CN101592673B - Method for forecasting wind speed along railway - Google Patents
Method for forecasting wind speed along railway Download PDFInfo
- Publication number
- CN101592673B CN101592673B CN2009100093024A CN200910009302A CN101592673B CN 101592673 B CN101592673 B CN 101592673B CN 2009100093024 A CN2009100093024 A CN 2009100093024A CN 200910009302 A CN200910009302 A CN 200910009302A CN 101592673 B CN101592673 B CN 101592673B
- Authority
- CN
- China
- Prior art keywords
- wind
- series
- forecasting
- wind speed
- decomposition
- 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
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a method for forecasting wind speed along a railway, which comprises the following steps: (1) performing multi-layer decomposition and restructuring calculation on a sequence for forecasting the wind speed by selecting the decomposition of a wavelet analysis method and a restructuring algorithm so as to decompose an original non-stationary wind speed sequence into a plurality of layers of stationary wind speed sequences; (2) respectively establishing a corresponding time sequence forecasting model for each layer of the wind speed sequences after the decomposition and reconstruction by using an improved time sequence analysis method; and (3) performing advanced multi-step forecasting calculation on the wind speed sequences of the decomposed layers by using the corresponding forecasting model established for each layer, and then performing weighted calculation on the wind speed forecasting value of each layer to acquire a corresponding forecasting value of the original wind speed sequence. The method can be applied to different forecasting occasions and can acquire a high-precision short-term forecasting optimization algorithm.
Description
Technical field
The invention belongs to the monitoring and the control technology field of railway safe driving under the strong wind rugged surroundings, be specifically related to a kind of original wind speed time series of measuring according to railway air measuring station along the line, carry out the method for forecasting wind speed in the short time accurately.
Background technology
Strong wind is one of major weather disaster that influences safety of railway traffic, and train streams the flow field and obviously changes under its effect, and aeroperformance worsens, and badly influences the lateral stability of train.At some special road sections such as the especially big bridge in zone, air port, high embankment, hills and curves, ambient wind causes train aerodynamic-force to enlarge markedly, and with train action mutual superposition firmly, the possibility of make derail, toppling increases greatly.
The driving accident that is caused by strong wind happens occasionally in countries in the world.Strong wind has constituted serious threat to railway operation safety.Be the harm that prevents that strong wind from bringing to train operating safety, the method that some transportation by railroad developed countries such as Japan, Germany, France mainly take is in the individual segment that high wind is often sent out observation station to be set, when the high wind speed of section surpasses a certain limit value, send train speed limit or the instruction of stopping transport.China's Lan-xing Railway has also tentatively been set up the gale monitoring system, but this system only can provide simple weather monitoring function, and can't satisfy under strong wind atmosphere provides an urgent demand that speed limit is instructed or command scheduling is made a strategic decision according to circuit and current running train information for safe driving in real time.
Wind speed along the line is subjected to all multifactor impacts such as special air-flow in plateau and topography and geomorphology, has the changeable uncertain feature of burst.Therefore, to really realize accurate early warning to train along the line, system not only will should consider the future trends of wind speed along the line simultaneously with information such as rain model load-carrying, road conditions parameter, air measuring station actual measurement wind speed as the deciding factor of sending the speed limit alarm command.Only in this way could improve the accuracy that early warning system is sent the train speed limit alarm command, give by the speed limit train with certain emergency processing time simultaneously.Therefore, the current Qinghai-Tibet Railway gale monitoring early warning system upgrading wind speed numerical forecasting function of building that presses for to grinding, the realization system can calculate this air measuring station k+1 and reach the wind speed predicted value in the moment later on automatically according to the actual measurement historical wind speed of air measuring station k along the line before the moment.
Other times sequences such as wind speed along railway time series and wind farm wind velocity, mechanical fault vibration, volume of rail freigh are same class non-stationary random time sequence.They have bigger similarity aspect modeling and forecasting, all relate to sequence non-stationary judgment processing, Model Distinguish, model and decide important modeling links such as rank, parameter estimation, prediction and calculation.
External generation is a lot of because the railway serious accident that strong wind caused has caused external correlative study mechanism and scholar's great attention.Multinational scholar such as Japan furthers investigate the mechanism of big wind effect train driving safety, and proposes to be provided with along the line measure such as windbreak to reduce wind speed harm.
Forecasting research to the wind speed object also develops very fast.Many scholars have proposed the effective ways of various forecasting wind speeds.German scholar Uwe Hopp mann proposes a kind of linear extrapolation and realizes forecasting wind speed along railway.This method is carried out linear extrapolation to the wind series of being gathered and is calculated to obtain following forecasting wind speed value constantly by collection real-time wind speed in weather station is set along the railway.American scholar Xie Lian proposes a kind of hurricane surface wind Forecasting Methodology based on nonlinear control theory.American scholar Milli-gan Michael utilizes time series method to set up the wind speed arma modeling, realizes the forecasting wind speed of certain wind energy turbine set.Greece scholar Zaphiro-poulos Yiorgos proposes the usage space correlation method certain mountain ridge air measuring station actual measurement wind speed is set up " measurement-correlation computations-prediction " model, successfully realizes this wind speed, wind direction prediction.American scholar Li Shu hui proposes to use neural network method and Kalman filtering method to set up the prediction improved model, realizes wind speed high-precision forecast in short-term.Its core procedure of setting up improved model is to utilize Kalman filtering method to determine the initial connection weights of circulation Multilayer Perception network (RMLP network).Canada scholar El-Fouly T.H.M. sets up grey GM (1,1) model wind farm wind velocity, wind power is realized short-term prediction.India scholar Basu S. proposes to use genetic algorithm that northern Indian Ocean measuring point ocean surface wind speed is realized prediction.American scholar Rife D.L. utilizes the climatology Approximation Theory successfully low-level wind speed to be set up the middle scale forecast model, realizes forecasting wind speed.
Domestic scholars is also carried out correlative study to wind speed along railway, as grinds the railway gale monitoring along the line system that builds.But the system of building of grinding only relates to functions such as simple wind speed collection monitoring, wind speed statistical study, does not carry out substantial forecasting wind speed along railway research work.Up to the present this direction still belongs to blank.Simultaneously, domestic forecasting research to the other types wind series is also at the early-stage, and its research wind speed object mainly concentrates on the wind farm wind velocity.
With regard to the predicting wind speed of wind farm field, domestic scholars has obtained original achievement, proposed to utilize time series method that China's wind farm wind velocity is carried out modeling and forecasting, earlier non-stationary wind speed time series is carried out difference processing, set up forecast model again after obtaining stationary sequence, finally rely on model to realize prediction and calculation.The neural network method of utilizing that also has is set up BP NEURAL NETWORK forecasting wind speed model to China's wind farm wind velocity.Another is based on time series method and neural network method, and proposition can obviously improve the hybrid modeling algorithm of forecasting wind speed precision.This method realizes preliminary modeling by time series method to wind series earlier, obtain the time series models equation of direct reflection wind speed feature, select the initial connection weights and the neuron number of neural network method BP model then according to model equation, make hybrid algorithm obtain the higher forecasting precision.Except forecasting wind speed, the forecasting research of other objects also develops rapidly, as to electric load amount, bridge vibration signal, equipment failure signal, volume of rail freigh equal time sequence.
Sum up domestic and international list of references as can be known, wind speed along railway is carried out forecasting research belong to the forward position direction, it relates to a plurality of subjects such as Communication and Transportation Engineering, environment weather, System Discrimination, Based Intelligent Control, signal Processing, and the research difficulty is bigger.Up to the present, the achievement in research that emerge in large numbers this aspect is very rare, and almost not having can be directly with reference to the ready-made algorithm that uses.
Summary of the invention
The invention provides the method for forecasting wind speed along railway that a kind ofly can be used for different forecast occasions, can obtain the optimized Algorithm of high precision short-term forecasting, its wavelet analysis method, time series analysis method, three kinds of intelligent algorithms of Kalman filtering method based on maturation, difficulty in computation is low, calculated amount is little, can carry out leading prediction and calculation more than a minute, and can obtain explicit predictive equation formula, carry out leading multistep high-precision forecast.
The technical solution adopted in the present invention is as follows:
A kind of method of forecasting wind speed along railway is used for the short-term forecasting of wind speed along railway, it is characterized in that described method comprises the steps:
(1) selects the decomposition and reconstruction algorithm of wavelet analysis method to carry out the calculating of multilayer decomposition and reconstruction, realize original non-stationary wind series is decomposed into multilayer wind series stably intending the prediction wind series;
(2) utilize the improvement time series analysis method that each layer wind series after decomposed and reconstituted set up corresponding time series forecasting model respectively;
(3) the corresponding forecast model that utilizes each layer to build carries out leading multi-step prediction calculating to the decomposition layer wind series, then to each layer forecasting wind speed value weighted calculation, obtains the correlation predictive value of original wind series.
Concretely, described step (2) also comprises the steps: afterwards
Set up the precision of prediction model by improving the mixing of time series analysis and Kalman filtering method, further improve and improve the jump ahead precision of prediction that time series method obtains.
Described original wind series is air measuring station to be set on along the railway, obtains wind speed and the time relationship sequence of this air measuring station in a certain short time.
Described step (1) comprise with original wind series v (t) (t=1,2,3 ...) carry out the wind velocity signal decomposition and reconstruction through the tower algorithm of wavelet analysis method Mallat, obtain high frequency wind series v respectively
n(t) (n=1,2,3 ...) and low frequency wind series ω (t).
Described step (2) comprising:
To high frequency wind series v
n(t) (n=1,2,3 ...), low frequency wind series ω (t) utilization improves time series method and sets up the time series forecasting model, calculates the leading k step high frequency wind series predicted value of each decomposition layer
Low frequency wind series predicted value
The foundation of described precision of prediction model comprises:
To high frequency wind series v
n(t) (n=1,2,3 ...), low frequency wind series ω (t) utilization Kalman time series method sets up the precision of prediction model, calculates the leading 1 step high frequency wind series predicted value of each decomposition layer
Low frequency wind series predicted value
Leading k step high frequency wind series predicted value to each decomposition layer
Low frequency wind series predicted value
Be weighted calculating, export final forecasting wind speed value
Described step (1) comprising: select for use Daubechies 6 small echos with time frequency compactly supported support and high regularity that original wind series v (t) is carried out signal decomposition, decomposing degree of depth n is 3-6; Select for use the tower algorithm of Mallat that the wind series after decomposing is carried out signal reconstruction on different scale.
Described step (2) comprising:
Certain one deck low frequency or the high fdrequency component sequence of carrying out after the signal decomposition with original wind series v (t) are the modeling sample sequence, get this sequence of layer before a plurality of data finish the modeling of tumbling-type time series method, and obtain with respect to the sampled point prediction air speed value in leading three steps.
Because the wind speed time series is a kind of non-linear, non-stationary sequence of variation abnormality complexity, be difficult to use single prediction algorithm to realize the degree of precision prediction usually.Simultaneously, the potential physics law of wind series is constantly to change, and forecast model only can this Changing Pattern of real-time follow-up, just might obtain high-precision predicting the outcome.The improvement time series analysis method that method adopted of forecasting wind speed along railway possesses the leading multi-step prediction ability of powerful short-term, and is outstanding unusually to the trace ability performance of wind series jump sampled point.The jump ahead precision of prediction is being required loose relatively occasion, and it is a kind of desirable selection that utilization improvement time series analysis method carries out modeling and forecasting.Kalman's time series analysis method obviously improves the improvement time series analysis method and obtains the jump ahead precision of prediction, embodies the outstanding algorithmic characteristic aspect the jump ahead prediction.The algorithm system of this optimization (comprise and improve time series analysis method and Kalman's time series analysis method) is for other algorithms, its hybrid modeling process is not significantly increased difficulty in computation and calculated amount, realize its computation optimization step by programming easily by relevant higher level lanquage, in engineering, can obtain leading wind speed predicted value in real time, can in relevant wind speed early warning system, promote the use of.
For further specifying beneficial effect of the present invention, the present invention is further illustrated below in conjunction with the drawings and specific embodiments.
Description of drawings
Fig. 1 is the original wind series figure that measures in the specific embodiment of the invention;
Fig. 2 is v in the specific embodiment of the invention (t) sequence wavelet decomposition and reconstruction calculations figure as a result; Fig. 2 (a) is { X
1tSequence; Fig. 2 (b) is { X
2tSequence; Fig. 2 (c) is { X
3tSequence; Fig. 2 (d) is { X
4tSequence;
Fig. 3 is that back 50 data of each layer wind series in the specific embodiment of the invention are carried out leading three step results of prediction and calculation figure; Fig. 3 (a) is { X
1tSequence leading three the step predict the outcome; Fig. 3 (b) is { X
2tSequence leading three the step predict the outcome; Fig. 3 (c) is { X
3tSequence leading three the step predict the outcome; Fig. 3 (d) is { X
4tSequence leading three the step predict the outcome;
Fig. 4 be in the specific embodiment of the invention leading three steps after the original wind series weighted calculation predict the outcome leading three steps of v (t) sequence of figure predict the outcome (1);
Fig. 5 is the leading nonsynchronous figure of predicting the outcome after the original wind series weighted calculation in the specific embodiment of the invention; Fig. 5 (a) is v (t) sequence jump ahead predict the outcome (1); Fig. 5 (b) is predict the outcome in leading five steps of v (t) sequence (1); Fig. 5 (c) is predict the outcome in leading ten steps of v (t) sequence (1);
Fig. 6 utilizes traditional time series analysis method to carry out leading nonsynchronous figure of predicting the outcome of original wind series; Fig. 6 (a) is v (t) sequence jump ahead predict the outcome (2); Fig. 6 (b) is predict the outcome in leading three steps of v (t) sequence (2); Fig. 6 (c) is predict the outcome in leading five steps of v (t) sequence (2); Fig. 6 (d) is predict the outcome in leading ten steps of v (t) sequence (2);
Fig. 7 is the figure that predicts the outcome that introduces Kalman filtering method in the specific embodiment of the invention.
Embodiment
The method of this forecasting wind speed along railway is surveyed original wind series v (t) (sampled point of per minute) with Qinghai-Tibet Railway Golmud-No. 18 air measuring stations of Lhasa section and is carried out modeling and prediction, and original series v (t) as shown in Figure 1.Get preceding 150 data and set up forecast model, back 50 data testing models.
At first, select the decomposition and reconstruction algorithm of wavelet analysis method to carry out the calculating of multilayer decomposition and reconstruction, realize original non-stationary wind series is decomposed into multilayer wind series stably intending the prediction wind series;
Select for use Daubechies 6 small echos that original series v (t) is carried out signal decomposition, decompose degree of depth n and get 3 with time frequency compactly supported support and high regularity.Select for use the tower algorithm of Mallat that the wind series after decomposing is carried out signal reconstruction on different scale.
v
n(t) (n=1,2,3) are illustrated respectively in and finish the high fdrequency component sequence that obtains after the reconstruction calculations on the 1st~3 layer.The low frequency component sequence that obtains after the 3rd layer of reconstruction processing of ω (t) expression.For the modeling convenience, with the 1st floor height frequency component sequence v
1(t) be designated as sequence { X
4t, the 2nd floor height frequency component sequence v
2(t) be designated as sequence { X
3t, the 3rd floor height frequency component sequence v
3(t) be designated as sequence { X
2t, the 3rd layer of low frequency component sequence ω (t) is designated as sequence { X
1t.Each sequence as shown in Figure 2.
Then, utilize the improvement time series analysis method that each layer wind series after decomposed and reconstituted set up corresponding time series forecasting model respectively;
With the 3rd layer of low frequency component sequence { X
1tBe the modeling sample sequence, the algorithm computation step of tumbling-type time series method is described, and the modeling procedure of other decomposition layer wind series can be with reference to finishing.Get sequence { X
1tPreceding 150 data finish the modeling of tumbling-type time series method, and obtain with respect to sampled point X
1t(200) the prediction air speed value in leading three steps
The specific algorithm calculation procedure is as follows:
(1) utilize time series analysis method to { X
1t(1), X
1t(2) ..., X
1t(150) } sequence is carried out the BJ modeling, selects minimal information criterion AIC to carry out model and decides rank, selects square to estimate to carry out model parameter estimation.Determine sequence { X
1tAppropriate model be AR IMA (6,1,0), obtain its corresponding predictive equation and be:
X
1t(t)=2.7423X
1t(t-1)-2.9532X
1t(t-2)
+1.8939X
1t(t-3)-1.3428X
1t(t-4)
+1.1826X
1t(t-5)-0.7913X
1t(t-6)
+0.2685X
1t(t-7)+a
t (1)
The equation (1) that uses a model is realized wind series { X
1tThe jump ahead prediction and calculation, the jump ahead predicted value obtained
(2) the appropriate model classification of maintenance step (1) institute identification is utilized
Sequence reappraises model parameter, obtains comprising predicted value
The new model equation of information characteristics is proceeded the jump ahead prediction and calculation, obtains with respect to sampled point X
1t(200) predicted value of super first two steps
X
1t(t)=2.4731X
1t(t-1)-2.1436X
1t(t-2)
+0.7840X
1t(t-3)-0.2292X
1t(t-4)
+0.2645X
1t(t-5)-0.3284X
1t(t-6)
+0.1796X
1t(t-7)+a
t (2)
(3) utilize
Sequence obtains comprising predicted value to model parameter estimation again
The new model equation of information characteristics is proceeded leading 1 step prediction, obtains with respect to sampled point X
1t(200) predicted value in leading three steps
Here it is, and the tumbling-type time series method is finished a complete computation cycle of leading three steps prediction.
X
1t(t)=2.4271X
1t(t-1)-2.1279X
1t(t-2)
+0.9923X
1t(t-3)-0.4392X
1t(t-4)
+0.1067X
1t(t-5)-0.0212X
1t(t-6)
+0.0621X
1t(t-7)+a
t (3)
(4) after finishing a computation period, restart the tumbling-type prediction and calculation according to up-to-date actual measurement wind series again, obtain sampled point X
1t(201) predicted value in leading three steps
Moreover, each decomposition layer is carried out leading multi-step prediction calculates,
With reference to tumbling-type time series method modeled example, respectively to sequence { X
1t, { X
2t, { X
3t, { X
4tSet up forecast model respectively, and then back 50 data of each layer wind series being carried out leading three step prediction and calculation, the result is as shown in Figure 3.
After the prediction and calculation of finishing each decomposition layer wind series, according to the leading three step predicted values of following formula weighted calculation sequence v (t)
Get weighting coefficient ρ
1=ρ
2=ρ
3=1, predict the outcome as shown in Figure 4.In addition, calculate the jump ahead predicted value of original series v (t)
Leading five step predicted values
Leading ten step predicted values
The result as shown in Figure 5.
Comparative example is in order further to show superperformance of the present invention, utilize traditional time series analysis method directly the same section wind speed sample of original series v (t) to be carried out modeling again, carry out jump ahead, three steps, five steps, ten step prediction and calculation then respectively, the result as shown in Figure 6.
Introduce general precision of prediction evaluation index shown in Figure 6 predicting the outcome assessed, the correlation computations formula of evaluation index is as follows:
(1) average error:
(2) mean absolute error:
(3) average relative error:
(4) root-mean-square error:
The result of calculation of index of correlation is as shown in table 1.
Table 1 analytical table (1) that predicts the outcome
As known from Table 1: the present invention based on wavelet analysis method and improvement time series analysis method possesses outstanding algorithmic characteristic, and its jump ahead, three steps, five steps, ten steps prediction average relative error only are respectively 0.30%, 0.75%, 1.15%, 1.65%; Comparison diagram 5-Fig. 6, in the prediction of jump sampled point, the present invention is more excellent than traditional time series analysis method, and there is the time-delay phenomenon in traditional time series analysis method prediction.
For further improving the resulting jump ahead precision of prediction of above-mentioned embodiment, introduce Kalman filtering method and time series analysis method hybrid modeling and improve.
The general linear discrete system can be expressed as:
X(t+1)=Φ(t+1,t)X(t)+Γ(t+1,t)w(t) (8)
Z(t+1)=H(t+1)X(t+1)+v(t+1) (9)
In the formula: X (t) is n dimension state vector; Z (t) is m dimension observation vector; W (t) maintains the system noise vector for p; V (t) measures noise vector for the m dimension; (t+1 t) is carved into t+1 state-transition matrix constantly for from t the time to Φ; (t+1 k) is carved into t+1 excitation transfer matrix constantly for from t the time to Γ; H (t+1) is a t+1 prediction output transition matrix constantly.Formula (8) is called state equation, and formula (9) is called the measurement equation.
The core advantage of Kalman filtering method is to have the characteristic of dynamic weighting correction according to the estimated value of up-to-date measured value correction previous moment.And to realize that the Kalman predicts, must derive correct state equation and measure equation.
If X
1(t)=X
1t(t), X
2(t)=X
1t(t+1) ..., X
7(t)=X
1t(t+6) formula (5) is rewritten into following form:
X
1t(t+1)=2.7423X
1t(t)-2.9532X
1t(t-1)
+1.8939X
1t(t-2)-1.3428X
1t(t-3) (10)
+1.1826X
1t(t-4)-0.7913X
1t(t-5)
+0.2685X
1t(t-6)+a
t
Then have
X
1(t+1)=2.7423X
1(t)-2.9532X
2(t)
+1.8939X
3(t)-1.3428X
4(t)
(11)
+1.1826X
5(t)-0.7913X
6(t)
+0.2685X
7(t)+a
t
Because X
2(t)=X
1(t+1), X
3(t)=X
2(t+1) ..., X
7(t+1)=X
6(t), then can get:
Yi Zhi, Z (k+1)=X (k+1)+v (k+1), wherein, v (k+1) is for measuring additional noise, and is convenient for modeling, can be assumed to be white noise, then measures equation and is:
Z(t+1)=[1,0,0,0,0,0,0]×[X
1(t+1),X
2(t+1),X
3(t+1),
(13)
X
4(t+1),X
5(t+1),X
6(t+1),X
7(t+1)]
T+v(t+1)
With formula (8), (9) and formula (12), (13) relatively, can obtain Φ (t+1, t), Γ (t+1, t), the concrete value of H (t+1).Consider convergent speed and with reference to engineering custom, it is as follows to get initial value: X (0|0)=[0], P (0|0)=10I.Get R (t)=[1] (t=1,2,3 ...),
Q(t)=[1](t=1,2,3,...)。Application card Germania prediction recurrence equation carries out the leading 1 step prediction of wind speed, as shown in Figure 7.
In like manner, utilization equation (4)-(7) predict the outcome to Fig. 7 and estimate, and correlation computations result is as shown in table 2.As known from Table 2, the effect of jump ahead prediction is significantly improved, and it optimizes the effect highly significant.Hence one can see that, and aspect the jump ahead prediction, the present invention obviously is better than traditional time series analysis method.
Table 2 analytical table (2) that predicts the outcome
Claims (8)
1. the method for a forecasting wind speed along railway is used for the short-term forecasting of wind speed along railway, it is characterized in that described method comprises the steps:
(1) selects the decomposition and reconstruction algorithm of wavelet analysis method to carry out the calculating of multilayer decomposition and reconstruction, realize original non-stationary wind series is decomposed into multilayer wind series stably intending the prediction wind series;
(2) utilize the improvement time series analysis method that each layer wind series after decomposed and reconstituted set up corresponding time series forecasting model respectively;
(3) the corresponding forecast model that utilizes each layer to build carries out leading multi-step prediction calculating to the decomposition layer wind series, then to each layer forecasting wind speed value weighted calculation, obtains the correlation predictive value of original wind series.
2. the method for forecasting wind speed along railway according to claim 1 is characterized in that described step (2) also comprises the steps: afterwards
Set up the precision of prediction model by improving the mixing of time series analysis and Kalman filtering method, further improve and improve the jump ahead precision of prediction that time series method obtains.
3. the method for forecasting wind speed along railway according to claim 1 is characterized in that described original wind series is air measuring station to be set on along the railway, obtains wind speed and the time relationship sequence of this air measuring station in a certain short time.
4. the method for forecasting wind speed along railway according to claim 1, it is characterized in that described step (1) comprises (t=1,2,3 with original wind series v (t),) carry out the wind velocity signal decomposition and reconstruction through the tower algorithm of wavelet analysis method Mallat, obtain high frequency wind series v respectively
n(t) (n=1,2,3 ...) and low frequency wind series ω (t).
5. the method for forecasting wind speed along railway according to claim 4 is characterized in that described step (2) comprising:
To high frequency wind series v
n(t) (n=1,2,3 ...), low frequency wind series ω (t) utilization improves time series method and sets up the time series forecasting model, calculates the leading k step high frequency wind series predicted value of each decomposition layer
Low frequency wind series predicted value
6. the method for forecasting wind speed along railway according to claim 2 is characterized in that the foundation of described precision of prediction model comprises:
With original wind series v (t) (t=1,2,3 ...) carry out the wind velocity signal decomposition and reconstruction through the tower algorithm of wavelet analysis method Mallat, obtain high frequency wind series v respectively
n(t) (n=1,2,3 ...) and low frequency wind series ω (t);
To high frequency wind series v
n(t) (n=1,2,3 ...), low frequency wind series ω (t) utilization Kalman time series method sets up the precision of prediction model, calculates the leading 1 step high frequency wind series predicted value of each decomposition layer
Low frequency wind series predicted value
7. the method for forecasting wind speed along railway according to claim 1, it is characterized in that described step (1) comprising: select for use Daubechies 6 small echos with time frequency compactly supported support and high regularity that original wind series v (t) is carried out signal decomposition, decomposing degree of depth n is 3-6; Select for use the tower algorithm of Mallat that the wind series after decomposing is carried out signal reconstruction on different scale.
8. the method for forecasting wind speed along railway according to claim 7 is characterized in that described step (2) comprising:
Certain one deck low frequency or the high fdrequency component sequence of carrying out after the signal decomposition with original wind series v (t) are the modeling sample sequence, get this sequence of layer before a plurality of data finish the modeling of tumbling-type time series method, and obtain with respect to the sampled point prediction air speed value in leading three steps.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2009100093024A CN101592673B (en) | 2009-02-18 | 2009-02-18 | Method for forecasting wind speed along railway |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2009100093024A CN101592673B (en) | 2009-02-18 | 2009-02-18 | Method for forecasting wind speed along railway |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101592673A CN101592673A (en) | 2009-12-02 |
CN101592673B true CN101592673B (en) | 2010-12-15 |
Family
ID=41407439
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2009100093024A Active CN101592673B (en) | 2009-02-18 | 2009-02-18 | Method for forecasting wind speed along railway |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101592673B (en) |
Families Citing this family (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101923605B (en) * | 2010-08-18 | 2012-05-23 | 北京佳讯飞鸿电气股份有限公司 | Wind pre-warning method for railway disaster prevention |
CN102478584B (en) * | 2010-11-26 | 2014-10-15 | 香港理工大学 | Wind power station wind speed prediction method based on wavelet analysis and system thereof |
CN102183802B (en) * | 2011-03-10 | 2013-07-10 | 西安交通大学 | Short-term climate forecast method based on Kalman filtering and evolution modeling |
CN102338808B (en) * | 2011-08-26 | 2012-11-21 | 天津理工大学 | Online hybrid forecasting method for short-term wind speed of wind power field |
CN102609766B (en) * | 2012-02-17 | 2014-03-12 | 中南大学 | Method for intelligently forecasting wind speed in wind power station |
CN102708305B (en) * | 2012-06-20 | 2016-10-19 | 天津工业大学 | A kind of wind series Forecasting Methodology merged based on Kalman filtering |
CN103617462B (en) * | 2013-12-10 | 2016-08-17 | 武汉大学 | A kind of wind farm wind velocity Spatiotemporal Data Modeling method based on geographical statistics |
CN103996085B (en) * | 2014-06-09 | 2015-03-11 | 中南大学 | Method for predicting near-ground wind field point domain mapping space along railway |
CN104951798A (en) * | 2015-06-10 | 2015-09-30 | 上海大学 | Method for predicting non-stationary fluctuating wind speeds by aid of LSSVM (least square support vector machine) on basis of EMD (empirical mode decomposition) |
CN105373641B (en) * | 2015-08-06 | 2018-04-06 | 南京信息工程大学 | A kind of Kalman filtering improved method for precipitation estimation |
CN106250611A (en) * | 2016-07-28 | 2016-12-21 | 中国路桥工程有限责任公司 | Railway wind section strong wind meteorological disaster monitoring system |
CN106372731B (en) * | 2016-11-14 | 2017-07-28 | 中南大学 | A kind of high wind line of high-speed railway wind speed spatial network structure forecast method |
CN109767043B (en) * | 2019-01-17 | 2020-11-24 | 中南大学 | Intelligent modeling and prediction method for big data of power load time sequence |
CN112498419B (en) * | 2020-11-25 | 2022-09-09 | 中铁第四勘察设计院集团有限公司 | Encryption method, device, equipment and storage medium |
CN117056735A (en) * | 2023-10-13 | 2023-11-14 | 云南电投绿能科技有限公司 | Short-term wind speed prediction method, device and equipment for wind farm and storage medium |
-
2009
- 2009-02-18 CN CN2009100093024A patent/CN101592673B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN101592673A (en) | 2009-12-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101592673B (en) | Method for forecasting wind speed along railway | |
CN103246943B (en) | Vehicle operating condition multi-scale predicting method based on Markov chain | |
CN102610088B (en) | Method for forecasting travel time between bus stops | |
CN107941537A (en) | A kind of mechanical equipment health state evaluation method | |
CN103903430B (en) | Dynamic fusion type travel time predicting method with multi-source and isomorphic data adopted | |
CN105809287A (en) | High-voltage transmission line icing process integrated prediction method | |
CN102063641A (en) | Method for forecasting wind speed of high speed railway line | |
CN104408900A (en) | Dynamic optimization based neural network flood warning device and method | |
CN103488869A (en) | Wind power generation short-term load forecast method of least squares support vector machine | |
CN102117383A (en) | Method for diagnosing river pollution in real time | |
CN106384170A (en) | Wavelet decomposition and reconstruction-based time sequence wind speed prediction method | |
CN105512741A (en) | Bus passenger traffic combined prediction method | |
CN102610092A (en) | Urban road speed predication method based on RBF (radial basis function) neural network | |
CN105160449A (en) | Electric automobile charging pile optimization layout method | |
CN105512447A (en) | Bus passenger volume prediction method based on Holt-Winters model | |
CN102789447A (en) | Method for analyzing ice and climate relationship on basis of grey MLR (Multiple Linear Regression) | |
CN106228808A (en) | City expressway travel time prediction method based on Floating Car space-time grid data | |
CN108427273A (en) | A kind of Feedback Control Design method reducing traffic congestion phenomenon | |
CN115063978B (en) | Bus arrival time prediction method based on digital twins | |
CN106887141A (en) | Continuous Traffic node congestion degree forecast model based on queueing theory, system and method | |
Gou et al. | A wind hazard warning system for safe and efficient operation of high-speed trains | |
Xu et al. | A combined model of random forest and multilayer perceptron to forecast expressway traffic flow | |
CN103605908A (en) | Wind speed sequence forecasting method based on Kalman filtering | |
CN103344395B (en) | A kind of confirmation method of bridge strengthening target bearing capacity and device | |
CN103020733A (en) | Method and system for predicting single flight noise of airport based on weight |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant |