CN103699800A - Ultrashort-period wind speed prediction method based on frequency-domain multi-scale wind speed signal predictability - Google Patents
Ultrashort-period wind speed prediction method based on frequency-domain multi-scale wind speed signal predictability Download PDFInfo
- Publication number
- CN103699800A CN103699800A CN201310750044.1A CN201310750044A CN103699800A CN 103699800 A CN103699800 A CN 103699800A CN 201310750044 A CN201310750044 A CN 201310750044A CN 103699800 A CN103699800 A CN 103699800A
- Authority
- CN
- China
- Prior art keywords
- wind speed
- frequency domain
- predictability
- frequency
- wind
- 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.)
- Pending
Links
Images
Landscapes
- Wind Motors (AREA)
Abstract
The invention belongs to the technical field of analysis and measurement control, and relates to an ultrashort-period wind speed prediction method based on frequency-domain multi-scale wind speed signal predictability, in order to solve the problems of low prediction accuracy and long model training time caused by the fact that that an existing prediction method does not take the problem about frequency-domain multi-scale predictability into consideration and data dimension selection of an input space in a statistical predication model needs to depend on experience. Adverse impact generated on predictions after superposition due to more steps of high-frequency component predication is prevented by addition of technical steps of predictability analysis and autocorrelation analysis, accuracy of ultrashort-period wind speed prediction is improved, and time for model training is reduced. The method is mainly used for predicating electric field power by a wind power plant so as to help a power grid to make a reasonable dispatch plan and determine spinning reserve, and operation of the power grid is guaranteed safely and economically.
Description
Technical field
The invention belongs to and analyze and survey control technology field, relate to the ultra-short term wind speed forecasting method based on the multiple dimensioned wind velocity signal predictability of frequency domain.
Background technology
Wind energy is one of new forms of energy of on the largest scaleization DEVELOPMENT PROSPECT in current generally acknowledged regenerative resource.China's Wind Power Development is started late but is developed rapidly; In recent years, a large amount of scale wind energy turbine set is put in actual operation.Wind-powered electricity generation is typical strong random uncertain power supply, and scale wind-powered electricity generation can bring a lot of problems to the safe and stable operation of electrical network; Therefore, scale wind-powered electricity generation is dissolved becomes the great realistic problem that China's electric system faces.Power Output for Wind Power Field prediction is accurately to solve one of important foundation of the safe and efficient grid-connected utilization of scale wind-powered electricity generation.Wind is the power source of wind-powered electricity generation unit output, and therefore, forecasting wind speed is the important prerequisite that realizes wind farm power prediction accurately.Wherein, take and be unit the ultra-short term forecast of 1～6 hour can help scheduling decision for 10 minutes or 15 minutes.Because the influence factor of wind speed is numerous, the mechanism of action is complicated, have very strong multiple dimensioned characteristic, therefore, wind series can be regarded a plurality of signal couplings result together as.Yet it is strong non-linear and non-stationary that this also makes wind velocity signal show, and causes the difficulty of wind speed Forecast model very large, is the most difficult meteorologic parameter of predicting of generally acknowledging.In recent years, Chinese scholars starts to pay close attention to the multiple dimensioned characteristic of frequency domain of wind speed successively, and wherein, widely used thinking is: first, original wind series is resolved into the subsequence of different frequency; Then, on each subsequence, set up regression model; Finally, then synthesize and provide final prediction effect.
Yet current Forecasting Methodology is not considered the predictability problem that frequency domain is multiple dimensioned, but each frequency domain wind speed is simple synthetic as final predicting the outcome using respectively predicting the outcome again after the identical length of prediction.In fact, each dimensions in frequency sequence is due to the not equal property difference of driving-energy of himself, and the time span that causes each frequency domain yardstick sequence to forecast is different; Especially for high-frequency fluctuation component, the quantity of information comprising in sequence is less, and its regularity is lower, does in the result after multi-step prediction that effectively step number is less; Low frequency component has stronger trend amount and Changing Pattern, and its autocorrelation is stronger; And the autocorrelation of high fdrequency component is very weak, its regularity a little less than.Therefore, after stack, to predicting the outcome, have a negative impact on the contrary the step number of high fdrequency component prediction is long, therefore it is inappropriate being done with the multi-step prediction of the same length of low frequency component, because by the identical step number of all scale component predict the outcome synthetic after, in high frequency series inclusion information amount less a few step results can be on affecting final prediction effect below.And air speed data is a kind of time series of implicit complex relationship, in current statistical forecast model, for the data dimension of the input space, choose generally and choose by experience, there is no unified standard.Therefore the model training time that just need to be longer, forecast precision is also limited simultaneously.
Summary of the invention
In order to solve the data dimension of the input space in the problem that has a negative impact to predicting the outcome on the contrary after the long stack of step number of the high fdrequency component prediction that existing Forecasting Methodology do not consider that the multiple dimensioned predictability of frequency domain causes and statistical forecast model, to choose the forecast precision that need to rely on experience to cause low, the problem that the model training time is long, and existing Forecasting Methodology is low to wind farm power prediction precision, the problem that wind energy turbine set security of operation is caused a hidden trouble, the invention provides a kind of ultra-short term wind speed forecasting method based on the multiple dimensioned wind velocity signal predictability of different frequency domains.
Ultra-short term wind speed forecasting method based on the multiple dimensioned wind velocity signal predictability of frequency domain of the present invention, concrete steps are as follows:
Step 1: the multiple dimensioned characteristic of frequency domain based on wind speed, by Mallat wavelet decomposition, original wind velocity signal is decomposed, resolve into the subsequence of 3～4 layers of different frequency domain yardstick;
Step 2: measure the predictability of each frequency domain subsequence by autocorrelation analytical approach, determine that with function threshold the prediction step number L(of each subsequence is prediction length according to the wind series predictability analysis result on each frequency domain yardstick);
Step 3: the magnitude signal that Mallat wavelet decomposition is obtained and each point in detail signal, by with its before constantly L put and L-1 Difference Terms forms the proper vector of each point, and do normalized; Utilize principal component analytical method to carry out dimension-reduction treatment to the input space of magnitude signal, obtain the input vector of dimension optimizing;
Step 4: the input vector of the dimension optimizing of the magnitude signal that the analysis result obtaining according to step 2 and step 3 obtain and the proper vector of detail signal, the statistical regression model of setting up respectively different prediction lengths on each frequency domain yardstick obtains predicting the outcome of each frequency domain yardstick simultaneously;
Advantage of the present invention:
The present invention is incorporated into the multiple dimensioned predictability problem of frequency domain among forecasting wind speed, and has provided a kind of method of utilizing autocorrelation analysis to measure predictability; By the predictability analysis of each frequency domain yardstick wind velocity signal, determine each frequency domain yardstick sequence signal predictable effective time of length; Then, to each frequency domain magnitude signal difference Modeling and Prediction; Finally, by each model Output rusults combination, provide final forecast result.Because the predictability of each dimensions in frequency subsequence is different, its autocorrelation is also different, therefore, can adopt the method for autocorrelation analysis to analyze the predictability problem of each frequency domain magnitude signal subsequence, the auto-correlation length calculating due to magnitude signal conventionally larger simultaneously, uses principal component analytical method, obtains and be hidden in data variable relation behind, its input space is carried out to dimensionality reduction, obtain the input vector of dimension optimizing.The result of actual wind speed prognostic experiment shows, the method that the present invention proposes is with directly synthetic each frequency domain identical forecast step-length method and use separately support vector regression method to compare square error (MSE) and low 10% left and right of mean absolute error (MAE), and shortened the model training time, improve wind farm power prediction precision, reduced the hidden danger that safe operation causes to wind energy turbine set.
Accompanying drawing explanation
Fig. 1 is the schematic flow sheet of the inventive method;
Fig. 2 is 3 layers of decomposition algorithm schematic diagram of Mallat small echo, is about to wind speed Time Series and obtains three floor height frequency sequences and the low frequency sequence of one deck;
Fig. 3 is 3 layers of restructing algorithm schematic diagram of Mallat small echo, by three floor height frequency sequences and the low frequency sequence of one deck are reconstructed into wind speed time series;
Fig. 4 is actual measurement original wind speed (time series) signal;
Fig. 5 is three layers of Multiscale Wavelet Decomposition result of wind velocity signal;
Fig. 6 is the autocorrelation analysis result of each dimensions in frequency vector sequence;
Fig. 7 is that the MSE of the inventive method in actual wind speed prediction is along with the change curve of prediction step;
Fig. 8 be on test set MAE along with the change curve of prediction step;
Fig. 9 is three kinds of forecasting wind speed effect contrast figures (a is the inventive method, and b is contrast test 1, and c is contrast test 2).
Embodiment
Embodiment one: the wind speed forecasting method based on the multiple dimensioned wind velocity signal predictability of frequency domain in present embodiment, concrete steps following (as shown in Figure 1):
Step 1: the multiple dimensioned characteristic of frequency domain based on wind speed, by wavelet decomposition, original wind velocity signal (as shown in Figure 4) is decomposed to (as shown in Figure 2), resolve into the subsequence of 3～4 layers of different frequency domain yardstick;
Step 2: measure the predictability of each frequency domain subsequence by autocorrelation analytical approach, determine the prediction length L of each subsequence with function threshold according to the wind series predictability analysis result on each frequency domain yardstick;
Step 3: the magnitude signal that wavelet decomposition is obtained and each point in detail signal, by with its before constantly L put and L-1 Difference Terms forms the proper vector of each point, and each vector is done to normalized; Utilization is carried out dimension-reduction treatment with principal component analytical method to the proper vector of magnitude signal, obtains the vector of dimension optimizing; Low frequency sequence in the subsequence that its Mesoscale signal is the different frequency domain yardsticks that obtain by wavelet decomposition, detail signal is high frequency series;
Step 4: the proper vector of the magnitude signal Dimensionality optimization vector sum detail signal that the analysis result obtaining according to step 2 and step 3 obtain, the statistical regression model of setting up respectively different prediction lengths on each frequency domain yardstick obtains predicting the outcome of each frequency domain yardstick simultaneously;
Step 5: utilize wavelet decomposition algorithm to synthesize (as shown in Figure 3) predicting the outcome of the statistical regression model of each frequency domain yardstick in step 4, finally obtain the result of forecasting wind speed.
Embodiment two: the wind speed forecasting method based on the multiple dimensioned wind velocity signal predictability of frequency domain in present embodiment and the difference of embodiment one are: in step 1, wavelet decomposition concrete grammar is: Mallat is used pyramid algorith, in conjunction with multiresolution analysis, carry out pyramid wavelet decomposition algorithm.
Signal f (t) is at metric space V
jwith wavelet space W
jbe projected as
d
j,k=< f (t), ψ
j,k(t) >, by
obtain
Two scaling Equations by scaling function can obtain
By scaling function orthogonality, can be obtained
By the two scaling Equations of wavelet function, can be obtained
Above three equations of simultaneous can obtain:
All the other are identical with embodiment one.
Embodiment three: the wind speed forecasting method based on the multiple dimensioned wind velocity signal predictability of frequency domain in present embodiment and the difference of embodiment one are: the autocorrelation analysis method in step 2 is specially:
Tolerance wind speed time series set { x
t}
t=1:nin x
tsample x with k step-length of its delay
t+kcoefficient of autocorrelation be defined as the covariance γ (k) of sample, that is:
Therefore autocorrelation function ρ (k) is:
In integrating step one, obtain each frequency domain yardstick subsequence and calculate auto-correlation function value.All the other are identical with embodiment one or two.
Embodiment
Choosing Ningxia wind energy turbine set in May, 2011 to three months July forecasts and illustrates by the air speed data of 10 minutes.
Step 2, by autocorrelation analytical approach, measure the auto-correlation function value of each frequency domain subsequence, result as shown in Figure 6; Low frequency component to the 3rd layer is designated as
autocorrelation function is dull downtrending, and getting threshold value is 0.8, the low frequency component of the 3rd layer
auto-correlation length predict that step number L is 24.In like manner, other each frequency components,
the upper auto-correlation length that obtains is respectively 4,2 and 1.
Step 3, by each point in magnitude signal (i.e. the 3rd layer of low frequency signal) and detail signal (being other layer of high-frequency signal), by with its before constantly L put and L-1 Difference Terms forms the proper vector of each point, and do normalized.
Step 4, according to the analysis result in step 2, on each frequency domain components, set up respectively a support vector regression (SVR) model, and the step number of each component Model prediction is respectively 24,4,2,1.The input vector of the support vector regression forecast model of every one deck is chosen according to auto-correlation length, gets current time and L-1 before air speed value constantly and forms input vector.The auto-correlation length calculating due to magnitude signal is conventionally larger, uses core principle component analysis method, obtains and is hidden in data variable relation behind, and its proper vector is carried out to dimensionality reduction, obtains dimension and be 18 vector.
Step 6, utilize Mallat wavelet reconstruction algorithm to synthesize predicting the outcome of each yardstick support vector regression model, obtain net result.
Compliance test result
Step 2, by autocorrelation analytical approach, measure the auto-correlation function value of each frequency domain subsequence, result as shown in Figure 6; Low frequency component to the 3rd layer is designated as
autocorrelation function is dull downtrending, and getting threshold value is 0.8, the low frequency component of the 3rd layer
auto-correlation length predict that step number L is 24.In like manner, other each frequency components,
the upper auto-correlation length that obtains is respectively 4,2 and 1.
Step 3, by each point in magnitude signal (i.e. the 3rd layer of low frequency signal) and detail signal (being other layer of high-frequency signal), by with its before constantly L put and L-1 Difference Terms forms the proper vector of each point, and do normalized.And use core principle component analysis method, and the proper vector of magnitude signal is carried out to dimensionality reduction, obtain dimension and be 18 vector.
Step 4, according to the analysis result in step 2, in each frequency component, set up respectively a support vector regression (SVR) model, and the step number of each component Model prediction is respectively 24,4,2 and 1.
In the proper vector of step 6, magnitude signal Dimensionality optimization vector sum detail signal that the analysis result obtaining according to step 2 and step 3 are obtained and substitution step 4, set up in support vector regression model (SVR) and predicted the outcome.
Step 7, utilize that Mallat wavelet reconstruction algorithm obtains each yardstick support vector regression model predict the outcome to synthesize finally and predicted the outcome.
Be to carry out, after wavelet decomposition, not carry out the analysis of predictability with the difference of test 1, on different subsequences, utilize the identical step-length of support vector regression model prediction, then predicting the outcome of each layer synthesized, obtain final predicting the outcome.
Contrast test 2
Be directly according to original wind speed time series, to carry out modeling with the difference of test 1 and contrast test 1, do not carry out wavelet decomposition process, then directly use a support vector regression (single SVR) model training, by air speed data substitution in August model, finally predicted the outcome.
Comparison of test results
Fig. 9 only, for the clarity of bandwagon effect, has therefore only provided predicting the outcome of part.Adding up the average error (as shown in table 1) of prediction in 4 hours can find, compare additive method square error (MSE) and mean absolute error (MAE) is lower, forecast precision is higher.
Table 1
Claims (4)
1. the ultra-short term wind speed forecasting method based on the multiple dimensioned wind velocity signal predictability of frequency domain, is characterized in that its concrete steps are as follows:
Step 1: the multiple dimensioned characteristic of frequency domain based on wind speed, by Mallat wavelet decomposition, original wind speed time series is decomposed, resolve into the subsequence of 3～4 layers of different frequency domain yardstick;
Step 2: measure the predictability of each frequency domain subsequence by autocorrelation analytical approach, determine the prediction step number L of each subsequence according to the wind series predictability analysis result on each frequency domain yardstick with function threshold;
Step 3: the magnitude signal that Mallat wavelet decomposition is obtained and each point in detail signal, by with its before constantly L put and L-1 Difference Terms forms the proper vector of each point, and do normalized; Utilize principal component analytical method to carry out dimension-reduction treatment to the input space of magnitude signal, obtain the input vector of dimension optimizing;
Step 4: the input vector of the dimension optimizing of the magnitude signal that the analysis result obtaining according to step 2 and step 3 obtain and the proper vector of detail signal, the statistical regression model of setting up respectively different prediction lengths on each frequency domain yardstick obtains predicting the outcome of each frequency domain yardstick simultaneously;
Step 5, by Mallat wavelet reconstruction algorithm, the predicting the outcome of statistical regression model of each frequency domain yardstick in step 4 synthesized, finally obtain the result of ultra-short term forecasting wind speed.
2. the ultra-short term wind speed forecasting method based on the multiple dimensioned wind velocity signal predictability of frequency domain according to claim 1, is characterized in that in step 1, the selected wavelet basis of wavelet decomposition is db10.
3. the ultra-short term wind speed forecasting method based on the multiple dimensioned wind velocity signal predictability of frequency domain according to claim 1 and 2, is characterized in that the autocorrelation analysis method described in step 2 is specially:
Tolerance wind speed time series set { x
t}
t=1:
nin x
tsample x with k step-length of its delay
t+kcoefficient of autocorrelation be defined as the covariance γ (k) of sample, that is:
Therefore autocorrelation function ρ (k) is:
In integrating step one, obtain each frequency domain yardstick subsequence and calculate auto-correlation function value.
4. the ultra-short term wind speed forecasting method based on the multiple dimensioned wind velocity signal predictability of frequency domain according to claim 3, is characterized in that the function threshold described in step 2 is 0.8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310750044.1A CN103699800A (en) | 2013-12-31 | 2013-12-31 | Ultrashort-period wind speed prediction method based on frequency-domain multi-scale wind speed signal predictability |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310750044.1A CN103699800A (en) | 2013-12-31 | 2013-12-31 | Ultrashort-period wind speed prediction method based on frequency-domain multi-scale wind speed signal predictability |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103699800A true CN103699800A (en) | 2014-04-02 |
Family
ID=50361327
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310750044.1A Pending CN103699800A (en) | 2013-12-31 | 2013-12-31 | Ultrashort-period wind speed prediction method based on frequency-domain multi-scale wind speed signal predictability |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103699800A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104239964A (en) * | 2014-08-18 | 2014-12-24 | 华北电力大学 | Ultra-short-period wind speed prediction method based on spectral clustering type and genetic optimization extreme learning machine |
| CN104333005A (en) * | 2014-08-18 | 2015-02-04 | 西南交通大学 | Electrical-power-system post-disturbance frequency dynamic-state prediction method based on support vector regression |
| CN105303051A (en) * | 2015-11-11 | 2016-02-03 | 中国科学院遥感与数字地球研究所 | Air pollutant concentration prediction method |
| CN105320995A (en) * | 2014-08-19 | 2016-02-10 | 上海交通大学 | Frequency domain decomposition based wind power generation short-term load prediction method and apparatus |
| CN105956252A (en) * | 2016-04-27 | 2016-09-21 | 哈尔滨工业大学 | Generative deep belief network-based multi-scale forecast modeling method for ultra-short-term wind speed |
| CN103902837B (en) * | 2014-04-16 | 2017-02-15 | 广西大学 | Method for wind speed prediction based on experience Copula function |
| CN107292446A (en) * | 2017-07-03 | 2017-10-24 | 西南交通大学 | A kind of mixing wind speed forecasting method based on consideration component relevance wavelet decomposition |
| CN110288150A (en) * | 2019-06-24 | 2019-09-27 | 华北电力大学（保定） | Wind speed forecasting method, terminal device and storage medium |
| CN111339092A (en) * | 2020-02-24 | 2020-06-26 | 江苏省南通环境监测中心 | Deep learning-based multi-scale air quality forecasting method |
| CN112132365A (en) * | 2020-11-24 | 2020-12-25 | 中国电力科学研究院有限公司 | Multi-scale feature-based forecast wind speed correction method and system |
| CN114912718A (en) * | 2022-07-13 | 2022-08-16 | 汉谷云智(武汉)科技有限公司 | Ultra-short-term solar irradiance prediction method based on discrete wavelet transform feature extraction |
-
2013
- 2013-12-31 CN CN201310750044.1A patent/CN103699800A/en active Pending
Non-Patent Citations (7)
| Title |
|---|
| M AHLSTROM等: "The future of Wind Forecasting and utility operations", 《IEEE POWER & ENERGY MAGAZINE》 * |
| 党香俊等: "基于相关分析和组合神经网络的退化预测", 《北京航空航天大学学报》 * |
| 朱跃龙等: "基于小波神经网络的水文时间序列预测", 《山东大学学报（工学版）》 * |
| 王妍等: "基于时间序列的相空间重构算法及验证（二）", 《山东大学学报（工学版）》 * |
| 王晓兰: "基于小波分解和最小二乘支持向量机的短期风速预测", 《电网技术》 * |
| 苏鹏宇等: "基于小波分解和ELMAN神经网络的短期风速多尺度预测方法", 《2012电站自动化信息化学术和技术交流会议论文集》 * |
| 陈盼等: "基于小波包和支持向量回归的风速预测", 《电网技术》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103902837B (en) * | 2014-04-16 | 2017-02-15 | 广西大学 | Method for wind speed prediction based on experience Copula function |
| CN104239964B (en) * | 2014-08-18 | 2017-10-10 | 华北电力大学 | Ultra-short term wind speed forecasting method based on spectral clustering and genetic optimization ExtremeLearningMachine |
| CN104333005A (en) * | 2014-08-18 | 2015-02-04 | 西南交通大学 | Electrical-power-system post-disturbance frequency dynamic-state prediction method based on support vector regression |
| CN104239964A (en) * | 2014-08-18 | 2014-12-24 | 华北电力大学 | Ultra-short-period wind speed prediction method based on spectral clustering type and genetic optimization extreme learning machine |
| CN105320995A (en) * | 2014-08-19 | 2016-02-10 | 上海交通大学 | Frequency domain decomposition based wind power generation short-term load prediction method and apparatus |
| CN105303051A (en) * | 2015-11-11 | 2016-02-03 | 中国科学院遥感与数字地球研究所 | Air pollutant concentration prediction method |
| CN105956252B (en) * | 2016-04-27 | 2019-04-02 | 哈尔滨工业大学 | Based on the multiple dimensioned Forecast model method of ultra-short term wind speed for generating moldeed depth belief network |
| CN105956252A (en) * | 2016-04-27 | 2016-09-21 | 哈尔滨工业大学 | Generative deep belief network-based multi-scale forecast modeling method for ultra-short-term wind speed |
| CN107292446A (en) * | 2017-07-03 | 2017-10-24 | 西南交通大学 | A kind of mixing wind speed forecasting method based on consideration component relevance wavelet decomposition |
| CN107292446B (en) * | 2017-07-03 | 2021-01-29 | 西南交通大学 | Hybrid wind speed prediction method based on component relevance wavelet decomposition |
| CN110288150A (en) * | 2019-06-24 | 2019-09-27 | 华北电力大学（保定） | Wind speed forecasting method, terminal device and storage medium |
| CN111339092A (en) * | 2020-02-24 | 2020-06-26 | 江苏省南通环境监测中心 | Deep learning-based multi-scale air quality forecasting method |
| CN111339092B (en) * | 2020-02-24 | 2023-09-08 | 江苏省南通环境监测中心 | Multi-scale air quality forecasting method based on deep learning |
| CN112132365A (en) * | 2020-11-24 | 2020-12-25 | 中国电力科学研究院有限公司 | Multi-scale feature-based forecast wind speed correction method and system |
| CN112132365B (en) * | 2020-11-24 | 2021-04-16 | 中国电力科学研究院有限公司 | Multi-scale feature-based forecast wind speed correction method and system |
| CN114912718A (en) * | 2022-07-13 | 2022-08-16 | 汉谷云智(武汉)科技有限公司 | Ultra-short-term solar irradiance prediction method based on discrete wavelet transform feature extraction |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103699800A (en) | Ultrashort-period wind speed prediction method based on frequency-domain multi-scale wind speed signal predictability | |
| CN103279804B (en) | The Forecasting Methodology of super short-period wind power | |
| Jiang et al. | Very short-term wind speed forecasting with Bayesian structural break model | |
| Alessandrini et al. | A novel application of an analog ensemble for short-term wind power forecasting | |
| Yang et al. | Probabilistic short-term wind power forecast using componential sparse Bayesian learning | |
| CN103117546B (en) | A kind of Ultrashort-term slide prediction method for wind power | |
| US20180240200A1 (en) | Method and device for modeling a long-time-scale photovoltaic output time sequence | |
| CN106875041A (en) | A kind of short-term wind speed forecasting method | |
| CN101793907A (en) | Short-term wind speed forecasting method of wind farm | |
| Shao et al. | Short‐term wind speed forecasting using wavelet transformation and AdaBoosting neural networks in Yunnan wind farm | |
| Lv et al. | Short-term wind speed forecasting based on non-stationary time series analysis and ARCH model | |
| CN107885951A (en) | A kind of Time series hydrological forecasting method based on built-up pattern | |
| CN106384170A (en) | Wavelet decomposition and reconstruction-based time sequence wind speed prediction method | |
| CN103020423A (en) | Copula-function-based method for acquiring relevant characteristic of wind power plant capacity | |
| CN103996084A (en) | Wind power probabilistic forecasting method based on longitudinal moment Markov chain model | |
| CN104615855A (en) | Day-ahead wind speed multistep prediction method fused with numerical weather prediction | |
| CN105354636A (en) | Wind power fluctuation probability density modeling method based on nonparametric kernel density estimation | |
| CN103065202A (en) | Wind power plant ultrashort term wind speed prediction method based on combination kernel function | |
| CN104182914A (en) | Wind power output time series modeling method based on fluctuation characteristics | |
| CN105373856A (en) | Wind electricity power short-term combined prediction method considering run detection method reconstruction | |
| CN106203732A (en) | Error in dipping computational methods based on ITD and time series analysis | |
| CN105356462A (en) | Wind farm harmonic evaluation prediction method | |
| CN102184335B (en) | Fire disaster time sequence prediction method based on ensemble empirical mode decomposition and phase space reconstruction | |
| CN112100911A (en) | Solar radiation prediction method based on deep BISLTM | |
| Xiyun et al. | Wind power probability interval prediction based on bootstrap quantile regression method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20140402 |
|
| RJ01 | Rejection of invention patent application after publication |