CN114912077B - Sea wave forecasting method integrating random search and mixed decomposition error correction - Google Patents
Sea wave forecasting method integrating random search and mixed decomposition error correction Download PDFInfo
- Publication number
- CN114912077B CN114912077B CN202210591957.2A CN202210591957A CN114912077B CN 114912077 B CN114912077 B CN 114912077B CN 202210591957 A CN202210591957 A CN 202210591957A CN 114912077 B CN114912077 B CN 114912077B
- Authority
- CN
- China
- Prior art keywords
- sea wave
- wave
- forecasting
- decomposition
- sequence
- 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
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 58
- 238000012937 correction Methods 0.000 title claims abstract description 25
- 238000013277 forecasting method Methods 0.000 title claims abstract description 24
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 10
- 230000008859 change Effects 0.000 claims abstract description 5
- 238000000556 factor analysis Methods 0.000 claims abstract description 5
- 238000012545 processing Methods 0.000 claims abstract description 4
- 230000035945 sensitivity Effects 0.000 claims abstract description 4
- 230000006870 function Effects 0.000 claims description 20
- 238000013528 artificial neural network Methods 0.000 claims description 18
- 238000000034 method Methods 0.000 claims description 18
- 238000004364 calculation method Methods 0.000 claims description 13
- 238000005070 sampling Methods 0.000 claims description 12
- 238000012360 testing method Methods 0.000 claims description 12
- 230000015654 memory Effects 0.000 claims description 11
- 238000010845 search algorithm Methods 0.000 claims description 10
- 238000012549 training Methods 0.000 claims description 10
- 230000003044 adaptive effect Effects 0.000 claims description 8
- 238000012795 verification Methods 0.000 claims description 8
- 238000010606 normalization Methods 0.000 claims description 7
- 238000010219 correlation analysis Methods 0.000 claims description 6
- VWDWKYIASSYTQR-UHFFFAOYSA-N sodium nitrate Chemical group [Na+].[O-][N+]([O-])=O VWDWKYIASSYTQR-UHFFFAOYSA-N 0.000 claims description 6
- 230000007787 long-term memory Effects 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 4
- 238000004458 analytical method Methods 0.000 claims description 4
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 claims description 3
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 230000003213 activating effect Effects 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 230000003111 delayed effect Effects 0.000 claims description 3
- 230000008034 disappearance Effects 0.000 claims description 3
- 238000004880 explosion Methods 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 claims description 3
- 238000001583 randomness test Methods 0.000 claims description 3
- 125000003275 alpha amino acid group Chemical group 0.000 claims 2
- 150000001875 compounds Chemical class 0.000 claims 2
- 230000006403 short-term memory Effects 0.000 claims 1
- 238000013135 deep learning Methods 0.000 abstract description 11
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 238000010801 machine learning Methods 0.000 description 8
- 230000008569 process Effects 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000007637 random forest analysis Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000004027 cell Anatomy 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000009022 nonlinear effect Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- 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
- 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
- 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
- G06Q50/10—Services
- G06Q50/26—Government or public services
-
- 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)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Tourism & Hospitality (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Development Economics (AREA)
- Computational Mathematics (AREA)
- Marketing (AREA)
- Mathematical Analysis (AREA)
- Operations Research (AREA)
- General Business, Economics & Management (AREA)
- Quality & Reliability (AREA)
- Algebra (AREA)
- Educational Administration (AREA)
- Primary Health Care (AREA)
- Probability & Statistics with Applications (AREA)
- Databases & Information Systems (AREA)
- Game Theory and Decision Science (AREA)
- Evolutionary Biology (AREA)
- Entrepreneurship & Innovation (AREA)
Abstract
A sea wave forecasting method integrating random search and mixed decomposition error correction relates to the technical field of artificial intelligence, and comprises the following steps: step 1, acquiring buoy historical observation data based on buoy observation, integrating and processing ocean and meteorological data closely related to ocean wave change, and constructing an ocean wave forecast database; step 2, sensitivity factor analysis; step 3, constructing a sea wave forecasting model; and 4, acquiring a sea wave forecast error time sequence, and adding the forecast error and the initial forecast wave height to obtain the corrected forecast sea wave height. According to the invention, the input of a numerical forecasting model is referred, data driving is carried out by using relevant elements of sea waves at forecasting time, meanwhile, automatic optimizing is carried out on super parameters of the model to optimize the model structure, and error correction is carried out on sea wave forecasting by combining a data decomposition algorithm and a deep learning mixed decomposition error correction model, so that the sea wave forecasting precision based on deep learning is improved, the forecasting timeliness is improved, and the sea wave forecasting time lag is reduced.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a sea wave forecasting method integrating random search and mixed decomposition error correction.
Background
The coastline is long, the coastline is densely populated, the offshore activities are frequent, and the coastline is one of countries which are affected by sea wave disasters. According to the publication of China ocean disaster bulletin of 2020 by the national ocean agency in 2021, the following is shown as [1]: sea wave disasters are still one of the main components of the ocean disasters in China, and various ocean disasters in China cause direct economic loss of 8.32 hundred million yuan in 2020, and death (including missing) of 6 people. Wherein, personnel death (containing missing) is caused by sea wave disasters. The effective wave height of the offshore ocean wave is more than 4.0 meters, and the disastrous ocean wave process is 36 times, and the ocean wave disasters occur 8 times, so that the ship is damaged by 22.
Therefore, timely and accurate sea wave forecasting has important significance in coastal engineering construction, offshore operation, ship navigation safety and the like. However, since the wave causes are complex and the waves themselves have strong nonlinearities, this makes accurate wave forecasting very difficult. The sea wave forecasting development has three stages of traditional sea wave forecasting, numerical sea wave forecasting and machine learning forecasting. The traditional sea wave forecasting method mainly comprises an empirical formula and a semi-empirical formula, and is widely applied to an effective wave forecasting method, a PNJ spectrum forecasting method, a Wilson empirical forecasting formula, an energy balance derived spectrum forecasting method proposed by the Saint and ordinary academy of China, and the like.
With the improvement of computer computing power and the continuous deepening of people's knowledge of wave mechanisms in the twentieth century, a numerical forecasting method of sea waves is proposed and paid attention to. The numerical forecasting mode continuously breaks through the technical bottleneck from the middle fifty years, so far develops into the third-generation wave numerical mode, breaks through the difficult problem that the wind field cannot be adapted to the rapid change in the past, fully considers the physical processes of wave generation and dissipation including the nonlinear effect between waves, and becomes the main stream mode of the current wave forecasting. The main modes which are widely used at present are WAM mode, SWAN (Surface Wave Models) mode and WAVEWATCHIII mode. The SWAN mode adopts a fully implicit calculation method and considers physical processes such as wave propagation, diffraction, crushing, reflection and the like in shallow water, so that the SWAN mode is usually applied to coastal areas with resolution of 500 meters to tens of kilometers, while WAM and WAVEWATCHIII use explicit or semi-implicit differential technology and consider processes such as surge and wind wave dissipation, wind energy input and the like, so that the SWAN mode is usually applied to ocean scales of more than 20 kilometers.
In recent years, along with the rapid development of machine learning methods, attention of a plurality of domestic and foreign scholars is brought, and a new thought is provided for sea wave forecasting. Machine learning is a cross discipline that includes several disciplines, such as mathematical statistics, higher mathematics, computers, and the like. The machine learning method has good self-adaptive learning and nonlinear mapping capability, and the physical mechanism of things does not need to be clearly understood, so that the machine learning method is suitable for dealing with nonlinear problems that the physical mechanism is complex, and causality and reasoning rules are difficult to determine. In addition, the types and the quantity of the marine meteorological observation data and marine meteorological re-analysis data are greatly increased along with the progress of science and the rapid and rapid progress of computer technology, and the data problem required by machine learning is well solved. Deep learning, which is the most popular subdivision of machine learning in the last decade, has strong advantages in the fields of speech recognition, image classification and the like, solves a plurality of problems which are very difficult in machine learning history for a long time, and has huge application potential in the marine field.
The current deep learning sea wave forecast is mainly divided into two types: one is to use the relevant elements of the sea wave as model input to forecast, such as Fan et al construct an LSTM-based wave prediction model with wave height, wind speed and wind direction as input. The results show that the LSTM algorithm is superior to BP neural networks, extreme Learning Machines (ELMs), support Vector Machines (SVMs), residual networks (res net) and Random Forest (RF) algorithms in terms of wave forecasting. The other is to combine the data decomposition method to take the decomposition subsequence of the history wave height as input, for example, deka et al carries out wave height prediction based on a mixed wave height prediction model combining discrete wavelet decomposition and artificial neural network ANN, and proves that the result is superior to the ANN model.
Although the traditional wave forecasting method plays a certain role in wave forecasting, the traditional wave forecasting method still has certain limitation on the universality and accuracy of forecasting areas and different wave processes. As a mainstream sea wave forecasting model at present, the numerical model is mainly solved based on an energy balance equation, and the method has the problems of large calculated amount, non-convergence and the like, and is difficult to solve in a short time. Although deep learning is widely applied to sea wave forecasting, the deep learning model taking sea wave elements as input usually takes the elements as elements at historical moments, the correlation and influence degree of the elements are far lower than the forecasting moment, and the prediction accuracy and timeliness of the deep learning forecasting model based on the method are lower than those of a forecasting model driven by wind wave elements with the numerical value of the forecasting moment. Although the deep learning forecasting model based on the data decomposition method reduces the nonlinearity of the sea waves, the influence of other factors cannot be considered, and the model can only be based on the self law of the historical sea waves, so that the accuracy of the model is insufficient. Meanwhile, many super parameters in the deep neural network model, such as model layer number, unit number, activation function, optimizer, learning rate and the like, need to be adjusted, and the super parameters cannot be continuously and iteratively optimized through the training process of the neural network, but need to be set manually. Currently, in deep learning research of sea wave forecasting, most of super-parameter adjustment is manually selected. The method for manually adjusting the super parameters not only wastes a great deal of time, but also can not fully exert the performance of the model. The related research of the current sea wave forecast correction model based on deep learning is still very limited, and the main application field is also in the error correction of the data forecast model.
Disclosure of Invention
The invention provides a sea wave forecasting method integrating random search and mixed decomposition error correction, which aims to refer to the input of a numerical forecasting model, data drive sea wave related elements at forecasting time, automatically optimize model super parameters to optimize a model structure, and combine a data decomposition algorithm and a mixed decomposition error correction model for deep learning to perform error correction on sea wave forecasting, so that sea wave forecasting precision based on deep learning is improved, forecasting timeliness is improved, and sea wave forecasting time lag is reduced.
In order to achieve the above purpose, the technical scheme of the invention is as follows:
a sea wave forecasting method integrating random search and mixed decomposition error correction comprises the following steps: step 1, acquiring buoy historical observation data based on buoy observation, integrating and processing ocean and meteorological data closely related to ocean wave change, and constructing an ocean wave forecast database; step 2, sensitivity factor analysis is carried out, and input parameters and step length of the sea wave forecasting model are determined; step 3, constructing a sea wave forecasting model; and 4, acquiring a sea wave forecast error time sequence, and adding the forecast error and the initial forecast wave height to obtain the corrected forecast sea wave height.
Preferably, in the step 1, the wind speed, wind direction, air temperature, air pressure, temperature, humidity, wave height, wave direction and wave period are used as input parameters to be evaluated of the sea wave forecasting model.
Preferably, the step 2 comprises the following specific steps: removing the abnormal values by using a Hampel filter, and defining an outlier as an element which is different from the local median by more than three times of local conversion MAD within the window length designated by window; filling the missing value by applying cubic polynomial interpolation, performing correlation analysis on the wave correlation elements by adopting a pearson correlation coefficient, and normalizing input parameters with different orders of magnitude; determining the optimal input characteristics and corresponding step sizes through an input step size test;
the calculation formula of the pearson correlation coefficient is as follows:
in the formula (1): n is the length of the sequence, o i For the value of the wave height sequence,
as the average value of the wave height sequence, u i For the input parameters to be evaluated, < >>
An average value of the input variables to be evaluated;
the calculation formula for normalizing the input parameters of different orders of magnitude is as follows:
Data nor ＝(Data nor,max -Data nor,min )·(Data-Data min )/(Data max -Data min )+Data nor,min (2)
in the formula (2): data nor For the normalized sequence, data nor，max To normalize the upper interval limit, data nor，min For normalizing the lower limit of the interval, data is the sequence before normalization, data max To normalize the maximum of the pre-sequence, data min Is the minimum value before normalization.
Preferably, in the step 2, the step of determining the input parameters of the sea wave forecast model is as follows:
(a) Carrying out correlation analysis on the wave related elements by adopting a Pearson correlation coefficient, selecting the wave related elements as input parameters to be evaluated according to analysis results, selecting the input parameters to be evaluated with 1-k step sizes and the effective wave height at the past moment as inputs, and inputting the input parameters to be evaluated and the effective wave height into a wave forecasting model;
(b) Training a sea wave forecasting model, and evaluating forecasting results;
(c) If the model accuracy is improved after the input parameters to be evaluated with k step sizes are added, the input parameters to be evaluated are reserved until the accuracy is reduced when the input parameters to be evaluated with k+1 step sizes are input;
(d) The last optimized variable is reserved, and then a new variable with highest correlation in the rest variables is added as an input parameter to be evaluated;
(e) Repeating steps (a) to (d) until all input parameter tests to be evaluated are completed;
(f) And obtaining the input parameters of the sea wave forecasting model according to the test result tested in the steps.
Preferably, the step 3 comprises the following specific steps: and (3) taking the long-short-period memory neural network as a sea wave forecasting model, dividing a training set and a verification set, optimizing super parameters of the sea wave forecasting model by adopting a random search algorithm, and outputting forecasting results.
Preferably, in the step 3, the long-short-term memory neural network solves the problems of gradient disappearance and gradient explosion through a gate unit structure, and the long-term memory neural network is connected with the input gate (i) t ) Forget door (f) t ) Output door (o) t ) The memory unit is used for controlling the information to be selectively stored in the memory unit or forgotten; each gate performs control output by activating a function Sigmoid;
the definition of Sigmoid function is:
i t ＝σ(W i ·[S t-1 ，x t ]+b i ) (4)
f t ＝σ(W f ·[S t-1 ，x t ]+b f ) (5)
o t ＝σ(W o ·[S t-1 ，x t ]+b o (6)
in the formulae (3) to (6), S t-1 Represents the output at time t-1, x t Representing the input of the current time t, sigma represents the Sigmoid activation function, W f 、W i 、W o Weight parameters of forgetting gate, input gate and output gate respectively, b f 、b i 、b o The bias coefficients of the forgetting gate, the input gate and the output gate are respectively.
Preferably, in the step 3, a random search algorithm is adopted to perform super-parameter adjustment to optimize a model structure including the number of layers of the neural network and the number of neurons, and a search strategy is as follows:
(a) Defining a search space;
(b) For the super parameter of which the search range is an interval, randomly sampling according to a given interval; for the super-parameters of a list with limited search range, sampling in a given list with equal probability;
(c) Traversing the niter group sampling result obtained in the step (b); if the given search ranges are all lists, not putting back the sampling niter times;
(d) And comparing the values of the objective functions of the points meeting the constraint conditions one by one, discarding the combination with large error, retaining the combination with small error, and finally obtaining the approximate solution of the optimal solution.
Preferably, the step 4 includes the following specific steps: the method comprises the steps of obtaining a sea wave forecast error time sequence, decomposing the error time sequence based on improved self-adaptive noise set empirical mode decomposition (ICEEMDAN), optimizing a mixed error correction model structure by adopting a random search algorithm, outputting a forecast error after training and verification, and adding the forecast error with an initial forecast wave height to obtain the corrected forecast sea wave height.
Preferably, in the step 4, the predicted value of the sea wave prediction model is set as
The measured data at the corresponding time is
Prediction error E i Expressed as:
performing pure randomness test, namely white noise test, on the error sequence to ensure that the error time sequence has own law; checking by using an autocorrelation coefficient, and if the sequence is white noise, the delayed non-zero period autocorrelation coefficient of the sequence is approximately subjected to an average value of 0; for time series { X t For }, x t And x t-k The definition of the autocorrelation coefficients between them is:
in the expression (8), cov represents covariance, and Var represents variance. When k=0, ρ 0 ＝1；x t And x t-k The meaning of the autocorrelation coefficient of (c) is to give the correlation of the function itself.
Preferably, in the step 4, the improved adaptive noise set empirical mode decomposition (icemdan) is an improved algorithm of Empirical Mode Decomposition (EMD); EMD is proposed by Huang et al in 1998, and is an algorithm for performing self-adaptive decomposition according to data, the method does not need to divide the number of decomposition layers manually, but performs self-adaptive decomposition, so that the study of students is facilitated, the precision of the EMD is also subjected to experimental verification of weather and other field prediction, and through EMD decomposition, the original data can be decomposed into a plurality of Intrinsic Mode Functions (IMFs) and a residual error term; the decomposition steps are as follows:
(a) Solving an upper envelope line and a lower envelope line according to the upper extreme point and the lower extreme point of the original sequence;
(b) Solving the average value of the upper envelope curve and the lower envelope curve to obtain an average value envelope curve;
(c) Subtracting the mean envelope curve from the original data to obtain an intermediate signal;
(d) Judging whether the intermediate signal meets 1) that the number of extreme points and the number of zero crossing points are equal or the difference between the extreme points and the zero crossing points cannot exceed one at most in the whole data segment; 2) At any moment, the average value of the upper envelope formed by the local maximum value points and the lower envelope formed by the local minimum value points is zero, namely the upper envelope and the lower envelope are locally symmetrical relative to a time axis;
(e) If the two conditions in (d) are not satisfied, repeating (a) - (d) based on the intermediate signal; if the condition is met, the IMF component is regarded as an IMF component, the IMF component is subtracted from the original signal to serve as a new original sequence, and the steps (a) - (d) are repeated to obtain IMF2 and … until the remainder contains less than 2 extreme values and no further decomposition is needed;
assuming x as the original signal, E k (. Cndot.) is the kth IMF component obtained by EMD decomposition of the signal, w (i) is the unit variance white noise with zero mean, and M (. Cndot.) is the signal local mean;
the calculation steps of the improved adaptive noise set empirical mode decomposition (icemdan) are as follows:
(a) EMD operation is carried out on the mixed original signals of L superimposed white noise:
x i ＝x+β 0 E 1 (w(i))，i＝1，...，L (9)
β 0 ＝λ 0 std(x)/std(E 1 (w(i))) (10)
in the formula (10), std (. Cndot.) represents the calculated standard deviation, recommended lambda 0 A value of 0.2;
the first remainder calculation formula is:
in the formula (11): r represents a residual component; (b) the 1 st modality component IMF may be expressed as follows:
IMF 1 (t)＝x-r 1 (t) (12)
in the formula (12): x is the original signal, r 1 (t) is a first remainder;
(c) Calculating the kth remainder:
in the formula (13): r is (r) k (t) is the kth remainder
(d) Calculate the kth IMF component:
IMF k (t)＝r k-1 (t)-r k (t)，k＝2，3，... (14)
(e) Repeating (c) and (d) until the remainder no longer requires further decomposition, the decomposition result of the original sequence being expressed as follows:
in formula (15): x (t) is the original signal, IMF is the modal component, r n (t) is the remainder, and n represents the number of IMFs.
The sea wave forecasting method integrating the random search and the mixed decomposition error correction has the beneficial effects that:
1. the invention adopts a sensibility factor analysis and random search algorithm to determine the input parameters and the model structure of the forecasting model, thereby realizing timely and accurate quick forecasting of the wave height of the sea wave.
2. The invention provides a sea wave prediction error correction technology based on mixed decomposition, which reduces the time lag of a model, improves the prediction precision and the prediction timeliness.
Drawings
Fig. 1, technical roadmap of the invention.
Detailed Description
The following detailed description of the embodiments of the present invention in a stepwise manner is provided merely as a preferred embodiment of the present invention, and is not intended to limit the scope of the present invention, but any modifications, equivalents, improvements, etc. within the spirit and principles of the present invention should be included in the scope of the present invention.
In the description of the present invention, it should be noted that, the positional or positional relationship indicated by the terms "upper", "lower", "left", "right", "top", "bottom", "inner", "outer", etc. are based on the positional or positional relationship shown in the drawings, are merely for describing the present invention and simplifying the description, and do not indicate or imply that the apparatus or elements referred to must have a specific orientation, and specific orientation configuration and operation, and thus should not be construed as limiting the present invention.
Please refer to fig. 1:
a sea wave forecasting method integrating random search and mixed decomposition error correction comprises the following steps: step 1, acquiring buoy historical observation data based on buoy observation, integrating and processing ocean and meteorological data closely related to ocean wave change, and constructing an ocean wave forecast database; step 2, sensitivity factor analysis is carried out, and input parameters and step length of the sea wave forecasting model are determined; the method comprises the steps of carrying out a first treatment on the surface of the Step 3, constructing a sea wave forecasting model; and 4, acquiring a sea wave forecast error time sequence to obtain the forecast sea wave height after correction.
In the step 1, the wind speed, wind direction, air temperature, air pressure, temperature, humidity, wave height, wave direction and wave period are used as input parameters to be evaluated of the sea wave forecasting model
The step 2 comprises the following specific steps: removing the abnormal values by using a Hampel filter, and defining an outlier as an element which is different from the local median by more than three times of local conversion MAD within the window length designated by window; filling the missing value by applying cubic polynomial interpolation, wherein the principle is to gradually approximate the function with the minimum point of the cubic curve, thereby obtaining an iterative algorithm of the approximate minimum point of the function to be solved; carrying out correlation analysis on the wave correlation elements by adopting the pearson correlation coefficient, and normalizing input parameters with different orders of magnitude; determining the optimal input characteristics and corresponding step sizes through an input step size test;
the calculation formula of the pearson correlation coefficient is as follows:
in the formula (1): n is the length of the sequence, o i For the value of the wave height sequence,
as the average value of the wave height sequence, u i For the input parameters to be evaluated, < >>
An average value of the input variables to be evaluated;
the calculation formula for normalizing the input parameters of different orders of magnitude is as follows:
Data nor ＝(Data nor,max -Data nor,min )·(Data-Data min )/(Data max -Data min )+Data nor,min (2)
in the formula (2): data nor For the normalized sequence, data nor，max To normalize the upper interval limit, data nor，min For normalizing the lower limit of the interval, data is the sequence before normalization, data max To normalize the maximum of the pre-sequence, data min Is the minimum value before normalization.
In the step 2, the step of determining the input parameters of the sea wave forecasting model is as follows:
(a) Carrying out correlation analysis on the wave related elements by adopting a Pearson correlation coefficient, selecting the wave related elements as input parameters to be evaluated according to analysis results, selecting the input parameters to be evaluated with 1-k step sizes and the effective wave height at the past moment as inputs, and inputting the input parameters to be evaluated and the effective wave height into a wave forecasting model;
(b) Training a sea wave forecasting model, and evaluating forecasting results;
(c) If the model accuracy is improved after the input parameters to be evaluated with k step sizes are added, the input parameters to be evaluated are reserved until the accuracy is reduced when the input parameters to be evaluated with k+1 step sizes are input;
(d) The last optimized variable is reserved, and then a new variable with highest correlation in the rest variables is added as an input parameter to be evaluated;
(e) Repeating steps (a) to (d) until all input parameter tests to be evaluated are completed;
(f) And obtaining the input parameters of the sea wave forecasting model according to the test result tested in the steps.
The step 3 comprises the following specific steps: and (3) taking the long-short-period memory neural network as a sea wave forecasting model, dividing a training set and a verification set, optimizing super parameters of the sea wave forecasting model by adopting a random search algorithm, and outputting forecasting results.
In the step 3, the long-term memory neural network is used as a variant of the circulating neural network, and the problems of gradient disappearance and gradient explosion in the circulating neural network are solved through the unique portal unit structure of the long-term memory neural network; the gate represents a way of screening information into the memory cell by inputting the gate (i t ) Forget door (f) t ) Output door (o) t ) The memory unit is used for controlling the information to be selectively stored in the memory unit or forgotten; each gate performs control output by activating a function Sigmoid; the definition of Sigmoid function is:
i t ＝σ(W i ·[S t-1 ，x t ]+b i ) (4)
f t ＝σ(W f ·[S t-1 ，x t ]+b f ) (5)
o t ＝σ(W o ·[S t-1 ，x t ]+ b o) (6)
in the formulae (3) to (6), S t-1 Indicating time t-1X is the output of (x) t Representing the input of the current time t, sigma represents the Sigmoid activation function, W f 、W i 、W o Weight parameters of forgetting gate, input gate and output gate respectively, b f 、b i 、b o The bias coefficients of the forgetting gate, the input gate and the output gate are respectively.
In the step 3, a random search algorithm is adopted to perform super-parameter adjustment so as to optimize a model structure including the number of layers of the neural network and the number of neurons, and a search strategy is as follows:
(a) Defining a search space;
(b) For the super parameter of which the search range is an interval, randomly sampling according to a given interval; for the super-parameters of a list with limited search range, sampling in a given list with equal probability;
(c) Traversing the niter group sampling result obtained in the step (b); if the given search ranges are all lists, not putting back the sampling niter times;
(d) And comparing the values of the objective functions of the points meeting the constraint conditions one by one, discarding the combination with large error, retaining the combination with small error, and finally obtaining the approximate solution of the optimal solution.
The step 4 comprises the following specific steps: the method comprises the steps of obtaining a sea wave forecast error time sequence, decomposing the error time sequence based on improved self-adaptive noise set empirical mode decomposition (ICEEMDAN), optimizing a mixed error correction model structure by adopting a random search algorithm, outputting a forecast error after training and verification, and adding the forecast error with an initial forecast wave height to obtain the corrected forecast sea wave height.
In the step 4, the predicted value of the sea wave prediction model is set as
The measured data at the corresponding time is +.>
Prediction error E i Expressed as:
performing pure randomness test, namely white noise test, on the error sequence to ensure that the error time sequence has own law; checking by using an autocorrelation coefficient, and if the sequence is white noise, the delayed non-zero period autocorrelation coefficient of the sequence is approximately subjected to an average value of 0; for time series { X t For }, x t And x t-k The definition of the autocorrelation coefficients between them is:
in the expression (8), cov represents covariance, and Var represents variance. When k=0, ρ 0 ＝1；x t And x t-k The meaning of the autocorrelation coefficient of (c) is to give the correlation of the function itself.
In the step 4, the improved adaptive noise set empirical mode decomposition (icemdan) is an improved algorithm of Empirical Mode Decomposition (EMD). EMD was proposed by Huang et al in 1998 as an algorithm for adaptive decomposition based on data. According to the method, the number of decomposition layers is not required to be divided manually, self-adaptive decomposition is adopted, so that the study of students is facilitated, and the accuracy of the method is also subjected to experimental verification of weather and other field prediction. Through EMD decomposition, raw data can be decomposed into many Intrinsic Mode Functions (IMFs) and one residual term. The decomposition steps are as follows:
(a) Solving an upper envelope line and a lower envelope line according to the upper extreme point and the lower extreme point of the original sequence;
(b) Solving the average value of the upper envelope curve and the lower envelope curve to obtain an average value envelope curve;
(c) Subtracting the mean envelope curve from the original data to obtain an intermediate signal;
(d) Judging whether the intermediate signal meets 1) that the number of extreme points and the number of zero crossing points are equal or the difference between the extreme points and the zero crossing points cannot exceed one at most in the whole data segment; 2) At any moment, the average value of the upper envelope formed by the local maximum value points and the lower envelope formed by the local minimum value points is zero, namely the upper envelope and the lower envelope are locally symmetrical relative to a time axis;
(e) If the two conditions in (d) are not satisfied, repeating (a) - (d) based on the intermediate signal; if the condition is met, the IMF component is regarded as an IMF component, the IMF component is subtracted from the original signal to serve as a new original sequence, and the steps (a) - (d) are repeated to obtain IMF2 and … until the remainder contains less than 2 extreme values and no further decomposition is needed;
assuming x as the original signal, E k (. Cndot.) is the kth IMF component obtained by EMD decomposition of the signal, w (i) is the unit variance white noise with zero mean, and M (. Cndot.) is the signal local mean; the calculation steps of the improved adaptive noise set empirical mode decomposition (icemdan) are as follows:
(a) EMD operation is carried out on the mixed original signals of L superimposed white noise:
in the formula (10), std (. Cndot.) represents the calculated standard deviation, recommended lambda 0 A value of 0.2;
the first remainder calculation formula is:
in the formula (11): r represents a residual component; (b) the 1 st modality component IMF may be expressed as follows:
IMF 1 (t)＝x-r 1 (t) (12)
in the formula (12): x is the original signal;
(c) Calculating the kth remainder:
in the formula (13): r is (r) k (t) is the kth remainder;
(d) Calculate the kth IMF component:
IMF k (t)＝r k-1 (t)-r k (t)，k＝2，3，... (14)
(e) Repeating (c) and (d) until the remainder no longer requires further decomposition, the decomposition result of the original sequence being expressed as follows:
in formula (15): x (t) is the original signal, IMF is the modal component, r n (t) is the remainder, and n represents the number of IMFs.
Claims (6)
1. A sea wave forecasting method integrating random search and mixed decomposition error correction is characterized by comprising the following steps: comprising the following steps: step 1, acquiring buoy historical observation data based on buoy observation, integrating and processing ocean and meteorological data closely related to ocean wave change, and constructing an ocean wave forecast database; step 2, sensitivity factor analysis is carried out, and input parameters and step length of the sea wave forecasting model are determined; step 3, constructing a sea wave forecasting model; step 4, acquiring a sea wave forecast error time sequence, and adding the forecast error and the initial forecast wave height to obtain a corrected forecast sea wave height;
in the step 1, the wind speed, wind direction, air temperature, air pressure, temperature, humidity, wave height, wave direction and wave period are used as input parameters to be evaluated of the sea wave forecasting model;
the step 2 comprises the following specific steps: removing the abnormal values by using a Hampel filter, and defining an outlier as an element which is different from the local median by more than three times of local conversion MAD within the window length designated by window; filling the missing value by applying cubic polynomial interpolation, performing correlation analysis on the wave correlation elements by adopting a pearson correlation coefficient, and normalizing input parameters with different orders of magnitude; determining the optimal input characteristics and corresponding step sizes through an input step size test;
the calculation formula of the pearson correlation coefficient is as follows:
in the formula (1): n is the length of the sequence and,
for wave height sequence value, +.>
Is the average value of the wave height sequence, +.>
For the input parameters to be evaluated, < >>
An average value of the input variables to be evaluated;
the calculation formula for normalizing the input parameters of different orders of magnitude is as follows:
in the formula (2):
for the sequence after normalization, ++>
For normalizing the upper limit of the interval, < >>
For the lower limit of the normalized interval, +.>
For the pre-normalization sequence,/->
For maximum value of normalized prosequence, +.>
Is the minimum value before normalization;
in the step 2, the step of determining the input parameters of the sea wave forecasting model is as follows:
(a) Carrying out correlation analysis on the wave related elements by adopting a pearson correlation coefficient, selecting the wave related elements as input parameters to be evaluated according to analysis results, selecting the input parameters to be evaluated with 1~k steps and the effective wave height at the past moment as inputs, and inputting the input parameters to be evaluated and the effective wave height into a wave forecasting model;
(b) Training a sea wave forecasting model, and evaluating forecasting results;
(c) If the model accuracy is improved after the input parameters to be evaluated with k step sizes are added, the input parameters to be evaluated are reserved until the accuracy is reduced when the input parameters to be evaluated with k+1 step sizes are input;
(d) The last optimized variable is reserved, and then a new variable with highest correlation in the rest variables is added as an input parameter to be evaluated;
(e) Repeating steps (a) to (d) until all input parameter tests to be evaluated are completed;
(f) Obtaining input parameters of the sea wave forecasting model according to the test result tested in the steps;
the step 4 comprises the following specific steps: and obtaining a sea wave forecast error time sequence, decomposing the error time sequence based on improved adaptive noise set empirical mode decomposition, optimizing a mixed error correction model structure by adopting a random search algorithm, outputting a forecast error after training and verification, and adding the forecast error and the initial forecast wave height to obtain the corrected forecast sea wave height.
2. The sea wave forecasting method integrating random search and mixed decomposition error correction as claimed in claim 1, wherein the sea wave forecasting method is characterized in that: the step 3 comprises the following specific steps: and (3) taking the long-short-period memory neural network as a sea wave forecasting model, dividing a training set and a verification set, optimizing super parameters of the sea wave forecasting model by adopting a random search algorithm, and outputting forecasting results.
3. The sea wave forecasting method integrating random search and mixed decomposition error correction as claimed in claim 2, wherein the sea wave forecasting method is characterized in that: in the step 3, the long-term and short-term memory neural network solves the problems of gradient disappearance and gradient explosion through a gate type unit structure, and the long-term memory neural network is imported into the gate type neural network
) Forgetting door (+)>
) Output door (+)>
) The memory unit is used for controlling the information to be selectively stored in the memory unit or forgotten; each gate performs control output by activating a function Sigmoid;
the definition of Sigmoid function is:
in the formulae (3) to (6),
represents the output at time t-1, +.>
Input representing the current time t +.>
Representing Sigmoid activation function,/->
、/>
、/>
Weight parameters of forget gate, input gate and output gate respectively, +.>
、/>
、/>
The bias coefficients of the forgetting gate, the input gate and the output gate are respectively.
4. A method of sea wave prediction incorporating a random search and hybrid solution error correction as claimed in claim 3, characterized by: in the step 3, a random search algorithm is adopted to perform super-parameter adjustment so as to optimize a model structure including the number of layers of the neural network and the number of neurons, and a search strategy is as follows:
(a) Defining a search space;
(b) For the super parameter of which the search range is an interval, randomly sampling according to a given interval; for the super-parameters of a list with limited search range, sampling in a given list with equal probability;
(c) Traversing the niter group sampling result obtained in the step (b); if the given search ranges are all lists, not putting back the sampling niter times;
(d) And comparing the values of the objective functions of the points meeting the constraint conditions one by one, discarding the combination with large error, retaining the combination with small error, and finally obtaining the approximate solution of the optimal solution.
5. The sea wave forecasting method integrating random search and mixed decomposition error correction as claimed in claim 4, wherein the sea wave forecasting method is characterized in that: in the step 4, the predicted value of the sea wave prediction model is set as
The measured data at the corresponding time is +.>
Prediction error->
Expressed as:
performing pure randomness test, namely white noise test, on the error sequence to ensure that the error time sequence has own law; checking by using an autocorrelation coefficient, and if the sequence is white noise, the delayed non-zero period autocorrelation coefficient of the sequence is approximately subjected to an average value of 0; for time series
For example, a +>
And->
The definition of the autocorrelation coefficients between them is:
in the formula (8), the amino acid sequence of the compound,
representing covariance +_>
Representing the variance; when->
When (I)>
；/>
And->
The meaning of the autocorrelation coefficient of (c) is to give the correlation of the function itself.
6. The sea wave forecasting method integrating random search and mixed decomposition error correction as claimed in claim 5, wherein the sea wave forecasting method is characterized in that: in the step 4, the improved adaptive noise set empirical mode decomposition is an improved algorithm of empirical mode decomposition, and the decomposition steps of the empirical mode decomposition are as follows:
(a) Solving an upper envelope line and a lower envelope line according to the upper extreme point and the lower extreme point of the original sequence;
(b) Solving the average value of the upper envelope curve and the lower envelope curve to obtain an average value envelope curve;
(c) Subtracting the mean envelope curve from the original data to obtain an intermediate signal;
(d) Judging whether the intermediate signal meets 1) that the number of extreme points and the number of zero crossing points are equal or the difference between the extreme points and the zero crossing points cannot exceed one at most in the whole data segment; 2) At any moment, the average value of the upper envelope formed by the local maximum value points and the lower envelope formed by the local minimum value points is zero, namely the upper envelope and the lower envelope are locally symmetrical relative to a time axis;
(e) If the two conditions in (d) are not satisfied, repeating (a) - (d) based on the intermediate signal; if the condition is met, the IMF component is regarded as an IMF component, the IMF component is subtracted from the original signal to serve as a new original sequence, and (a) - (d) are repeated to obtain IMF2, … until the remainder contains less than 2 extreme values and no further decomposition is needed;
let x be the original signal and,
is the kth IMF component of the signal after EMD decomposition,/I>
Is unit variance white noise with mean value zero, < >>
Is a local average value of the signal;
the calculation steps for improving the adaptive noise set empirical mode decomposition are as follows:
(a) EMD operation is carried out on the mixed original signals of L superimposed white noise:
in the formula (10), the amino acid sequence of the compound,
representing calculated standard deviation>
A value of 0.2;
the first remainder calculation formula is:
in the formula (11): r represents a residual component;
(b) The 1 st modality component IMF is expressed as follows:
in the formula (12): x is the original signal and is the signal,
is the first remainder;
(c) Calculating the kth remainder:
in the formula (13):
is the kth remainder;
(d) Calculate the kth IMF component:
(e) Repeating (c) and (d) until the remainder no longer requires further decomposition, the decomposition result of the original sequence being expressed as follows:
in formula (15):
for the original signal, IMF is the modal component, < ->
As the remainder, n represents the number of IMFs.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210591957.2A CN114912077B (en) | 2022-05-27 | 2022-05-27 | Sea wave forecasting method integrating random search and mixed decomposition error correction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210591957.2A CN114912077B (en) | 2022-05-27 | 2022-05-27 | Sea wave forecasting method integrating random search and mixed decomposition error correction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114912077A CN114912077A (en) | 2022-08-16 |
| CN114912077B true CN114912077B (en) | 2023-06-30 |
Family
ID=82768852
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210591957.2A Active CN114912077B (en) | 2022-05-27 | 2022-05-27 | Sea wave forecasting method integrating random search and mixed decomposition error correction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114912077B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115711612B (en) * | 2022-11-02 | 2024-06-11 | 中国人民解放军国防科技大学 | Prediction method for effective wave height of waves |
| CN115983141B (en) * | 2023-03-21 | 2023-07-14 | 国家海洋局北海预报中心((国家海洋局青岛海洋预报台)(国家海洋局青岛海洋环境监测中心站)) | Deep learning-based method, medium and system for inverting wave height of ocean waves |
| CN117421601B (en) * | 2023-12-19 | 2024-03-01 | 山东省科学院海洋仪器仪表研究所 | Sea surface evaporation waveguide near-future rapid forecasting method |
| CN117909666A (en) * | 2024-03-19 | 2024-04-19 | 青岛哈尔滨工程大学创新发展中心 | Intelligent sea wave correction method and system integrating numerical mode and deep learning |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8423487B1 (en) * | 2010-08-11 | 2013-04-16 | The United States Of America As Represented By The Secretary Of The Navy | Machine learning approach to wave height prediction |
| CN108038577A (en) * | 2017-12-26 | 2018-05-15 | 国家海洋局北海预报中心 | A kind of single station more key element modification methods of wave significant wave height numerical forecast result |
| CN110598170A (en) * | 2019-08-06 | 2019-12-20 | 天津大学 | Data prediction method based on FEEMD decomposition time sequence |
| CN114445634A (en) * | 2022-02-28 | 2022-05-06 | 南京信息工程大学 | Sea wave height prediction method and system based on deep learning model |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107886351A (en) * | 2017-10-18 | 2018-04-06 | 中国地质大学（武汉） | A kind of forecast of crude oil price method and system based on CEEMD PSO BP models and error compensation |
| CN112116162B (en) * | 2020-09-26 | 2022-11-29 | 国家电网公司华中分部 | Power transmission line icing thickness prediction method based on CEEMDAN-QFAOA-LSTM |
| CN112307676B (en) * | 2020-11-04 | 2022-10-14 | 国家海洋局北海预报中心((国家海洋局青岛海洋预报台)(国家海洋局青岛海洋环境监测中心站)) | Wave height numerical prediction model result correction method |
| CN113205226B (en) * | 2021-05-28 | 2022-10-14 | 河北工业大学 | Photovoltaic power prediction method combining attention mechanism and error correction |
| CN113283588B (en) * | 2021-06-03 | 2022-04-19 | 青岛励图高科信息技术有限公司 | Near-shore single-point wave height forecasting method based on deep learning |
| CN113961613A (en) * | 2021-08-25 | 2022-01-21 | 国网上海市电力公司 | Linear prediction method based on periodic filtering |
-
2022
- 2022-05-27 CN CN202210591957.2A patent/CN114912077B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8423487B1 (en) * | 2010-08-11 | 2013-04-16 | The United States Of America As Represented By The Secretary Of The Navy | Machine learning approach to wave height prediction |
| CN108038577A (en) * | 2017-12-26 | 2018-05-15 | 国家海洋局北海预报中心 | A kind of single station more key element modification methods of wave significant wave height numerical forecast result |
| CN110598170A (en) * | 2019-08-06 | 2019-12-20 | 天津大学 | Data prediction method based on FEEMD decomposition time sequence |
| CN114445634A (en) * | 2022-02-28 | 2022-05-06 | 南京信息工程大学 | Sea wave height prediction method and system based on deep learning model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114912077A (en) | 2022-08-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114912077B (en) | Sea wave forecasting method integrating random search and mixed decomposition error correction | |
| Tian | Short-term wind speed prediction based on LMD and improved FA optimized combined kernel function LSSVM | |
| CN111563706A (en) | Multivariable logistics freight volume prediction method based on LSTM network | |
| CN110163433B (en) | Ship flow prediction method | |
| CN111767517B (en) | BiGRU multi-step prediction method, system and storage medium applied to flood prediction | |
| CN113705864A (en) | Weather drought prediction method and device based on VMD-CNN-BilSTM-ATT mixed model | |
| CN112434848B (en) | Nonlinear weighted combination wind power prediction method based on deep belief network | |
| CN111665575B (en) | Medium-and-long-term rainfall grading coupling forecasting method and system based on statistical power | |
| CN113283588B (en) | Near-shore single-point wave height forecasting method based on deep learning | |
| CN114676822B (en) | Multi-attribute fusion air quality forecasting method based on deep learning | |
| CN111626785A (en) | CNN-LSTM network fund price prediction method based on attention combination | |
| CN107292446A (en) | A kind of mixing wind speed forecasting method based on consideration component relevance wavelet decomposition | |
| CN113705877A (en) | Real-time monthly runoff forecasting method based on deep learning model | |
| CN111222992A (en) | Stock price prediction method of long-short term memory neural network based on attention mechanism | |
| Lauguico et al. | A fuzzy logic-based stock market trading algorithm using bollinger bands | |
| CN116187835A (en) | Data-driven-based method and system for estimating theoretical line loss interval of transformer area | |
| CN111141879B (en) | Deep learning air quality monitoring method, device and equipment | |
| CN117239722A (en) | System wind load short-term prediction method considering multi-element load influence | |
| CN115186923A (en) | Photovoltaic power generation power prediction method and device and electronic equipment | |
| Sylaios et al. | A fuzzy inference system for wind-wave modeling | |
| CN113222234A (en) | Gas demand prediction method and system based on integrated modal decomposition | |
| CN117371303A (en) | Prediction method for effective wave height under sea wave | |
| Luo et al. | A novel nonlinear combination model based on support vector machine for stock market prediction | |
| CN116842358A (en) | Soft measurement modeling method based on multi-scale convolution and self-adaptive feature fusion | |
| CN116933152A (en) | Wave information prediction method and system based on multidimensional EMD-PSO-LSTM neural network |
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 |