CN105894090B - A kind of tide intelligence Real-time Forecasting Method based on TSP question particle group optimizing - Google Patents
A kind of tide intelligence Real-time Forecasting Method based on TSP question particle group optimizing Download PDFInfo
- Publication number
- CN105894090B CN105894090B CN201610255994.0A CN201610255994A CN105894090B CN 105894090 B CN105894090 B CN 105894090B CN 201610255994 A CN201610255994 A CN 201610255994A CN 105894090 B CN105894090 B CN 105894090B
- Authority
- CN
- China
- Prior art keywords
- particle
- msub
- mrow
- network
- sapso
- 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.)
- Expired - Fee Related
Links
Classifications
-
- 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
- G06N3/10—Interfaces, programming languages or software development kits, e.g. for simulating neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Theoretical Computer Science (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Genetics & Genomics (AREA)
- Physiology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of tide intelligence Real-time Forecasting Method based on TSP question particle group optimizing, comprise the following steps:It is loaded into tide measured data;Build SAPSO BP network forecasting models;Calculation error function;Loop iteration optimizing;The network parameter of BP neural network is set.The present invention uses for reference the variation thought in Genetic Algorithms, mutation operation is introduced in PSO algorithms, the population search space constantly reduced in an iterative process is expanded, so that particle can jump out the optimal location that prior searches arrive, deploy search in bigger search space, maintain the diversity of population again simultaneously, improve algorithm and search out the more figure of merit and obtain possibility.Therefore, the present invention has higher search precision and search efficiency relative to traditional PSO BP algorithms, has higher precision of prediction relative to traditional PSO BP models and harmonic analysis model.
Description
Technical field
The present invention relates to a kind of forecasting technique of tide, particularly a kind of tide intelligence Real-time Forecasting Method.
Background technology
The learning process of BP (Back Propagation) neutral net, i.e. error backpropagation algorithm, by information just
Formed to propagating with two processes of the backpropagation of error.Each neuron of input layer is responsible for receiving the input information from the external world,
And pass to each neuron in intermediate layer;Intermediate layer is internal information process layer, is responsible for information conversion, according to information change ability
Demand, intermediate layer can be designed as single hidden layer or more hidden layer configurations;Last hidden layer is delivered to each neuron of output layer
Information, after further treatment after, complete the forward-propagating processing procedure that once learns, outwardly output information is handled by output layer
As a result.When reality output and desired output are not inconsistent, into the back-propagation phase of error.Error is by output layer, by error
Each layer weights of mode amendment that gradient declines, to the successively anti-pass of hidden layer, input layer.The information forward-propagating to go round and begin again and error
Back-propagation process, is the process that each layer weights constantly adjust, and the process of neural network learning training, this process one are straight
Untill row is reduced to acceptable degree, or study number set in advance to the error of network output.Although BP networks
Be widely used, but itself there is also some defects and deficiency:Because learning rate is fixed, therefore receipts of network
Speed is held back slowly, it is necessary to the longer training time;The selection of the number of plies and unit number of network hidden layer there is no theoretic guidance;Net
The learning and memory of network has unstability.
Particle swarm optimization algorithm (Particle Swarm Optimization, PSO) is a kind of based on colony's intelligence method
Evolutionary computing, it is mainly used to seek globally optimal solution.Its basic thought is that the potential solution of each optimization problem is that search is empty
Between in a particle, all particles have the adaptive value that an optimised function determines, each particle also has a speed
Degree vector determines their directions for circling in the air and distance, and then particles just follow current optimal particle searching in solution space
Rope.PSO is initialized as a group random particles, then finds optimal solution by iteration.In each iteration, particle passes through tracking
Two extreme values update oneself, and first is exactly preferably solution that particle finds untill current time in itself, and this solution is referred to as
Body best values, another extreme value are exactly the preferably solution that whole population is found to current time, this value be global best fitness (i.e. most
Excellent solution).
PSO-BP particle swarm optimization algorithms are a kind of evolutionary computation techniques, and PSO is similar with genetic algorithm, are a kind of are based on repeatedly
The optimization tool in generation.System initialization is one group of RANDOM SOLUTION, passes through iterated search optimal value.But it does not have the friendship of genetic algorithm
Fork and variation, but particle is followed optimal particle in solution space and scanned for.PSO optimization neural networks include 2 aspects:
1) topological structure of optimization neural network;2) neural network model is trained, mainly between each layer of optimization neural network model
Connection weight and threshold values.Traditional PSO-BP forecast models are using PSO to the connection weight and threshold values between each layer of neutral net
Optimize, with the training quality accelerated the training speed of neutral net with improve neutral net.
System of harmonic analysis:Tide is sea level cyclical upturn and downturn campaign.Motive power caused by tide is celestial body
The vector sum of centrifugal inertial force needed for gravitation and the earth-celestial body relative motion of power to lead tide, i.e. celestial body.Have perhaps in practical application
A variety of methods can be with the lifting information of observational record tidal level, but a kind of most general method is exactly system of harmonic analysis, it
The tide of complexity is resolved into some periodically variable parts, each of which part is all to draw tide by the celestial body of a hypothesis
Partial tide ripple caused by power.Analyzed according to the tide level data of actual observation, can be in the hope of each normal in tide mediation model
Number.Then further according to the obtained harmonic constant of tide, it will be appreciated that the size of partial tide wave component, and can be used to calculate tide simultaneously
Foundation is provided for tidal wave numerical computations.In theory, the part of tide is complicated, the partial tide quantity of tide be it is numerous,
But in engineering calculation, when the mean amplitude of tide of partial tide nighttide is relatively small and the Greenwich delay angle of partial tide (partial tide celestial body process
Angle value corresponding to partial tide high water time occurs to somewhere for somewhere meridian upper transit) it is relatively long when, have significant component of
Tide partial tide composition is negligible.In practical application calculating, the actual tide level in somewhere is represented by:
Wherein:H0For the mean sea level height during analysis, n is tide partial tide quantity, hkFor the amplitude of partial tide, ωkWith
φkIt is the frequency and phase of partial tide respectively, t is any time.
System of harmonic analysis is the most traditional technology of tide prediction aspect, and it is based on tidal static and dynamics, warp
After improvement and development for many years, having been carried out substantially can be to the stabilization forecast of tidal level.However, system of harmonic analysis needs largely
Long-term observation tide level data analysis, can just access relatively accurate harmonic analysis model.But due to long-term scene
Data observation record spends cost too high, therefore typically hardly results in the long-term observation data of these needs.And harmonic analysis
The average prediction error of model is about 20-30cm.In addition, harmonic analysis model only only accounts for the shadow in terms of Between Celestial Tide-generating Forces
Ring, but the generation of tide is influenceed by many time-varying factors such as wind-force, wind direction, air pressure, ocean temperature.Therefore reconcile and divide
Analysis model have ignored time-varying some effects caused by tide, because the factor of tide generation is complicated and changeable, so the entirety of tide
Change shows certain non-linear and uncertainty, traditional static structure harmonic analysis model be difficult carry out high accuracy and
Real-time tidal level forecast.
Traditional PSO optimized algorithms are there is easy Premature Convergence, search precision is relatively low, later stage iteration is inefficient etc. lacks
Point.Neutral net (BP neural network) based on error back propagation is a kind of current most popular, most practical neutral net, but
It is due to that BP neural network is highly susceptible to the topological structure of network and the influence of network size, thus causes the emulation of BP networks
Result of calculation fluctuating change is larger.Meanwhile the shadow that the convergence rate of BP neural network can be selected by network initial weight threshold value
Ring, if initial network parameter selection is improper, it is possible to cause the simulation training of BP neural network can be absorbed in local optimum.For
This, it is proposed that the PSO optimized algorithm forecasting models based on BP neural network.
Traditional PSO-BP mixing forecasting models.It is by particle group optimizing i.e. without improved traditional PS O-BP forecast models
Algorithm (particle swarm optimism algorithm) is applied to the net of BP (back propagation) neutral net
The optimization of network parameter, i.e. weights and threshold value to Optimized BP Neural Network.PSO-BP models are by the weights threshold of BP neural network
It is worth and carries out random initializtion as the population particle position of PSO optimized algorithms, is then counted by error function (fitness function)
Calculate the fitness value of each particle.Judge whether iterative algorithm meets the requirement of iteration optimizing using the fitness value of calculating, most
The best initial weights threshold value through PSO iteration optimizing is obtained afterwards, and best initial weights threshold value is assigned to BP neural network, and it is real to carry out emulation
Test.
But traditional PSO-BP forecasting models still have, and search precision is low, search efficiency is low low with forecast precision asks
Topic.
Document related to the present invention is as follows:
[1]C.P.Tsai and T.L.Lee,Back-propagation neural network in tidal-
level forecasting.Journal of Waterways,Port,Coastal,and Ocean Engineering,125
(4):195-202,1999。
[2]Y.Guo,J.P.Zhang and R.Dai,Marine Navigation,Dalian:Dalian Maritime
University Press,2014。
[3]J.Kennedy and R.Eberhart,Particle Swarm Optimization,Proceedings
of IEEE International Conference on Neural Networks,1942-1948,1995。
[4]J.C.Yin,J.Z.Zou,and F.Xu,Sequential learning radial basis function
network for real-time tidal level predictions,Ocean Engineering,57:49-55,
2013。
[5]Y.Guo,J.P.Zhang and R.Dai,Marine Navigation,Dalian:Dalian Maritime
University Press,2014。
[6]G.Li.Y.L.Hao,and Y.X.Zhao,Research of neural network to tidal
prediction,Proceedings of International Joint Conference on Computational
Science and Optimization,282-284,2009。
[7]S.Yu,K.Zhu,F.Diao,A dynamic all parameters adaptive BP neural
networks model and its application on oil reservoir prediction,Applied
Mathematic and Computation,195:66-75,2008。
The content of the invention
To solve above mentioned problem existing for prior art, the present invention to design that a kind of search precision is high, search efficiency is high and
The high tide intelligence Real-time Forecasting Method based on TSP question particle group optimizing of forecast precision.
To achieve these goals, technical scheme is as follows:It is a kind of based on TSP question particle group optimizing
Tide intelligence Real-time Forecasting Method, comprises the following steps:
A, it is loaded into tide measured data
Described tide measured data derives from the real-time monitored record value of each tidal observation point, and tide is surveyed into number
According to normalized.
B, SAPSO-BP network forecasting models are built
Based on the tide measured data after normalized, create BP neural network model and SAPSO optimized algorithms be set,
Build SAPSO-BP network forecasting models:Mutation operator SA that will be adaptive is introduced into PSO in traditional particle swarm optimization algorithm,
Then the network parameter of BP neural network model is included into weights and threshold value, is initialized as adaptive particle swarm optimization algorithm
SAPSO population particle position.The power that optimal particle position is BP neural network model is obtained by SAPSO iteration optimizing
Value and threshold value, the optimal network parameter that optimizing obtains is assigned to BP neural network model and carries out final network simulation forecast.
BP neural network hidden layer output valve is calculated by equation below:
Wherein:θjFor the threshold value of hidden layer, f () is the nonlinear transfer function of hidden layer node.wijFor input layer with
Weights between hidden layer, n are input layer number, and l is node in hidden layer, xiFor input data.BP network output layers are defeated
Go out value to be calculated by equation below:
Wherein:akOutput layer threshold value, wjkWeights between hidden layer and output layer, m are output layer nodes, HjTo be hidden
Output valve containing layer.
The error calculation formula of SAPSO-BP network forecasting models is as follows:
Wherein YkFor the tide measured data of SAPSO-BP network forecasting models, OkFor SAPSO-BP network forecasting models
Simulation data data, m are output layer nodes.
For SAPSO-BP network forecasting models in iterative process each time, the position of particle and speed more new formula are as follows:
vi(t+1)=ω * vi(t)+c1*r1*(pi-xi(t))+c2*r2*(pg-xi(t)) (4)
xi(t+1)=xi(t)+vi(t+1) j=1,2 ..., n (5)
Wherein ω is inertia weight, and k is current iteration number, xiFor particle position, viFor particle rapidity, PiFor individual pole
Value, PgFor colony's extreme value, c1And c2For nonnegative constant, r1And r2The random number to be between 0 and 1.To prevent particle blindly
Search, is limited the initial position and speed of particle.The network parameter c1=c2=of SAPSO-BP network forecasting models
1.55, iteration optimizing number is 200, and population scale 20, the initial velocity of each particle is limited between [- 3,3], each grain
The initial position of son is limited between [- 5,5], and the adaptive mutation rate formula in SAPSO-BP network forecasting models is as follows:
Pop (j, pos)=λ * rands (1,1) (6)
Wherein:J is number of particles, and pos is a uniform Discrete Stochastic integer.Pop is particle populations quantity, and λ is grain
The maximum of sub- population quantity.
The network parameter of BP neural network is initialized as to the particle populations position of SAPSO optimized algorithms, BP nerve nets C,
The network parameter on road includes:The weights between weights, hidden layer threshold value, hidden layer and output layer between input layer and hidden layer
And output layer threshold value.The initial fitness function value of particle is calculated according to error function formula.Error function calculation formula is as follows:
Error=| Yk-Ok| (7)
Wherein:YkFor the actual observation data of forecasting model, OkFor the simulation data data of forecasting model.
D, in iterative process each time, particle updates speed and the position of itself according to more new formula (4) and (5).And
New fitness function value is calculated according to error function calculation formula (7).Then the variation thought in genetic algorithm is used for reference,
In PSO algorithms introduce TSP question operation, then according to TSP question formula (6) calculate and more new particle individual extreme value with
Colony's extreme value.
E, judge whether the error function value that optimum individual fitness function value i.e. error function formula (7) calculates meets to miss
Whether poor setting requirements, or iteration optimizing number reach setting requirements, and step F is performed if meeting to require, is otherwise returned
Step D continues loop iteration optimizing.
F, the iteration optimizing of SAPSO optimized algorithms is terminated, the optimal network parameter assignment that optimizing is obtained gives BP nerve nets
Network carries out emulation experiment.
The network parameter of BP neural network is set G,:Iterative cycles number is arranged between [1,500], learning rate and study
Target is all disposed between [0,1], and optimal network parameter then is assigned into BP neural network carries out tide real-time prediction emulation in fact
Test.
Iterative cycles number described in step G of the present invention is 100, learning rate 0.1, learning objective 0.00001.
Compared with prior art, the invention has the advantages that:
Particle swarm optimization algorithm PSO convergences are fast, have a very strong versatility, but simultaneously there is easy Premature Convergence, search
The shortcomings of Suo Jingdu is low, later stage iteration is inefficient.The present invention uses for reference the change in Genetic Algorithms (Genetic Algorithm)
Different thought, mutation operation is introduced in PSO algorithms, i.e., some variables are reinitialized with certain probability.Mutation operation is opened up
The population search space constantly reduced in an iterative process is opened up, so that particle can jump out the optimal position that prior searches arrive
Put, deploy search in bigger search space, while maintain the diversity of population again, improve algorithm and search out the more figure of merit
Obtain possibility.Therefore, mutation operator is introduced on the basis of ordinary particle group's algorithm, basic thought is updated every time in particle
Afterwards, particle is reinitialized with certain probability.The real-time observed data at same harbour is selected as network forecasting model
Input data is trained, and emulation experiment is carried out with harmonic analysis model, PSO-BP network models, and SAPSO-BP models.
It can show that the fitness curve decrease speed of SAPSO-BP forecast models is substantially faster than PSO-BP by Fig. 2 com-parison and analysis, and
And fall is also relatively large.Therefore, SAPSO-BP algorithms have higher search essence relative to traditional PSO-BP algorithms
Degree and search efficiency.By Fig. 3 com-parison and analysis SAPSO-BP models can be drawn relative to traditional PSO-BP models and mediation
Analysis model has higher precision of prediction.
Brief description of the drawings
The shared accompanying drawing 3 of the present invention is opened, wherein:
Fig. 1 is the schematic flow sheet of the present invention.
Fig. 2 be two kinds of Forecasting Methodologies fitness function value curve ratio compared with.
Fig. 3 is the precision of prediction contrast of three kinds of different Forecasting Methodologies.
Embodiment
The present invention is further described through below in conjunction with the accompanying drawings.
As shown in Fig. 2 calculated according to the flow shown in Fig. 1, SAPSO-BP network forecasting model fitness function values
Fall off rate it is substantially faster without improved PSO-BP networks forecasting model than tradition, that is to say, that SAPSO-BP networks forecast mould
The error reduction speed of type is very fast, later stage iteration efficiency high.And its error amount when tending towards stability also is significantly less than tradition not
The error amount of improved PSO-BP network forecasting models.In addition can from the fitness value curve of two kinds of network forecasting models
Go out, the local optimum that iteration optimizing obtains before the particle swarm optimization algorithm with mutation operator can be jumped out rapidly, so as to
To more excellent result.
As shown in Figure 3:Three kinds of heterogeneous networks forecasting models are using identical training data under identical simulated environment
Precision of prediction contrasts.The prediction error that system of harmonic analysis is can obtain by Fig. 3 changes fluctuation, Er Qieqi between -0.05 to -0.25
Change fluctuating range is larger and mean error is larger, and it is difficult that stable progress is high to illustrate traditional system of harmonic analysis forecasting model
The tide prediction of precision.The prediction error of traditional PSO-BP network forecasting models changes fluctuation, phase between 0.1 to -0.05
There is very big decline than system of harmonic analysis to fluctuating range, and its fluctuating error scope there has also been certain diminution, it is average to miss
Difference is gradually to 0 convergence.The prediction error of SAPSO-BP network forecasting models changes fluctuation between 0.05 to -0.05, and it is fluctuated
Amplitude has obvious diminution relative to first two network forecasting model, and fluctuating range is compared to more steady, mean error base
Originally 0 is tended to.It can be inferred that:The prediction error of the SAPSO-BP network forecasting models of SAPSO-BP network forecasting models is
It is minimum in three kinds of forecasting models, and its prediction error amplitude of variation is also minimum, therefore improved TSP question grain
Subgroup Optimized BP Neural Network model can carry out stable high-precision tide intelligence real-time prediction.
Claims (2)
- A kind of 1. tide intelligence Real-time Forecasting Method based on TSP question particle group optimizing, it is characterised in that:Including following Step:A, it is loaded into tide measured data:Described tide measured data derives from the real-time monitored record value of each tidal observation point, and tide measured data is returned One change is handled;B, SAPSO-BP network forecasting models are built:Based on the tide measured data after normalized, create BP neural network model and SAPSO optimized algorithms are set, build SAPSO-BP network forecasting models:Mutation operator SA that will be adaptive is introduced into PSO in traditional particle swarm optimization algorithm, then The network parameter of BP neural network model is included into weights and threshold value, is initialized as adaptive particle swarm optimization algorithm SAPSO's Population particle position;The weights and threshold that optimal particle position is BP neural network model are obtained by SAPSO iteration optimizing Value, the optimal network parameter that optimizing obtains is assigned to BP neural network model and carries out final network simulation forecast;BP neural network hidden layer output valve is calculated by equation below:<mrow> <msub> <mi>H</mi> <mi>j</mi> </msub> <mo>=</mo> <mi>f</mi> <mrow> <mo>(</mo> <mover> <mo>&Sigma;</mo> <mi>n</mi> </mover> <msub> <mi>w</mi> <mrow> <mi>i</mi> <mi>j</mi> </mrow> </msub> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>-</mo> <msub> <mi>&theta;</mi> <mi>j</mi> </msub> <mo>)</mo> </mrow> <mo>,</mo> <mi>j</mi> <mo>=</mo> <mn>1</mn> <mo>,</mo> <mn>2</mn> <mo>,</mo> <mn>...</mn> <mo>,</mo> <mi>l</mi> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>)</mo> </mrow> </mrow>Wherein:θjFor the threshold value of hidden layer, f () is the nonlinear transfer function of hidden layer node;wijBe input layer with it is implicit Weights between layer, n are input layer number, and l is node in hidden layer, xiFor input data;BP network output layer output valves Calculated by equation below:<mrow> <msub> <mi>O</mi> <mi>k</mi> </msub> <mo>=</mo> <munderover> <mo>&Sigma;</mo> <mrow> <mi>j</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>t</mi> </munderover> <msub> <mi>H</mi> <mi>j</mi> </msub> <msub> <mi>w</mi> <mrow> <mi>j</mi> <mi>k</mi> </mrow> </msub> <mo>-</mo> <msub> <mi>a</mi> <mi>k</mi> </msub> <mo>,</mo> <mi>k</mi> <mo>=</mo> <mn>1</mn> <mo>,</mo> <mn>2</mn> <mo>,</mo> <mn>...</mn> <mo>,</mo> <mi>m</mi> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>2</mn> <mo>)</mo> </mrow> </mrow>Wherein:akOutput layer threshold value, wjkWeights between hidden layer and output layer, m are output layer nodes, HjFor hidden layer Output valve;The error calculation formula of SAPSO-BP network forecasting models is as follows:<mrow> <msub> <mi>e</mi> <mi>k</mi> </msub> <mo>=</mo> <mfrac> <mn>1</mn> <mn>2</mn> </mfrac> <munderover> <mo>&Sigma;</mo> <mi>k</mi> <mi>m</mi> </munderover> <msup> <mrow> <mo>(</mo> <msub> <mi>Y</mi> <mi>k</mi> </msub> <mo>-</mo> <msub> <mi>O</mi> <mi>k</mi> </msub> <mo>)</mo> </mrow> <mn>2</mn> </msup> <mo>,</mo> <mi>k</mi> <mo>=</mo> <mn>1</mn> <mo>,</mo> <mn>2</mn> <mo>,</mo> <mn>...</mn> <mo>,</mo> <mi>m</mi> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>3</mn> <mo>)</mo> </mrow> </mrow>Wherein YkFor the tide measured data of SAPSO-BP network forecasting models, OkFor the emulation of SAPSO-BP network forecasting models Output data, m are output layer nodes;For SAPSO-BP network forecasting models in iterative process each time, the position of particle and speed more new formula are as follows:vi(t+1)=ω * vi(t)+c1*r1*(pi-xi(t))+c2*r2*(pg-xi(t)) (4)xi(t+1)=xi(t)+vi(t+1) j=1,2, n (5)Wherein ω is inertia weight, and k is current iteration number, xiFor particle position, viFor particle rapidity, PiFor individual extreme value, Pg For colony's extreme value, c1And c2For nonnegative constant, r1And r2The random number to be between 0 and 1;To prevent particle blind search, The initial position and speed of particle are limited;The network parameter c1=c2=1.55 of SAPSO-BP network forecasting models, repeatedly It is 200 for optimizing number, population scale 20, the initial velocity of each particle is limited between [- 3,3], at the beginning of each particle For beginning position limitation between [- 5,5], the adaptive mutation rate formula in SAPSO-BP network forecasting models is as follows:Pop (j, pos)=λ * rands (1,1) (6)Wherein:J is number of particles, and pos is a uniform Discrete Stochastic integer;Pop is particle populations quantity, and λ is particle kind The maximum of group's quantity;The network parameter of BP neural network is initialized as to the particle populations position of SAPSO optimized algorithms C, BP neural network Network parameter includes:Weights between weights, hidden layer threshold value, hidden layer and output layer between input layer and hidden layer and Output layer threshold value;The initial fitness function value of particle is calculated according to error function formula;Error function calculation formula is as follows:Error=| Yk-Ok| (7)Wherein:YkFor the actual observation data of forecasting model, OkFor the simulation data data of forecasting model;D, in iterative process each time, particle updates speed and the position of itself according to more new formula (4) and (5);And according to Error function calculation formula (7) calculates new fitness function value;Then the variation thought in genetic algorithm is used for reference, is calculated in PSO TSP question operation is introduced in method, then according to TSP question formula (6) calculating and more new particle individual extreme value and colony Extreme value;E, judge whether the error function value that optimum individual fitness function value i.e. error function formula (7) calculates meets that error is set Requirement is put, or whether iteration optimizing number reaches setting requirements, step F is performed if meeting to require, otherwise return to step D Continue loop iteration optimizing;F, terminate the iteration optimizing of SAPSO optimized algorithms, the optimal network parameter assignment that optimizing obtains is entered to BP neural network Row emulation experiment;The network parameter of BP neural network is set G,:Iterative cycles number is arranged between [1,500], learning rate and learning objective It is all disposed between [0,1], optimal network parameter then is assigned into BP neural network carries out tide real-time prediction emulation experiment.
- 2. a kind of tide intelligence Real-time Forecasting Method based on TSP question particle group optimizing according to claim 1, It is characterized in that:In step G, described iterative cycles number is 100, learning rate 0.1, learning objective 0.00001.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610255994.0A CN105894090B (en) | 2016-04-22 | 2016-04-22 | A kind of tide intelligence Real-time Forecasting Method based on TSP question particle group optimizing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610255994.0A CN105894090B (en) | 2016-04-22 | 2016-04-22 | A kind of tide intelligence Real-time Forecasting Method based on TSP question particle group optimizing |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105894090A CN105894090A (en) | 2016-08-24 |
CN105894090B true CN105894090B (en) | 2018-02-06 |
Family
ID=56704506
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610255994.0A Expired - Fee Related CN105894090B (en) | 2016-04-22 | 2016-04-22 | A kind of tide intelligence Real-time Forecasting Method based on TSP question particle group optimizing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105894090B (en) |
Families Citing this family (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106960259B (en) * | 2017-03-27 | 2020-07-28 | 宁波市水利水电规划设计研究院有限公司 | Tidal river reach bidirectional wave water-withdrawal process forecasting method, device and system |
CN109086933A (en) * | 2018-08-03 | 2018-12-25 | 中山大学 | A kind of salty tide Medium-long Term Prediction method |
CN109190270B (en) * | 2018-09-12 | 2022-12-27 | 北京化工大学 | APSO-BP-based double-counterweight-disc automatic balance control method |
CN111414807B (en) * | 2020-02-28 | 2024-02-27 | 浙江树人学院(浙江树人大学) | Tidal water identification and crisis early warning method based on YOLO technology |
CN113158556B (en) * | 2021-03-31 | 2023-08-08 | 山东电力工程咨询院有限公司 | Short-time high-precision forecasting method for regional water level |
CN113777000B (en) * | 2021-10-09 | 2024-04-12 | 山东科技大学 | Dust concentration detection method based on neural network |
CN114236401B (en) * | 2021-12-20 | 2023-11-28 | 上海正泰电源系统有限公司 | Battery state estimation method based on self-adaptive particle swarm algorithm |
CN117574213B (en) * | 2024-01-15 | 2024-03-29 | 南京邮电大学 | APSO-CNN-based network traffic classification method |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102214262A (en) * | 2010-04-02 | 2011-10-12 | 上海海洋大学 | Tide predicting method |
CN103871002A (en) * | 2014-03-25 | 2014-06-18 | 上海电机学院 | Wind power forecast method and device based on self-adaptation bee colony algorithm |
CN104376230A (en) * | 2014-12-03 | 2015-02-25 | 大连海事大学 | Tidal prediction method |
-
2016
- 2016-04-22 CN CN201610255994.0A patent/CN105894090B/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102214262A (en) * | 2010-04-02 | 2011-10-12 | 上海海洋大学 | Tide predicting method |
CN103871002A (en) * | 2014-03-25 | 2014-06-18 | 上海电机学院 | Wind power forecast method and device based on self-adaptation bee colony algorithm |
CN104376230A (en) * | 2014-12-03 | 2015-02-25 | 大连海事大学 | Tidal prediction method |
Also Published As
Publication number | Publication date |
---|---|
CN105894090A (en) | 2016-08-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105894090B (en) | A kind of tide intelligence Real-time Forecasting Method based on TSP question particle group optimizing | |
CN114625151B (en) | Underwater robot obstacle avoidance path planning method based on reinforcement learning | |
CN103744428A (en) | Unmanned surface vehicle path planning method based on neighborhood intelligent water drop algorithm | |
CN114077256A (en) | Overwater unmanned ship path planning method | |
CN110278571A (en) | It is a kind of based on simple forecast-correction link distributed signal tracking | |
CN116151102A (en) | Intelligent determination method for space target ultra-short arc initial orbit | |
Vasechkina et al. | Evolving polynomial neural network by means of genetic algorithm: some application examples | |
CN114037050B (en) | Robot degradation environment obstacle avoidance method based on internal plasticity of pulse neural network | |
Das et al. | Evolving fuzzy reasoning approach using a novel nature-inspired optimization tool | |
Mojarrad et al. | Particle swarm optimization with chaotic velocity clamping (CVC-PSO) | |
Panda et al. | Model reduction of linear systems by conventional and evolutionary techniques | |
Brintaki et al. | Coordinated UAV path planning using differential evolution | |
Lin et al. | Hybrid differential evolution and particle swarm optimization algorithm based on random inertia weight | |
La et al. | Optimal flocking control for a mobile sensor network based a moving target tracking | |
Sreejeth et al. | Tuning A PID Controller using Metaheuristic Algorithms | |
Wang et al. | Parameters tuning of fuzzy controller for rotated pendulum based on improved particle swarm optimization | |
CN115016248A (en) | Motor PID control method for optimizing RBF neural network based on PSO algorithm | |
KhalafAnsar et al. | Adaptive Inverse Deep Reinforcement Lyapunov learning control for a floating wind turbine | |
Wang et al. | Parameters optimization of fuzzy controller based on improved particle swarm optimization | |
Doudkin et al. | Spacecraft Telemetry Time Series Forecasting With Ensembles of Neural Networks | |
Yang et al. | Research on fire rescue path optimization of unmanned equipment based on improved Slime mould Algorithm | |
Wang et al. | Prediction of Chlorophyll-a content using hybrid model of least squares support vector regression and radial basis function neural networks | |
Xu et al. | A combined training algorithm for RBF neural network based on particle swarm optimization and gradient descent | |
Shim et al. | Crack identification of a planar frame structure based on a synthetic artificial intelligence technique | |
Jihong et al. | Groundwater table prediction based on improved PSO Algorithm and RBF Neural Network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180206 Termination date: 20210422 |