CN111353640B - Method for constructing wind speed prediction model by combination method - Google Patents
Method for constructing wind speed prediction model by combination method Download PDFInfo
- Publication number
- CN111353640B CN111353640B CN202010120902.4A CN202010120902A CN111353640B CN 111353640 B CN111353640 B CN 111353640B CN 202010120902 A CN202010120902 A CN 202010120902A CN 111353640 B CN111353640 B CN 111353640B
- Authority
- CN
- China
- Prior art keywords
- wind speed
- prediction
- prediction model
- model
- neural network
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/067—Enterprise or organisation modelling
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Biophysics (AREA)
- Economics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- General Health & Medical Sciences (AREA)
- General Business, Economics & Management (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Game Theory and Decision Science (AREA)
- Evolutionary Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Development Economics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Tourism & Hospitality (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Educational Administration (AREA)
- Physiology (AREA)
- Genetics & Genomics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method for constructing a wind speed prediction model by a combination method, which comprises the following steps: s1: performing wavelet decomposition on the original wind speed to obtain a decomposed wind speed value, obtaining a decomposition predicted value, combining the predicted values, and finally obtaining an integral wind speed predicted value based on a wavelet decomposition time series method; s2: EEMD decomposition is carried out on the original wind speed to obtain a plurality of data, one-step prediction is carried out in advance to obtain each subsequence prediction value, and the prediction values are overlapped to obtain an integral wind speed prediction value; s3: taking the prediction results of two models, namely a wavelet decomposed time sequence wind speed prediction result and an EEMD decomposed GA-BP wind speed prediction result, as input values of a BP neural network, taking an actually measured wind speed value at a corresponding moment as an output value of the network, and obtaining a combined method prediction model through training; by comprehensively utilizing the wind speed information provided by various single prediction results and combining the single prediction results, a new prediction model with higher precision is constructed.
Description
Technical Field
The invention relates to the field of high-speed railway wind speed prediction, in particular to a method for constructing a wind speed prediction model by a combination method.
Background
Under the background of continuous expansion of the network coverage area of the high-speed railway, the mileage of the high-speed railway in a strong wind area is longer and longer, higher requirements are also put forward on the running safety of the train due to continuous improvement of the running speed of the high-speed train, and when the train passes through the strong wind area, the air pressure on the surface of the train is changed under the action of wind force, so that the aerodynamic performance of the train is deteriorated, and the transverse stability and the running safety of the train are seriously influenced. Aiming at the running safety accidents of trains, high-speed rail disaster prevention and reduction systems with 73 lines are built in China, wherein a high-speed rail disaster prevention network capable of monitoring wind, rain, snow and earthquake conditions along the railway in real time is formed at a strong wind monitoring point 2612, a rainfall monitoring point 1432, a snow depth monitoring point 126 and an earthquake monitoring point 1180. At present, a great amount of wind, rain, snow and earthquake original monitoring data are accumulated in a high-speed railway disaster prevention network, however, the data are mainly used for analyzing disaster accidents, and are not more effectively analyzed and utilized, for example, a high-speed railway disaster forecast and early warning system is perfected.
The wind speed prediction models adopted in the prior art mainly comprise a time series wind speed prediction model and a neural network wind speed prediction model, and at present, the two wind speed prediction models are independently predicted. In wind speed prediction, if a single prediction model is selected for prediction, the selected model must adapt to the characteristics of the current data, otherwise, there is a risk of selection error, which may result in prediction failure. Meanwhile, different single prediction models can extract useful information from raw data from different angles, so that if a certain result with a large prediction error is discarded in prediction, a part of the useful information can be lost. Under the circumstances, how to establish an accurate wind speed prediction model on the basis of keeping all useful information becomes an urgent problem to be solved in the field of high-speed railway wind speed prediction.
Disclosure of Invention
The invention aims to overcome the defect that a single wind speed prediction model in the prior art possibly loses part of useful original information to cause poor prediction result precision, and provides a combined method for constructing a wind speed prediction model.
The purpose of the invention is mainly realized by the following technical scheme:
a method for constructing a wind speed prediction model by using a combination method comprises the following steps:
s1: performing wavelet decomposition on the original wind speed to obtain a decomposed wind speed value, predicting each decomposed wind speed by one step in advance by using a model to obtain a decomposed predicted value, and combining the predicted values to finally obtain an integral wind speed predicted value based on a time series method of wavelet decomposition;
s2: EEMD decomposition is carried out on the original wind speed to obtain a plurality of eigenmode functions imf and residual information r, a BP neural network optimized by a genetic algorithm is used for carrying out one-step prediction on each decomposed wind speed in advance to obtain each subsequence predicted value, and the predicted values are overlapped to obtain an integral wind speed predicted value;
s3: the results of the wavelet-time sequence prediction model and the EEMD-GA-BP neural network prediction model are used as input vectors of a neural network; smoothing the prediction data of the wavelet-time sequence prediction model and the EEMD-GA-BP neural network prediction model, namely averaging the two prediction results to be used as the input vector of the combined prediction model; and processing the predicted value and establishing a GA-BP neural network combined prediction model.
In the prior art, a time series in a time series wind speed prediction model is a sequence formed by arranging numerical values of a concerned index variable at different moments according to the sequence of time, the sequence comprises sequence information of data and also covers the inherent characteristics of the data, and mathematical statistics relations exist among the data, so that future values of the sequence can be predicted through past values and present values of the sequence, and a currently common stationary time series analysis model comprises the following steps: the system comprises an autoregressive model, a moving average model and an autoregressive moving average model, wherein a non-stationary time series analysis model is a common summation autoregressive moving average model; the BP neural network wind speed prediction model is constructed based on an artificial neural network, the artificial neural network is an algorithmic mathematical model with information processing capability based on the principle of a biological neural network, the artificial neural network is similar to the biological neural network, the basic unit of the artificial neural network is also a neuron, a neuron structure is generally provided with a plurality of inputs and an output, each input is provided with a corresponding weight, the output of the neuron is obtained by acting a weighted summation result on a transfer function and then is transferred to the next neuron structure, a large number of neurons are connected with each other to form the neural network, the BP neural network mainly comprises two processes of forward transfer of information and backward transfer of errors when in operation, in the forward transfer process of the information, the information enters from an input layer and is output by an output layer after being processed by an implicit layer, when the output result of the output layer is greatly different from the expected result, performing a back propagation process of errors, firstly calculating the errors of the expected result and the output result, then returning to the input layer from the output layer through the hidden layer, adjusting the weight of each connection one by one in the process, continuously adjusting each connection weight through multiple times of information forward transmission and error back propagation until the output result meets the requirement of error precision or the number of times of execution reaches a preset number of times, and stopping the training and learning of the neural network at this moment; in the invention, the data of the original wind speed is decomposed into two parts of low-frequency information and high-frequency information through wavelet decomposition in the step S1, the low-frequency information changes slowly, the high-frequency information changes rapidly, and the decomposed wind speed value is obtained through sorting; the EEMD is ensemble empirical mode decomposition, provides a noise auxiliary data analysis method aiming at the defects of the EMD method, and has the principle that when additional white noise is uniformly distributed in the whole time-frequency space, the time-frequency space is composed of different scale components divided by a filter bank; the results of the wavelet-time sequence prediction model and the EEMD-GA-BP neural network prediction model are used as input vectors of a neural network; smoothing the prediction data of the wavelet-time sequence prediction model and the EEMD-GA-BP neural network prediction model, namely averaging the two prediction results to be used as the input vector of the combined prediction model; the method comprises the steps of processing a predicted value, establishing a GA-BP neural network combined prediction model, and enabling the predicted value of wind speed prediction obtained through the combined model to be closer to a true value than the predicted values obtained through two single prediction models, namely the combined prediction model established by the invention can improve prediction accuracy to a certain extent.
Further, the step S1 includes:
s1.1: obtaining raw wind speed data
Decomposing and reconstructing by using db3 wavelet to obtain a trend signal a3
Represents; detail signals d1, d2, d3, respectively
Represents;
s1.2: respectively calculating an autocorrelation function and a partial autocorrelation function of the trend signal a3 and the detail signals d1, d2 and d3, checking the stationarity of the autocorrelation function and the partial autocorrelation function, and if the signals are non-stationary signals, carrying out differential processing on the signals to obtain stationary time series signals;
s1.3: according to the tailing and truncation of the autocorrelation function and the partial autocorrelation function obtained by each trend signal and each detail signal, respectively selecting a proper model for each trend signal and each detail signal and determining the order of the model by combining with the AIC criterion;
s1.4: respectively carrying out parameter estimation on parameters in each model by using a least square estimation method;
s1.5: checking the effectiveness of each model, if the model does not meet the requirements, returning to the step S1.3, and reselecting a proper model and a proper order until an optimal model is selected;
s1.6: and (4) carrying out wind speed prediction on a fitting model established for each trend signal and each detail signal to obtain a wind speed predicted value
S1.7: each wind speed predicted value is superposed and combined to obtain a final wind speed predicted value
And comparing the measured wind speed data with the measured wind speed data to obtain a prediction error.
Further, the step S2 includes:
s2.1: decomposing the original wind speed time series signal into a series of eigenmode functions and a residual information by utilizing integrated empirical mode decomposition;
s2.2: respectively establishing three layers of BP neural network structures based on the decomposed sequences, optimizing the initial weight and the threshold of the neural network by using a genetic algorithm, searching an optimal individual to initialize each layer of neurons of the BP neural network by using a training error as a fitness function of the genetic algorithm when optimizing the initial weight and the threshold of the BP neural network by using the genetic algorithm, and then carrying out network training;
s2.3: and superposing the prediction results of each sequence to obtain a final wind speed prediction value.
Further, in step S2, the EEMD decomposition algorithm includes the following steps:
A. after the original time sequence is obtained, a Gaussian white noise sequence is added to obtain a new sequence;
B. performing EMD decomposition on the obtained new sequence to obtain a decomposition result;
C. sorting the decomposition results into k imf air quantities and one residual information;
D. and averaging the decomposed sequences and obtaining a result.
When EEMD decomposition is adopted, when the attached white noise is uniformly distributed in the whole time-frequency space, the time-frequency space is composed of components with different scales which are divided by a filter bank, and when the white noise background which is uniformly distributed is added to the signal, signal areas with different scales are automatically mapped to a proper scale related to the background white noise; of course, each individual test may produce very noisy results because each additive noise component includes both the signal and the additive white noise; since the noise is different in each individual test, when using the ensemble of means sufficient for the test, the noise will be cancelled, the ensemble of means will eventually be considered as the real result, with more and more tests, the additional noise is cancelled, the only persistent and robust part is the signal itself, so the wind speed time series signal processed by the EEMD decomposition algorithm is more accurate and stable.
Further, the step S3 includes:
s3.1: the results of the wavelet-time sequence prediction model and the EEMD-GA-BP neural network prediction model are used as input vectors of a neural network;
s3.2: smoothing the prediction data of the wavelet-time sequence prediction model and the EEMD-GA-BP neural network prediction model, namely averaging the two prediction results to be used as the input vector of the combined prediction model;
s3.3: and processing the predicted value and establishing a GA-BP neural network combined prediction model.
When combined prediction is carried out, firstly, input and output parameters of a neural network need to be determined, and in order to fully utilize wind speed information extracted by each single prediction model, the results of all the single prediction models are generally used as input vectors of the neural network; the single prediction model selected in the invention is the wavelet-time sequence prediction model and the EEMD-GA-BP neural network prediction model; considering that the prediction error of a certain prediction model may be relatively large at a certain time, in order to reduce such adverse effects, the two single prediction data are smoothed, that is, the two prediction results are averaged and also used as one of the inputs of the combined prediction model.
In conclusion, compared with the prior art, the invention has the following beneficial effects:
(1) the invention constructs a new prediction model with higher precision by comprehensively utilizing the wind speed information provided by various single prediction results and giving reasonable weight to each single prediction model and combining the single prediction results, thereby comprehensively utilizing various single prediction results and extracting more useful information from the original data.
(2) When EEMD decomposition is adopted, when the attached white noise is uniformly distributed in the whole time-frequency space, the time-frequency space is composed of components with different scales which are divided by a filter bank, and when the white noise background which is uniformly distributed is added to the signal, the signal areas with different scales are automatically mapped to the proper scale related to the background white noise, so that the wind speed time sequence signal processed by the EEMD decomposition algorithm is more accurate and stable.
(3) The prediction data of the wavelet-time sequence prediction model and the EEMD-GA-BP neural network prediction model are smoothed, namely the average value of the two prediction results is obtained and is also used as one of the inputs of the combined prediction model, so that the prediction error is effectively reduced.
Drawings
The accompanying drawings, which are included to provide a further understanding of the embodiments of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the principles of the invention. In the drawings:
FIG. 1 is a flow chart of a time-series wind speed prediction model based on wavelet decomposition according to the present invention;
FIG. 2 is a flow chart of EEMD-GA-BP neural network wind speed prediction modeling of the present invention;
FIG. 3 is a flow chart of the EEMD decomposition algorithm of the present invention;
FIG. 4 is a flow chart of wind speed combination prediction according to the present invention;
FIG. 5 is a block diagram of the wind speed combination prediction of the present invention;
FIG. 6 is a waveform diagram of a time-series wind speed prediction based on wavelet decomposition according to an embodiment of the present invention;
FIG. 7 is a waveform diagram illustrating an EEMD-GA-BP neural network wind speed prediction embodiment of the present invention;
FIG. 8 is a diagram of combined predicted results according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to examples and accompanying drawings, and the exemplary embodiments and descriptions thereof are only used for explaining the present invention and are not meant to limit the present invention.
Example (b):
in the embodiment, wind speed data of 14-19 days in 4 months in 2017 of a Lanxin high-speed railway is taken as a sample, the wind speed data monitored on site is subjected to average processing for 10min by referring to a sampling time interval of basic wind speed, so that original wind speed data for modeling is obtained, and a corresponding simulation operation test is carried out.
As shown in fig. 1 to 8, the present embodiment relates to a method for constructing a wind speed prediction model by using a combination method, including the following steps:
s1: performing wavelet decomposition on the original wind speed to obtain a decomposed wind speed value, predicting each decomposed wind speed by one step in advance by using a model to obtain a decomposed predicted value, and combining the predicted values to finally obtain an integral wind speed predicted value based on a time series method of wavelet decomposition;
step S1 includes:
s1.1: obtaining raw wind speed data
Decomposing and reconstructing by using db3 wavelet to obtain a trend signal a3
Represents; detail signals d1, d2, d3, respectively
Represents;
s1.2: respectively calculating an autocorrelation function and a partial autocorrelation function of the trend signal a3 and the detail signals d1, d2 and d3, checking the stationarity of the autocorrelation function and the partial autocorrelation function, and if the signals are non-stationary signals, carrying out differential processing on the signals to obtain stationary time series signals;
s1.3: according to the tailing and truncation of the autocorrelation function and the partial autocorrelation function obtained by each trend signal and each detail signal, respectively selecting a proper model for each trend signal and each detail signal and determining the order of the model by combining with the AIC criterion;
s1.4: respectively carrying out parameter estimation on parameters in each model by using a least square estimation method;
s1.5: checking the effectiveness of each model, if the model does not meet the requirements, returning to the step S1.3, and reselecting a proper model and a proper order until an optimal model is selected;
s1.6: and (4) carrying out wind speed prediction on a fitting model established for each trend signal and each detail signal to obtain a wind speed predicted value
S1.7: each wind speed predicted value is superposed and combined to obtain a final wind speed predicted value
And comparing the measured wind speed data with the measured wind speed data to obtain a prediction error.
The predicted value is obtained based on the established wavelet decomposition time series wind speed prediction model and is compared with the actual value, and the prediction process is shown in fig. 6.
S2: EEMD decomposition is carried out on the original wind speed to obtain a plurality of eigenmode functions imf and residual information r, a BP neural network optimized by a genetic algorithm is used for carrying out one-step prediction on each decomposed wind speed in advance to obtain each subsequence predicted value, and the predicted values are overlapped to obtain an integral wind speed predicted value;
step S2 includes:
s2.1: decomposing the original wind speed time series signal into a series of eigenmode functions and a residual information by utilizing integrated empirical mode decomposition;
s2.2: respectively establishing three layers of BP neural network structures based on the decomposed sequences, optimizing the initial weight and the threshold of the neural network by using a genetic algorithm, searching an optimal individual to initialize each layer of neurons of the BP neural network by using a training error as a fitness function of the genetic algorithm when optimizing the initial weight and the threshold of the BP neural network by using the genetic algorithm, and then carrying out network training;
s2.3: and superposing the prediction results of each sequence to obtain a final wind speed prediction value.
The training related parameters of the BP neural network are shown in table 1, and the values of the related parameters during operation are shown in table 2 when a genetic algorithm is adopted.
TABLE 1 parameters associated with training BP neural networks
| Parameter name | Value of parameter |
| Number of input layer neuron nodes | 7 are provided with |
| Number of hidden |
10 are provided with |
| Number of neuron nodes in |
1 is provided with |
| Hidden layer transfer function | Sigmoid function |
| Output layer transfer function | Linear function of |
| Learning rate | 0.09 |
| Coefficient of momentum term | 0.95 |
| Error accuracy | 0.00001 |
TABLE 2 values of relevant parameters during the operation of genetic algorithms
| Parameter name | Value of parameter |
| Initial population number | 80 |
| Length of chromosome | 91 |
| Evolutionary algebra of |
100 |
| Probability of cross operation | 0.25 |
| Probability of mutation operation | 0.1 |
And in step S2 the EEMD decomposition algorithm comprises the steps of:
A. after the original time sequence is obtained, a Gaussian white noise sequence is added to obtain a new sequence;
B. performing EMD decomposition on the obtained new sequence to obtain a decomposition result;
C. sorting the decomposition results into k imf air quantities and one residual information;
D. and averaging the decomposed sequences and obtaining a result.
The prediction is carried out based on the EEMD-GA-BP neural network wind speed prediction model and is compared with an actual value, and the prediction process is shown in FIG. 7.
S3: the results of the wavelet-time sequence prediction model and the EEMD-GA-BP neural network prediction model are used as input vectors of a neural network; smoothing the prediction data of the wavelet-time sequence prediction model and the EEMD-GA-BP neural network prediction model, namely averaging the two prediction results to be used as the input vector of the combined prediction model; processing the predicted value and establishing a GA-BP neural network combined prediction model; step S3 includes:
s3.1: the results of the wavelet-time sequence prediction model and the EEMD-GA-BP neural network prediction model are used as input vectors of a neural network;
s3.2: smoothing the prediction data of the wavelet-time sequence prediction model and the EEMD-GA-BP neural network prediction model, namely averaging the two prediction results to be used as the input vector of the combined prediction model;
s3.3: and processing the predicted value and establishing a GA-BP neural network combined prediction model.
And training the combined prediction model according to the prediction values obtained by the wavelet-time series prediction model and the EEMD-GA-BP neural network prediction model to obtain a combined predicted wind speed value, and comparing the combined predicted wind speed value with an actual wind speed value, wherein the prediction result is shown in FIG. 8.
TABLE 3 comparison graph of prediction result accuracy for various prediction methods
The accuracy of the prediction results of various prediction methods is compared and sorted into a table 3, and the predicted value obtained by the finally obtained combined model is closer to a true value compared with the predicted values obtained by two single prediction models, namely the combined prediction model established by the invention can improve the prediction accuracy to a certain extent.
The above-mentioned embodiments are intended to illustrate the objects, technical solutions and advantages of the present invention in further detail, and it should be understood that the above-mentioned embodiments are merely exemplary embodiments of the present invention, and are not intended to limit the scope of the present invention, and any modifications, equivalent substitutions, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (5)
1. A method for constructing a wind speed prediction model by using a combination method is characterized by comprising the following steps:
s1: performing wavelet decomposition on the original wind speed to obtain a decomposed wind speed value, predicting each decomposed wind speed by one step in advance by using a model to obtain a decomposed predicted value, and combining the predicted values to finally obtain an integral wind speed predicted value based on a time series method of wavelet decomposition;
s2: EEMD decomposition is carried out on the original wind speed to obtain a plurality of eigenmode functions imf and residual information r, a BP neural network optimized by a genetic algorithm is used for carrying out one-step prediction on each decomposed wind speed in advance to obtain each subsequence predicted value, and the predicted values are overlapped to obtain an integral wind speed predicted value;
s3: the results of the wavelet-time sequence prediction model and the EEMD-GA-BP neural network prediction model are used as input vectors of a neural network; smoothing the prediction data of the wavelet-time sequence prediction model and the EEMD-GA-BP neural network prediction model, namely averaging the two prediction results to be used as the input vector of the combined prediction model; and processing the predicted value and establishing a GA-BP neural network combined prediction model.
2. The method for building a combined method wind speed prediction model according to claim 1, wherein the step S1 includes:
s1.1: obtaining raw wind speed data
Decomposing and reconstructing by using db3 wavelet to obtain a trend signal a3
Represents; detail signals d1, d2, d3, respectively
Represents;
s1.2: respectively calculating an autocorrelation function and a partial autocorrelation function of the trend signal a3 and the detail signals d1, d2 and d3, checking the stationarity of the autocorrelation function and the partial autocorrelation function, and if the signals are non-stationary signals, carrying out differential processing on the signals to obtain stationary time series signals;
s1.3: according to the tailing and truncation of the autocorrelation function and the partial autocorrelation function obtained by each trend signal and each detail signal, respectively selecting a proper model for each trend signal and each detail signal and determining the order of the model by combining with the AIC criterion;
s1.4: respectively carrying out parameter estimation on parameters in each model by using a least square estimation method;
s1.5: checking the effectiveness of each model, if the model does not meet the requirements, returning to the step S1.3, and reselecting a proper model and a proper order until an optimal model is selected;
s1.6: needleCarrying out wind speed prediction on a fitting model established by each trend signal and each detail signal to obtain a wind speed prediction value
S1.7: each wind speed predicted value is superposed and combined to obtain a final wind speed predicted value
And comparing the measured wind speed data with the measured wind speed data to obtain a prediction error.
3. The method for building a combined method wind speed prediction model according to claim 1, wherein the step S2 includes:
s2.1: decomposing the original wind speed time series signal into a series of eigenmode functions and a residual information by utilizing integrated empirical mode decomposition;
s2.2: respectively establishing three layers of BP neural network structures based on the decomposed sequences, optimizing the initial weight and the threshold of the neural network by using a genetic algorithm, searching an optimal individual to initialize each layer of neurons of the BP neural network by using a training error as a fitness function of the genetic algorithm when optimizing the initial weight and the threshold of the BP neural network by using the genetic algorithm, and then carrying out network training;
s2.3: and superposing the prediction results of each sequence to obtain a final wind speed prediction value.
4. The combined method wind speed prediction model building method according to claim 3, wherein the EEMD decomposition algorithm in step S2 comprises the following steps:
A. after the original time sequence is obtained, a Gaussian white noise sequence is added to obtain a new sequence;
B. performing EMD decomposition on the obtained new sequence to obtain a decomposition result;
C. sorting the decomposition results into k imf air quantities and one residual information;
D. and averaging the decomposed sequences and obtaining a result.
5. The method for constructing a wind speed prediction model according to any one of claims 1 to 4, wherein the step S3 comprises:
s3.1: the results of the wavelet-time sequence prediction model and the EEMD-GA-BP neural network prediction model are used as input vectors of a neural network;
s3.2: smoothing the prediction data of the wavelet-time sequence prediction model and the EEMD-GA-BP neural network prediction model, namely averaging the two prediction results to be used as the input vector of the combined prediction model;
s3.3: and processing the predicted value and establishing a GA-BP neural network combined prediction model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010120902.4A CN111353640B (en) | 2020-02-26 | 2020-02-26 | Method for constructing wind speed prediction model by combination method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010120902.4A CN111353640B (en) | 2020-02-26 | 2020-02-26 | Method for constructing wind speed prediction model by combination method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111353640A CN111353640A (en) | 2020-06-30 |
| CN111353640B true CN111353640B (en) | 2022-04-08 |
Family
ID=71197910
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010120902.4A Active CN111353640B (en) | 2020-02-26 | 2020-02-26 | Method for constructing wind speed prediction model by combination method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111353640B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112561180B (en) * | 2020-12-21 | 2022-05-17 | 深圳大学 | Short-term wind speed prediction method and device based on meta-learning, computer equipment and storage medium |
| CN112749792B (en) * | 2021-02-02 | 2023-07-07 | 南京信息工程大学 | BP algorithm-based wind speed prediction method |
| CN116029435B (en) * | 2023-01-10 | 2023-07-07 | 盐城工学院 | Environmental comfort early warning system is bred to live pig facility |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102063641A (en) * | 2010-10-14 | 2011-05-18 | 北京大学 | Method for forecasting wind speed of high speed railway line |
| CN102338808A (en) * | 2011-08-26 | 2012-02-01 | 天津理工大学 | Online hybrid forecasting method for short-term wind speed of wind power field |
| CN106372731A (en) * | 2016-11-14 | 2017-02-01 | 中南大学 | Strong-wind high-speed railway along-the-line wind speed space network structure prediction method |
| CN109376897A (en) * | 2018-08-29 | 2019-02-22 | 广东工业大学 | A kind of short-term wind power forecast method based on hybrid algorithm |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102628876B (en) * | 2012-02-13 | 2013-07-31 | 甘肃省电力公司风电技术中心 | Ultra-short term prediction method comprising real-time upstream and downstream effect monitoring |
-
2020
- 2020-02-26 CN CN202010120902.4A patent/CN111353640B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102063641A (en) * | 2010-10-14 | 2011-05-18 | 北京大学 | Method for forecasting wind speed of high speed railway line |
| CN102338808A (en) * | 2011-08-26 | 2012-02-01 | 天津理工大学 | Online hybrid forecasting method for short-term wind speed of wind power field |
| CN106372731A (en) * | 2016-11-14 | 2017-02-01 | 中南大学 | Strong-wind high-speed railway along-the-line wind speed space network structure prediction method |
| CN109376897A (en) * | 2018-08-29 | 2019-02-22 | 广东工业大学 | A kind of short-term wind power forecast method based on hybrid algorithm |
Non-Patent Citations (2)
| Title |
|---|
| 基于小波-BP 神经网络的风电场短期风速预测;胡晓虎;《铜陵学院学报》;20120831;第107-109页 * |
| 基于经验模态分解和遗传算法改进的;贺坤；李星野;《电子测量技术》;20180430;第146-149页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111353640A (en) | 2020-06-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111353640B (en) | Method for constructing wind speed prediction model by combination method | |
| CN106779151B (en) | A kind of line of high-speed railway wind speed multi-point multi-layer coupling prediction method | |
| CN108764520B (en) | Water quality parameter prediction method based on multilayer cyclic neural network and D-S evidence theory | |
| CN109472110A (en) | A kind of aero-engine remaining life prediction technique based on LSTM network and ARIMA model | |
| CN111160520A (en) | BP neural network wind speed prediction method based on genetic algorithm optimization | |
| AU2020103741A4 (en) | Feed forward neural networks combined with extreme learning machine approach for large weather data | |
| CN101739819A (en) | Method and device for predicting traffic flow | |
| CN112396234A (en) | User side load probability prediction method based on time domain convolutional neural network | |
| CN113610188B (en) | Bow net contact force non-section abnormality identification method and device | |
| CN108415884B (en) | Real-time tracking method for structural modal parameters | |
| CN106526710A (en) | Haze prediction method and device | |
| CN112116198A (en) | Data-driven process industrial state perception network key node screening method | |
| Shabri | A hybrid wavelet analysis and adaptive neuro-fuzzy inference system for drought forecasting | |
| CN111695288B (en) | Transformer fault diagnosis method based on Apriori-BP algorithm | |
| CN111415008B (en) | Ship flow prediction method based on VMD-FOA-GRNN | |
| CN114881204B (en) | Road surface flatness prediction method based on road element splitting and GA-BP neural network model | |
| Li et al. | Data-driven structural condition assessment for high-speed railway bridges using multi-band FIR filtering and clustering | |
| CN111930601A (en) | Deep learning-based database state comprehensive scoring method and system | |
| CN114563671A (en) | High-voltage cable partial discharge diagnosis method based on CNN-LSTM-Attention neural network | |
| CN117521512A (en) | Bearing residual service life prediction method based on multi-scale Bayesian convolution transducer model | |
| CN117892084B (en) | Bridge support static displacement probability prediction model construction method and early warning method | |
| CN117744495A (en) | Method for predicting service life of extra-large bearing driven by multiple models in different degradation stages | |
| CN109426857B (en) | Water quality index prediction method based on state pool network | |
| CN114492507B (en) | Bearing residual life prediction method under digital-analog cooperative driving | |
| CN114390376A (en) | Fire big data remote detection and early warning system |
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 |