CN103345585A - Wind power prediction correction method and system based on support vector machine - Google Patents
Wind power prediction correction method and system based on support vector machine Download PDFInfo
- Publication number
- CN103345585A CN103345585A CN2013102940958A CN201310294095A CN103345585A CN 103345585 A CN103345585 A CN 103345585A CN 2013102940958 A CN2013102940958 A CN 2013102940958A CN 201310294095 A CN201310294095 A CN 201310294095A CN 103345585 A CN103345585 A CN 103345585A
- Authority
- CN
- China
- Prior art keywords
- wind power
- support vector
- vector machine
- data
- power prediction
- 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
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a wind power prediction correction method. The method includes the steps that 1, the total capacity of a selected wind power plant is obtained, and wind power prediction data and wind power actual measurement data of the whole field in a recent civil year of the wind power plant are obtained; 2, normalization processing is carried out on the wind power data, obtained from the step1, of the wind power plant by means of the total capacity of the wind power plant; 3, an input and output data set is formed according to the wind power prediction data and the wind power actual measurement data obtained after preprocessing in the step 2; 4, 2 / 3 of the input and output data set obtained in the step 3 is selected randomly to serve as a training set, and the remaining 1 / 3 serves as a testing set; 5, a kernel function and training parameters of the support vector machine are selected, training is carried out by means of the training set obtained from the step 4, and the testing set is used for testing; 6, a grid searching method is utilized to correct the parameters of the support vector machine, and an average absolute percentage error and a root-mean-square relative error of a correction result are utilized to serve as evaluation criteria to obtain a local optimum support vector machine training model, namely a local optimum wind power prediction correction model.
Description
Technical field
The present invention relates to generation of electricity by new energy and control field, relate in particular to a kind of wind power prediction bearing calibration and system based on support vector machine (SVM).
Background technology
Entered since the new century, the situation of fossil energy shortage and environmental pollution is more and more serious, impel power industry to seek the reproducible clean energy resource of exploitation and substitute existing chemical energy source, optimize energy structure, wherein, wind-powered electricity generation begins to be subject to people's attention gradually as a kind of reproducible clean energy resource of extensive existence.On the other hand, the primary energy wind energy of wind-powered electricity generation has very big undulatory property and intermittence, can cause bigger interference to electric system, therefore need carry out certain prediction to wind power, includes wind-powered electricity generation in the conventional power generation usage plan, better the management and use wind-powered electricity generation.According to the requirement of National Energy Board, wind energy turbine set should report predicted data () and real-time estimate () a few days ago at following 24 hours, 96 points at following 4 hours, 15 minutes points.But in the reality, the error of predicted data a few days ago that wind energy turbine set reports is very big, can not be used for the calculating of generation schedule a few days ago, otherwise the unusual out of true of generation schedule a few days ago, even may cause safety issue.Therefore, need exploitation one cover simply, method efficiently, the original predicted data of wind power that wind energy turbine set reports is proofreaied and correct, to satisfy the needs that generation schedule a few days ago calculates, this also meets the development trend of generation of electricity by new energy and control.
Summary of the invention
(1) technical matters that will solve
The technical problem to be solved in the present invention is: how a kind of simple and convenient, economical and practical wind power prediction bearing calibration is provided.
(2) technical scheme
For addressing the above problem, the invention provides and the invention discloses a kind of wind power prediction bearing calibration, comprise step: S1. obtains selected wind energy turbine set total volume, obtains this wind energy turbine set interior whole audience wind power prediction data of a nearest calendar year and wind power measured data; S2. utilize the wind energy turbine set total volume, the wind energy turbine set wind power data that step S1 is obtained carries out normalized; S3. pretreated wind power prediction data and the wind power measured data that obtains according to step S2 forms input, output data set; S4. picked at random step S3 imported, output data set 2/3 as training set, remaining 1/3 as inspection set; S5. choose support vector machine kernel function and training parameter, the training set that utilizes step S4 to obtain is trained, and utilizes inspection set to test; S6. utilize the grid search method, revise the parameter of support vector machine, and utilize to proofread and correct result's mean absolute percentage error and root mean square relative error as evaluation criterion, obtain local optimum support vector machine training pattern, i.e. local optimum wind power prediction calibration model.
Preferably, described step S6 comprises: S6.1 sets the discrete value set of regularization parameter and the discrete value set of RBF function, and one has 121 combinations; S6.2 makes up for each, carries out step S4 and step S5, and records mean absolute percentage error and root mean square relative error that each combination obtains down; S6.3 chooses a pair of combination of mean absolute percentage error and root mean square relative error minimum, as the parameter of local optimum support vector machine, forms local optimum wind power prediction calibration model.
On the other hand, the present invention also provides a kind of wind power prediction corrective system, it is characterized in that, comprising: first module, be used for obtaining selected wind energy turbine set total volume, and obtain this wind energy turbine set interior whole audience wind power prediction data of a nearest calendar year and wind power measured data; Second module is used for utilizing the wind energy turbine set total volume, and the wind energy turbine set wind power data that step S1 is obtained carries out normalized; The 3rd module for the pretreated wind power prediction data and the wind power measured data that obtain according to step S2, forms input, output data set; Four module, be used for that picked at random step S3 is imported, output data set 2/3 as training set, remaining 1/3 as inspection set; The 5th module is used for choosing support vector machine kernel function and training parameter, and the training set that utilizes step S4 to obtain is trained, and utilizes inspection set to test; The 6th module, be used for utilizing the grid search method, revise the parameter of support vector machine, and utilize the mean absolute percentage error of proofreading and correct the result and root mean square relative error as evaluation criterion, obtain local optimum support vector machine training pattern, i.e. local optimum wind power prediction calibration model.
Preferably, described the 6th module comprises: first submodule, be used for setting the discrete value set of regularization parameter and the discrete value set of RBF function, and one has 121 combinations; Second submodule is used for for each combination, carries out step S4 and step S5, and records mean absolute percentage error and root mean square relative error that each combination obtains down; The 3rd submodule is used for choosing a pair of combination of mean absolute percentage error and root mean square relative error minimum, as the parameter of local optimum support vector machine, forms local optimum wind power prediction calibration model.
(3) beneficial effect
The present invention trains wind power prediction historical data and wind power actual measurement historical data by support vector machine, can proofread and correct the wind power prediction data, obtain wind power prediction data more accurately, thereby realized system's wind power is better controlled, had following beneficial effect:
1) can improve the control technology level of electrical network, overcome factors such as existing wind power prediction means deficiency, precision of prediction be relatively poor relatively, scheduling and the running quality of electrical network have been improved, prediction electrical network development in future can very big limit ground be improved economy and the quality of power supply of the green electric power supply system operation that contains high wind-powered electricity generation permeability;
2) computing velocity is fast, can satisfy the requirement in line computation;
3) simple and convenient, do not need to obtain numerical weather forecast information, can be applied to multiple different occasion.
Description of drawings
Fig. 1 is the process flow diagram according to the wind power prediction bearing calibration of one embodiment of the present invention.
Embodiment
Below in conjunction with drawings and Examples, the specific embodiment of the present invention is described in further detail.Following examples are used for explanation the present invention, but are not used for limiting the scope of the invention.
Method of the present invention is introduced electric system according to the recessive characteristics that comprise the numerical weather forecast result among the conventional wind power prediction result with the thought of classification prediction in the data mining theories.As shown in Figure 1, the wind power prediction bearing calibration based on support vector machine according to one embodiment of the present invention comprises step:
S1. obtain selected wind energy turbine set total volume P
Specified, obtain this wind energy turbine set interior whole audience wind power prediction data of a nearest calendar year and wind power measured data, be designated as respectively
T=1 wherein ..., N represent wind power prediction data and wind power measured data in 1 year the markers ordered series of numbers (wherein, the total number of target when N represents, N=96 * D, D represent the total fate in a year, but according to actual conditions value 365 or 366).According to the regulation of China energy office, the wind power data that wind energy turbine set reports is data point of 15min, and therefore there are 96 data points in a nature sky, and there be 96 * D data point a calendar year;
S2. utilize the wind energy turbine set total volume, the wind energy turbine set wind power data that step S1 is obtained carries out normalized, and the data after the processing are designated as P respectively
Prediction(T) and P
Actual measurement(T).Wherein, the normalization formula is as (S2-1) with (S2-2);
S3. pretreated wind power prediction data and the wind power measured data that obtains according to step S2 forms input, output data set.Wherein, each group input data comprises 9 data unit, is respectively P
Prediction(t+1), P
Prediction(t+2), P
Prediction(t+3), P
Prediction(t+4), P
Prediction(t+5), P
Actual measurement(t+1), P
Actual measurement(t+2), P
Actual measurement(t+3) and P
Actual measurement(t+4), each group output data comprises 1 data unit, is P
Actual measurement(t+5), namely utilize the wind power prediction value of previous hour wind power prediction value and real output value and current time as input, output valve is the corrected value of current time wind power prediction value.According to this rule, can form the input of N-4 group, output data groups altogether, be used for subsequent calculations;
S4. picked at random step S3 obtain the data group 2/3 as training set, remaining 1/3 as inspection set.Then input, output data groups are organized in one total [2 (N-4)/3] in the training set, total (N-4)-[2 (N-4)/a 3] group input, output data groups in the inspection set, and [*] is that rounding operation accords with;
S5. choose suitable support vector machine kernel function and training parameter, wherein kernel function is used radial basis function (RBF), and regularization parameter γ is initially set 1, RBF function parameter σ
2Be initially set 1, the training set that utilizes step S4 to obtain is trained.Then, utilize inspection set to test, respectively calculation correction result's mean absolute percentage error e
MAPEWith root mean square relative error e
MSEWherein, e
MAPEAnd e
MSEComputing method suc as formula (S5-1) with (S5-2);
Wherein, Y
iThe ideal output data that exist in the expression inspection set,
The actual output data through the support vector computer that expression is corresponding, n represents the quantity of input, output data groups in the inspection set.
S6. utilize the grid search method, revise parameter γ and the σ of support vector machine
2, and utilize the mean absolute percentage error e that proofreaies and correct the result
MAPEWith root mean square relative error e
MSEAs evaluation criterion, obtain local optimum support vector machine training pattern, i.e. local optimum wind power prediction calibration model.
Wherein, step S6 further comprises:
S6.1 set the γ value be [10-5,10-4 ..., 104,105], set σ
2Value be [10-5,10-4 ..., 104,105], one has 121 (γ, σ
2) combination;
S6.2 is for each (γ, σ
2) combination, carry out step S4 and step S5, and record the mean absolute percentage error e that each combination obtains down
MAPEWith root mean square relative error e
MSE
S6.3 chooses mean absolute percentage error e
MAPEWith root mean square relative error e
MSEMinimum a pair of (γ, σ
2) combination, as the parameter of local optimum support vector machine, form local optimum wind power prediction calibration model.
Wind power prediction correcting algorithm based on support vector machine (SVM) of the present invention can be used among the economic load dispatching automated system of each provincial electric system of China and local wind electric field, can improve security and the accuracy of output of wind electric field prediction, have great economic and social benefit.
The above only is preferred implementation of the present invention; should be pointed out that for those skilled in the art, under the prerequisite that does not break away from the technology of the present invention principle; can also make some improvement and replacement, these improvement and replacement also should be considered as protection scope of the present invention.
Claims (4)
1. a wind power prediction bearing calibration is characterized in that, comprises step:
S1. obtain selected wind energy turbine set total volume, obtain this wind energy turbine set interior whole audience wind power prediction data of a nearest calendar year and wind power measured data;
S2. utilize the wind energy turbine set total volume, the wind energy turbine set wind power data that step S1 is obtained carries out normalized;
S3. pretreated wind power prediction data and the wind power measured data that obtains according to step S2 forms input, output data set;
S4. picked at random step S3 imported, output data set 2/3 as training set, remaining 1/3 as inspection set;
S5. choose support vector machine kernel function and training parameter, the training set that utilizes step S4 to obtain is trained, and utilizes inspection set to test;
S6. utilize the grid search method, revise the parameter of support vector machine, and utilize to proofread and correct result's mean absolute percentage error and root mean square relative error as evaluation criterion, obtain local optimum support vector machine training pattern, i.e. local optimum wind power prediction calibration model.
2. the method for claim 1 is characterized in that, described step S6 comprises:
S6.1 sets the discrete value set of regularization parameter and the discrete value set of RBF function, and one has 121 combinations;
S6.2 makes up for each, carries out step S4 and step S5, and records mean absolute percentage error and root mean square relative error that each combination obtains down;
S6.3 chooses a pair of combination of mean absolute percentage error and root mean square relative error minimum, as the parameter of local optimum support vector machine, forms local optimum wind power prediction calibration model.
3. a wind power prediction corrective system is characterized in that, comprising:
First module is used for obtaining selected wind energy turbine set total volume, obtains this wind energy turbine set interior whole audience wind power prediction data of a nearest calendar year and wind power measured data;
Second module is used for utilizing the wind energy turbine set total volume, and the wind energy turbine set wind power data that step S1 is obtained carries out normalized;
The 3rd module for the pretreated wind power prediction data and the wind power measured data that obtain according to step S2, forms input, output data set;
Four module, be used for that picked at random step S3 is imported, output data set 2/3 as training set, remaining 1/3 as inspection set;
The 5th module is used for choosing support vector machine kernel function and training parameter, and the training set that utilizes step S4 to obtain is trained, and utilizes inspection set to test;
The 6th module, be used for utilizing the grid search method, revise the parameter of support vector machine, and utilize the mean absolute percentage error of proofreading and correct the result and root mean square relative error as evaluation criterion, obtain local optimum support vector machine training pattern, i.e. local optimum wind power prediction calibration model.
4. system as claimed in claim 3 is characterized in that, described the 6th module comprises:
First submodule is used for setting the discrete value set of regularization parameter and the discrete value set of RBF function, and one has 121 combinations;
Second submodule is used for for each combination, carries out step S4 and step S5, and records mean absolute percentage error and root mean square relative error that each combination obtains down;
The 3rd submodule is used for choosing a pair of combination of mean absolute percentage error and root mean square relative error minimum, as the parameter of local optimum support vector machine, forms local optimum wind power prediction calibration model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013102940958A CN103345585A (en) | 2013-07-12 | 2013-07-12 | Wind power prediction correction method and system based on support vector machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2013102940958A CN103345585A (en) | 2013-07-12 | 2013-07-12 | Wind power prediction correction method and system based on support vector machine |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103345585A true CN103345585A (en) | 2013-10-09 |
Family
ID=49280380
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2013102940958A Pending CN103345585A (en) | 2013-07-12 | 2013-07-12 | Wind power prediction correction method and system based on support vector machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103345585A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103761580A (en) * | 2013-12-31 | 2014-04-30 | 北华大学 | Energy consumption supervision method capable of achieving energy dynamic prediction for beer brewing enterprises |
| CN103955755A (en) * | 2014-04-18 | 2014-07-30 | 国家电网公司 | Wind electricity power short-term prediction method by adopting composite data source based on self-learning polynomial kernel function support vector machine |
| CN105046057A (en) * | 2015-06-24 | 2015-11-11 | 上海大学 | LSSVM (Least Square Support Vector Machine) pulsation wind speed prediction method based on Morlet wavelet kernel |
| CN105302848A (en) * | 2014-10-11 | 2016-02-03 | 山东鲁能软件技术有限公司 | Evaluation value calibration method of equipment intelligent early warning system |
| CN105701562A (en) * | 2016-01-05 | 2016-06-22 | 上海思源弘瑞自动化有限公司 | Training method, suitable method of predicating generated power and respective systems |
| CN107609697A (en) * | 2017-09-06 | 2018-01-19 | 南京邮电大学 | A kind of Wind power forecasting method |
| CN108304976A (en) * | 2018-03-06 | 2018-07-20 | 西安交通大学 | A kind of electric system fining load prediction and analysis method |
| CN109978204A (en) * | 2017-12-27 | 2019-07-05 | 北京金风慧能技术有限公司 | The method and apparatus for determining the predictablity rate of the wind power of wind power plant |
| CN114462710A (en) * | 2022-02-11 | 2022-05-10 | 华润电力技术研究院有限公司 | Short-term prediction method, device and medium for fan generated power |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101916998A (en) * | 2010-07-12 | 2010-12-15 | 东北电力科学研究院有限公司 | Support vector machine-based wind electric powder prediction device and method |
| CN102102626A (en) * | 2011-01-30 | 2011-06-22 | 华北电力大学 | Method for forecasting short-term power in wind power station |
| CN102938093A (en) * | 2012-10-18 | 2013-02-20 | 安徽工程大学 | Wind power forecasting method |
-
2013
- 2013-07-12 CN CN2013102940958A patent/CN103345585A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101916998A (en) * | 2010-07-12 | 2010-12-15 | 东北电力科学研究院有限公司 | Support vector machine-based wind electric powder prediction device and method |
| CN102102626A (en) * | 2011-01-30 | 2011-06-22 | 华北电力大学 | Method for forecasting short-term power in wind power station |
| CN102938093A (en) * | 2012-10-18 | 2013-02-20 | 安徽工程大学 | Wind power forecasting method |
Non-Patent Citations (4)
| Title |
|---|
| 史洁: "《基于支持向量机的风电场功率短期预测方法研究》", 《中国优秀硕士论文全文数据库》 * |
| 叶徐静: "《基于最小二乘支持向量机的风电功率超短期预测》", 《电源学报》 * |
| 周松林 等: "《基于预测误差校正的支持向量机短期风速预测》", 《***仿真学报》 * |
| 王兴玲 等: "《基于网格搜索的支持向量机核函数参数的确定》", 《中国海洋大学学报》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103761580A (en) * | 2013-12-31 | 2014-04-30 | 北华大学 | Energy consumption supervision method capable of achieving energy dynamic prediction for beer brewing enterprises |
| CN103955755A (en) * | 2014-04-18 | 2014-07-30 | 国家电网公司 | Wind electricity power short-term prediction method by adopting composite data source based on self-learning polynomial kernel function support vector machine |
| CN105302848A (en) * | 2014-10-11 | 2016-02-03 | 山东鲁能软件技术有限公司 | Evaluation value calibration method of equipment intelligent early warning system |
| CN105302848B (en) * | 2014-10-11 | 2018-11-13 | 山东鲁能软件技术有限公司 | A kind of assessed value calibration method of device intelligence early warning system |
| CN105046057A (en) * | 2015-06-24 | 2015-11-11 | 上海大学 | LSSVM (Least Square Support Vector Machine) pulsation wind speed prediction method based on Morlet wavelet kernel |
| CN105046057B (en) * | 2015-06-24 | 2019-04-02 | 上海大学 | LSSVM fluctuating wind speed prediction technique based on Morlet Wavelet Kernel |
| CN105701562A (en) * | 2016-01-05 | 2016-06-22 | 上海思源弘瑞自动化有限公司 | Training method, suitable method of predicating generated power and respective systems |
| CN105701562B (en) * | 2016-01-05 | 2019-12-17 | 上海思源弘瑞自动化有限公司 | Training method, applicable method for predicting generated power and respective system |
| CN107609697A (en) * | 2017-09-06 | 2018-01-19 | 南京邮电大学 | A kind of Wind power forecasting method |
| CN109978204A (en) * | 2017-12-27 | 2019-07-05 | 北京金风慧能技术有限公司 | The method and apparatus for determining the predictablity rate of the wind power of wind power plant |
| CN109978204B (en) * | 2017-12-27 | 2021-04-06 | 北京金风慧能技术有限公司 | Method and device for determining prediction accuracy of wind power plant |
| CN108304976A (en) * | 2018-03-06 | 2018-07-20 | 西安交通大学 | A kind of electric system fining load prediction and analysis method |
| CN114462710A (en) * | 2022-02-11 | 2022-05-10 | 华润电力技术研究院有限公司 | Short-term prediction method, device and medium for fan generated power |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Esteban et al. | 100% renewable energy system in Japan: Smoothening and ancillary services | |
| CN103345585A (en) | Wind power prediction correction method and system based on support vector machine | |
| Liu et al. | Forecasting power output of photovoltaic system using a BP network method | |
| CN102269124B (en) | Ultra-short term wind power station generated power forecasting system | |
| Shi et al. | Hybrid forecasting model for very-short term wind power forecasting based on grey relational analysis and wind speed distribution features | |
| Kondziella et al. | Flexibility requirements of renewable energy based electricity systems–a review of research results and methodologies | |
| Foley et al. | Wind power forecasting & prediction methods | |
| Zeng et al. | Short-term wind power prediction using a wavelet support vector machine | |
| CN107609697B (en) | A kind of Wind power forecasting method | |
| CN102479347B (en) | Method and system for forecasting short-term wind speed of wind farm based on data driving | |
| CN102509173B (en) | A kind of based on markovian power system load Accurate Prediction method | |
| CN109978242A (en) | The photovoltaic power generation cluster power forecasting method and device of scale are risen based on statistics | |
| Chuang et al. | Analyzing major renewable energy sources and power stability in Taiwan by 2030 | |
| Mao et al. | A novel short-term wind speed prediction based on MFEC | |
| CN106897793B (en) | Genetic algorithm-based wind power plant wind driven generator arrangement optimization method capable of guaranteeing safe distance | |
| Mohan et al. | Solar energy disaggregation using whole-house consumption signals | |
| Majumder et al. | Selection of features for analysis of reliability of performance in hydropower plants: a multi-criteria decision making approach | |
| CN103366225A (en) | Wind power prediction error identification method | |
| Alanazi et al. | Wind energy assessment using Weibull distribution with different numerical estimation methods: a case study | |
| Eltohamy et al. | Technical investigation for power system flexibility | |
| CN115577903A (en) | Method, device, equipment and medium for evaluating influence of new energy on power grid peak regulation | |
| CN104657787A (en) | Wind power time series combined prediction method | |
| Sarkar | Minding the P's and Q's: Real and reactive power assessment of hybrid energy conversion systems with wind and solar resources | |
| Ding et al. | Optimal intra-day coordination of wind farm and pumped-hydro-storage plant | |
| CN110874611B (en) | Wind power cluster power ultra-short term prediction error grade grading compensation 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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20131009 |