflows in the grid, and drawing an oil overflow grid chart according to the existence condition. The traditional cellular automaton oil spill prediction model has the advantages that dynamic oil spill track simulation analysis can be realized, and the oil spill quality in a cellular grid can be output in real time, which is consistent with the data output form of each time period in the whole data driving process, so that the traditional cellular automaton model is selected to realize physical knowledge driving. Constructing a traditional cellular automaton model, judging whether oil overflows from each cell or not by the mass of each cell, and determining the mass of each cell (i, j, k) at the time of t+1The oil spill quality calculation formula is:wherein: m represents the transmission coefficients of four positive directions of east, west, south and north; md represents the bevel angle correction coefficients of four bevel angle directions of southeast, northeast, southwest and northwest, < ->Is the quality of the spilled oil in the cell (i, j, k) at time t,/i>Representing the oil spill quality of the cell (i, j, k) at time t+1; />Respectively representing the oil spill quality of the cells (i, j-1, k), (i, j+1, k), (i+1, j, k) and (i-1, j, k) at the time t; />Representing the oil spill quality of the cells (i+1, j+1, k), (i+1, j-1, k), (i-1, j-1, k) and (i-1, j+1, k) at time t, respectively.
1.2. Comprehensively considering the problems of cell space, cell state, cell neighbor, cell rule and initial oil spill quality of cells, and constructing a sea surface oil spill track prediction model based on a cellular automaton.
Step 2: the specific rule of the cellular automaton model is determined, and the specific steps are as follows:
2.1. the cellular space is a space formed by all the ocean oil spill cells according to a certain rule, the selection of different shapes of a triangle, a quadrangle or a hexagon is also important except the dimension of the cellular space, the cellular space is coupled with gridding divided data according to the need of a cellular automaton model, the vertical diffusion of the oil spill has a transformation influence on the oil spill, the ocean oil spill cell simulation adopts a three-dimensional cellular space, a research area is divided into N multiplied by N areas, and N is a natural number.
2.2. The cell state refers to two states of uncontaminated and contaminated of each cell, M is set 0 For the state threshold, the mass of the cell (i, j, k) at time t is expressed asThen at +.>In the case of a cell state being contaminated;the lower cell state is uncontaminated.
2.3. The cell neighbors refer to surrounding cells which play a role in determining the state of a cell in a cellular automaton model, the cell neighbors adopt a molar neighbor type, and the number of neighbors is 3 d -1, d is the dimension, i.e. each central cell considers the influence of four positive directions around, east, west, south, north and four oblique directions, northeast, southeast, northwest, southwest, the working distance is 1 cell distance.
2.4. The transformation rule is that the influence of the self and external factors received by the spilled oil in the diffusion process influences the track, quality and the like of the spilled oil, and the factors are considered to correct the spilled oil quality output by the cellular automaton model.
Step 3: according to the analysis of the factors actually influencing the sea surface oil spill, the influence on the oil spill quality due to the nature of the oil spill and external factors is determined, and the specific steps are as follows:
3.1. and (3) analyzing influence of wind and flow states, wherein an influence coefficient W represents the influence of different neighbor cells on a central cell in the wind and flow states, and the wind influence coefficient is as follows: when the wind speed is from west to east, only the neighboring cells (i-1, j, k) affect the central cells (i, j, k), and the influence coefficient W generated by the wind at t F Introducing the average value of the wind speeds at (i, j, k) and (i-1, j, k), and taking the ratio of the average value to the maximum value of the wind speed as an influence coefficient of wind, wherein the formula is as follows:wherein->And->East wind velocity at cells (i, j, k) and (i-1, j, k) at t, respectively; FV (FV) max Is the maximum wind speed; r is R W Is the wind conversion coefficient, and the influence coefficients in other directions are calculated sequentially.
W s To the influence coefficient of water flow in a certain direction at t, W s Correcting the ratio of the water flow speed in the certain direction to the maximum flow speed in the river basin, wherein the formula is as follows:wherein->The velocity of the water flow at the cell (i, j, k) in the east direction at time t; SV (SV) max For this purpose the maximum flow rate observed in the basin, the influence coefficients in the other directions are calculated in sequence.
Calculating the influence coefficient of wind and flow as W=W according to the wind influence coefficient and the water flow influence coefficient F +W S The influence coefficient specific to a certain grid is。
3.2. The quality consumption influence analysis of the sea surface spilled oil in the vertical direction is carried out, the spilled oil has vertical diffusion phenomenon in the vertical direction in the process of the diffusion and drifting of the sea surface spilled oil, when the cells are all sea surface cells, wind power is the main power for the vertical transmission of the spilled oil, and the transmission quantity of the spilled oil in the vertical direction is as followsThe calculation formula is as follows:wherein->Representing the relative wind speed of the cells (i, j, k) at time t, anWherein->Representing the wind speed, WV, of a cell (i, j, k) at time t max Represents the maximum wind speed value once observed, +.>Representing a cell (i, j, k) at tMoment oil spill quality, R W Is the transmission rate in the vertical direction of the oil.
3.3. Analysis of the influence of evaporation by means of an evaporation model driver&Mackay mode to obtain evaporation quality of spilled oil The corrected calculation formula is as follows: />Wherein F is the evaporation rate, and the equation is expressed as:wherein a and b are constants, T is oil temperature, T G Is the gradient of the boiling point curve, T 0 Is the initial boiling point temperature of spilled oil, K is the evaporation coefficient, S is the contact area of oil and seawater, V 0 Is the initial volume of spilled oil and t is the time.
3.4. Analysis of diffusion influence, the sea surface spilled oil is influenced by physical properties, chemical properties and environmental factors of the spilled oil in the diffusion process, including a dissolution process and an emulsification process, wherein the dissolution process refers to the process that part of spilled oil is dissolved in sea water, and the mass M of spilled oil dissolved in sea water r The calculation formula of (2) is as follows:wherein D is a diffusion coefficient, (-) ->For a gradient of spilled oil concentration in sea water, the negative sign indicates that the direction of the increase in spilled oil mass is different from the direction of the diffusion of spilled oil. />
The emulsifying process refers to diffusion change of the quality influence on the central cell, and the quality of emulsified oil spillIs calculated according to the formula: />Emulsion moisture content Y W The calculation formula is that:/>Wherein, K A =4.5×10 -6 ,u w For wind speed>,/>The final water content was obtained.
3.5. The analysis of the influence of dissolution loss is that a small part of water-melting substances are dissolved in the sea water in the process of floating the spilled oil in the sea, and the quality parameters that the spilled oil is dissolved in the sea water and lost are set as The specific calculation formula is as follows:wherein->Is the contact area of spilled oil and sea water.
Step 4: the case reasoning system for constructing the hybrid drive prediction model comprises the following specific steps:
4.1. and (3) case retrieval: the attribute variable subset x= { X of the case 1 ,X 2 ,…,X n As node variables for case retrieval { temperature, air pressure, salinity, }, respectively. Then, carrying out case retrieval on a case base built by the actual oil spill data to obtain and predict a case C D Most similar case group of (2). Once the most similar cases to the problem to be solved are recovered from the case library, the information of the most similar cases is used as a model input in the case reuse and correction stage to obtain the predicted case +.>Spill ratio +.>. And finally, combining the oil spill ratios of the oil spill grids of each prediction grid at different moments to obtain the development condition of the oil spill track.
4.2. The case reasoning system reuses and corrects cases, and gives each oil spilling grid C in 0-T time i,j (I, j=1, 2, …, n) spill-over ratio (intensity) information Y i,j,0-T And marine element behavior information for each oil spill grid at all timesRespectively { temperature, air pressure, salinity, } > Total 10 behavior information variables, as shown in table 1.
TABLE 1 Marine element behavior information parameter Table
The specific steps for carrying out case correction are as follows:
4.2.1 Pure data driving, marine element behavior information of each oil spill grid based on 0-T momentAnd the oil spilling ratio intensity Y corresponding to the next moment i,j,t+1 The predictive neural network NN (X, T; θ) can be trained, where θ is the set of trainable weights ω and bias b, where the pure data-driven error is。
4.2.2 Purely physical model driven, based on each oil spill grid C i,j We can use the traditional cellular automaton model, and the physical process includes oil spill driftMigration, weathering and other factors to obtain oil spill grid C driven by pure physical knowledge i,j Predicted oil spill quality at next time. Wherein the cell or grid spill mass is denoted as M i,j,k Where k represents the different depths of the grid area in the sea, the invention predicts the sea surface oil spill mass, so k is typically 0, the oil spill mass at time t+1 +.>This can be obtained by the following formula,wherein the influence coefficient of wind and flow w=w F +W S The method comprises the steps of carrying out a first treatment on the surface of the Vertical oil spill transport quantity +.>The method comprises the steps of carrying out a first treatment on the surface of the Evaporation quality of spilled oil->The method comprises the steps of carrying out a first treatment on the surface of the Amount of emulsification loss->The method comprises the steps of carrying out a first treatment on the surface of the Amount of dissolution loss M r Both derived and calculated in step 3. In general, the meshing accuracy with data driving (5 km×5km for example) is much smaller than that with physical models (500 m×500m for example). Therefore, we can take the large grids as the reference of research to convert the oil spill quality data output by the physical model into the oil spill percentage, namely each large grid comprises 100 small grids, wherein the oil spill quality of each small grid is set with a threshold M 0 If the oil spilling quality of the small grid at a certain moment +.>The small grid is considered to have oil spills. Finally, large grid oil spill ratio output +.>Wherein n is oil Is the number of small cells in the large cell for which oil spills are deemed to exist. At this time, the residual error of the physical driving result and the data driving result +.>;
4.2.3 After obtaining the residual error of data driving and physical driving and the result of data driving, the case correction is carried out to the two items according to regression theory, and finally the predicted value is obtained. Wherein beta is 1 And beta 2 Is the weight used to balance the interaction between the data driven result and the residual. Finally, by minimizing the loss +.>The optimal weight parameter beta is obtained, a sea surface oil spill track data driving model which carries out residual correction based on physical priori knowledge is obtained, and the sea surface oil spill track data driving model is used as a core model in the whole case reuse and revision stage.
4.3. After the case learning is performed to obtain the predicted value of the target case, the characteristic attribute vector of the target case and the predicted value thereof form a new vector sequence together. The new cases are updated into the case base for expansion, so that the long-term development of the follow-up reasoning work is facilitated. However, the update of the case library should conform to a certain rule, if the similarity between the new case and the existing case in the case library is high, the meaning of the case for improving the retrieval efficiency is smaller, and even the retrieval efficiency is reduced due to too many similar cases in the library, so that the similarity determination needs to be performed at this stage.
Step 5: according to the content of the steps, a sea surface oil spill track prediction model under hybrid driving is obtained, and in order to improve the accuracy of prediction of a case-based reasoning system, the hybrid system comprises different artificial intelligence technologies so as to achieve the aim of each CBR stage. As shown in fig. 4, each CBR phase uses artificial intelligence techniques to obtain its solution, and the phases of the related art will be explained in the following steps.
5.1. The data preprocessing and the case retrieval are carried out, in the data preprocessing stage, different dimensions are often considered among different indexes, the numerical difference is large, and geographic grid division standards of different data source systems are inconsistent. The system needs to carry out dimension consistency and grid unification on related data from different data sources, and the invention selects a minimum-maximum normalization method and a bilinear interpolation method to carry out.
5.1.1 Minimum-maximum normalization method, which performs linear transformation on original data and can map data value to [0,1 ]]Between, convert the formula into。
5.1.2 The bilinear interpolation method performs linear interpolation once in two directions, respectively. The interpolation process follows the interpolation rules shown in FIG. 5, assuming Q 11 =(x 1, y 1 )、Q 12 =(x 1, y 2 )、Q 21 =(x 2, y 1 )、Q 22 =(x 2, y 2 ) As is known, to obtain the value of the unknown function f at the point p= (x, y), first a linear interpolation is performed in the x direction:,wherein R is 1 =(x 1, y 1 ),R 2 =(x 2, y 2 )。
Then, linear interpolation is carried out in the y direction, and the following results are obtained:。
further, bilinear interpolation results are obtained:
5.2. and after the case storage and the data preprocessing are finished, a K-Means clustering algorithm is applied to store historical data clusters in a case library to serve other stages in the CBR period. When a new problem or a predicted problem occurs, the classification search algorithm calculates the similarity of the new problem with different clustering centers of the stored cases, the similarity being represented by the multidimensional distances between the case structure variables. And taking the case cluster group with the smallest calculated distance as the most similar case group for the next stage.
Considering that the influence of different variables on the change of the oil spill track is different, the similarity measurement is carried out by adopting a weighted Euclidean distance, and a weight coefficient is introduced to represent the importance degree of each characteristic variable. Similarity between target case and source case clustering center The method comprises the following steps: />European distance->The calculation formula of (2) is as follows:wherein->Is a source case clustering center, wherein->Numbering each source case in the case library; />Is a target case; />The method is used for representing the influence degree of each characteristic variable on sea surface oil spill drift; />Is->Individual sourcesCase->A characteristic variable value, wherein->For feature vector number, ++>;/>First->A characteristic vector value; />And->Respectively +.>Maximum and minimum values of the feature vectors.
The rest of the procedure has been described in step 4.
Step 6: the actual case analysis and algorithm effect verification, the local time of 20 days of 4 months in 2010 is about 22:00, the oil drilling platform of the "deep water horizon" (Deepwater Horizon) is ignited and exploded until 15 days of 7 months in 2010, after the oil leakage event occurs for 3 months, the new oil control device successfully covers the underwater oil leakage point, and no crude oil flows into the bay. The whole deepwater horizon event is greatly focused by people in all communities of the whole society and has more complete and massive basic data, so the technical scheme takes the deepwater horizon event as a case to be analyzed and displayed for approving a prediction result in the application, and the specific steps are as follows:
6.1. The method for collecting and arranging the real oil spill data comprises the following specific steps:
6.1.1 Oil spill image data, the present solution collects correlated processed image data of cumulative oil stain areas collected during the north-in-the-bay DWH response during the period from 4/23/2010 to 8/11/2010.
6.1.2 Flow field data, the technical scheme obtains flow field data from a bay area ocean circulation model (GOM HYCOM) provided by an HYCOM model website: seawater salinity, water surface temperature, eastern water flow rate, northbound water flow rate.
The model uses a coupled marine data assimilation (NCOMA) system for data assimilation, absorbing available satellite altimeter observations and in situ Sea Surface Temperatures (SST) as well as sea surface temperature and salinity profile from XBT, argo buoy and anchoring buoy. The use of the NCOMA assimilation technique can control the forecast error in a smaller range, so that the forecast result has higher accuracy. After reanalysis, the flow field data are all unified into a grid (gomu 0.04) with a resolution of 1/25 ° (0.04), and converted into NetCDF format, providing the following 5 variables/fields: sea level height, water temperature, salinity, east-to-sea water flow rate, and north-to-sea water flow rate.
6.1.3 Wind farm data, wind farm data sources are wind farms (http:// RDA. Ucar. Edu) of environmental forecast system analysis data (CFSR) provided by the environmental forecast center (NCEP) of CIRL Research Data Archive (RDA), CFSR is initialized 4 times per day (0000, 0600, 1200 and 1800 UTC), i.e. data is collected every 6 hours. The horizontal resolutions of the 6 hour atmospheric, marine and land surface analysis products were 0.3, 0.5, 1.0, 1.9 and 2.5 degree horizontal resolutions (0.01 degree approximately equal to 1000 meters). According to the technical scheme, 0.3 horizontal resolution data with the most clear resolution is selected, and the resolution of the data is kept consistent with that of the image data through a double-line interpolation method. Data type: the water surface air pressure, the warp direction, the wind force, the weft direction and the wind force.
6.1.4 In the deep water horizon event (Deepwater Horizon oil spill) experiment, other relevant data of the cellular automaton model are set as follows: cell size 500 m, time step 1800 s, assume that oil spills 15000 barrels consecutively ten days before oil leaks, followed by 80000 barrels consecutively daily as input.
6.2. Based on the data and the model setting rules, the prediction result is obtained by using python and java codes, and as the related work before, the method carries out result display on two aspects of the whole sea surface oil spill track prediction image comparison and the conventional prediction evaluation index, and the specific steps are as follows:
6.2.1 And (5) carrying out overall track prediction of sea surface spilled oil. Considering that the oil spilling time of the deepwater horizon event is too long and the period is subject to artificial interference, and the accuracy of the satellite identification technology is limited at the time. Therefore, the technical scheme utilizes the established case-based reasoning system to predict and display the whole track of the same day for 5 months and 1 day after the occurrence of the deepwater horizon event. For convenience of comparison, the simulation result is displayed in a white grid for the grid with oil spill and in a black grid for the grid without oil spill along with a display method of a cellular automaton model, and the specific prediction result is shown in fig. 6.
6.2.2 And (3) checking the overall prediction effect, wherein in order to further quantify the overall track prediction effect, a Kappa coefficient method is adopted for checking. The calculation formula of Kappa coefficient method is:wherein->For the actual agreement of the simulation result with the actual result, < >>To simulate the actual agreement with the actual results.
Dividing a study area intoGrid space->For simulation and net number of non-spilled oil as actual result, +.>To simulate as a spill grid and actually be non-spillGrid number of oil grid->Grid number actually being the oil spill grid for simulation as non-oil spill grid, +. >For the simulation and the actual result are the grid number of the oil spill grid, the +.>And->The calculation formula of (2) is as follows: />,/>Thus, a specific calculation method of Kappa coefficients according to the present invention can be expressed as: />。
Different Kappa coefficients indicate different levels of consistency. When Kappa coefficient is more than 0.8, the simulation result and the actual result are proved to have excellent consistency; the Kappa coefficient is more than 0.6 and less than 0.8, which shows that the simulation result has higher consistency with the actual result; the Kappa coefficient is more than 0.4 and less than 0.6, which shows that the simulation result is moderately consistent with the actual result; kappa coefficient < 0.4 indicates poor consistency. The prediction simulation result obtained by using the prediction model formed by the prediction algorithm of the application is shown in table 2, and the Kappa coefficient reaches 0.625, which marks that the simulation result has better consistency with the actual result.
TABLE 2 Kappa coefficient simulation calculation results
6.2.3 Conventional predictor evaluations, using three different techniques, were performed using a different number of case libraries, ranging from 10000 to 30000 cases, with specific results shown in table 3.
TABLE 3 presentation of conventional predictive evaluation indicators for different technologies and different number of case libraries
From table 3 above, we can conclude that: (1) When only using RBF neural network as a case reuse stage model and not predicting the structural design of the case library, simply increasing the scale of the case library can reduce the prediction accuracy; (2) The predicting effect of the CBR (KMeans+RBF) column is better than that of the CBR (RBF) column, because a KMeans algorithm is utilized to construct a case library, the organization characteristics of the case library are enhanced, useless data are eliminated before model training is used, and the purpose of improving model precision is achieved; (3) The hybrid drive model results are significantly better than the other two techniques, which illustrates that the residuals of the cellular automaton model prediction results and the data drive prediction results, which are introduced herein and driven by physical knowledge, have significant correction effects during the case correction stage.
The above embodiments are only for illustrating the inventive concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly, and are not intended to limit the scope of the present invention. All equivalent changes or modifications made in accordance with the essence of the present invention should be included in the scope of the present invention.
Claims (9)
1. The prediction algorithm for predicting the physical knowledge and data hybrid driving of the sea surface oil spill track is characterized by comprising the following steps:
step one: carding and theoretical analysis are carried out on a common model for solving the problem of predicting the sea surface oil spill track by using a cellular automaton theory;
step two: confirming influence factors of sea surface oil spill tracks, analyzing the influence factors of the sea surface oil spill tracks on the basis of the first step, establishing a basis for rules of a cellular automaton model, and carrying out supplementary correction on results obtained by a traditional cellular automaton model;
step three: fusing the influence factors of the sea surface oil spill track calculated in the second step, constructing an improved cellular automaton physical prediction model, and calculating a prediction result driven by physical knowledge;
step four: analyzing the parameters used in the data-driven prediction process, determining a data-driven prediction model, and calculating a prediction value of the data-driven prediction model, wherein the method comprises the following steps:
Step 4-1: training the data prediction grid for changes in oil slivers through the RBF neural network,
step 4-2: selecting a radial base center of the RBF neural network by using a fixed center method, and selecting original data points from an input data set as a center to obtain sample distribution conditions of data samples of the same cluster group after being clustered;
step 4-3: the Gaussian function is used as an activation function to ensure the training speed of the neural network, the RBF neural network local approximation is realized, and when the input data is far from the center of the activation function, the output value of the hidden layer is ignored;
step 4-4: each oil spill grid C in a given 0-T moment i,j (I, j=1, 2, …, n) the spilled oil has a specific intensity Y i,j,0-T Ocean element behavior information based on each oil spill grid at 0-T momentAnd the oil spilling ratio intensity Y corresponding to the next moment i,j,t+1 The predictive neural network NN (X, T; θ) is trained, where θ is the set of trainable weights ω and bias b, where the pure data driven error is +.>Wherein->Driving a prediction result for data at the next moment;
step five: summarizing and analyzing the results calculated by the physical driving model and the data driving model calculated in the second step and the third step, and constructing a sea surface oil spill track prediction model under mixed driving;
Step six: on the basis of the sea surface oil spill track prediction model constructed in the step five, carrying out dimension consistency and grid unification on related data of different data sources, and carrying out hybrid driving on physical knowledge and data in a residual error fusion mode;
step seven: and step six, carrying out operation through mixed driving to obtain a sea surface oil spill track prediction result.
2. The prediction algorithm for predicting the sea surface oil spill track based on the mixed driving of the physical knowledge and the data according to claim 1, wherein the first step comprises the steps of constructing a sea surface oil spill track prediction model based on a cellular automaton, and realizing a driving model of the physical knowledge, and the specific steps are as follows:
step 1-1: constructing a traditional cellular automaton model, judging whether oil overflows from each cell or not by the mass of each cell, and determining the mass of each cell (i, j, k) at the time of t+1The oil spill quality calculation formula is:
wherein: m represents the transmission coefficients of four positive directions of east, west, south and north; md represents bevel angle correction coefficients of four bevel angle directions of southeast, northeast, southwest and northwest,is the quality of the spilled oil in the cell (i, j, k) at time t,/i>Representing the time t+1 of the cell (i, j, k)Oil spilling quality; />Respectively representing the oil spill quality of the cells (i, j-1, k), (i, j+1, k), (i+1, j, k) and (i-1, j, k) at the time t; Respectively representing the oil spill quality of the cells (i+1, j+1, k), (i+1, j-1, k), (i-1, j-1, k) and (i-1, j+1, k) at the time t;
step 1-2: comprehensively considering the problems of cell space, cell state, cell neighbor, cell rule and initial oil spill quality of cells, and constructing a sea surface oil spill track prediction model based on a cellular automaton.
3. The prediction algorithm for predicting the physical knowledge and data hybrid driving of a sea surface oil spill trajectory according to claim 2, wherein,
the cellular space is a space formed by all the ocean oil spill cells according to a certain rule, the selection of different shapes of a triangle, a quadrangle or a hexagon is also important except the dimension of the cellular space, the cellular space is coupled with gridding divided data according to the need of a cellular automaton model, the vertical diffusion of the oil spill has a transformation effect on the oil spill, the cellular simulation of the ocean oil spill adopts a three-dimensional cellular space, a research area is divided into N multiplied by N areas, and N is a natural number;
the cell state refers to two states of uncontaminated and contaminated of each cell, and is setFor the state threshold, the mass of the cell (i, j, k) at time t is denoted +.>Then at->In the case of a cellThe state is pollution; / >The lower cell state is pollution-free;
the cell neighbors refer to surrounding cells which play a role in determining the state of a cell in a cellular automaton model, the cell neighbors adopt a molar neighbor type, and the number of neighbors is 3 d -1, d is the dimension, i.e. each central cell considers four positive directions around, east, west, south, north and four oblique directions, the influence of north-east, south-east, north-west, south-west, the working distance is 1 cell distance;
the cell rule is that the track and quality of the spilled oil are influenced by the spilled oil and the influence of external factors in the diffusion process, and the cell rule is used for correcting the spilled oil quality output by a cell automaton model.
4. The prediction algorithm for predicting a physical knowledge and data hybrid drive of a sea surface spilled oil trajectory according to claim 2, wherein the second step comprises:
step 2-1: and (3) analyzing the influence of wind and flow states, wherein a wind flow influence coefficient W represents the influence of different neighbor cells on a central cell in the wind and flow states, and the wind influence coefficient is as follows: when the wind speed is from west to east, only the neighboring cells (i-1, j, k) affect the central cells (i, j, k), and the influence coefficient W generated by the wind at t F Introducing the average value of the wind speeds at (i, j, k) and (i-1, j, k), and taking the ratio of the average value to the maximum value of the wind speed as an influence coefficient of wind, wherein the formula is as follows:
in the method, in the process of the invention,and->East wind velocity at cells (i, j, k) and (i-1, j, k) at t, respectively; FV (FV) max Is the maximum wind speed; r is R W Is the wind conversion coefficient, the influence coefficients in other directions are calculated in turn,
W s to the influence coefficient of water flow in a certain direction at t, W s Correcting the ratio of the water flow speed in the certain direction to the maximum flow speed in the river basin, wherein the formula is as follows:
in the method, in the process of the invention,the velocity of the water flow at the cell (i, j, k) in the east direction at time t; SV (SV) max For this purpose, the maximum flow rate observed in the basin, the influence coefficients in the other directions are calculated in sequence,
calculating the influence coefficient of wind and flow as W=W according to the wind influence coefficient and the water flow influence coefficient F +W S The influence coefficient specific to a certain grid isStep 2-2: in the analysis of the influence of the quality consumption of the spilled oil on the sea surface in the vertical direction, the spilled oil has the vertical diffusion phenomenon in the vertical direction in the diffusion and drifting processes of the spilled oil on the sea, when the cells are all sea surface cells, the wind power is the main power for the vertical transmission of the spilled oil, and the transmission quantity of the spilled oil in the vertical direction is ∈ - >The calculation formula is as follows:
in the method, in the process of the invention,representing the relative wind speed of the cell (i, j, k) at time t, and +.>Wherein->Representing the wind speed, WV, of a cell (i, j, k) at time t max Represents the maximum wind speed value once observed, +.>Representing the oil spill quality of the cell (i, j, k) at time t, R W Is the transmission rate in the vertical direction of the oil;
step 2-3: analysis of influence of evaporation, obtaining evaporation quality of spilled oil by evaporation modelThe corrected calculation formula is as follows:
wherein F is the evaporation rate, and the equation is expressed as:
wherein a and b are constants, T is oil temperature, T G Is the gradient of the boiling point curve, T 0 Is the initial boiling point temperature of spilled oil, K is the evaporation coefficient, S is the contact area of oil and seawater, V 0 Is the initial volume of spilled oil, t is the time;
step 2-4: analysis of diffusion influence, the spilled oil on the sea surface can be subjected to physical property, chemical property and environment of the spilled oil in the diffusion processThe influence of factors includes dissolution process and emulsification process, wherein the dissolution process refers to the process of oil spilling, and a part of oil spilling is dissolved in seawater, and the mass M of oil spilling dissolved in seawater r The calculation formula of (2) is as follows:
wherein D is the diffusion coefficient,for concentration gradients in seawater, the negative sign indicates that the direction of the increase of the oil spill mass is different from the direction of the oil spill diffusion,
The emulsifying process refers to diffusion change of the quality influence on the central cell, and the quality of emulsified oil spillIs calculated according to the formula:
emulsion moisture content Y W The calculation formula is as follows:
wherein K is A Is constant, u w For the wind speed of the wind, the final water content is constant;
step 2-5: analysis of the influence of dissolution loss, a small part of water-melting substances can be generated in the process of floating spilled oil in the oceanIs dissolved in seawater, and the quality parameter of spilled oil which is lost when dissolved in seawater is set asThe specific calculation formula is as follows:
5. The prediction algorithm for predicting a physical knowledge and data hybrid drive of a sea surface spilled oil trajectory of claim 4, wherein the third step comprises:
step 3-1: based on each oil spill grid C i,j The marine element behavior information of (2) is used for obtaining the oil spill grid C driven by pure physical knowledge by using the traditional cellular automaton model i,j Predicted oil spill quality at next timeThe metamorphic oil spill quality is expressed as M i,j,k K represents different depths of the grid area in the sea water according to the wind current influence coefficient W F Vertical oil spill transfer>Evaporation quality of spilled oil- >Amount of emulsification loss->Amount of dissolution loss->And calculating the oil spill quality at the time t+1 according to a traditional cellular automaton model oil spill quality calculation formula>Obtained by the following formula:
step 3-2: rasterizing the prediction region, wherein a large grid of 5km×5km and a small grid of 500m×500M are arranged, so that the oil spill quality data output by the physical model is converted into oil spill percentages, namely each large grid comprises 100 small grids, wherein the oil spill quality of each small grid is provided with a threshold value M 0 If the oil spilling quality of the small grid at a certain momentThen the small grid is considered to have oil spill, and finally, the large grid oil spill duty ratio output under the physical knowledge driving modelI.e. the predicted result of the physical driving model, where n oil Is the number of small cells in the large cell for which oil spills are deemed to exist.
6. The prediction algorithm for predicting the physical knowledge and data hybrid driving of a sea surface spilled oil trajectory according to claim 1, wherein the fifth step comprises:
step 5-1: the prediction result of the physical driving model obtained by calculation is Y i,j,PHYS The prediction result of the data driving model obtained by calculation is Y' i,j,t+1 According to the related concept of the residual error, the residual error of the physical driving result compared with the data driving result is calculated as follows: l (L) i,j,t+1 =Y i,j,t+1 -Y i,j,t+1,PHYS ;
Step 5-2: after obtaining the residual error of data driving and physical driving and the result of data driving, respectively processing the two items of data according to regression theory, and finally predicting the value H (Y i,j,t+1 )=βY+ε=β 1 Y′ i,j,t+1 +β 2 L i,j,t+1 +ε wherein β 1 And beta 2 Is a weight for balancing the interaction between the data driven result and the residual by minimizing the lossThe optimal weight parameter beta is obtained, and the sea surface oil spill track data driving model based on residual error correction of physical priori knowledge is obtained.
7. The prediction algorithm for predicting sea surface oil spill trajectories according to claim 6, wherein the step six uses a min-max normalization method and a bilinear interpolation method to perform dimension consistency and grid unification on related data from different data sources, and the specific steps are as follows:
step 6-1: minimum-maximum normalization method for linear transformation of original data and mapping of data value to [0,1 ]]Between, convert the formula into
Step 6-2: bilinear interpolation, which performs linear interpolation once in two directions, respectively, at Q 11 =(x 1 ,y 1 )、Q 12 =(x 1 ,y 2 )、Q 21 =(x 2 ,y 1 )、Q 22 =(x 2 ,y 2 ) When known, to obtain the value of the unknown function f at the point p= (x, y), first, a linear interpolation is performed in the x direction:
wherein R is 1 =(x 1 ,y 1 ),R 2 =(x 2 ,y 2 ),
Then, linear interpolation is carried out in the y direction, and the following results are obtained:
Further, bilinear interpolation results are obtained:
step 6-3: the weighted Euclidean distance is adopted to carry out similarity measurement, and a weight coefficient is introduced to represent the importance degree of each characteristic variable, so that different influences of different variables on the change of the oil spilling track are eliminated, and the similarity S (C 0 ,C ij ) The method comprises the following steps:
wherein C is ij The method is a source case clustering center, wherein ij is the number of each source case in a case library; c (C) 0 Is a target case; omega i The method is used for representing the influence degree of each characteristic variable on sea surface oil spill drift;the kth feature variable value for the jth source case, where k is the feature vector number, k=1, 2,..10; />A kth feature vector value for the target case; max (max) k And min k The maximum and minimum of the kth feature vector, respectively.
8. The prediction algorithm for predicting sea surface oil spill trajectories according to any one of claims 1 to 6, wherein the seventh step comprises validity analysis, wherein calculation examples are set based on publicly published global meteorological data, the calculation examples are respectively solved and predicted by a physical driving model, a data driving model and a hybrid driving model, and validity of the model and the algorithm is verified by comparing the predicted trajectories obtained by solving the real calculation examples with the real oil spill diffusion trajectories.
9. The prediction algorithm for predicting the physical knowledge and data hybrid driving of a sea surface spillover trajectory according to claim 8, wherein the validity analysis is verified by using a Kappa coefficient method, and the calculation formula is as follows:wherein P is A To simulate the actual consistency of the result and the actual result, P e For the practical consistency of the simulation and the practical result, the research area is divided into n grid spaces, a is the grid number of the simulation and the practical result which are both non-oil spilling, b is the grid number of the simulation which is the oil spilling grid and the practical result which is the non-oil spilling grid, c is the grid number of the simulation which is the non-oil spilling grid and the practical result which is the oil spilling grid, d is the grid number of the simulation and the practical result which is the oil spilling grid, and P is A And P e The calculation formula of (2) is as follows:
Binding P A And P e The specific calculation method of Kappa coefficient can be expressed as:
wherein, the value of the Kappa coefficient indicates the consistency degree of the simulation result and the actual result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310416873.XA CN116151487B (en) | 2023-04-19 | 2023-04-19 | Physical knowledge and data hybrid-driven prediction algorithm for predicting sea surface oil spill track |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310416873.XA CN116151487B (en) | 2023-04-19 | 2023-04-19 | Physical knowledge and data hybrid-driven prediction algorithm for predicting sea surface oil spill track |
Publications (2)
Publication Number | Publication Date |
---|---|
CN116151487A CN116151487A (en) | 2023-05-23 |
CN116151487B true CN116151487B (en) | 2023-07-07 |
Family
ID=86358539
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310416873.XA Active CN116151487B (en) | 2023-04-19 | 2023-04-19 | Physical knowledge and data hybrid-driven prediction algorithm for predicting sea surface oil spill track |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116151487B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116499532B (en) * | 2023-06-27 | 2023-09-01 | 中建三局集团华南有限公司 | Complex marine environment deep water pile group construction monitoring system constructed based on hydrologic model |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2014203039A1 (en) * | 2013-06-19 | 2014-12-24 | Aselsan Elektronik Sanayi Ve Ticaret Anonim Sirketi | System and method for implementing reservoir computing using cellular automata |
CN112115598A (en) * | 2020-09-09 | 2020-12-22 | 南京大学 | Data and model coupling analysis method for marine drift trajectory prediction |
CN115544916A (en) * | 2022-10-19 | 2022-12-30 | 中国农业大学 | Method and device for predicting pollutant diffusion of vegetation-containing open channel |
CN115564150A (en) * | 2022-12-06 | 2023-01-03 | 中交天津港湾工程研究院有限公司 | Oil spill track prediction method based on stokes drift and tide mixing mechanism |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
AR109623A1 (en) * | 2018-02-16 | 2019-01-09 | Pescarmona Enrique Menotti | PROCESS AND SYSTEM OF ANALYSIS AND HYDROLOGICAL MANAGEMENT FOR BASINS |
-
2023
- 2023-04-19 CN CN202310416873.XA patent/CN116151487B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2014203039A1 (en) * | 2013-06-19 | 2014-12-24 | Aselsan Elektronik Sanayi Ve Ticaret Anonim Sirketi | System and method for implementing reservoir computing using cellular automata |
CN112115598A (en) * | 2020-09-09 | 2020-12-22 | 南京大学 | Data and model coupling analysis method for marine drift trajectory prediction |
CN115544916A (en) * | 2022-10-19 | 2022-12-30 | 中国农业大学 | Method and device for predicting pollutant diffusion of vegetation-containing open channel |
CN115564150A (en) * | 2022-12-06 | 2023-01-03 | 中交天津港湾工程研究院有限公司 | Oil spill track prediction method based on stokes drift and tide mixing mechanism |
Non-Patent Citations (1)
Title |
---|
基于元胞自动机的海洋溢油模型;满苗苗;柳林;程鹏;魏国忠;张省;;海洋科学(09);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN116151487A (en) | 2023-05-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN114926749B (en) | Near-surface atmospheric pollutant inversion method and system based on remote sensing image | |
Li et al. | Decision-making on process risk of Arctic route for LNG carrier via dynamic Bayesian network modeling | |
CN102222313B (en) | Urban evolution simulation structure cell model processing method based on kernel principal component analysis (KPCA) | |
CN116151487B (en) | Physical knowledge and data hybrid-driven prediction algorithm for predicting sea surface oil spill track | |
CN105653826A (en) | Improved maritime search and rescue region predicting method and system | |
CN111859779B (en) | Method and device for early warning of third party construction damage risk of gas pipe network | |
CN116187210B (en) | Time-space multisource offshore area water quality time sequence prediction method of LSTM coupling mechanism model | |
KR102365072B1 (en) | Apparatus and Method for Improving Wave Prediction Model Accuracy by Advancing Correction of Wind Prediction Data | |
Chiri et al. | Mid-long term oil spill forecast based on logistic regression modelling of met-ocean forcings | |
CN117556197A (en) | Typhoon vortex initialization method based on artificial intelligence | |
Xiong et al. | AI-GOMS: Large AI-Driven Global Ocean Modeling System | |
Dinçer et al. | Enhancing wind turbine site selection through a novel wake penalty criterion | |
Chock et al. | The impact of an 8 h ozone air quality standard on ROG and NOx controls in Southern California | |
CN114385611A (en) | Precipitation prediction method and system based on artificial intelligence algorithm and knowledge graph | |
Zhou et al. | A decision framework of offshore wind power station site selection using a MULTIMOORA method under Pythagorean hesitant fuzzy environment | |
CN113962426A (en) | Offshore water navigation safety intelligent forecasting method and device | |
Gordon et al. | Separating internal and forced contributions to near term SST predictability in the CESM2-LE | |
CN116863097A (en) | Space environment data model construction method and device | |
Tian et al. | Risk assessment of nautical navigational environment based on grey fixed weight cluster | |
Yu et al. | i4OilSpill, an operational marine oil spill forecasting model for Bohai Sea | |
Shahriar et al. | Spatial-temporal prediction of algal bloom | |
Foreback et al. | A new implementation of FLEXPART with Enviro-HIRLAM meteorological input, and a case study during a heavy air pollution event | |
Bogdanović et al. | Winds of Change: A Study on the Resource Viability of Offshore Wind Energy in Montenegro | |
McAllister et al. | Deep learning for wind vector determination | |
CN117828490B (en) | Typhoon disaster forecasting method and system based on ensemble learning |
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 |