CN106960257A - A kind of wind power combination Forecasting Methodology of noise auxiliary signal decomposition method and Elman neutral nets - Google Patents
A kind of wind power combination Forecasting Methodology of noise auxiliary signal decomposition method and Elman neutral nets Download PDFInfo
- Publication number
- CN106960257A CN106960257A CN201710183612.2A CN201710183612A CN106960257A CN 106960257 A CN106960257 A CN 106960257A CN 201710183612 A CN201710183612 A CN 201710183612A CN 106960257 A CN106960257 A CN 106960257A
- Authority
- CN
- China
- Prior art keywords
- formula
- wind power
- elman
- imf
- data
- 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
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
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The combination forecasting method of a kind of noise auxiliary signal decomposition method based on complex data empirical mode decomposition and Elman neutral nets carries out short-term forecast to wind power, belongs to wind power prediction technical field.Including:Step one:White noise is mixed in original signal sequence;Step 2 is to seven:Solve and obtain IMF components and surplus;Step 8:Elman neural network prediction models are built using obtained IMF components and surplus and are predicted, and are finally always superimposed, are finally predicted the outcome.The present invention provides the wind power combination Forecasting Methodology of a kind of noise auxiliary signal decomposition method and Elman neutral nets, and this algorithm can further reduce modal overlap problem present in old EEMD decomposition methods, improve short-term wind power prediction precision.
Description
Technical field
A kind of noise auxiliary signal decomposition method based on complex data empirical mode decomposition and Elman god are proposed in the present invention
Combination forecasting method through network carries out short-term forecast to wind power, belongs to wind power prediction technical field.
Background technology
Because the randomness of wind speed causes wind power to have nonlinear feature, therefore, it is possible to handle nonlinear problem
Elman neural networks possess some superiority.But it is existing studies have found that, wind power prediction essence is carried out using Individual forecast method
Degree need to be improved, because the precision of prediction of Individual forecast method is easily disturbed by data high frequency components, so as to cause
Precision of prediction is reduced.In order to solve the non-stationary interference to predicting the outcome of wind power, there is scholar to propose combined prediction
Thought.Mainly on the basis of existing Individual forecast method, the preprocessing process that increase data are decomposed, first by wind power number
Forecasting Methodology is then recycled to be predicted it according to multiple independent component is decomposed into.
The content of the invention
The deficiency existed for existing composite prediction technology, the present invention provide a kind of noise auxiliary signal decomposition method with
The wind power combination Forecasting Methodology of Elman neutral nets, this algorithm can be reduced further to be existed in old EEMD decomposition methods
Modal overlap problem, improve short-term wind power prediction precision.
The technical scheme that the present invention takes is:
The wind power combination Forecasting Methodology of a kind of noise auxiliary signal decomposition method and Elman neutral nets, including following step
Suddenly:
Step 1:White noise, the complex signal x of composition are mixed in primary signalc(t), as shown in formula (1):
xc(t)=xo(t)+ixv(t) (1)
In formula:xo(t) it is primary signal, xv(t) it is the white noise of limited extent, xc(t) it is complex signal.
Step 2:Determine projecting directionAnd by complex signal xc(t) project toOn:
In formula:Represent xc(t) projection in all directions,Projecting direction is represented, and 1≤n≤N, n are iteration time
Number, i is imaginary unit.
Step 3:Euler's formula e is substituted into (2)-ix=cosx-isinx, and carry out abbreviation and obtain:
Step 4:ExtractLocal maximumAgain to setEnter row interpolation, obtain in side
ToOn maximum envelope
Step 5:Calculate maximum envelope barycenter m (t) on all directions:
In formula:N represents maximum iteration.
Step 6:After setting is decomposed by NACEMD, it is intrinsic mode function (IMF) each to decompose obtained component,
And its following two conditions must are fulfilled for, actual conditions setting comprises the following steps:
Step 6.1:In whole signal sequence, the number of extreme point and zero crossing must be equal or at most differs one;
Step 6.2:The lower envelope determined at any point, the coenvelope and local minimum determined by local maximum
Average value is necessary for zero.
Step 7:Judge whether h (t)=x (t)-m (t) meets IMF condition, the IMF1=h (t) if meeting, if not being inconsistent
Close and require then by h (t) as new signal x (t), and repeat step 2 is to step 7, and each rank IMF and surplus are asked for successively,
Until x (t) meets the stop condition in formula (5).
In formula:SD is screening criteria value, is typically taken between 0.2 to 0.3, T is time constant, hn(t) it is test signal sequence
Row, wherein n is iterations, and t is time independent variable.
Step 8:The step of NACEMD decomposition methods are combined with Elman neural network prediction methods, is accorded with after step 7 end
Each IMF vector sequences and a surplus of standardization, then build Elman neural network prediction models pair for each IMF components
It is predicted, and the result predicted always is superimposed, obtain it is final predict the outcome, and it is compared with reality output
Compared with.
Comprise the following steps in step 8:
Step 8.1:Using the IMF sequences obtained in step 7, Elman neural network prediction models are built to it respectively and are entered
Row prediction, what it is due to NACEMD methods structure is 2-D data, it is therefore desirable to which different magnitude of data are normalized,
The feature for preventing some numerical value low is submerged.On the other hand data are normalized, invalid data can be reduced to mould
The influence of type precision, can accelerate model convergence rate.The normalization specifically used uses following formal approach:
Normalization:
Renormalization:
xi=(xmax-xmin)yi+xmin (7)
In formula:yiIt is the result of a certain data normalization in training sample T;xmaxAnd xminIt is to change in training sample T respectively
The maximum and minimum value of group variable data.
Step 8.2:Determine the non-linear state space expression of Elman neutral nets:
In formula:D represents moment variable, and y represents one-dimensional output node vector, and x represents m dimension hidden layer node vectors, and u is represented
N dimensional input vectors, xcRepresent m dimension feedback state vectors, p3, p2, p1Be connection weight matrix, respectively connect hidden layer with it is defeated
Go out layer, input layer and hidden layer and articulamentum and hidden layer.F () is hidden layer transmission function expression formula, b1With b2Represent defeated
Enter the threshold value of layer and hidden layer.
The present invention is the wind power combination Forecasting Methodology of a kind of noise auxiliary signal decomposition method and Elman neutral nets, its
Advantage is:
1st, the thought of combined prediction is introduced in short-term wind power prediction, and simulation results show combinatorial forecast is pre-
Survey precision and be higher than Individual forecast method.
2nd, decomposed using NACEMD and complicated wind power sequence is decomposed into multiple independent IMF components, effectively mitigated
While the non-stationary interference of wind power, additionally it is possible to reduce in EEMD decomposable processes due to newly being produced using integrated average thought
Raw modal overlap phenomenon, then improves the precision of short-term wind power prediction.This demonstrate that being used in short-term wind power prediction
NACEMD-Elman neural network ensemble predicted methods are better than EEMD-Elman neural network ensemble predicted methods, and combinatorial forecast
Precision of prediction be higher than Individual forecast method.
Brief description of the drawings
Fig. 1 is the wind power chosen.
Fig. 2 is component and surplus figure of the wind power after NACEMD is decomposed.
Fig. 3 is component and surplus figure of the wind power after EEMD is decomposed.
Fig. 4 is NACEMD-Elman neural network prediction result figures.
Fig. 5 is EEMD-Elman neural network prediction result figures.
Fig. 6 is Elman neural network prediction result figures.
Embodiment
The wind power combination Forecasting Methodology of a kind of noise auxiliary signal decomposition method and Elman neutral nets, including following step
Suddenly:
Step 1:White noise, the complex signal x of composition are mixed in primary signalc(t), as shown in formula (1):
xc(t)=xo(t)+ixv(t) (9)
In formula:xo(t) it is primary signal, xv(t) it is the white noise of limited extent, xc(t) it is complex signal.
Step 2:Determine projecting directionAnd by complex signal xc(t) project toOn:
In formula:Represent xc(t) projection in all directions,Projecting direction is represented, and 1≤n≤N, n are iteration time
Number, i is imaginary unit.
Step 3:Euler's formula e is substituted into (2)-ix=cosx-isinx, and carry out abbreviation and obtain:
Step 4:ExtractLocal maximumAgain to setEnter row interpolation, obtain in side
ToOn maximum envelope
Step 5:Calculate maximum envelope barycenter m (t) on all directions:
In formula:N represents maximum iteration.
Step 6:After setting is decomposed by NACEMD, it is intrinsic mode function (IMF) each to decompose obtained component,
And its following two conditions must are fulfilled for, actual conditions setting comprises the following steps:
Step 6.1:In whole signal sequence, the number of extreme point and zero crossing must be equal or at most differs one;
Step 6.2:The lower envelope determined at any point, the coenvelope and local minimum determined by local maximum
Average value is necessary for zero.
Step 7:Judge whether h (t)=x (t)-m (t) meets IMF condition, the IMF1=h (t) if meeting, if not being inconsistent
Close and require then by h (t) as new signal x (t), and repeat step 2 is to step 7, and each rank IMF and surplus are asked for successively,
Until x (t) meets the stop condition in formula (5).
In formula:SD is screening criteria value, is typically taken between 0.2 to 0.3, T is time constant, hn(t) it is test signal sequence
Row, wherein n is iterations, and t is time independent variable.
Step 8:The step of NACEMD decomposition methods are combined with Elman neural network prediction methods, is accorded with after step 7 end
Each IMF vector sequences and a surplus of standardization, then build Elman neural network prediction models pair for each IMF components
It is predicted, and the result predicted always is superimposed, obtain it is final predict the outcome, and it is compared with reality output
Compared with.
Comprise the following steps in step 8:
Step 8.1:Using the IMF sequences obtained in step 7, Elman neural network prediction models are built to it respectively and are entered
Row prediction, what it is due to NACEMD methods structure is 2-D data, it is therefore desirable to which different magnitude of data are normalized,
The feature for preventing some numerical value low is submerged.On the other hand data are normalized, invalid data can be reduced to mould
The influence of type precision, can accelerate model convergence rate.The normalization specifically used uses following formal approach:
Normalization:
Renormalization:
xi=(xmax-xmin)yi+xmin (15)
In formula:yiIt is the result of a certain data normalization in training sample T;xmaxAnd xminIt is to change in training sample T respectively
The maximum and minimum value of group variable data.
Step 8.2:Determine the non-linear state space expression of Elman neutral nets:
In formula:D represents moment variable, and y represents one-dimensional output node vector, and x represents m dimension hidden layer node vectors, and u is represented
N dimensional input vectors, xcRepresent m dimension feedback state vectors, p3, p2, p1Be connection weight matrix, respectively connect hidden layer with it is defeated
Go out layer, input layer and hidden layer and articulamentum and hidden layer.F () is hidden layer transmission function expression formula, b1With b2Represent defeated
Enter the threshold value of layer and hidden layer.
From Fig. 2, Fig. 3, NACEMD and EEMD both approaches can effectively be decomposed to wind-powered electricity generation time series,
The IMF1-IMF2 components of the two are respectively provided with high-frequency, and fluctuation is big and periodically indefinite feature.Component IMF3- in Fig. 2
IMF5 frequencies are relatively low, and with periodically, surplus r (t) is a more stable curve.In Fig. 3, component IMF3-IMF6 frequencies
Relatively low to have periodically, surplus r (t) is the curve of a monotone decreasing.
Modal overlap phenomenon is reduced by comparison diagram 2 and Fig. 3, NACEMD and EEMD, and can be isolated
High frequency intermittent vibrates and low frequency base signal, but both approaches also have certain difference:First, different time-frequencies are special in figure 3
Property white noise carry out it is integrated can averagely cause EEMD obvious modal overlap phenomenon occur in low frequency part, be mainly reflected in phase
The signal of nearly yardstick is appeared in IMF4 and IMF5.Secondly, by comparing the two IMF quantity decomposited, NACEMD points
The IMF quantity of solution is considerably less than what EEMD was decomposited, it was demonstrated that NACEMD is decomposed and be can be derived that more effective IMF components, and then
Existing in the IMF that proof EEMD is decomposed can not be by the integrated average noise signal to eliminate, therefore NACEMD is decomposed better than EEMD
Decompose.
As shown in Figure 4, the prediction output of NACEMD-Elman short-term wind powers combinatorial forecast can be immediately following real data
Variation tendency, with relatively low error, it was demonstrated that the correctness and validity of the forecast model proposed in the present invention.By Fig. 5 with
Fig. 4 is compared to knowable to, and prediction output and the real data of EEMD-Elman combinatorial forecasts have certain error, and prediction output
The variation tendency for the data that can not effectively gear to actual circumstances.By Fig. 6 compared with Fig. 4 it is visible, the prediction of Individual forecast method output with it is actual
Data are complete to have obvious deviation.
Embodiment:
The actual measurement wind power data that the present invention chooses a certain unit in the wind power plant of somewhere carries out wind power prediction as example,
Its sampling period is 10min, and the rated power of unit is 850kW.To reduce human intervention, therefore to the greatest extent may be used using downtime point
Data segment that can be few carries out simulation analysis, chooses 360 continuous power data points, and first 288 are used to train, and latter 72 are used to survey
Examination and analysis.
The important component that quantitative assessment is prediction effect analysis is carried out for the precision and reliability predicted the outcome.This
Elman neural network prediction models are built in invention based on each IMF components, wherein each IMF components correspondence forecast model ginseng
As shown in table 1, wherein MSE is that mean square error, MAPE are mean absolute percentage error and MSPE to be square for number and its error
Percentage error.
Each IMF prediction model parameterses of table 1 and error
Tab.1 The IMF prediction model parameters and errors
The model performance index of table 2
Tab.2 Model performance
For checking the present invention proposed in the combinatorial forecast based on NACEMD-Elman neutral nets validity and
Feasibility, by three kinds of forecast models and its result, using MSE, MAPE, the MSPE and the degree of correlation being noted above, these four refer to
Mark is analyzed.
It can be found that by based on NACEMD-Elman neutral nets and the short-term wind work(based on EEMD-Elman neutral nets
Rate forecast model is compared, and the error amount based on NACEMD-Elman neural network prediction models is low and degree of correlation highest, reaches
To 0.9608.It follows that its model performance index is better than the former, it was demonstrated that NACEMD decomposition methods are better than EEMD under the present conditions
Decomposition method.
The model performance index of Individual forecast method and combinatorial forecast is compared, it can be found that the mistake of combinatorial forecast
Difference is lower, and the degree of correlation is higher, and its model performance index is better than Individual forecast method.It follows that combinatorial forecast is in conditions present
It is better than Individual forecast method down.
As shown in Table 2, model accuracy is carried compared with the performance indications of each other forecast model, in the present invention higher,
With certain advance.
Short-term wind power prediction method based on NACEMD-Elman neutral nets is divided original wind power sequence first
Solution, obtains component and surplus, then builds Elman forecast models to it respectively, and the superposition of each forecast model finally is obtained into final pre-
Survey result.From Fig. 4, Fig. 5, Fig. 6 and table 2, method provided by the present invention can avoid the non-stationary to pre- of wind power
The influence of survey is both the influence for solving the problems, such as modal overlap to the full extent to precision of prediction, compared to EEMD-Elman nerve nets
Network combinatorial forecast, Elman neutral net Individual forecast methods, its precision of prediction is higher.
The present invention is illustrated according to the preferred embodiment, but above-described embodiment does not limit the present invention in any form, all
The technical scheme obtained using the form of equivalent or equivalent transformation, in the range of all falling within technical solution of the present invention.
Claims (2)
1. a kind of wind power combination Forecasting Methodology of noise auxiliary signal decomposition method and Elman neutral nets, it is characterised in that bag
Include following steps:
Step 1:White noise, the complex signal x of composition are mixed in primary signalc(t), as shown in formula (1):
xc(t)=xo(t)+ixv(t) (1)
In formula:xo(t) it is primary signal, xv(t) it is the white noise of limited extent, xc(t) it is complex signal;
Step 2:Determine projecting directionAnd by complex signal xc(t) project toOn:
In formula:Represent xc(t) projection in all directions,Projecting direction is represented, and 1≤n≤N, n are iterations, i
For imaginary unit.
Step 3:Euler's formula e is substituted into (2)-ix=cosx-isinx, and carry out abbreviation and obtain:
Step 4:ExtractLocal maximumAgain to setEnter row interpolation, obtain in direction
On maximum envelope
Step 5:Calculate maximum envelope barycenter m (t) on all directions:
In formula:N represents maximum iteration.
Step 6:After setting is decomposed by NACEMD, it is intrinsic mode function (IMF) each to decompose obtained component, and its
Following two conditions must are fulfilled for, actual conditions setting comprises the following steps:
Step 6.1:In whole signal sequence, the number of extreme point and zero crossing must be equal or at most differs one;
Step 6.2:The lower envelope that is determined at any point, the coenvelope and local minimum determined by local maximum is averaged
Value is necessary for zero;
Step 7:Judge whether h (t)=x (t)-m (t) meets IMF condition, the IMF1=h (t) if meeting will if not meeting
Ask then by h (t) as new signal x (t), and repeat step 2 is to step 7, each rank IMF and surplus is asked for successively, until x
(t) stop condition in formula (5) is met:
In formula:SD is screening criteria value, is typically taken between 0.2 to 0.3, T is time constant, hn(t) it is test signal sequence, its
Middle n is iterations, and t is time independent variable.
Step 8:The step of NACEMD decomposition methods are combined with Elman neural network prediction methods, obtains meeting mark after step 7 end
Accurate each IMF vector sequences and a surplus, then it is entered for each IMF components structure Elman neural network prediction models
Row prediction, the result predicted always is superimposed, obtain it is final predict the outcome, and it is compared with reality output.
2. the wind power combination of a kind of noise auxiliary signal decomposition method and Elman neutral nets is predicted according to claim 1
Method, it is characterised in that comprise the following steps in step 8:
Step 8.1:Using the IMF sequences obtained in step 7, Elman neural network prediction models are built to it respectively and are carried out in advance
Survey, what it is due to NACEMD methods structure is 2-D data, it is therefore desirable to different magnitude of data are normalized, prevented
The low feature of some numerical value is submerged.On the other hand data are normalized, invalid data can be reduced to model essence
The influence of degree, can accelerate model convergence rate.The normalization specifically used uses following formal approach:
Normalization:
Renormalization:
xi=(xmax-xmin)yi+xmin (7)
In formula:yiIt is the result of a certain data normalization in training sample T;xmaxAnd xminIt is to reorganize variable in training sample T respectively
The maximum and minimum value of data;
Step 8.2:Determine the non-linear state space expression of Elman neutral nets:
In formula:D represents moment variable, and y represents one-dimensional output node vector, and x represents m dimension hidden layer node vectors, and u represents that n is tieed up
Input vector, xcRepresent m dimension feedback state vectors, p3, p2, p1Connection weight matrix is, hidden layer and output are connected respectively
Layer, input layer and hidden layer and articulamentum and hidden layer.F () is hidden layer transmission function expression formula, b1With b2Represent input
The threshold value of layer and hidden layer.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710183612.2A CN106960257A (en) | 2017-03-24 | 2017-03-24 | A kind of wind power combination Forecasting Methodology of noise auxiliary signal decomposition method and Elman neutral nets |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710183612.2A CN106960257A (en) | 2017-03-24 | 2017-03-24 | A kind of wind power combination Forecasting Methodology of noise auxiliary signal decomposition method and Elman neutral nets |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106960257A true CN106960257A (en) | 2017-07-18 |
Family
ID=59471272
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710183612.2A Pending CN106960257A (en) | 2017-03-24 | 2017-03-24 | A kind of wind power combination Forecasting Methodology of noise auxiliary signal decomposition method and Elman neutral nets |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106960257A (en) |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104899656A (en) * | 2015-06-05 | 2015-09-09 | 三峡大学 | Wind power combined predication method based on ensemble average empirical mode decomposition and improved Elman neural network |
CN106295798A (en) * | 2016-08-29 | 2017-01-04 | 江苏省电力试验研究院有限公司 | Empirical mode decomposition and Elman neural network ensemble wind-powered electricity generation Forecasting Methodology |
-
2017
- 2017-03-24 CN CN201710183612.2A patent/CN106960257A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104899656A (en) * | 2015-06-05 | 2015-09-09 | 三峡大学 | Wind power combined predication method based on ensemble average empirical mode decomposition and improved Elman neural network |
CN106295798A (en) * | 2016-08-29 | 2017-01-04 | 江苏省电力试验研究院有限公司 | Empirical mode decomposition and Elman neural network ensemble wind-powered electricity generation Forecasting Methodology |
Non-Patent Citations (2)
Title |
---|
曲建岭等: "基于复数据经验模态分解的噪声辅助信号分解方法", 《物理学报》 * |
阮晴: "超声无损检测缺陷识别方法研究", 《中国优秀硕士学位论文全文数据库(信息科技辑)》 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Aly | An intelligent hybrid model of neuro Wavelet, time series and Recurrent Kalman Filter for wind speed forecasting | |
CN109299430A (en) | The short-term wind speed forecasting method with extreme learning machine is decomposed based on two stages | |
Nfaoui et al. | A stochastic Markov chain model for simulating wind speed time series at Tangiers, Morocco | |
Hur | Short-term wind speed prediction using Extended Kalman filter and machine learning | |
Ghoushchi et al. | An extended new approach for forecasting short-term wind power using modified fuzzy wavelet neural network: A case study in wind power plant | |
US8041597B2 (en) | Self-calibrating outlier model and adaptive cascade model for fraud detection | |
CN107765347A (en) | A kind of Gaussian process returns and the short-term wind speed forecasting method of particle filter | |
CN107563067A (en) | Analysis of structural reliability method based on Adaptive proxy model | |
CN105868853B (en) | Method for predicting short-term wind power combination probability | |
CN107704953A (en) | The short-term wind-electricity power probability density Forecasting Methodology of EWT quantile estimate forests | |
CN104915534B (en) | Electric power tower deformation analysis based on Sequence Learning and decision-making technique | |
CN106846173A (en) | Short-term wind power forecast method based on EWT ESN | |
CN106295798A (en) | Empirical mode decomposition and Elman neural network ensemble wind-powered electricity generation Forecasting Methodology | |
CN106443285A (en) | Multi-harmonic-source harmonic contribution quantitative analysis method based on total-least square method | |
CN104573876A (en) | Wind power plant short-period wind speed prediction method based on time sequence long memory model | |
CN102622510A (en) | System and method for quantitative management of software defects | |
CN110417011A (en) | A kind of online dynamic secure estimation method based on mutual information Yu iteration random forest | |
CN101825622A (en) | Water quality prediction method and device | |
Arias-Rosales et al. | Wind turbine selection method based on the statistical analysis of nominal specifications for estimating the cost of energy | |
CN104269867A (en) | Node disturbance power transfer distribution balance degree analyzing method | |
CN107132762A (en) | Online Static security assessment method based on forecast failure collection automatic screening | |
CN104036328A (en) | Self-adaptive wind power prediction system and prediction method | |
CN108647807A (en) | The prediction technique of river discharge | |
CN104156542A (en) | Implicit-projection-based method for simulating stability of active power distribution system | |
Akyurek et al. | Tesla: Taylor expanded solar analog forecasting |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170718 |