CN108898251B - Offshore wind farm power prediction method considering meteorological similarity and power fluctuation - Google Patents
Offshore wind farm power prediction method considering meteorological similarity and power fluctuation Download PDFInfo
- Publication number
- CN108898251B CN108898251B CN201810696196.0A CN201810696196A CN108898251B CN 108898251 B CN108898251 B CN 108898251B CN 201810696196 A CN201810696196 A CN 201810696196A CN 108898251 B CN108898251 B CN 108898251B
- Authority
- CN
- China
- Prior art keywords
- prediction
- power
- power fluctuation
- data
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 26
- 238000013528 artificial neural network Methods 0.000 claims abstract description 14
- 238000005096 rolling process Methods 0.000 claims abstract description 10
- 238000012937 correction Methods 0.000 claims abstract description 9
- 238000004458 analytical method Methods 0.000 claims abstract description 3
- 238000012549 training Methods 0.000 claims description 11
- 238000012847 principal component analysis method Methods 0.000 claims description 4
- 239000013598 vector Substances 0.000 claims description 4
- 238000004422 calculation algorithm Methods 0.000 claims description 3
- 230000001186 cumulative effect Effects 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 2
- 230000008901 benefit Effects 0.000 abstract description 3
- 230000008859 change Effects 0.000 description 4
- 230000009194 climbing Effects 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- PEDCQBHIVMGVHV-UHFFFAOYSA-N Glycerine Chemical compound OCC(O)CO PEDCQBHIVMGVHV-UHFFFAOYSA-N 0.000 description 3
- 238000005311 autocorrelation function Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000001174 ascending effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000000053 physical method Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 101100129500 Caenorhabditis elegans max-2 gene Proteins 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000007621 cluster analysis Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000013213 extrapolation Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000013179 statistical model Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
-
- H02J3/386—
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A30/00—Adapting or protecting infrastructure or their operation
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E10/00—Energy generation through renewable energy sources
- Y02E10/70—Wind energy
- Y02E10/76—Power conversion electric or electronic aspects
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Physics & Mathematics (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Power Engineering (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)
- Supply And Distribution Of Alternating Current (AREA)
- Wind Motors (AREA)
Abstract
The invention relates to an offshore wind farm power prediction method considering meteorological similarity and power fluctuation, which comprises the following steps: 1) acquiring wind energy data of an offshore wind farm within a certain time, and taking the wind energy data as an original wind energy sample; 2) performing meteorological similarity clustering analysis on the original wind energy sample data to obtain a meteorological similarity classification result; 3) classifying the power fluctuation, and acquiring the category of the power fluctuation range by adopting an Extreme Learning Machine (ELM) to obtain a classification result of the power fluctuation; 4) an Elman neural network is adopted to establish a preliminary prediction model based on sample similarity, short-term wind speed rolling prediction of a wind power plant is carried out at a future moment, and an initial prediction result is obtained; 5) and (4) establishing a NWP correction model by adopting a multi-layer perceptron MLP to correct the initial prediction result to obtain a corrected final result. Compared with the prior art, the method has the advantages of considering meteorological similarity and power fluctuation, being accurate and comprehensive in prediction and the like.
Description
Technical Field
The invention relates to the technical field of offshore power prediction, in particular to an offshore wind power plant power prediction method considering meteorological similarity and power fluctuation.
Background
The offshore wind power is a key development direction of the future wind power industry due to the fact that wind power is stable, the offshore wind power is close to a load center, land resources are not occupied, and the like. However, as installed capacity increases, the volatility and randomness of offshore wind power output present challenges to the stable operation of the system. Therefore, the power system is scheduled through accurate short-term power prediction, and the method has important significance for stable operation of the power system and improvement of economic benefits. The traditional wind power prediction method can be mainly divided into a physical method for describing the outline of the whole area of the wind power plant in detail; and a statistical method for searching a potential statistical relationship between the meteorological information and the power generation amount of the wind power plant for prediction by analyzing a large amount of data. Numerical weather forecast information (NWP) information, whether physical or statistical, is often used as an important set of input data for the predictive models described above. This is because NWP data reflects the nature of atmospheric motion for predictions of duration longer than three hours. On the physical method level, the offshore physical modeling computation amount is complicated and the modeling is difficult due to the reasons of complex sea area environment, wide wake flow influence range, lack of climate data and the like; and due to the difference of geographic environments, the flexibility and generalization of the physical model are poor. The extrapolation model method based on the statistical viewpoint can be trained by using given meteorological conditions, so that the intermediate complex physical modeling step is avoided; on the other hand, the accuracy of the statistical model is directly influenced by the low accuracy and the large wind power output change of offshore NWP information in China.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide an offshore wind farm power prediction method considering meteorological similarity and power fluctuation.
The purpose of the invention can be realized by the following technical scheme:
an offshore wind farm power prediction method considering meteorological similarity and power fluctuation comprises the following steps:
1) acquiring wind energy data of an offshore wind farm within a certain time, and taking the wind energy data as an original wind energy sample;
2) performing meteorological similarity clustering analysis on the original wind energy sample data by combining a principal component analysis method and a self-organizing neural network to obtain a meteorological similarity classification result;
3) classifying the power fluctuation, and acquiring the category of the power fluctuation range by adopting an Extreme Learning Machine (ELM) to obtain a classification result of the power fluctuation;
4) an Elman neural network is adopted to establish a preliminary prediction model based on sample similarity, training samples related to the prediction time are selected according to the meteorological similarity classification result and the power fluctuation classification result to conduct short-term wind speed rolling prediction on the wind power plant at the future time, and an initial prediction result is obtained;
5) and (4) establishing a NWP correction model by adopting a multi-layer perceptron MLP to correct the initial prediction result to obtain a corrected final result.
SaidIn step 1), the vector space of the original wind energy sample at the time t is represented as StThe data structure is as follows:
wherein WS, WP, WD, T', W, P represent wind speed, power, wind direction, temperature, humidity, and barometric pressure data, respectively, subscripts represent time values of the data, WStFor predicting the wind speed forecast data at the time t, WSt-T1The historical wind speed value at the T1 moment before the predicted moment T is obtained.
The step 2) specifically comprises the following steps:
21) extracting principal components by a principal component analysis method, obtaining the contribution rate of the corresponding principal components by a cross-correlation matrix, and selecting the corresponding principal components with the cumulative variance contribution rate exceeding 90%;
22) and taking the screened principal components as input variables of the self-organizing neural network, setting the number of clustering centers, and performing similarity clustering to obtain primary weather similarity classification.
The step 3) specifically comprises the following steps:
31) the power fluctuation is divided into 5 types according to the power fluctuation amplitude, specifically:
the first type: the power fluctuation is below-15% Pr,
the second type: the power fluctuation is between-15% Pr and 0,
in the third category: the power fluctuation is 0 to 10% Pr,
the fourth type: the power fluctuation is between 10% Pr and 20% Pr,
the fifth type: the power fluctuation is above 20% Pr,
pr is the rated installed capacity of the wind turbine generator;
32) construction of a multivariate time series model DtSorting the variables from big to small after calculating the characteristic correlation indexes MIV of the variables in the multivariate time series model, selecting the variables corresponding to the first 7 MIVs as prediction variables, and using the multivariate time series model DtThe data structure of (1) is:
wherein PRR is the power variation in unit time, STD is the standard deviation of wind speed, Max and Min are respectively the maximum value and the minimum value of wind speed, Mean is the average value of wind speed, and subscript represents the time value of data;
33) and (3) taking the extreme learning machine ELM as a classifier, taking the screened predicted variable data as the input quantity of the extreme learning machine ELM, taking the 5-class power fluctuation range as the output quantity of the extreme learning machine ELM, and predicting and classifying the power fluctuation at the future moment.
The step 4) specifically comprises the following steps:
41) selecting primary selection data samples related to wind power prediction moments from the classification results of the meteorological similarities;
42) selecting data corresponding to the power fluctuation classification from the initially selected data samples as a training set of the Elman neural network;
43) and training the network weight of the Elman neural network by adopting a BP algorithm, determining a network structure, forming a preliminary prediction model based on sample similarity, and performing preliminary prediction by adopting a rolling prediction mode to obtain an initial prediction result.
Compared with the prior art, the invention has the following advantages:
the invention provides a power prediction method capable of being applied to offshore wind power aiming at the prediction difficulty of offshore wind power, and provides a PCA-SOM method for carrying out data preprocessing on actual data of an offshore wind farm, and carrying out preliminary classification on samples on the premise of retaining key information of sample data as much as possible, wherein the obtained classification result has clear classification boundary and is beneficial to improving the model precision.
And secondly, after the preliminary classification, extracting fluctuation features, determining the wind power fluctuation grade at the prediction time by using an extreme learning machine technology, selecting data with the same fluctuation grade as the prediction time on the basis of the preliminary classification, obtaining historical data samples which are further similar to the target time, and performing power prediction based on an Elman network to obtain an initial prediction model. Compared with a single Elman network, the initial prediction model shows stability and superiority of prediction accuracy at the moment of severe power fluctuation.
And aiming at the problem of inaccurate NWP information, the difference between the rough wind direction prediction and the rough wind speed prediction and the initial model prediction is used as the input quantity of the MLP, the prediction error is used as the output quantity to establish the final offshore prediction model, and the prediction abnormity caused by the large error of the NWP information is effectively improved. The results of the examples show that the proposed model can meet the prediction requirements in various environments and can be applied to actual production.
And fourthly, when the fluctuation amplitude of the wind speed is large, the Elman model can accurately follow the power fluctuation trend to increase or decrease, but the amplitude is not accurate. The provided model predicts the category of the wind power fluctuation amplitude accurately and uses the data of the same category, and the prediction curve can still well follow the fluctuation of the actual power curve when the wind power fluctuation is large, so that the model prediction precision is effectively improved.
Drawings
FIG. 1 is a flow chart of the offshore wind power prediction according to the present invention.
FIG. 2 is a diagram showing the PCA-SOM clustering results.
FIG. 3 is a graph of the relative importance of predictor variables.
FIG. 4 is the normalized error of the initial prediction model.
FIG. 5 is a graph of the predicted results of EWSRP.
Fig. 6 is a graph of the EWDRP prediction results.
FIG. 7 is a graph comparing single model prediction results.
FIG. 8 is a graph comparing the proposed model with the single Elman model prediction results.
FIG. 9 is a diagram showing the effect of NWP model correction.
Detailed Description
The invention is described in detail below with reference to the figures and specific embodiments.
Examples
For example, as shown in fig. 1, the method for predicting the power of an offshore wind farm provided by the present invention includes the following steps:
(1) data of 2016 (3 month, 8 month and 12 month) with typical characteristics are selected as research objects due to differences of offshore monthly wind energy characteristics. Taking historical meteorological data and wind power data of 21 days before each month as a training set of the model, and taking data of 9 days after each month as a test set of the model to carry out prediction and inspection, specifically:
(101) the vector space of the selected original wind energy sample at the time t is represented as StThe data structure is as follows:
in the formula, WS, WP, WD, T', W, P represent wind speed, power, wind direction, temperature, humidity, and air pressure, respectively. The subscripts indicating the time-of-day value of the data, e.g. WStWind speed forecasts provided for NWP. WSt-T1The historical wind speed value at the T1 moment before the predicted moment T is obtained.
(102) The values of T1 and T2 are obtained from the autocorrelation function equation based on the correlation of the sample itself, as shown in equation (3). The autocorrelation function is used for representing the dependency relationship between the value of the current time and the value of the previous time, and for the WStThe autocorrelation function is as follows:
in the formula, gammakRepresents the autocorrelation coefficient, mu and sigma respectively represent the mean value and standard deviation of the sequence; e (-) denotes the expected value. Finally, the time series T1 and T2 are found to be 3;
(2) when data preprocessing is carried out, a method of combining PCA and SOM neural networks is adopted for cluster analysis of original sample data. The method is characterized in that the main characteristics of the sample data are kept, and meanwhile, preliminary meteorological similarity classification is carried out on the sample data, and the method specifically comprises the following steps:
(201) after the original sample is obtained, principal components are first extracted from the sample using principal component analysis to form a low-dimensional data set. And establishing a cross-correlation matrix through sample data, calculating the eigenvalue and the eigenvector, and obtaining the contribution degree of the principal component corresponding to each eigenvalue.
(202) In the PCA process, the principal component with the variance contribution ratio of the first three is selected as the reduction result of the original data set. The cumulative contribution rates were 78.11%, 89.39%, and 94.92%, respectively.
(203) And taking the selected principal component as an input variable of the SOM network, and setting the clustering center of the SOM to be 4. The wind energy sample sequence clustering result is shown in fig. 2.
(3) On the basis of dividing the historical wind power fluctuation into ranges, predicting the category of the power fluctuation range at the future moment by using an Extreme Learning Machine (ELM) technology, specifically:
(301): the power fluctuation is classified into 5 classes by the power fluctuation amplitude. Because the influence factors of the ascending slope and the descending slope are different, the ascending slope and the descending slope are classified separately in the classification, and the fluctuation quantity of the wind power climbing grade in unit time is classified into the following 5 types as the output variable of the extreme learning machine:
the first type: the power fluctuation is below-15% Pr;
the second type: the power fluctuation is between-15% Pr and 0;
in the third category: the power fluctuation is 0-10% Pr;
the fourth type: the power fluctuation is between 10% Pr and 20% Pr;
the fifth type: the power fluctuation is more than 20% Pr;
and Pr is the rated installed capacity of the wind turbine generator.
(302): and (4) predicting and classifying the power fluctuation at the future moment by using the ELM as a classifier. The 5-class power fluctuation range is used as the output quantity of the ELM, the input quantity adopts a data mining method, and 7 different parameters are used for establishing a multivariate time series model: including the average, standard deviation, maximum, minimum of wind speed, and power, ramp rate, NWP information. Selecting PRR and wind speed related first 5 time histories for further mining data characteristicsHistory value, its data structure uses DtAs shown in formula (7):
wherein PRR represents the amount of power change per unit time; STD represents the standard deviation of wind speed; max and Min respectively represent the maximum value and the minimum value of the wind speed; mean represents the Mean wind speed; the remaining symbols have the same meanings as above.
(303): in order to reduce the calculation cost and improve the classification accuracy, only important features in the initial feature set should be selected as the model input quantity. The MIV is one of the best indicators for evaluating the characteristic correlation and is selected for characteristic screening to obtain a variable having a large influence on the fluctuation degree of the wind power. MIViIs defined by formula (1):
the sign of the MIV represents the correlation direction and the absolute value represents the relative importance of the influence. And (3) ranking the MIV values of each variable to obtain the relative influence of each variable, and finally selecting 7 predicted variables with the highest relative influence values. The arrangement is from big to small: power-2, Power-4, Max-0, PRR-2, Max-2, Power-2, and Mean-1. FIG. 3 shows the significance indices of 30 predictor variables arranged from large to small;
(4) an Elman neural network is used for establishing a preliminary prediction model based on sample similarity, training samples similar to the characteristics of the prediction time are selected for the historical wind speed time sequence, and the short-term wind speed rolling prediction of the wind power plant is carried out at the future time, and the method specifically comprises the following steps:
(401) 4 types of data classified by PCA and SOM clustering models are represented by vectors (Z1, Z2, Z3 and Z4), each type of data comprises data corresponding to 5 fluctuation ranges, and 20 to-be-selected data subsets in total can be used as training data of the Elman network;
(402) when wind power prediction is performed, it is first determined which data type the prediction time belongs to, and it is first assumed that the prediction time belongs to class Z1, and the wind power type at the prediction time is determined from the power fluctuation level model, and is set as class 3. Then reversely extracting historical power with a 3 rd type fluctuation grade from the Z1 data type as a training set to train a single Elman network for prediction, and obtaining an initial model prediction result;
(403) and training the network Elman network weight by adopting a BP algorithm, and determining that the network structure is 7-15-1-1. The Elman network is a typical dynamic regression network. The structure of the network is a traditional BP network, and the output of the hidden layer is self-connected to the input of the hidden layer through the delay and storage of the accepting layer. The predicted results are shown in fig. 4 by data clustering results.
(404) The proposed rolling prediction is an hourly run-time process based on rolling. This means that a power prediction result can be obtained at each prediction time, and when performing the prediction at the next time, the power prediction result at the previous time is taken as input data, and the data farthest from the prediction time is rejected. The specific method comprises the following steps:
let the history sequence utilized for the first prediction be x0(1)，x0(2)，...，x0(n), the sequence of the (j + 1) th rolling prediction is x0(1+j)，x0(2+j)，…x0(n)，x0(n+1)，…，x0(n + j) wherein x0(1)，x0(2)，...x0(n) is history data, x0(n+1)，...，x0(n + j) is the predicted value obtained by the previous i rolling predictions. (ii) a
(5) On the basis of a preliminary model, the method excavates the internal relation between the accuracy of (NWP) information and the model error through a multilayer perceptron (MLP), and finally establishes an NWP correction model;
the specific process of NWP error prediction correction based on MLP comprises the following steps:
step (501) power prediction is directly carried out on the basis of an Elman network by respectively utilizing a wind speed prediction value and a wind direction prediction value in the NWP information, and the results are defined as a wind speed rough prediction value (WSRP) and a wind direction rough prediction value (WDRP);
step (502) defines the WSRP and WDRP as the WPP, and the resulting differences are regarded as the degree of deviation of NWP information from the actual state, as NWP wind speed prediction Error (EWSRP) and NWP wind direction prediction Error (EWDRP). The results are shown in FIGS. 5 and 6;
step (503) comparing fig. 5, 6 and 4, the errors between the wind speed and wind direction information predicted in NWP and the initial prediction model are regular as a whole. The WPR is in a non-linear positive correlation with the changes of the EWSRP and the EWDRP, namely the EWSRP and the EWDRP have obvious influence on the prediction error of the initial model. The relation is expressed by adopting a multilayer perceptron, and the NWP sequence is corrected by utilizing the self-learning function of the MLP neural network, so that the NWP prediction precision is improved.
Data of 2016 (3 months, 8 months and 12 months) with typical characteristics are selected as research objects; the data characteristics of the ensemble of test sets are shown in table 1:
TABLE 1 test set data characteristics for three months
TABLE 23 prediction of wind power over the moon
TABLE 38 prediction of wind power over moon
TABLE 412 prediction of wind power over moon
Fig. 7 shows the results of predicting the screened data as the input quantities of three single models. The tendency of the change of the wind-out electric power can be well predicted by the Elman model, and the historical data in the input quantity can be dynamically recurred due to the internal feedback connection of the related layer, so that the network generalization capability is enhanced. However, the overall tendency of the Elman model is consistent with the reality in the two periods of 15-18 h downward slope climbing events and 25-35 h continuous power fluctuation, but the accuracy of some points is inferior to that of the BP model and the RBF model. This is because the model does not take into account the fluctuation amplitude and the hill climbing event, resulting in similar fluctuation trends but different hill climbing amplitudes.
Partial results using the Elman model and the proposed marine prediction model are shown in fig. 8. When the wind speed changes more slowly, the two models can better track the actual wind speed curve, and the prediction effect is better. When the wind speed fluctuation amplitude becomes large, the Elman model can accurately follow the power fluctuation trend to increase or decrease, but the amplitude is not accurate. The provided model predicts the category of the wind power fluctuation amplitude accurately and uses the data of the same category, and the prediction curve can still well follow the fluctuation of the actual power curve when the wind power fluctuation is large, so that the model prediction precision is effectively improved.
Finally, the data of 12 months in the test set are selected for NWP forecast error correction and verification, as shown in fig. 9. The NWP forecast has larger error in the period, and the average absolute error reaches 3.13 m/s. The initial prediction model and the final model modified by the NWP modification module, which are provided in section 3.3, are used for prediction respectively. In fig. 9, the prediction error is normalized by the installed capacity of the wind farm, and it can be seen that the initial prediction model has more abnormal prediction points, and the error of some points even reaches 79%. This is because the overall wind speed is large in 12 months, and at medium and high wind speeds, a small error between the wind speed forecast and the wind direction forecast according to the NWP wind speed curve of the wind turbine brings a large power prediction error. In addition, a small difference in wind direction forecast at high wind speeds also increases the prediction error. In fig. 9, the final predicted prediction result has a significant effect of correcting the abnormal point and the abrupt change point, and is also improved in a relatively gentle wind power stage. It should be noted that at some points where the prediction accuracy is higher, the error will increase after the correction by the correction model, but the increased range is within an acceptable range.
Claims (3)
1. An offshore wind farm power prediction method considering meteorological similarity and power fluctuation is characterized by comprising the following steps:
1) acquiring wind energy data of an offshore wind farm within a certain time, and taking the wind energy data as an original wind energy sample, wherein the vector space of the original wind energy sample at the moment t is represented as StThe data structure is as follows:
wherein WS, WP, WD, T', W, P represent wind speed, power, wind direction, temperature, humidity, and barometric pressure data, respectively, subscripts represent time values of the data, WStFor predicting the wind speed forecast data at the time t, WSt-T1The historical wind speed value at the T1 moment before the predicted moment T is obtained;
2) performing meteorological similarity clustering analysis on the original wind energy sample data by combining a principal component analysis method and a self-organizing neural network to obtain a meteorological similarity classification result;
3) classifying the power fluctuation, and acquiring the category of the power fluctuation range by using an Extreme Learning Machine (ELM) to obtain a classification result of the power fluctuation, wherein the classification result specifically comprises the following steps:
31) the power fluctuation is divided into 5 types according to the power fluctuation amplitude, specifically:
the first type: the power fluctuation is below-15% Pr,
the second type: the power fluctuation is between-15% Pr and 0,
in the third category: the power fluctuation is 0 to 10% Pr,
the fourth type: the power fluctuation is between 10% Pr and 20% Pr,
the fifth type: the power fluctuation is above 20% Pr,
pr is the rated installed capacity of the wind turbine generator;
32) construction of a multivariate time series model DtSorting the variables from big to small after calculating the characteristic correlation indexes MIV of the variables in the multivariate time series model, selecting the variables corresponding to the first 7 MIVs as prediction variables, and using the multivariate time series model DtThe data structure of (1) is:
wherein PRR is the power variation in unit time, STD is the standard deviation of wind speed, Max and Min are respectively the maximum value and the minimum value of wind speed, Mean is the average value of wind speed, and subscript represents the time value of data;
33) taking the ELM as a classifier, taking the screened predicted variable data as the input quantity of the ELM, taking the 5-class power fluctuation range as the output quantity of the ELM, and predicting and classifying the power fluctuation at the future moment;
4) an Elman neural network is adopted to establish a preliminary prediction model based on sample similarity, training samples related to the prediction time are selected according to the meteorological similarity classification result and the power fluctuation classification result to conduct short-term wind speed rolling prediction on the wind power plant at the future time, and an initial prediction result is obtained;
5) and (4) establishing a NWP correction model by adopting a multi-layer perceptron MLP to correct the initial prediction result to obtain a corrected final result.
2. The method for predicting the power of the offshore wind farm by considering meteorological similarity and power fluctuation according to claim 1, wherein the step 2) comprises the following steps:
21) extracting principal components by a principal component analysis method, obtaining the contribution rate of the corresponding principal components by a cross-correlation matrix, and selecting the corresponding principal components with the cumulative variance contribution rate exceeding 90%;
22) and taking the screened principal components as input variables of the self-organizing neural network, setting the number of clustering centers, and performing similarity clustering to obtain primary weather similarity classification.
3. The method for predicting the power of the offshore wind farm by considering meteorological similarity and power fluctuation according to claim 1, wherein the step 4) comprises the following steps:
41) selecting primary selection data samples related to wind power prediction moments from the classification results of the meteorological similarities;
42) selecting data corresponding to the power fluctuation classification from the initially selected data samples as a training set of the Elman neural network;
43) and training the network weight of the Elman neural network by adopting a BP algorithm, determining a network structure, forming a preliminary prediction model based on sample similarity, and performing preliminary prediction by adopting a rolling prediction mode to obtain an initial prediction result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810696196.0A CN108898251B (en) | 2018-06-29 | 2018-06-29 | Offshore wind farm power prediction method considering meteorological similarity and power fluctuation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810696196.0A CN108898251B (en) | 2018-06-29 | 2018-06-29 | Offshore wind farm power prediction method considering meteorological similarity and power fluctuation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108898251A CN108898251A (en) | 2018-11-27 |
| CN108898251B true CN108898251B (en) | 2021-10-08 |
Family
ID=64347021
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810696196.0A Active CN108898251B (en) | 2018-06-29 | 2018-06-29 | Offshore wind farm power prediction method considering meteorological similarity and power fluctuation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108898251B (en) |
Families Citing this family (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109784473A (en) * | 2018-12-13 | 2019-05-21 | 天津大学 | A kind of short-term wind power prediction method based on Dual Clocking feature learning |
| CN110276472A (en) * | 2019-05-21 | 2019-09-24 | 南方电网调峰调频发电有限公司 | A kind of offshore wind farm power ultra-short term prediction method based on LSTM deep learning network |
| CN110942170A (en) * | 2019-08-29 | 2020-03-31 | 华北电力大学（保定） | Short-term wind speed prediction method and system based on information processing |
| CN111008725B (en) * | 2019-10-23 | 2022-06-24 | 中国农业大学 | Meteorological factor fluctuation feature extraction method for short-term wind power prediction |
| CN111030106B (en) * | 2019-12-26 | 2021-04-27 | 福州大学 | Wind power fluctuation quality assessment method based on waveform similarity theory |
| CN111626473A (en) * | 2020-04-17 | 2020-09-04 | 国网上海市电力公司 | Two-stage photovoltaic power prediction method considering error correction |
| CN112149879B (en) * | 2020-09-02 | 2022-10-18 | 上海电力大学 | New energy medium-and-long-term electric quantity prediction method considering macroscopic volatility classification |
| CN112288164B (en) * | 2020-10-29 | 2023-04-07 | 四川大学 | Wind power combined prediction method considering spatial correlation and correcting numerical weather forecast |
| CN112216181A (en) * | 2020-10-30 | 2021-01-12 | 哈尔滨飞机工业集团有限责任公司 | System and method capable of realizing single-shot training of double-shot helicopter |
| CN112418346B (en) * | 2020-12-07 | 2022-10-25 | 天津大学 | Numerical weather forecast total radiation system error classification calculation method |
| CN113449847B (en) * | 2021-06-17 | 2022-08-19 | 江苏方天电力技术有限公司 | Offshore wind power rolling prediction method considering second-level time series wind speed change |
| CN114297186B (en) * | 2021-12-30 | 2024-04-26 | 广西电网有限责任公司 | Power consumption data preprocessing method and system based on deviation coefficient |
| CN114465231B (en) * | 2022-02-14 | 2023-06-13 | 南京邮电大学盐城大数据研究院有限公司 | Power access method and system for offshore wind power generation |
| CN117937638A (en) * | 2024-03-12 | 2024-04-26 | 山西城市动力新能源有限公司 | Wind power plant cluster power generation control method and device, electronic equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3113312A1 (en) * | 2015-07-01 | 2017-01-04 | General Electric Company | Predictive control for energy storage on a renewable energy system |
| CN107609697A (en) * | 2017-09-06 | 2018-01-19 | 南京邮电大学 | A kind of Wind power forecasting method |
| CN107944622A (en) * | 2017-11-21 | 2018-04-20 | 华北电力大学 | Wind power forecasting method based on continuous time cluster |
| CN108205717A (en) * | 2017-12-30 | 2018-06-26 | 国网江苏省电力公司无锡供电公司 | A kind of photovoltaic generation power Multiple Time Scales Forecasting Methodology |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2011124226A1 (en) * | 2010-04-08 | 2011-10-13 | Vestas Wind Systems A/S | Method and system for forecasting wind energy |
| CN105894106B (en) * | 2015-01-05 | 2020-06-16 | 国家电网公司 | Integrated coupling method for marine mode and meteorological mode |
| CN105117788A (en) * | 2015-07-22 | 2015-12-02 | 河南行知专利服务有限公司 | Wind power generation power prediction method |
| CN105930900B (en) * | 2016-05-09 | 2018-02-27 | 华北电力大学 | The Forecasting Methodology and system of a kind of hybrid wind power generation |
| CN107067099B (en) * | 2017-01-25 | 2020-06-19 | 清华大学 | Wind power probability prediction method and device |
| CN108133279B (en) * | 2017-08-29 | 2020-12-22 | 甘肃省电力公司风电技术中心 | Wind power probability prediction method, storage medium and device |
-
2018
- 2018-06-29 CN CN201810696196.0A patent/CN108898251B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3113312A1 (en) * | 2015-07-01 | 2017-01-04 | General Electric Company | Predictive control for energy storage on a renewable energy system |
| CN107609697A (en) * | 2017-09-06 | 2018-01-19 | 南京邮电大学 | A kind of Wind power forecasting method |
| CN107944622A (en) * | 2017-11-21 | 2018-04-20 | 华北电力大学 | Wind power forecasting method based on continuous time cluster |
| CN108205717A (en) * | 2017-12-30 | 2018-06-26 | 国网江苏省电力公司无锡供电公司 | A kind of photovoltaic generation power Multiple Time Scales Forecasting Methodology |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108898251A (en) | 2018-11-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108898251B (en) | Offshore wind farm power prediction method considering meteorological similarity and power fluctuation | |
| CN108539738B (en) | Short-term load prediction method based on gradient lifting decision tree | |
| CN107944622B (en) | Wind power prediction method based on continuous time interval clustering | |
| CN110728411B (en) | High-low altitude area combined rainfall prediction method based on convolutional neural network | |
| CN111753893A (en) | Wind turbine generator power cluster prediction method based on clustering and deep learning | |
| CN111027775A (en) | Step hydropower station generating capacity prediction method based on long-term and short-term memory network | |
| CN111626473A (en) | Two-stage photovoltaic power prediction method considering error correction | |
| CN115374995A (en) | Distributed photovoltaic and small wind power station power prediction method | |
| CN117272850B (en) | Elastic space analysis method for safe operation scheduling of power distribution network | |
| CN113505938A (en) | Ultra-short-term wind power combined prediction method and system | |
| CN112001537A (en) | Short-term wind power prediction method based on gray model and support vector machine | |
| Gama et al. | Stream-based electricity load forecast | |
| Barve et al. | Air quality index forecasting using parallel dense neural network and LSTM cell | |
| CN115099511A (en) | Photovoltaic power probability estimation method and system based on optimized copula | |
| CN108830405B (en) | Real-time power load prediction system and method based on multi-index dynamic matching | |
| CN115310648A (en) | Medium-and-long-term wind power combination prediction method based on multi-meteorological variable model identification | |
| CN113095547B (en) | Short-term wind power prediction method based on GRA-LSTM-ICE model | |
| CN110222910A (en) | A kind of active power distribution network Tendency Prediction method and forecasting system | |
| CN117200223A (en) | Day-ahead power load prediction method and device | |
| CN110110784B (en) | Transformer fault identification method based on transformer related operation data | |
| CN110163437B (en) | Day-ahead photovoltaic power generation power prediction method based on DPK-means | |
| CN114971005B (en) | Bay water temperature combination prediction method based on LSTM and differential regression model dynamic weighting | |
| CN116663393A (en) | Random forest-based power distribution network continuous high-temperature fault risk level prediction method | |
| CN115907228A (en) | Short-term power load prediction analysis method based on PSO-LSSVM | |
| Aji et al. | Short-term solar power forecasting using svr on hybrid pv power plant in indonesia |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |