CN109886488B - Distributed wind power plant layered hybrid short-term prediction method considering wind speed time lag - Google Patents
Distributed wind power plant layered hybrid short-term prediction method considering wind speed time lag Download PDFInfo
- Publication number
- CN109886488B CN109886488B CN201910129892.8A CN201910129892A CN109886488B CN 109886488 B CN109886488 B CN 109886488B CN 201910129892 A CN201910129892 A CN 201910129892A CN 109886488 B CN109886488 B CN 109886488B
- Authority
- CN
- China
- Prior art keywords
- wind speed
- prediction
- historical
- wind
- target
- 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 40
- 238000005311 autocorrelation function Methods 0.000 claims abstract description 22
- 238000012549 training Methods 0.000 claims description 46
- 238000012937 correction Methods 0.000 claims description 28
- 238000004422 calculation algorithm Methods 0.000 claims description 25
- 238000004364 calculation method Methods 0.000 claims description 22
- 230000009467 reduction Effects 0.000 claims description 14
- 238000013528 artificial neural network Methods 0.000 claims description 12
- 238000004590 computer program Methods 0.000 claims description 6
- 238000010845 search algorithm Methods 0.000 claims description 5
- 238000012216 screening Methods 0.000 claims description 4
- 239000002245 particle Substances 0.000 description 19
- 230000006870 function Effects 0.000 description 9
- 238000010248 power generation Methods 0.000 description 8
- 230000008569 process Effects 0.000 description 5
- 230000009471 action Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000011218 segmentation Effects 0.000 description 3
- 238000013517 stratification Methods 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000000053 physical method Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012843 least square support vector machine Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000035699 permeability Effects 0.000 description 1
- 238000013179 statistical model Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- 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
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application discloses a distributed wind power plant layered mixed short-term prediction method, a system and a device considering wind speed time lag, wherein a target historical wind speed value time lag and a target historical wind power value time lag are respectively obtained by utilizing a historical wind speed time sequence, a historical wind power time sequence and a partial autocorrelation function, a target wind speed prediction period is obtained by utilizing a numerical weather prediction model to predict the historical wind speed time sequence, deviation of actual data is avoided, the target wind speed prediction period is corrected to obtain a target corrected wind speed prediction period, accuracy is improved, the target corrected wind speed prediction period is layered, data are classified and analyzed, the target historical wind power value time lag is utilized, each layer of wind speed range is respectively substituted into a plurality of preset initial prediction models to obtain a plurality of groups of initial wind power prediction values, finally, the preset mixed prediction model is utilized to obtain a wind power prediction value, and the more accurate prediction value is obtained by mixing a plurality of prediction models.
Description
Technical Field
The invention relates to the field of new energy power generation of a power system, in particular to a distributed wind power plant layered mixed short-term prediction method, system and device considering wind speed time lag.
Background
With the development of science and technology, new energy power supply is gradually popularized, wherein wind power generation is one of new energy which is vigorously developed at present, and wind power resources are different from water energy and are widely distributed, so in order to maximally utilize the wind power resources, the wind power generation is often distributed, and due to the unstable characteristic of the wind power generation and the distributed arrangement of distributed wind power, the grid-connected management of the distributed wind power is an important research subject at present.
Increasing wind power generation requires solving the problems of market integration and design, real-time grid operation, interconnection standards, power quality, transmission capacity upgrades, power system dynamics, stability and reliability, etc. Wind unpredictability is the biggest challenge in prediction and is also a major obstacle to further development of wind-powered iontophoresis. Researches around the world show that an accurate and reliable wind power prediction system can improve the permeability of wind power, and the importance of the research topic is highlighted.
In order to accurately predict wind power generation, researchers have developed some wind power generation prediction methods, which may be classified into physical methods, statistical and learning methods, probabilistic methods, and hybrid methods. Physical methods, such as numerical weather prediction models, typically require a lot of information to achieve optimal prediction accuracy. This method has the advantage of being able to operate in a newly commissioned wind farm without historical data, and has the disadvantage of being computationally expensive. Conventional statistical models are time series based models, including autoregressive models or autoregressive integral moving average models. In addition, various artificial intelligence models are also widely used for wind power prediction, such as neural networks, fuzzy logic, adaptive wavelet neural networks and the like; probabilistic predictive methods are currently more popular methods, however they have the disadvantage of being time consuming.
However, when conventional independent models and non-hierarchical hybrid models are applied to predict wind power, they are generally unable to make accurate predictions due to the complexity of the wind power generation system, including randomness and intermittency.
For this reason, a more accurate wind power prediction method is required.
Disclosure of Invention
In view of the above, the present invention provides a method, a system, and a device for a distributed wind farm layered hybrid short-term prediction in consideration of wind speed time lag, so as to improve the prediction accuracy. The specific scheme is as follows:
a decentralized wind farm tiered mixing short term prediction system taking into account wind speed time lag, comprising:
the data acquisition interface is used for receiving the historical wind speed time sequence and the historical wind power time sequence;
the wind speed time lag generation module is used for obtaining a target historical wind speed value time lag meeting a preset condition by utilizing the historical wind speed time sequence and the partial autocorrelation function;
the wind speed forecasting period module is used for forecasting the historical wind speed time sequence by utilizing a numerical weather forecasting model to obtain a target wind speed forecasting period meeting preset conditions; the wind speed prediction period comprises a plurality of predicted wind speeds measured and recorded in a preset time period, and the numerical weather prediction model is formed by utilizing historical wind speed time sequence training data in advance;
the prediction period correction module is used for correcting the target wind speed prediction period to obtain a target corrected wind speed prediction period;
the wind power time lag generation module is used for obtaining a target historical wind power value time lag meeting a preset condition by utilizing the historical wind power time sequence and the partial autocorrelation function;
forecast period layering module for utilizingThe target corrected wind speed prediction period is layered according to a layering rule to obtain a multilayer wind speed range with the number of target layers;
the multi-model prediction module is used for substituting the wind speed range of each layer into a plurality of preset initial prediction models respectively by utilizing the value time lag of the target historical wind power to obtain a plurality of groups of initial wind power prediction values; each initial prediction model is different and is formed by utilizing the historical wind power value time lag and the historical wind speed range in advance for training;
the hybrid model prediction module is used for obtaining a wind power prediction value by utilizing a plurality of groups of initial wind power prediction values and a preset hybrid prediction model; the hybrid prediction model is formed by training a plurality of groups of historical initial wind power predicted values.
Optionally, the wind speed prediction period module includes:
the wind speed prediction period unit is used for predicting the historical wind speed time sequence by utilizing the numerical weather prediction model to obtain a plurality of groups of wind speed prediction periods;
the error calculation unit is used for comparing each group of wind speed prediction period with the corresponding historical actual wind speed to obtain a prediction average error;
and the error screening unit is used for selecting the wind speed prediction period with the minimum prediction average error as the target wind speed prediction period.
Optionally, the error calculating unit includes:
the individual error calculation subunit is used for comparing each group of wind speed prediction period with the corresponding historical actual wind speed to obtain individual prediction errors of each wind speed recorded in each group of wind speed prediction period and the corresponding historical actual wind speed;
and the average error calculation subunit is used for obtaining the predicted average error of each group of wind speed prediction periods by using all the individual errors in each group of wind speed prediction periods.
Optionally, the forecast period correcting module is specifically configured to obtain the corrected target forecast wind speed group by using the target historical wind speed value time lag, the target forecast wind speed forecast period, and the forecast wind speed correction model; the predicted wind speed correction model is formed by training by utilizing a historical wind speed data value range and a predicted wind speed group in advance.
Optionally, the predicted wind speed correction model is formed by training a model established based on an artificial neural network algorithm by using a value range of historical wind speed data and a predicted wind speed group in advance.
Optionally, the forecast period layering module includes:
the occurrence frequency counting unit is used for counting the occurrence frequency of the wind speed of each preset wind speed section in the target correction wind speed prediction period; the preset wind speed section is a plurality of preset wind speed sections at equal intervals;
the cubic root calculating unit is used for obtaining the cubic root of the occurrence frequency of each wind speed section by utilizing the occurrence frequency of the wind speed of each wind speed section;
the cubic root sum calculating unit is used for obtaining the cubic root sum of each wind speed section by utilizing the cubic root of the occurrence frequency of each wind speed section; the sum of the cubic roots is the sum of the cubic root of the occurrence frequency of each wind speed section and the cubic root of the occurrence frequency of the wind speed section smaller than the wind speed;
the prediction period layering unit is used for layering the target correction wind speed prediction period in sequence from small to large according to the number of layers to obtain a plurality of groups of layering results with different numbers of layers;
the variance reduction rate calculation unit is used for obtaining the variance reduction rate of two adjacent layering results by utilizing each layering result;
the layering result determining unit is used for selecting a layering result with the variance reduction rate firstly lower than that of the last layering result as a target layering result to obtain a multilayer wind speed range;
wherein, the layering result comprises a multi-layer wind speed range equal to the layering number.
Optionally, the multi-model prediction module is specifically configured to utilize the value time lag of the target historical wind power to substitute each layer of wind speed range into a preset first prediction model, a preset second prediction model and a preset third prediction model respectively to obtain multiple groups of initial wind power prediction values;
the first prediction model is formed by training an artificial neural network model established based on a BP algorithm and an LM algorithm by utilizing the historical wind power value time lag and the historical wind speed range;
the second prediction model is formed by training a model established based on an ANFIS algorithm by utilizing the historical wind power value time lag and the historical wind speed range;
and the third prediction model is formed by training a model established based on an LC-SVM algorithm by utilizing the historical wind power value time lag and the historical wind speed range.
Optionally, the hybrid prediction model is formed by training a model established based on a charged system search algorithm by using multiple groups of historical initial wind power prediction values;
the hybrid prediction model is as follows: p fc =α 1 ·P f,1 +α 2 ·P f,2 +…+α n ·P f,n ;
In the formula, P fc Representing the wind power predicted value, n representing the number of prediction models, and alpha representing the weight coefficient of each prediction model.
The invention also discloses a distributed wind power plant layered mixed short-term prediction method considering the wind speed time lag, which comprises the following steps:
acquiring a historical wind speed time sequence and a historical wind power time sequence in advance;
obtaining a target historical wind speed value time lag meeting a preset condition by utilizing the historical wind speed time sequence and the partial autocorrelation function;
predicting the historical wind speed time sequence by using a numerical weather prediction model to obtain a target wind speed prediction period meeting preset conditions; the wind speed prediction period comprises a plurality of predicted wind speeds measured and recorded in a preset time period, and the numerical weather prediction model is formed by utilizing historical wind speed time sequence training data in advance;
correcting the target wind speed prediction period to obtain a target corrected wind speed prediction period;
obtaining a target historical wind power value time lag meeting a preset condition by using the historical wind power time sequence and the partial autocorrelation function;
by usingThe target corrected wind speed prediction period is layered according to a layering rule to obtain a multilayer wind speed range with the number of target layers;
utilizing the value time lag of the target historical wind power, and substituting each layer of wind speed range into a plurality of preset initial prediction models respectively to obtain a plurality of groups of initial wind power prediction values; each initial prediction model is different and is formed by utilizing the historical wind power value time lag and the historical wind speed range in advance for training;
obtaining a wind power predicted value by utilizing a plurality of groups of initial wind power predicted values and a preset hybrid prediction model; the hybrid prediction model is formed by training a plurality of groups of historical initial wind power predicted values.
The invention also discloses a device for layered mixed short-term prediction of the distributed wind power plant in consideration of the time lag of the wind speed, which comprises the following steps:
a memory for storing a computer program;
a processor for executing the computer program to implement a decentralized wind farm tiered mixing short term prediction method taking into account wind speed time lag as described above.
The invention relates to a distributed wind power plant layered hybrid short-term prediction system considering wind speed time lag, which comprises a data acquisition interface, a data processing interface and a data processing interface, wherein the data acquisition interface is used for receiving a historical wind speed time sequence and a historical wind power time sequence; the wind speed time lag generation module is used for obtaining a target historical wind speed value time lag meeting a preset condition by utilizing the historical wind speed time sequence and the partial autocorrelation function; the wind speed forecasting period module is used for forecasting the historical wind speed time sequence by utilizing a numerical weather forecasting model to obtain a target wind speed forecasting period meeting preset conditions; wherein the wind speedThe forecasting period comprises a plurality of forecasting wind speeds measured and recorded in a preset time period, and the numerical weather forecasting model is trained by utilizing historical wind speed time sequence training data in advance; the prediction period correction module is used for correcting the target wind speed prediction period to obtain a target corrected wind speed prediction period; the wind power time lag generation module is used for obtaining a target historical wind power value time lag meeting a preset condition by utilizing the historical wind power time sequence and the partial autocorrelation function; forecast period layering module for utilizingThe target corrected wind speed prediction period is layered according to the layering rule to obtain a multilayer wind speed range with the number of target layers; the multi-model prediction module is used for substituting a wind speed range of each layer into a plurality of preset initial prediction models respectively by utilizing the value time lag of the target historical wind power to obtain a plurality of groups of initial wind power prediction values; each initial prediction model is different and is formed by utilizing the historical wind power value time lag and the historical wind speed range in advance for training; the hybrid model prediction module is used for obtaining a wind power prediction value by utilizing a plurality of groups of initial wind power prediction values and a preset hybrid prediction model; the hybrid prediction model is formed by training multiple groups of historical initial wind power prediction values.
According to the method, a historical wind speed time sequence and a partial autocorrelation function are utilized to obtain a target historical wind speed value time lag meeting a preset condition, a numerical weather prediction model is utilized to predict the historical wind speed time sequence to obtain a target wind speed prediction period meeting the preset condition, deviation of actual data is avoided, more regular prediction data is obtained, in order to further improve accuracy, the target wind speed prediction period is corrected to obtain a target corrected wind speed prediction period, the target corrected wind speed prediction period is layered to analyze data in a classified mode, input quantity of a single group of prediction models is reduced to improve accuracy, a historical wind power time sequence and a partial autocorrelation function are utilized to obtain a target historical wind power time lag meeting the preset value, then the target historical wind power time lag is utilized to substitute a wind speed range of each layer into a plurality of preset initial prediction models respectively to obtain a plurality of groups of initial wind power prediction values, finally a plurality of groups of initial wind power prediction values and a preset mixed prediction model are utilized to predict wind power to obtain a more accurate prediction value.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
FIG. 1 is a schematic structural diagram of a distributed wind power plant layered hybrid short-term prediction system with consideration of wind speed time lag according to an embodiment of the invention;
fig. 2 is a schematic flow chart of a distributed wind farm layered hybrid short-term prediction method considering wind speed time lag according to the embodiment of the invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The embodiment of the invention discloses a distributed wind power plant layered hybrid short-term prediction system considering wind speed time lag, and as shown in figure 1, the device comprises:
the data acquisition interface 11 is used for receiving a historical wind speed time sequence and a historical wind power time sequence;
specifically, historical data is needed for supporting the prediction of the wind power, so that the historical wind speed time series and the historical wind power time series are obtained through the data obtaining interface 11, and are used for the processor to perform subsequent processing.
And the wind speed time lag generating module 12 is configured to obtain a target historical wind speed value time lag meeting a preset condition by using the historical wind speed time series and the partial autocorrelation function.
Specifically, a time series of historical wind speeds is analyzed by utilizing a partial autocorrelation function, a peak value is selected as an input of a subsequent model, the time series of the historical wind speeds is calculated by utilizing the partial autocorrelation function to obtain a PACF value, and the PACF value is judged to be remarkably reduced after the first lag output, so that a target historical wind speed value time lag is selected, for example, the PACF value is remarkably reduced after the first lag output, so that the historical wind speed of (t-1) hour is selected, and if the PACF value is remarkably reduced after the third lag output, the historical wind speeds of (t-1), (t-2) hour and (t-3) hour can be selected as inputs, so that the target historical wind speed value time lag is selected.
And the wind speed prediction period module 13 is configured to predict the historical wind speed time series by using a numerical weather prediction model, so as to obtain a target wind speed prediction period meeting a preset condition.
Specifically, a plurality of groups of wind speed prediction periods can be predicted by using the numerical weather prediction model, each group of wind speed prediction period can comprise a plurality of measured and recorded wind speeds in a period, for example, the numerical weather prediction model can perform prediction once every 6 hours to obtain a group of predicted wind speeds comprising 6 hours, and meanwhile, the numerical weather prediction model can continuously predict the wind speed of 3 days, for example, the wind speed is predicted for three days by taking 6 hours as a time interval, and respectively taking 1-6 hours, 7-12h,13-18h,19-24h,25-30h, 8230, and 67-72h as a prediction time range.
Specifically, since the wind speed may have a sudden change in a part of time period during the wind speed measurement period, for example, when the wind speed changes in the day, the wind speed changes violently and unstably due to rapid changes in air temperature, such sudden change data is small in quantity and difficult to predict, and is not suitable for being used as a prediction basis, after a plurality of groups of wind speed prediction periods obtained from the numerical weather prediction model, a target wind speed prediction period satisfying a preset condition is selected from the groups of wind speed prediction periods, for example, the preset condition may be a wind speed prediction period satisfying a predicted wind speed curve, a wind speed prediction period in which fluctuation is within a certain range, or a wind speed prediction period in which a predicted wind speed error from a historical actual wind speed error is within a certain range in the wind speed prediction period.
The numerical weather prediction model is trained by utilizing historical wind speed time series training data in advance.
And the prediction period correction module 14 is configured to correct the target wind speed prediction period to obtain a target corrected wind speed prediction period.
Specifically, since the numerical weather prediction model has a certain error, in order to improve the accuracy of subsequent prediction, the predicted wind speed in the target wind speed prediction period is further corrected, so as to improve the prediction accuracy, for example, the target wind speed prediction period is corrected by using the historical actual wind speed or corrected by using a preset correction algorithm.
And the wind power time lag generation module 15 is configured to obtain a target historical wind power value time lag meeting a preset condition by using the historical wind power time series and the partial autocorrelation function.
Specifically, a time series of historical wind power is analyzed by utilizing a partial autocorrelation function to obtain a PACF value, and a target historical wind power value time lag is selected by judging that the PACF value is remarkably reduced after the first lag output, for example, the PACF value is remarkably reduced after the first lag output, so that the historical wind speed in (t-1) hour is selected, and if the PACF value is remarkably reduced after the third lag output, the historical wind power in (t-1) hour, (t-2) hour and (t-3) hour can be selected as input, so that the target historical wind power value time lag is selected.
Forecast period layering module 16 for utilizingAnd (4) layering the target corrected wind speed prediction period according to the layering rule to obtain a multilayer wind speed range.
Specifically, for analyzing and classifying the predicted wind speed in the target corrected wind speed prediction period, the method adoptsThe layering rules are layered, the predicted wind speed in the target corrected wind speed prediction period is divided into a plurality of layers according to the wind speed range, each layer is a layer of wind speed range, a plurality of layers of wind speed ranges are obtained, each layer of wind speed range comprises all the predicted wind speeds belonging to the wind speed range in the target corrected wind speed prediction period, for example, the maximum wind speed in the target corrected wind speed prediction period is 15m/s, and the maximum wind speed in the target corrected wind speed prediction period is obtained byThe layering rule can finally divide the target corrected wind speed prediction period into 3 layers, namely a first layer wind speed range of 0m/s-6m/s, a second layer wind speed range of 6m/s-10m/s and a third layer wind speed range of 10m/s-15m/s, wherein each layer wind speed range comprises all predicted wind speeds belonging to the wind speed range interval in the target corrected wind speed prediction period, and boundary points of the two layers of wind speed ranges fall within a low wind speed range, for example, 6m/s in the layering belongs to the first layer wind speed range, but not the second layer wind speed range.
And the multi-model prediction module 17 is configured to utilize the value time lag of the target historical wind power and substitute each layer of wind speed range into a plurality of preset initial prediction models respectively to obtain a plurality of groups of initial wind power prediction values.
And the hybrid model prediction module 18 is configured to obtain a wind power prediction value by using multiple sets of initial wind power prediction values and a preset hybrid prediction model.
Specifically, in order to improve the accuracy of prediction, a plurality of initial prediction models are used for prediction to obtain a plurality of groups of initial wind power prediction values, then a final hybrid prediction model comprehensively considers the plurality of groups of initial wind power prediction values obtained by the plurality of initial prediction models to obtain a wind power prediction value finally, and the finally obtained wind power prediction value compensates errors of the plurality of initial prediction models through the hybrid prediction model, so that the advantages of the plurality of initial prediction models are taken, and therefore the prediction result is more accurate.
Each initial prediction model is different and is formed by utilizing the historical wind power value time lag and the historical wind speed range in advance for training; the hybrid prediction model is formed by training a plurality of groups of historical initial wind power prediction values;
therefore, the embodiment of the invention obtains the target historical wind speed value time lag meeting the preset condition by using the historical wind speed time sequence and the partial autocorrelation function, predicts the historical wind speed time sequence by using the numerical weather prediction model to obtain the target wind speed prediction period meeting the preset condition, avoids the deviation of actual data to obtain more regular prediction data, corrects the target wind speed prediction period to obtain the target corrected wind speed prediction period in order to further improve the accuracy, performs layering on the target corrected wind speed prediction period to classify and analyze the data, simultaneously reduces the input quantity of a single group of the prediction model to improve the accuracy, obtains the target historical wind power value time lag meeting the preset condition by using the historical wind power time sequence and the partial autocorrelation function, obtains a plurality of groups of initial wind power value time lags by using the target historical wind power value time lag to substitute a plurality of preset initial prediction models in each layer of wind speed range respectively to obtain a plurality of groups of initial wind power prediction values, and finally obtains the wind power prediction value by using a plurality of groups of initial wind power prediction models and a preset mixed prediction model to predict the wind power accurately to obtain a more predicted value.
It can be understood that the prediction duration can be set according to the actual application requirements, and the number of results output by each model needs to be correspondingly changed.
The embodiment of the invention discloses a specific distributed wind power plant layered hybrid short-term prediction system considering wind speed time lag, and compared with the previous embodiment, the embodiment further explains and optimizes the technical scheme. Specifically, the method comprises the following steps:
specifically, the wind speed prediction period module 13 may specifically include a wind speed prediction period unit, an error calculation unit, and an error screening unit; wherein,
and the wind speed prediction period unit is used for predicting the historical wind speed time sequence by using the numerical weather prediction model to obtain a plurality of groups of wind speed prediction periods.
And the error calculation unit is used for comparing each group of wind speed prediction period with the corresponding historical actual wind speed to obtain a prediction average error.
Specifically, the error calculation unit may further specifically include an individual error calculation subunit and an average error calculation subunit; wherein,
the individual error calculation subunit is used for comparing each group of wind speed prediction period with the corresponding historical actual wind speed to obtain individual prediction errors of each wind speed recorded in each group of wind speed prediction period and the corresponding historical actual wind speed;
and the average error calculation subunit is used for obtaining the predicted average error of each group of wind speed prediction periods by using all the individual errors in each group of wind speed prediction periods.
Specifically, each group of wind speed prediction periods is compared with corresponding historical actual wind speeds, and each group of wind speed prediction period comprises a plurality of predicted wind speeds at different times, so that each predicted wind speed needs to be compared with the corresponding historical actual wind speed to obtain an individual prediction error, for example, one group of predicted wind speeds can comprise the predicted wind speed 3m/s at 8 points 00 on the same day of No. 7 in No. 5/7 in 2017, the predicted wind speed 4m/s at 8 points 10 and the predicted wind speed 3.5m/s at 8 points 20, the corresponding historical actual wind speed is the historical actual wind speed 3.1m/s at 8 points 00 on the same day of No. 7 in No. 5/7 in 2017, the historical actual wind speed 3.8m/s at 8 points 10 and the historical actual wind speed 3.8m/s at 8 points 20, and the individual error is 0.1m/s,0.2m/s and 0.3m/s respectively; and then, obtaining the predicted average error of each group of wind speed prediction period by using all the individual errors in each group of wind speed prediction period and calculating the average value of all the individual errors.
And the error screening unit is used for selecting the wind speed prediction period with the minimum prediction average error as the target wind speed prediction period.
Specifically, a wind speed prediction period with the minimum prediction average error is selected from a plurality of wind speed prediction periods to serve as a target wind speed prediction period, so that the accuracy of subsequent prediction is improved.
Specifically, the forecast period correction module 14 is specifically configured to obtain a corrected target forecast wind speed group by using the target historical wind speed value-taking time lag, the target forecast wind speed period and the forecast wind speed correction model.
The method comprises the steps that a wind speed forecasting correction model is formed by utilizing a historical wind speed data value range and a wind speed forecasting group in advance for training, furthermore, the wind speed forecasting correction model can utilize the historical wind speed data value range and the wind speed forecasting group as the input of an artificial neural network training model, after back propagation, hidden nodes and iteration times are determined in a heuristic mode, training of the wind speed forecasting correction model is completed, and a forecasting network of the wind speed forecasting correction model is forecasted.
Specifically, the prediction period stratification module 16 may specifically include an occurrence frequency statistics unit, a cubic root calculation unit, a cubic root sum calculation unit, a prediction period stratification unit, a variance reduction rate calculation unit, and a stratification result determination unit; wherein,
the occurrence frequency counting unit is used for counting the occurrence frequency of the wind speed of each preset wind speed section in the target correction wind speed prediction period; the preset wind speed section is a plurality of preset wind speed sections with equal intervals.
Specifically, the wind speed in the target corrected wind speed prediction period may be as shown in a wind speed data table in table 1, and include 100 pieces of wind speed data, after a plurality of wind speed segments are set, the wind speed in the target corrected wind speed prediction period is one-to-one corresponding to each wind speed segment, and the wind speed belonging to each wind speed segment is counted to obtain the frequency of occurrence of the wind speed in the target corrected wind speed prediction period in each wind speed segment, as shown in the tableThe frequency f of occurrence of each wind speed segment in the quantity statistical table is shown.
TABLE 1 wind speed data sheet
A cubic root calculating unit for obtaining cubic root of the occurrence frequency of each wind speed segment by using the occurrence frequency of the wind speed of each wind speed segment, as shown in Table 2As shown.
A cube root sum calculating unit, configured to obtain a cube root sum of each wind speed segment by using the appearance frequency cube root of each wind speed segment, where the cube root sum is a sum of the appearance frequency cube root of each wind speed segment and a cube root of a wind speed segment with an appearance frequency smaller than that of the wind speed segment, such as a tableCum in the quantitative statistics TableIt is shown, for example, that the sum of the cubic roots in the wind speed range 3m/s-4m/s is the frequency of occurrence of the cubic root 1.9129 in the range 3m/s-4m/s plus the frequency of occurrence of the cubic roots in the wind speed ranges 2m/s-3m/s, 1m/s-2m/s and 0m/s-1m/s, respectively, 1.8171, 0 and 0, equal to 3.7300.
The prediction period layering unit is used for layering the target correction wind speed prediction period in sequence from small to large according to the number of layers to obtain a plurality of groups of layering results with different numbers of layers;
the variance reduction rate calculation unit is used for obtaining the variance reduction rate of two adjacent layering results by utilizing each layering result;
and the layering result determining unit is used for selecting a layering result with the variance reduction rate firstly lower than the previous layering result as a target layering result to obtain a multi-layer wind speed range, wherein the layering result comprises the multi-layer wind speed range with the number equal to that of layering.
Specifically, after a target correction wind speed prediction period is divided into a plurality of wind speed sections, carrying out layering operation, wherein the number of the layering sections is sequentially set to be L =1,2,3 \8230; the weighted variance is calculated by summing the products of the weight coefficients and the variance of each layer, for example, L =2, and the weight coefficients in the two layers are respectively L 1 ,L 2 Variance is Var 1 And Var 2 Then the weighted variance of the segments of the two layers can be calculated as (Var) 1 ×L 1 +Var 2 ×L 2 ) And then calculating the proportional variance between two adjacent segments, wherein the proportional variance is expected to decrease along with the increase of the number of the layering segments, and the number L of the segments when the variance reduction rate begins to decrease is the optimal number of the segmentation blocks, namely the target layering result.
As shown in the hierarchical table in table 3, as the number of segments reaches three, i.e., L =3, the variance reduction rate starts to decrease, and the optimal number of divided blocks is set to 3.
The method comprises the following steps of obtaining a unit layering range by eradicating the total cube of a wind speed section according to the current layering number, taking the wind speed section closest to the unit layering range in the cube roots of the wind speed section as a reference, and layering the wind speed section, which is closest to the unit layering range, in the cube roots of the wind speed section, for example, the total cube root of the wind speed section in the table 2 is 24.7711, if the current layering number is 2, the unit layering range is 24.7711/2=12.38555, and the wind speed section closest to 12.38555 is 12.6748 of the wind speed range 7m/s-8m/s, so as shown in table three, the wind speed range of 0m/s-8m/s is taken as a first layer, the wind speed range of 8m/s-15m/s is taken as a second layer, wherein 8m/s belongs to the first layer of 0m/s-8 m/s; if the current number of the layering layers is 3, the unit layering range is 24.7711/3=8.257, the wind speed range of the first segment is between 5m/s and 6m/s, the first-layer segmentation point r1 is set to be 6m/s, the wind speed range of the second segment is between 9m/s and 10m/s, and the second-layer segmentation point r2 is set to be 10m/s, so as shown in table three, the wind speed range of 0m/s-6m/s is taken as the first layer, the wind speed range of 6m/s-10m/s is taken as the second layer, and the wind speed range of 10m/s-15m/s is taken as the third layer.
TABLE 3 hierarchical tables
Specifically, the multi-model prediction module 17 is specifically configured to utilize a value time lag of a target historical wind power, and substitute each layer of wind speed range into a preset first prediction model, a preset second prediction model, and a preset third prediction model respectively to obtain multiple groups of initial wind power prediction values;
the first prediction model is formed by training an artificial neural network model established based on a BP (Back Propagation, multilayer feedforward Back Propagation) algorithm and an LM (Levenberg-Marquardt) algorithm by utilizing a historical wind power value time lag and a historical wind speed range;
the second prediction model is formed by training a model established based on an ANFIS algorithm by utilizing a historical wind power value time lag and a historical wind speed range;
and the third prediction model is formed by training a model established based on an LC-SVM algorithm by utilizing the value time lag of historical wind power and the range of historical wind speed.
Specifically, a target historical wind power dereferencing time lag is utilized, wind speed ranges of each layer are simultaneously and respectively input into an improved artificial neural network, a fuzzy inference system of a self-adaptive network and a least square support vector machine model to obtain a plurality of groups of initial wind power prediction values, for example, the wind speed ranges are three layers, namely, 0m/s-6m/s of a first layer, 6m/s-10m/s of a second layer and 10m/s-15m/s of a third layer, the wind speed data of each layer are respectively substituted into three prediction models according to the target historical wind power dereferencing time lag, so that each prediction model obtains 3 groups of initial wind power prediction values, for example, the wind speed ranges are three layers, namely, 0m/s-6m/s of the first layer, 6m/s-10m/s of the second layer and 10m/s-15m/s of the third layer, the wind speed dereferencing can be firstly according to the target historical wind power dereferencing, for example, the wind speeds corresponding to the previous unit time of the current time and the previous two unit times are respectively input into the first prediction model, the first group of the initial wind power prediction values are obtained, the second prediction model, the sixth prediction value of the third prediction model is input into the third prediction model, and the third prediction model, the third prediction value of the initial wind power prediction model is obtained, and the third prediction value of the initial wind power prediction model is obtained, and the third prediction value of the third prediction model is obtained, and the initial prediction model is input into the third prediction model, and the third prediction value of the initial prediction model, and the initial prediction model is obtained; the unit time can be 1 hour, 30 minutes and the like, and can be set according to the actual application requirements.
Specifically, the first prediction model is an artificial neural network model trained by using a multi-layer feedforward Back Propagation (BP) network as a training algorithm and using a modified LM (Levenberg-Marquardt) method, wherein the modified LM algorithm uses a steepest descent algorithm to improve an initial guess of parameters in an initial stage and then becomes Newton's method when the minimum value of the function is approached.
Specifically, the second prediction model is an Adaptive Network-based Fuzzy Inference System (ANFIS) model, the Network of which is a five-layer feedforward artificial neural Network including a Fuzzy layer, a rule layer, a normalization layer, a deblurring layer and a single summation neuron, and the membership function is a gaussian function.
Specifically, the third prediction model is a least squares support vector machine (LS-SVM) model, and the calculation formula is:
wherein alpha is i And alpha i * Is the lagrange multiplier and the kernel function is the inner product of two phi (x) functions, where phi (x) uses the radial basis function kernel, for convolution calculations.
Specifically, the hybrid prediction model in the hybrid model prediction module 18 is formed by training a model established based on a charged system search algorithm by using multiple groups of historical initial wind power prediction values;
the hybrid prediction model is: p fc =α 1 ·P f,1 +α 2 ·P f,2 +…+α n ·P f,n ;
In the formula, P fc Representing the wind power prediction value, n representing the number of prediction models, and alpha representing each predictionThe weight coefficients of the model.
Each prediction model comprises a plurality of groups of initial wind power prediction values, each group of initial wind power prediction values can comprise a plurality of initial wind power prediction values, and the final wind power prediction value can be a group of wind power prediction values.
Specifically, the training process of the hybrid prediction model comprises setting an objective function to minimize an absolute prediction error, wherein the absolute prediction error epsilon is defined as | P fc -P a L, wherein P a Is the actual value of the wind power. Initial weight α 1 ,α 2 ,α 3 Given in a random manner, the initial weight determines the initial position (x, y, z) of the charged particle in 3-D space, and the error ε is inversely proportional to the charge on the particle.
Wherein the magnitude of the charge on the particle is defined as:
where fittest and fitworst fitness for all particles, respectively, fit (i) represents the value of the objective function, i.e., the absolute prediction error, and N is the total number of charged particles.
Specifically, each charged particle is subjected to the resultant force of other charged particles, so that each charged particle moves to a new position, and the new position (x ', y ', z ') of each charged particle corresponds to a new weight of a prediction model, which is a new weight corresponding to the first prediction model, the second prediction model and the third prediction model in the embodiment of the present invention. All particles will gradually move to the position with the smallest prediction error, the final position (x) * ,y * ,z * ) The final positions respectively correspond to the final weights of the first prediction model, the second prediction model and the third prediction model.
The method comprises the following steps of constructing a wind power hybrid prediction model based on a charged system search algorithm, wherein the steps comprise S11 to S18; wherein,
s11: the charge amount of each charged particle is calculated using a fitness function.
S12: an initial position and velocity of each charged particle is determined.
S13: the probability of attraction between charged particles is determined.
S14: the resultant power acting on each charged particle is calculated.
S15: a new position and velocity of each charged particle is determined.
S16: charge storage is created to store the best charged particles to date.
S17: by using the data stored in the charge storage, the position of the charged particles is corrected if the charged particles violate the boundary constraint.
S18: if the charged particles stop moving or the program reaches the maximum iteration, the algorithm is finished; otherwise, return to S13.
Correspondingly, the embodiment of the invention discloses a distributed wind power plant layered hybrid short-term prediction method considering wind speed time lag, and referring to fig. 2, the method comprises the following steps:
s21: acquiring a historical wind speed time sequence and a historical wind power time sequence in advance;
s22: obtaining a target historical wind speed value time lag meeting a preset condition by utilizing the historical wind speed time sequence and the partial autocorrelation function;
s23: predicting the historical wind speed time sequence by using a numerical weather prediction model to obtain a target wind speed prediction period meeting preset conditions; the wind speed prediction period comprises a plurality of predicted wind speeds measured and recorded in a preset time period, and the numerical weather prediction model is formed by utilizing historical wind speed time sequence training data in advance;
s24: correcting the target wind speed prediction period to obtain a target corrected wind speed prediction period;
s25: obtaining a target historical wind power value time lag meeting a preset condition by utilizing the historical wind power time sequence and the partial autocorrelation function;
s26: by usingThe target corrected wind speed prediction period is layered according to the layering rule to obtain a multilayer wind speed range with the number of target layers;
s27: utilizing the value time lag of the target historical wind power, and substituting each layer of wind speed range into a plurality of preset initial prediction models respectively to obtain a plurality of groups of initial wind power prediction values; each initial prediction model is different and is formed by utilizing the historical wind power value time lag and the historical wind speed range in advance for training;
s28: obtaining a wind power predicted value by utilizing a plurality of groups of initial wind power predicted values and a preset hybrid prediction model; the hybrid prediction model is formed by training multiple groups of historical initial wind power prediction values.
Specifically, the step S23 of predicting the historical wind speed time series by using the numerical weather prediction model to obtain a target wind speed prediction period satisfying a preset condition specifically includes steps S231 to S233; wherein,
s231: predicting the historical wind speed time sequence by using a numerical weather prediction model to obtain a plurality of groups of wind speed prediction periods;
s232: comparing each group of wind speed prediction period with corresponding historical actual wind speed to obtain a prediction average error;
s233: and selecting a wind speed prediction period with the minimum prediction average error as a target wind speed prediction period.
Specifically, the step S232 of comparing each group of wind speed prediction period with the corresponding historical actual wind speed to obtain a prediction average error may specifically include steps S2321 and S2322; wherein,
s2321: comparing each group of wind speed prediction period with the corresponding historical actual wind speed to obtain individual prediction errors of each wind speed recorded in each group of wind speed prediction period and the corresponding historical actual wind speed;
s2322: and obtaining the predicted average error of each group of wind speed prediction period by using all individual errors in each group of wind speed prediction period.
Specifically, the predicted wind speed correction model in S24 is formed by training a model established based on an artificial neural network algorithm by using a value range of historical wind speed data and a predicted wind speed group in advance.
Specifically, the above-mentioned S26 usesThe process of layering the target corrected wind speed prediction period by the layering rules to obtain a multilayer wind speed range with the number of target layering numbers specifically comprises S261 to S266; wherein,
s261: counting the occurrence frequency of the wind speed of each preset wind speed section in the target correction wind speed prediction period; the preset wind speed section is a plurality of preset wind speed sections at equal intervals;
s262: obtaining the cubic root of the occurrence frequency of each wind speed section by using the occurrence frequency of the wind speed of each wind speed section;
s263: obtaining the sum of cubic roots of each wind speed section by using the occurrence frequency cubic roots of each wind speed section; the sum of the cubic roots is the sum of the cubic root of the occurrence frequency of each wind speed section and the cubic root of the occurrence frequency of the wind speed section smaller than the wind speed;
s264: layering the target corrected wind speed prediction period from small to large according to the layer number to obtain a plurality of groups of layering results with different layer numbers;
s265: obtaining the variance reduction rate of two adjacent layering results by utilizing each layering result;
s266: selecting a layering result with the variance reduction rate firstly lower than the previous layering result as a target layering result to obtain a multilayer wind speed range;
wherein the layering result comprises a multi-layer wind speed range equal to the layering number.
Specifically, in the step S27, the target historical wind power value time lag is used to substitute each layer of wind speed range into a plurality of preset initial prediction models, so as to obtain a plurality of groups of initial wind power prediction values, and specifically, the target historical wind power value time lag is used to substitute each layer of wind speed range into a first preset prediction model, a second preset prediction model, and a third preset prediction model, so as to obtain a plurality of groups of initial wind power prediction values;
the first prediction model is formed by training a model established based on a BP algorithm and an LM algorithm by utilizing a historical wind power value time lag and a historical wind speed range;
the second prediction model is formed by training a model established based on an ANFIS algorithm by utilizing a historical wind power value time lag and a historical wind speed range;
and the third prediction model is formed by training a model established based on an LC-SVM algorithm by utilizing the value time lag of historical wind power and the range of historical wind speed.
Specifically, the hybrid prediction model in S28 is formed by training a model established based on a search algorithm of the charged system, using multiple groups of historical initial wind power prediction values;
the hybrid prediction model is: p is fc =α 1 ·P f,1 +α 2 ·P f,2 +…+α n ·P f,n ;
In the formula, P fc Representing the wind power predicted value, n representing the number of prediction models, and alpha representing the weight coefficient of each prediction model.
In addition, the embodiment of the invention also discloses a device for layering and mixing short-term prediction of the distributed wind power plant in consideration of the time lag of the wind speed, which comprises the following steps:
a memory for storing a computer program;
a processor for executing a computer program to implement a decentralized wind farm tiered mixing short term prediction method taking into account wind speed time lag as described above.
Finally, it should also be noted that, in this document, relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrases "comprising a," "8230," "8230," or "comprising" does not exclude the presence of additional like elements in a process, method, article, or apparatus that comprises the element.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
The method, the system and the device for the layered hybrid short-term prediction of the distributed wind power plant considering the time lag of the wind speed are introduced in detail, a specific example is applied in the method to explain the principle and the implementation mode of the invention, and the description of the embodiment is only used for helping to understand the method and the core idea of the invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
Claims (10)
1. A decentralized wind farm tiered hybrid short-term prediction system that takes into account wind speed time-lag, comprising:
the data acquisition interface is used for receiving the historical wind speed time sequence and the historical wind power time sequence;
the wind speed time lag generation module is used for obtaining a target historical wind speed value time lag meeting a preset condition by utilizing the historical wind speed time sequence and the partial autocorrelation function;
the wind speed forecasting period module is used for forecasting the historical wind speed time sequence by using a numerical weather forecasting model to obtain a target wind speed forecasting period meeting preset conditions; the wind speed forecasting period comprises a plurality of forecasting wind speeds which are measured and recorded in a preset time period, and the numerical weather forecasting model is formed by utilizing historical wind speed time sequence training data in advance;
the prediction period correction module is used for correcting the target wind speed prediction period to obtain a target corrected wind speed prediction period;
the wind power time lag generation module is used for obtaining a target historical wind power value time lag meeting a preset condition by utilizing the historical wind power time sequence and the partial autocorrelation function;
forecast period layering module for utilizingThe target corrected wind speed prediction period is layered according to a layering rule to obtain a multilayer wind speed range with the number of target layers; wherein,cubic roots representing the frequency of occurrence of each wind speed segment;
the multi-model prediction module is used for substituting the wind speed range of each layer into a plurality of preset initial prediction models respectively by utilizing the value time lag of the target historical wind power to obtain a plurality of groups of initial wind power prediction values; each initial prediction model is different and is formed by utilizing the historical wind power value time lag and the historical wind speed range in advance for training;
the hybrid model prediction module is used for obtaining a wind power prediction value by utilizing a plurality of groups of initial wind power prediction values and a preset hybrid prediction model; the hybrid prediction model is formed by training a plurality of groups of historical initial wind power predicted values.
2. The decentralized wind farm hierarchical hybrid short-term prediction system according to claim 1 taking into account wind speed time lag, wherein the wind speed prediction period module comprises:
the wind speed prediction period unit is used for predicting the historical wind speed time sequence by utilizing the numerical weather prediction model to obtain a plurality of groups of wind speed prediction periods;
the error calculation unit is used for comparing each group of wind speed prediction period with the corresponding historical actual wind speed to obtain a prediction average error;
and the error screening unit is used for selecting the wind speed prediction period with the minimum prediction average error as the target wind speed prediction period.
3. The decentralized wind farm hierarchical hybrid short term prediction system according to claim 2, wherein the error calculation unit comprises:
the individual error calculation subunit is used for comparing each group of wind speed prediction period with the corresponding historical actual wind speed to obtain individual prediction errors of each wind speed recorded in each group of wind speed prediction period and the corresponding historical actual wind speed;
and the average error calculation subunit is used for obtaining the predicted average error of each group of wind speed prediction periods by using all the individual errors in each group of wind speed prediction periods.
4. The distributed wind farm layered hybrid short-term prediction system considering wind speed time lag according to claim 3, wherein the prediction period correction module is specifically configured to obtain a corrected target predicted wind speed group by using the target historical wind speed value time lag, the target corrected wind speed prediction period, and a predicted wind speed correction model; the predicted wind speed correction model is formed by training in advance by utilizing a historical wind speed data value range and a predicted wind speed group.
5. The distributed wind power plant layered hybrid short-term prediction system considering wind speed time lag according to claim 4, wherein the predicted wind speed correction model is formed by training a model established based on an artificial neural network algorithm by utilizing a historical wind speed data value range and a predicted wind speed group in advance.
6. The decentralized wind farm tiered hybrid short term prediction system according to claim 4 that considers wind speed time lag, wherein the forecast period tiering module comprises:
the occurrence frequency counting unit is used for counting the occurrence frequency of the wind speed of each preset wind speed section in the target correction wind speed prediction period; the preset wind speed section is a plurality of preset wind speed sections at equal intervals;
the cubic root calculating unit is used for obtaining the cubic root of the occurrence frequency of each wind speed section by utilizing the occurrence frequency of the wind speed of each wind speed section;
the cubic root sum calculating unit is used for obtaining the cubic root sum of each wind speed section by utilizing the cubic root of the occurrence frequency of each wind speed section; the sum of the cubic roots is the sum of the cubic root of the occurrence frequency of each wind speed section and the cubic root of the occurrence frequency of the wind speed section smaller than the wind speed;
the prediction period layering unit is used for layering the target correction wind speed prediction period from small layer number to large layer number in sequence to obtain a plurality of groups of layering results with different layer numbers;
the variance reduction rate calculation unit is used for obtaining the variance reduction rate of two adjacent layering results by utilizing each layering result;
the layering result determining unit is used for selecting a layering result with the variance reduction rate firstly lower than that of the last layering result as a target layering result to obtain a multilayer wind speed range;
wherein, the layering result comprises a multi-layer wind speed range equal to the layering number.
7. The distributed wind power plant layered hybrid short-term prediction system considering wind speed time lag according to claim 6, wherein the multi-model prediction module is specifically configured to utilize the target historical wind power value time lag to respectively substitute a wind speed range of each layer into a preset first prediction model, a preset second prediction model and a preset third prediction model to obtain multiple groups of initial wind power prediction values;
the first prediction model is formed by training an artificial neural network model established based on a BP algorithm and an LM algorithm by utilizing the historical wind power value time lag and the historical wind speed range;
the second prediction model is formed by training a model established based on an ANFIS algorithm by utilizing the historical wind power value time lag and the historical wind speed range;
and the third prediction model is formed by training a model established based on an LC-SVM algorithm by utilizing the historical wind power value time lag and the historical wind speed range.
8. The distributed wind power plant layered hybrid short-term prediction system considering wind speed time lag according to any one of claims 1 to 6, wherein the hybrid prediction model is trained on a model established based on a charged system search algorithm by using multiple groups of historical initial wind power predicted values;
the hybrid prediction model is as follows: p fc =α 1 ·P f,1 +α 2 ·P f,2 +…+α n ·P f,n ;
In the formula, P fc Representing the wind power predicted value, n representing the number of prediction models, and alpha representing the weight coefficient of each prediction model.
9. A distributed wind power plant layered hybrid short-term prediction method considering wind speed time lag is characterized by comprising the following steps:
acquiring a historical wind speed time sequence and a historical wind power time sequence in advance;
obtaining a target historical wind speed value time lag meeting a preset condition by using the historical wind speed time sequence and the partial autocorrelation function;
predicting the historical wind speed time sequence by using a numerical weather prediction model to obtain a target wind speed prediction period meeting preset conditions; the wind speed prediction period comprises a plurality of predicted wind speeds measured and recorded in a preset time period, and the numerical weather prediction model is formed by utilizing historical wind speed time sequence training data in advance;
correcting the target wind speed prediction period to obtain a target corrected wind speed prediction period;
obtaining a target historical wind power value time lag meeting a preset condition by using the historical wind power time sequence and the partial autocorrelation function;
by usingLayering the target correction wind speed prediction period according to a layering rule to obtain a multilayer wind speed range with the number of target layering; wherein,cubic roots representing the frequency of occurrence of each wind speed segment;
utilizing the value time lag of the target historical wind power, and substituting each layer of wind speed range into a plurality of preset initial prediction models respectively to obtain a plurality of groups of initial wind power prediction values; each initial prediction model is different and is formed by utilizing historical wind power value time lag and historical wind speed range training in advance;
obtaining a wind power predicted value by utilizing a plurality of groups of initial wind power predicted values and a preset hybrid prediction model; the hybrid prediction model is formed by training multiple groups of historical initial wind power prediction values.
10. A distributed wind farm layered hybrid short-term prediction device considering wind speed time lag is characterized by comprising:
a memory for storing a computer program;
a processor for executing said computer program to implement the decentralized wind farm tiered hybrid short term prediction method taking into account wind speed time lag as defined in claim 9.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910129892.8A CN109886488B (en) | 2019-02-21 | 2019-02-21 | Distributed wind power plant layered hybrid short-term prediction method considering wind speed time lag |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910129892.8A CN109886488B (en) | 2019-02-21 | 2019-02-21 | Distributed wind power plant layered hybrid short-term prediction method considering wind speed time lag |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109886488A CN109886488A (en) | 2019-06-14 |
CN109886488B true CN109886488B (en) | 2022-10-18 |
Family
ID=66928760
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910129892.8A Active CN109886488B (en) | 2019-02-21 | 2019-02-21 | Distributed wind power plant layered hybrid short-term prediction method considering wind speed time lag |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109886488B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112232566A (en) * | 2020-10-16 | 2021-01-15 | 华中科技大学 | Method and system for predicting short-term water consumption rate of hydropower station |
CN112613674B (en) * | 2020-12-29 | 2024-03-08 | 国能日新科技股份有限公司 | Medium-and-long-term wind power generation capacity prediction method and device, electronic equipment and storage medium |
CN116760006A (en) * | 2023-04-28 | 2023-09-15 | 觐联(上海)工业技术有限公司 | Wind power prediction method, device, equipment and storage medium |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105184391A (en) * | 2015-08-19 | 2015-12-23 | 国网山东省电力公司电力科学研究院 | Method for predicting wind speed and power of wind farm based on wavelet decomposition and support vector machine |
CN107133695A (en) * | 2017-04-28 | 2017-09-05 | 北京天诚同创电气有限公司 | A kind of wind power forecasting method and system |
CN107507097A (en) * | 2017-07-03 | 2017-12-22 | 上海电力学院 | A kind of short-term wind power prediction method |
CN107765347A (en) * | 2017-06-29 | 2018-03-06 | 河海大学 | A kind of Gaussian process returns and the short-term wind speed forecasting method of particle filter |
CN108549962A (en) * | 2018-06-04 | 2018-09-18 | 中国农业大学 | Wind power forecasting method based on history fragment sequence search and sequential rarefaction |
CN109242204A (en) * | 2018-09-30 | 2019-01-18 | 淮阴工学院 | Ultra-short term wind speed forecasting method based on optimal VMD and Synchronous fluorimetry |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI476430B (en) * | 2010-08-23 | 2015-03-11 | Inst Nuclear Energy Res Atomic Energy Council | A wind energy forecasting method with extreme wind speed prediction function |
US9230219B2 (en) * | 2010-08-23 | 2016-01-05 | Institute Of Nuclear Energy Research Atomic Energy Council, Executive Yuan | Wind energy forecasting method with extreme wind speed prediction function |
-
2019
- 2019-02-21 CN CN201910129892.8A patent/CN109886488B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105184391A (en) * | 2015-08-19 | 2015-12-23 | 国网山东省电力公司电力科学研究院 | Method for predicting wind speed and power of wind farm based on wavelet decomposition and support vector machine |
CN107133695A (en) * | 2017-04-28 | 2017-09-05 | 北京天诚同创电气有限公司 | A kind of wind power forecasting method and system |
CN107765347A (en) * | 2017-06-29 | 2018-03-06 | 河海大学 | A kind of Gaussian process returns and the short-term wind speed forecasting method of particle filter |
CN107507097A (en) * | 2017-07-03 | 2017-12-22 | 上海电力学院 | A kind of short-term wind power prediction method |
CN108549962A (en) * | 2018-06-04 | 2018-09-18 | 中国农业大学 | Wind power forecasting method based on history fragment sequence search and sequential rarefaction |
CN109242204A (en) * | 2018-09-30 | 2019-01-18 | 淮阴工学院 | Ultra-short term wind speed forecasting method based on optimal VMD and Synchronous fluorimetry |
Also Published As
Publication number | Publication date |
---|---|
CN109886488A (en) | 2019-06-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Adedeji et al. | Wind turbine power output very short-term forecast: A comparative study of data clustering techniques in a PSO-ANFIS model | |
CN108564192B (en) | Short-term photovoltaic power prediction method based on meteorological factor weight similarity day | |
Pousinho et al. | A hybrid PSO–ANFIS approach for short-term wind power prediction in Portugal | |
CN109886488B (en) | Distributed wind power plant layered hybrid short-term prediction method considering wind speed time lag | |
CN103268366A (en) | Combined wind power prediction method suitable for distributed wind power plant | |
CN110969290A (en) | Runoff probability prediction method and system based on deep learning | |
Kolhe et al. | GA-ANN for short-term wind energy prediction | |
CN117498400B (en) | Distributed photovoltaic and energy storage data processing method and system | |
CN114362175B (en) | Wind power prediction method and system based on depth certainty strategy gradient algorithm | |
CN116596044B (en) | Power generation load prediction model training method and device based on multi-source data | |
Eseye et al. | Short-term wind power forecasting using a double-stage hierarchical hybrid GA-ANN approach | |
CN114298140A (en) | Wind power short-term power prediction correction method considering unit classification | |
CN114118596A (en) | Photovoltaic power generation capacity prediction method and device | |
CN115186923A (en) | Photovoltaic power generation power prediction method and device and electronic equipment | |
CN113722980A (en) | Ocean wave height prediction method, system, computer equipment, storage medium and terminal | |
CN112906935A (en) | Method for predicting ultra-short-term power of wind power plant | |
Kumar et al. | An ensemble model for short-term wind power forecasting using deep learning and gradient boosting algorithms | |
CN116826710A (en) | Peak clipping strategy recommendation method and device based on load prediction and storage medium | |
CN108038518A (en) | A kind of photovoltaic generation power based on meteorological data determines method and system | |
CN116681154A (en) | Photovoltaic power calculation method based on EMD-AO-DELM | |
Li et al. | A short-term wind power forecasting method based on NWP wind speed fluctuation division and clustering | |
CN114444763A (en) | Wind power prediction method based on AFSA-GNN | |
CN112183814A (en) | Short-term wind speed prediction method | |
Abdullah et al. | A hybrid neuro-fuzzy & bootstrap prediction system for wind power generation | |
CN116632842B (en) | Clustering characteristic-based method and system for predicting distribution type photovoltaic load probability of platform |
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 |