CN115912334A - Method for establishing prediction model of output guarantee rate of wind power plant and prediction method - Google Patents
Method for establishing prediction model of output guarantee rate of wind power plant and prediction method Download PDFInfo
- Publication number
- CN115912334A CN115912334A CN202211389698.1A CN202211389698A CN115912334A CN 115912334 A CN115912334 A CN 115912334A CN 202211389698 A CN202211389698 A CN 202211389698A CN 115912334 A CN115912334 A CN 115912334A
- Authority
- CN
- China
- Prior art keywords
- meteorological
- output
- simulation result
- rate
- guarantee rate
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Abstract
The invention relates to the technical field of wind power generation, discloses a method for creating a prediction model of a wind power plant output guarantee rate and a prediction method, and aims to solve the problem of poor accuracy of the existing output guarantee rate prediction method, and the scheme mainly comprises the following steps: acquiring an actual output guarantee rate of a produced wind power plant based on a time sequence in a first preset time period, and acquiring first meteorological observation data of a site area of the wind power plant in the first preset time period; carrying out a four-dimensional data assimilation test on the first meteorological observation data by using a WRF-FDDA assimilation system to obtain a first meteorological simulation result based on a time sequence; training a prediction model of the output assurance rate according to the actual output assurance rate and the first meteorological simulation result; and predicting the output guarantee rate based on the prediction model. The invention improves the accuracy of the output guarantee rate prediction, and is particularly suitable for the development of wind power plants.
Description
Technical Field
The invention relates to the technical field of wind power generation, in particular to a method for creating a prediction model of a guarantee rate of output of a wind power plant and a prediction method.
Background
The output guarantee rate of the proposed wind power plant based on the time sequence is needed in the planning stage of the wind power plant, and the wind power output guarantee rate has the characteristics of volatility and randomness, so that the difference between the wind power output guarantee rate and the traditional thermal power and hydroelectric power output characteristics is large, and the analysis of the wind power characteristics by adopting a traditional output curve method is difficult. Along with the change of time, the output guarantee rate of the wind power project has stronger fluctuation characteristics in the year and the day. In the prior art, the hourly output guarantee rate of the wind power plant is calculated by using mesoscale historical meteorological data in engineering, but the hourly output guarantee rate calculated by using the mesoscale meteorological data is greatly different from the actual operation condition, so that the errors of the annual output guarantee rate obtained by calculation and the annual output guarantee rate obtained by actual operation data of the wind power plant are large at low output probability and high output probability.
The prior art mainly adopts the following method to determine the output guarantee rate of new energy power generation: the method comprises the steps of firstly, carrying out cluster analysis on historical output data of the new energy station by utilizing the space-time correlation of new energy power generation under different resource environmental conditions, taking different cluster center curves as typical output curves, constructing a new energy differential output model, and calculating an output guarantee rate based on the new energy differential output model. And on the basis of the measured data of the anemometer tower for one year, on the basis of the arrangement scheme of the positions of the wind power plant, a fluid computation CFD model is adopted, a wind power effective output computation model based on a wind resource assessment inaccurate analysis method is provided, and then the output guarantee rate is computed based on the wind power effective output computation model.
The prior art has at least the following defects:
firstly, the meteorological data and the operation data of the wind power plant have a missing phenomenon. Meteorological data, actual operating data of a wind farm, and the like are typical spatio-temporal data, i.e., there are significant temporal and spatial features. Due to the fact that monitoring stations are unevenly distributed and limited in number, and various mechanical problems, environmental problems and thought factors cause the monitoring data to have a missing phenomenon, when the traditional method fills up the missing phenomenon of the data, most of the traditional methods only aim at analysis results of space or time, and influence of space-time correlation of variables and other related variables is ignored.
Secondly, the simulation precision of the existing output model is not enough. For the existing power supply planning, in order to solve the situation that no actual measurement wind power data exists, engineers often use mesoscale historical wind field data of Greenwich power to calculate the hourly output guarantee rate of a wind power plant, but the hourly output guarantee rate obtained by mesoscale meteorological data is greatly different from the actual operation situation, and the three aspects are mainly shown, namely, the hourly output probability obtained by mesoscale data calculation is low and is generally lower than 10%, but the annual output probability is higher than 10% in the actual operation process; secondly, in the weather condition of clear no wind or good quiet light wind, the guarantee rate (about 80%) that the output process is 0 exists in the actual operation process of the wind power plant, but the low output probability obtained by the calculation of the mesoscale data has obvious errors; thirdly, hourly output guarantee rates obtained by taking mesoscale data as input through calculation are distributed smoothly within a day, and the output process of actual operation of the wind power plant is influenced by snowfall, rainfall, strong wind and the like, so that the fluctuation within a day is strong.
Thirdly, the space-time resolution of the existing mesoscale meteorological data is not enough. For the existing mesoscale meteorological numerical mode, the WRF mode is widely applied, but the result of the WRF mode has some uncertainties, and the sources of the uncertainties mainly comprise errors caused by parameterization, errors of initial conditions and boundary conditions, meteorological field input errors and the like.
And fourthly, the existing output model obtains the typical annual output guarantee rate, and when water-wind-light complementary planning is carried out, the wind-power and photovoltaic output time sequence which needs many years is measured and calculated through water-wind-light complementary measurement due to large difference of abundance between water-power and wind-power interplanetals.
Disclosure of Invention
The invention aims to solve the problem that the existing method for predicting the output guarantee rate of a wind power plant is poor in accuracy, and provides a method for creating a prediction model of the output guarantee rate of the wind power plant and a prediction method.
The technical scheme adopted by the invention for solving the technical problems is as follows:
in a first aspect, a method for creating a prediction model of a wind power plant output guarantee rate is provided, which includes the following steps:
step 1, acquiring an actual output guarantee rate of a produced wind power plant based on a time sequence in a first preset time period, and acquiring first meteorological observation data of a site area of the wind power plant in the first preset time period;
step 2, performing a four-dimensional data assimilation test on the first meteorological observation data by using a WRF-FDDA assimilation system to obtain a first meteorological simulation result based on a time sequence;
and 3, training a prediction model of the output guarantee rate according to the actual output guarantee rate and the first meteorological simulation result.
Further, after obtaining the first weather simulation result, the method further comprises:
calculating a first correlation coefficient, a first average deviation, a first normalized average deviation and a first root mean square error between the first meteorological observation data and a first meteorological simulation result;
and judging whether the first correlation coefficient, the first average deviation, the first standardized average deviation and the first root mean square error are all in corresponding preset ranges, if so, entering a step 3, otherwise, entering a step 2 after adjusting assimilation test parameters.
Further, the calculation formula of the first correlation coefficient R is as follows:
the calculation formula of the first average deviation MB is as follows:
the first normalized mean deviation NMB is calculated as follows:
the first root mean square error RSME is calculated as follows:
in the formula, m o The observation value corresponding to the first meteorological observation data,
is the mean value of the observed values, is>
m f For the analog value corresponding to the first weather simulation result, is selected>
Is the mean value of the analog value>
And N is the number of samples of the first meteorological observation data.
Further, training a prediction model of the output assurance rate according to the actual output assurance rate and the first weather simulation result specifically includes:
step 31, establishing an LSTM model according to the model function and the meteorological element combination mode in the first meteorological simulation result;
step 32, dividing the actual output guarantee rate based on the time sequence and the first meteorological simulation result into a training set and a verification set, and training an LSTM model according to the actual output guarantee rate in the training set and the corresponding first meteorological simulation result;
step 33, determining the average relative error, the average absolute error and the prediction accuracy of the LSTM model according to the actual output guarantee rate of the verification set and the corresponding first meteorological simulation result;
and step 34, judging whether the average relative error, the average absolute error and the prediction accuracy are all in corresponding preset ranges, if so, taking the LSTM model as a prediction model of the output guarantee rate, otherwise, after adjusting a model function and/or a meteorological element combination mode in a first meteorological simulation result, entering step 32.
Further, the step 33 specifically includes:
and inputting the first meteorological simulation result in the verification set into an LSTM model to obtain a predicted output guarantee rate, and calculating an average relative error, an average absolute error and a prediction accuracy according to the predicted output guarantee rate and an actual output guarantee rate.
Further, the average relative error
The calculation formula of (a) is as follows:
the mean absolute error
The calculation formula of (a) is as follows:
the calculation formula of the prediction accuracy P is as follows:
in the formula,. DELTA. i ＝|X i -Y i |，X i Predicting the guarantee rate of output, Y, for the ith i For the ith actual contribution rate, Δ i Is the absolute error of the ith data, and M is the number of samples in the validation set.
Further, the meteorological elements in the first meteorological simulation result comprise temperature, wind direction, wind speed and humidity.
Further, the first meteorological observation data are obtained according to FNL reanalyzed meteorological data and observation data of meteorological observation stations in a site area of the wind power plant.
In a second aspect, a method for predicting a wind farm output guarantee rate is provided, including the method for creating a prediction model of a wind farm output guarantee rate according to the first aspect, and further including the following steps:
step 4, obtaining second meteorological observation data of a site area of the wind power plant to be developed in a second preset time period;
step 5, performing a four-dimensional data assimilation test on the second meteorological observation data by using a WRF-FDDA assimilation system to obtain a second meteorological simulation result based on a time sequence;
and 6, determining the output guarantee rate of the site area in a second preset time period according to the second meteorological simulation result and on the basis of the prediction model.
Further, after obtaining the second meteorological simulation result, the method further comprises:
calculating a second correlation coefficient, a second average deviation, a second normalized average deviation and a second root mean square error between the second meteorological observation data and a second meteorological simulation result;
and judging whether the second correlation coefficient, the second average deviation, the second normalized average deviation and the second root mean square error are all in corresponding preset ranges, if so, entering a step 6, otherwise, entering a step 5 after adjusting assimilation test parameters.
The invention has the beneficial effects that: the method for establishing the prediction model of the wind power plant output assurance rate and the prediction method thereof carry out data assimilation test on meteorological observation data based on a WRF-FDDA assimilation system, combine the continuous constraint of physical variables and the time change rule of state variables, and input the combined result into the mode simulation, thereby greatly making up the uncertainty of the pure mode simulation, effectively improving the meteorological background field, solving the problem of low spatial and temporal resolution precision and insufficient precision of medium-scale meteorological data, and simultaneously establishing the output wind power plant output assurance rate prediction model by using the meteorological simulation result after the data assimilation test, improving the accuracy of the output assurance rate prediction model, wherein the output assurance rate calculated based on the prediction model has lower error and the annual output assurance rate is more in line with the actual situation.
Drawings
FIG. 1 is a schematic flow chart of a method for creating a prediction model of a guarantee rate of output of a wind farm according to an embodiment of the present invention;
FIG. 2 is another schematic flow chart of a method for creating a prediction model of a wind farm output guarantee rate according to an embodiment of the present invention;
FIG. 3 is a schematic flow chart of a predictive model training method according to an embodiment of the invention;
FIG. 4 is a schematic flow chart of a method for predicting a guarantee rate of output of a wind farm according to an embodiment of the present invention;
FIG. 5 is another schematic flow chart of a method for predicting a guarantee rate of output of a wind farm according to an embodiment of the present invention;
FIG. 6 is a graphical illustration of annual contribution guarantee rate as determined by the prior art;
FIG. 7 is a graphical representation of annual contribution guarantee rates determined by an embodiment of the present invention.
Detailed Description
Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
The invention aims to improve the accuracy of the prediction of the output guarantee rate of a wind power plant, and provides a method for establishing a prediction model of the output guarantee rate of the wind power plant and a prediction method, wherein the main technical scheme comprises the following steps: acquiring an actual output guarantee rate of a produced wind power plant based on a time sequence in a first preset time period, and acquiring first meteorological observation data of a site area of the wind power plant in the first preset time period; carrying out a four-dimensional data assimilation test on the first meteorological observation data by using a WRF-FDDA assimilation system to obtain a first meteorological simulation result based on a time sequence; training a prediction model of the output assurance rate according to the actual output assurance rate and the first meteorological simulation result; and predicting the output guarantee rate based on the prediction model.
Specifically, the method includes the steps that firstly, a site area with a built wind power plant is selected, the actual output guarantee rate of the wind power plant based on a time sequence in a preset time period is obtained, for example, the hourly actual output guarantee rate in multiple years is obtained, the hourly actual output guarantee rate is the average value of output active power in each hour, and meteorological observation data in the site area in multiple years are obtained, and in actual application, the meteorological observation data are possibly lost, namely the preset time period cannot be completely covered; then, a WRF-FDDA assimilation system is used for carrying out a four-dimensional data assimilation test on the first meteorological observation data, and an assimilation meteorological result, namely a first meteorological simulation result (mainly comprising temperature, wind direction, wind speed, humidity and the like), is obtained through simulation; then, a wind power plant output guarantee rate prediction model with dominant meteorological conditions (wind speed) is established by using different functions and meteorological element combination methods in the LSTM, and the output guarantee rate can be predicted based on the prediction model.
Examples
Terms to be presented in the present embodiment will be explained below.
WRF model: the WRF (Weather Research and learning Model) mode is a new generation mesoscale Forecasting mode and assimilation system jointly developed by a plurality of Research units such as the national environmental Forecasting center (NCEP) and the national atmospheric Research center (NCAR). The WRF mode can meet the requirements of multi-scale atmospheric research simulation experiments and meteorological forecast services of a horizontal grid from a few meters to a few kilometers, is a fully compressible and non-static mode, and can provide a plurality of dynamic modes and different physical parameters. The mode is widely applied to the fields of atmosphere, ocean, environment and the like, and plays an important role in scientific research and business forecast.
Four-dimensional data assimilation (FDDA): nudging is a Four-Dimensional Data Assimilation FDDA (finite-Dimensional Data analysis), which is a continuous form of Data Assimilation applied at each time step over a period of time. The four-dimensional data assimilation is to introduce observation data at irregular moments into an initial field and a forecast initial field of a numerical mode in time to form a numerical weather analysis-forecast system, so that the analysis and forecast initial fields are simultaneously and continuously updated to obtain continuous weather analysis and forecast images.
The hourly output is as follows: the average value of the output active power in each hour is ranked according to 8760/8761 hours per year.
Long short term memory network (LSTM): the long-short term memory network is a time-cycle neural network, and is specially designed to solve the long-term problems of the general RNN (recurrent neural network). Due to the unique design structure, LSTM is suitable for handling and predicting significant events in time series that are very long-spaced and delayed.
The method for creating the prediction model of the wind power plant output guarantee rate, disclosed by the embodiment of the invention, as shown in fig. 1 and 2, comprises the following steps of:
step 1, acquiring an actual output guarantee rate of a put-in-production wind power plant based on a time sequence in a first preset time period, and acquiring first meteorological observation data of a site area of the wind power plant in the first preset time period;
in this embodiment, the hourly actual output guarantee rate of the produced wind farm for many years can be obtained, and hourly observation Data of a multi-year FNL (Final reality Data) Reanalysis meteorological station in a site area and a ground automatic meteorological observation station are used as the first meteorological observation Data.
Step 2, performing a four-dimensional data assimilation test on the first meteorological observation data by using a WRF-FDDA assimilation system to obtain a first meteorological simulation result based on a time sequence;
specifically, the physical parameterization schemes for setting the WRF-FDD mode in this step include micro-physics, cloud parameterization, planetary boundary layers, land surface models, and atmospheric radiation physics (long wave, short wave radiation), initial conditions, boundary conditions, and meteorological fields. WRF modes can set single-layer, double-layer, and multi-layer region nesting. And carrying out continuous four-dimensional data assimilation tests on the first meteorological observation data by using a WRF-FDDA assimilation system to obtain a first meteorological simulation result based on a time sequence. Wherein, the meteorological elements in the first meteorological simulation result mainly comprise temperature, wind direction, wind speed and humidity.
In order to further improve the accuracy of the prediction model, the embodiment further includes, after obtaining the first weather simulation result:
calculating a first correlation coefficient, a first average deviation, a first normalized average deviation and a first root mean square error between the first meteorological observation data and a first meteorological simulation result;
and judging whether the first correlation coefficient, the first average deviation, the first standardized average deviation and the first root mean square error are all in corresponding preset ranges, if so, entering a step 3, otherwise, entering a step 2 after adjusting assimilation test parameters.
In this embodiment, the calculation formula of the first correlation coefficient R is as follows:
the calculation formula of the first average deviation MB is as follows:
the first normalized mean deviation NMB is calculated as follows:
the first root mean square error RSME is calculated as follows:
in the formula, m o Is a firstAn observation value corresponding to the meteorological observation data,
is the mean value of the observed values, is>
m f For the analog value corresponding to the first weather simulation result, is selected>
Is the mean value of the analog value>
And N is the number of samples of the first meteorological observation data.
Wherein the magnitude of the first correlation coefficient represents a linear correlation between the simulation result and the observation data. The positive and negative of the first average deviation indicate that the simulation result is relative to the observation data, the positive of the first average deviation indicates that the simulation result is larger than the observation data, and the negative of the first average deviation indicates that the simulation result is smaller than the observation data. The first normalized mean deviation represents the average degree of dispersion of the simulation results from the observed data. The magnitude of the first root mean square error represents the deviation degree of the mode result and the observed data, the larger the first root mean square error result is, the larger the deviation is, the worse the stability of the simulation result is, and otherwise, the better the simulation result is.
It can be understood that when the first correlation coefficient, the first average deviation, the first normalized average deviation and the first root mean square error are all in the corresponding preset ranges, the simulation result is better, at this time, the creation of the prediction model is performed, when the first correlation coefficient, the first average deviation, the first normalized average deviation and the first root mean square error are not all in the corresponding preset ranges, the simulation result is worse, at this time, the assimilation parameter is adjusted to perform the assimilation test again. Therefore, the accuracy of the initial data is ensured, and the accuracy of the prediction model is further improved.
The preset range may be set according to actual requirements, and this embodiment does not limit this.
And 3, training a prediction model of the output guarantee rate according to the actual output guarantee rate and the first meteorological simulation result.
In this embodiment, the step of training the prediction model specifically includes:
step 31, establishing an LSTM model according to the model function and the meteorological element combination mode in the first meteorological simulation result;
the step of establishing the LSTM model in this embodiment includes: setting parameters of the neural network, including the number of hidden layers, the number of neurons of the hidden layers, function training and the like; and based on the first meteorological simulation result, trying different functions and meteorological element combination methods to establish a meteorological condition-dominant wind power plant output guarantee rate prediction model.
Step 32, dividing the actual output guarantee rate based on the time sequence and the first meteorological simulation result into a training set and a verification set, and training an LSTM model according to the actual output guarantee rate in the training set and the corresponding first meteorological simulation result;
in this embodiment, the actual output assurance rate of the time sequence 80% before the wind farm is put into production and the first weather simulation result are used as a training set, the actual output assurance rate of the time sequence 20% before the wind farm is put into production and the first weather simulation result are used as a verification set, and the LSTM model established in step 31 is trained according to the training set, and please refer to fig. 3 for a specific training flow.
Step 33, determining the average relative error, the average absolute error and the prediction accuracy of the LSTM model according to the actual output guarantee rate of the verification set and the corresponding first meteorological simulation result;
specifically, the first meteorological simulation result in the verification set is input into an LSTM model to obtain a predicted output guarantee rate, and an average relative error, an average absolute error and a predicted accuracy rate are calculated according to the predicted output guarantee rate and an actual output guarantee rate.
In this embodiment, the average relative error
The calculation formula of (a) is as follows:
the mean absolute error
The calculation formula of (a) is as follows:
the calculation formula of the prediction accuracy P is as follows:
in the formula,. DELTA. i ＝|X i -Y i |，X i Predicting the guarantee rate of output, Y, for the ith i For the ith actual contribution rate, Δ i Is the absolute error of the ith data, and M is the number of samples in the validation set.
Wherein, the smaller the average relative error is, the more similar the two groups of data are, the closer the predicted value is to the actual value. The smaller the average absolute error is, the more similar the two groups of data are, and the closer the predicted value is to the actual value. The larger the prediction accuracy is, the closer the representative predicted value is to the actual value, and the better the prediction effect of the representative prediction model is.
And step 34, judging whether the average relative error, the average absolute error and the prediction accuracy are all in corresponding preset ranges, if so, taking the LSTM model as a prediction model of the output guarantee rate, otherwise, after adjusting a model function and/or a meteorological element combination mode in a first meteorological simulation result, entering step 32.
It can be understood that when the average relative error, the average absolute error and the prediction accuracy are all in the corresponding preset ranges, the accuracy of the prediction model is high, the prediction model can be used for predicting the output guarantee rate, when the average relative error, the average absolute error and the prediction accuracy are not in the corresponding preset ranges, the accuracy of the prediction model is poor, and at the moment, the model function and/or the meteorological element combination mode in the first meteorological simulation result is adjusted to rebuild the model. Thus, the accuracy of the prediction model is further improved.
Based on the above technical solution, this embodiment further provides a method for predicting a wind farm output guarantee rate, as shown in fig. 4 and 5, including the method for creating a prediction model of the wind farm output guarantee rate, further including the following steps:
step 4, obtaining second meteorological observation data of a site area of the wind power plant to be developed in a second preset time period;
step 5, performing a four-dimensional data assimilation test on the second meteorological observation data by using a WRF-FDDA assimilation system to obtain a second meteorological simulation result based on a time sequence;
and 6, determining the output guarantee rate of the site area in a second preset time period according to the second meteorological simulation result and based on the prediction model.
It can be understood that when the output guarantee rate of the wind power plant to be developed needs to be predicted, first obtaining second meteorological observation data of a site area in a second preset time period needing to be predicted, then performing a four-dimensional data assimilation test on the second meteorological observation data by using a WRF-FDDA assimilation system, and simulating to obtain an assimilation meteorological result, namely a second meteorological simulation result (mainly comprising temperature, wind direction, wind speed, humidity and the like); and finally, inputting the second meteorological simulation result into the prediction model to obtain the corresponding output guarantee rate. The method for acquiring the second meteorological observation data and the method for data assimilation test of the second meteorological observation data are the same as those of the first meteorological observation data, and are not repeated herein.
Similarly, when the output guarantee rate is predicted, the embodiment may also evaluate the second meteorological simulation result, which mainly includes:
after obtaining a second meteorological simulation result, calculating a second correlation coefficient, a second average deviation, a second normalized average deviation and a second root mean square error between the second meteorological observation data and the second meteorological simulation result;
and judging whether the second correlation coefficient, the second average deviation, the second normalized average deviation and the second root mean square error are all in corresponding preset ranges, if so, entering a step 6, otherwise, entering a step 5 after adjusting assimilation test parameters.
The purpose of the above steps is to ensure the accuracy of the meteorological simulation result for predicting the output guarantee rate, when the second correlation coefficient, the second average deviation, the second normalized average deviation and the second root mean square error are all in the corresponding preset ranges, the simulation result is better, at this moment, the output guarantee rate is predicted, when the second correlation coefficient, the second average deviation, the second normalized average deviation and the second root mean square error are not in the corresponding preset ranges, the simulation result is worse, at this moment, the assimilation parameters are adjusted to perform the assimilation test again. Therefore, the accuracy of initial data is guaranteed, and the accuracy of output guarantee rate prediction is further improved.
Fig. 6 shows a schematic diagram of the output assurance rate predicted by the prior art, and fig. 7 shows a schematic diagram of the output assurance rate predicted by this embodiment, which can be seen that, by using the prediction model created by the method provided by this embodiment and the prediction method of the output assurance rate, the error between the predicted output assurance rate and the actual operating data under the conditions of high output (i.e., the output coefficient is greater than 0.8) and low output (the output coefficient is 0) is small, and the accuracy of the output assurance rate prediction is improved.
In summary, according to the prediction model creation method and the prediction method for the wind farm output assurance rate provided by this embodiment, multi-year FNL re-analysis meteorological data in a site area and hourly observation data of a ground automatic meteorological observation site are utilized, a WRF-FDDA assimilation system is utilized to develop a continuous four-dimensional data assimilation test, an assimilated meteorological simulation result (mainly including temperature, wind direction, wind speed, humidity and the like) is obtained through simulation, and the meteorological simulation result and the meteorological observation result are evaluated. And tries to establish a wind farm output assurance rate prediction model with dominant meteorological conditions (wind speed) by using different functions and meteorological element combination methods in the LSTM. The method includes the steps that hourly output of a built wind power plant for many years is collected to form a sample data set, 80% of time sequence samples in the front of the built wind power plant are selected for model training, 20% of time sequence samples are subjected to model verification to obtain a prediction model for predicting the output guarantee rate, then the output guarantee rate prediction is carried out on a site area of the wind power plant to be developed based on the prediction model, and the accuracy of the output guarantee rate prediction is improved.
Claims (10)
1. The method for creating the prediction model of the output guarantee rate of the wind power plant is characterized by comprising the following steps of:
step 1, acquiring an actual output guarantee rate of a produced wind power plant based on a time sequence in a first preset time period, and acquiring first meteorological observation data of a site area of the wind power plant in the first preset time period;
step 2, performing a four-dimensional data assimilation test on the first meteorological observation data by using a WRF-FDDA assimilation system to obtain a first meteorological simulation result based on a time sequence;
and 3, training a prediction model of the output guarantee rate according to the actual output guarantee rate and the first meteorological simulation result.
2. The method for creating a predictive model of wind farm output assurance rate according to claim 1, further comprising, after obtaining the first weather simulation result:
calculating a first correlation coefficient, a first average deviation, a first normalized average deviation and a first root mean square error between the first meteorological observation data and the first meteorological simulation result;
and judging whether the first correlation coefficient, the first average deviation, the first standardized average deviation and the first root mean square error are all in corresponding preset ranges, if so, entering a step 3, otherwise, entering a step 2 after adjusting assimilation test parameters.
3. The method for creating a prediction model for ensuring output of a wind farm according to claim 2, wherein the first correlation coefficient R is calculated as follows:
the calculation formula of the first average deviation MB is as follows:
the first normalized mean deviation NMB is calculated as follows:
the first root mean square error RSME is calculated as follows:
in the formula, m o The observation value corresponding to the first meteorological observation data,
is the mean value of the observed values, is>
m f For the analog value corresponding to the first weather simulation result, is selected>
Is a mean value of analog values>
N being observation data of first meteorological phenomenaThe number of samples.
4. The method for creating the prediction model of the wind farm output assurance rate according to claim 1, wherein training the prediction model of the output assurance rate according to the actual output assurance rate and the first weather simulation result specifically comprises:
step 31, establishing an LSTM model according to the model function and the meteorological element combination mode in the first meteorological simulation result;
step 32, dividing the actual output guarantee rate based on the time sequence and the first meteorological simulation result into a training set and a verification set, and training an LSTM model according to the actual output guarantee rate in the training set and the corresponding first meteorological simulation result;
step 33, determining the average relative error, the average absolute error and the prediction accuracy of the LSTM model according to the actual output guarantee rate of the verification set and the corresponding first meteorological simulation result;
and step 34, judging whether the average relative error, the average absolute error and the prediction accuracy are all in corresponding preset ranges, if so, taking the LSTM model as a prediction model of the output guarantee rate, otherwise, after adjusting a model function and/or a meteorological element combination mode in a first meteorological simulation result, entering step 32.
5. The method for creating a prediction model for wind farm output assurance rate according to claim 4, wherein the step 33 specifically comprises:
and inputting the first meteorological simulation result in the verification set into an LSTM model to obtain a predicted output guarantee rate, and calculating an average relative error, an average absolute error and a prediction accuracy according to the predicted output guarantee rate and an actual output guarantee rate.
6. The method for creating a predictive model of wind farm output assurance rate according to claim 5, characterized in that said average relative error is
The calculation formula of (a) is as follows:
the mean absolute error
The calculation formula of (a) is as follows:
the calculation formula of the prediction accuracy P is as follows:
in the formula,. DELTA. i ＝|X i -Y i |，X i Predicting the guarantee rate of output, Y, for the ith i For the ith actual contribution rate, Δ i Is the absolute error of the ith data, and M is the number of samples in the validation set.
7. The method for creating a predictive model of wind farm output assurance rate according to claim 4, characterized in that the meteorological elements in the first meteorological simulation result include temperature, wind direction, wind speed and humidity.
8. The method for creating a prediction model for guarantee rate of wind farm output according to claim 1, wherein the first meteorological observed data is obtained according to FNL re-analyzed meteorological data and meteorological observation site observed data in a site area of the wind farm.
9. Method for predicting a wind farm output assurance rate, characterized in that it comprises a method for creating a prediction model of a wind farm output assurance rate according to any one of claims 1 to 8, and also comprises the following steps:
step 4, obtaining second meteorological observation data of a site area of the wind power plant to be developed in a second preset time period;
step 5, performing a four-dimensional data assimilation test on the second meteorological observation data by using a WRF-FDDA assimilation system to obtain a second meteorological simulation result based on a time sequence;
and 6, determining the output guarantee rate of the site area in a second preset time period according to the second meteorological simulation result and on the basis of the prediction model.
10. The method for predicting wind farm output assurance rate of claim 9, further comprising, after obtaining the second meteorological simulation result:
calculating a second correlation coefficient, a second average deviation, a second normalized average deviation and a second root mean square error between the second meteorological observation data and a second meteorological simulation result;
and judging whether the second correlation coefficient, the second average deviation, the second normalized average deviation and the second root mean square error are all in corresponding preset ranges, if so, entering a step 6, otherwise, entering a step 5 after adjusting assimilation test parameters.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211389698.1A CN115912334A (en) | 2022-11-08 | 2022-11-08 | Method for establishing prediction model of output guarantee rate of wind power plant and prediction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211389698.1A CN115912334A (en) | 2022-11-08 | 2022-11-08 | Method for establishing prediction model of output guarantee rate of wind power plant and prediction method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115912334A true CN115912334A (en) | 2023-04-04 |
Family
ID=86473693
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211389698.1A Pending CN115912334A (en) | 2022-11-08 | 2022-11-08 | Method for establishing prediction model of output guarantee rate of wind power plant and prediction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115912334A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116565864A (en) * | 2023-07-11 | 2023-08-08 | 上海融和元储能源有限公司 | Photovoltaic power generation power forecasting method based on PCA-RBF algorithm |
-
2022
- 2022-11-08 CN CN202211389698.1A patent/CN115912334A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116565864A (en) * | 2023-07-11 | 2023-08-08 | 上海融和元储能源有限公司 | Photovoltaic power generation power forecasting method based on PCA-RBF algorithm |
| CN116565864B (en) * | 2023-07-11 | 2023-10-20 | 上海融和元储能源有限公司 | Photovoltaic power generation power forecasting method based on PCA-RBF algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Stathopoulos et al. | Wind power prediction based on numerical and statistical models | |
| CN111310889B (en) | Evaporation waveguide profile estimation method based on deep neural network | |
| CN108388956B (en) | Photovoltaic power prediction method considering radiation attenuation | |
| CN102945508B (en) | Model correction based wind power forecasting method | |
| CN110110912B (en) | Photovoltaic power multi-model interval prediction method | |
| Zucatelli et al. | Short-term wind speed forecasting in Uruguay using computational intelligence | |
| CN109726867B (en) | High-resolution electric power weather forecasting method based on multi-mode set | |
| CN112100921A (en) | Method for acquiring wind resource and wind speed based on WRF and random forest | |
| CN109784563B (en) | Ultra-short-term power prediction method based on virtual anemometer tower technology | |
| Dimitrov et al. | From SCADA to lifetime assessment and performance optimization: how to use models and machine learning to extract useful insights from limited data | |
| CN112700068B (en) | Reservoir dispatching rule optimization method based on machine learning fusion of multi-source remote sensing data | |
| CN112785035A (en) | Medium-short term hydrological forecasting method and system integrating multivariate information | |
| CN115912334A (en) | Method for establishing prediction model of output guarantee rate of wind power plant and prediction method | |
| CN111985712A (en) | Multi-step wind speed combined prediction model building method | |
| Cabezon et al. | Comparison of methods for power curve modelling | |
| CN117526274A (en) | New energy power prediction method, electronic equipment and storage medium in extreme climate | |
| CN112884601B (en) | Power system operation risk assessment method based on weather division strategy | |
| Corscadden et al. | The impact of variable wind shear coefficients on risk reduction of wind energy projects | |
| CN113095547B (en) | Short-term wind power prediction method based on GRA-LSTM-ICE model | |
| CN112070320B (en) | Ultra-short-term wind power prediction method and system based on dynamic harmonic regression | |
| Zhang et al. | A data-driven method to characterize turbulence-caused uncertainty in wind power generation | |
| CN115563848A (en) | Distributed photovoltaic total radiation prediction method and system based on deep learning | |
| TW201740296A (en) | Method and system for predicting power generation capacity of renewable energy using a neural network to accurately calculate the power generation capacity of renewable energy | |
| CN113052352A (en) | New energy station wind speed forecast correction method and system | |
| JP7114162B1 (en) | Wind condition estimation device and wind condition estimation method |
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 |