CN113554203A - Wind power prediction method and device based on high-dimensional gridding and LightGBM - Google Patents
Wind power prediction method and device based on high-dimensional gridding and LightGBM Download PDFInfo
- Publication number
- CN113554203A CN113554203A CN202010332939.3A CN202010332939A CN113554203A CN 113554203 A CN113554203 A CN 113554203A CN 202010332939 A CN202010332939 A CN 202010332939A CN 113554203 A CN113554203 A CN 113554203A
- Authority
- CN
- China
- Prior art keywords
- power
- wind speed
- wind
- data
- grid
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 27
- 238000009826 distribution Methods 0.000 claims abstract description 21
- 238000012549 training Methods 0.000 claims abstract description 21
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 230000002159 abnormal effect Effects 0.000 claims description 12
- 238000005457 optimization Methods 0.000 claims description 10
- 238000012216 screening Methods 0.000 claims description 7
- 230000003247 decreasing effect Effects 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 2
- 238000010248 power generation Methods 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000005684 electric field Effects 0.000 description 3
- 238000013179 statistical model Methods 0.000 description 3
- 238000003066 decision tree Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 239000013598 vector Substances 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Wind Motors (AREA)
- Control Of Eletrric Generators (AREA)
Abstract
The invention provides a wind power prediction method and a wind power prediction device based on high-dimensional gridding and LightGBM.A wind power unit or a wind power plant is extracted from meteorological element data of W grid points around a central coordinate position, and various meteorological elements are processed into time series meteorological data with the same time resolution as power data through linear interpolation to form multi-dimensional meteorological forecast data; preprocessing the actual power data based on a power curve of density distribution to obtain effective actual power data; and taking the effective actual power data and the multidimensional weather forecast data in the corresponding time period as training data, calling a LightGBM algorithm, and establishing a prediction model to obtain a power prediction result. According to the invention, high-dimensional input characteristics are constructed through meteorological information around a wind turbine generator or a wind power plant, and the correlation information of the meteorological information and the wind power is mined by utilizing the strong learning capacity of the lightGBM, so that the wind power can be more accurately predicted.
Description
Technical Field
The invention belongs to the field of wind power, and particularly relates to a wind power prediction method and device based on high-dimensional gridding and LightGBM.
Background
Wind power generation is a new energy power generation mode with great development potential, is clean and pollution-free, and has great commercial development prospect, but because the wind power generation has obvious intermittency and volatility, the power quality of a power grid is seriously influenced after grid connection, meanwhile, along with the gradual increase of the wind power grid connection capacity, the influence of the wind power generation on the power grid is increasingly serious, and in order to reduce the impact of wind power on the power grid and protect the stable operation of the power grid, a dispatching mechanism has to take a measure for limiting grid connection to a wind power plant when necessary, so that the problem of wind energy consumption is caused. The wind power is accurately predicted in advance, the peak load regulation pressure of a power system can be relieved, and the wind power grid-connection capacity is effectively improved. Therefore, wind power prediction plays an important role in the sustainable development of wind power generation.
A great deal of research is carried out on wind power prediction methods at home and abroad, and the mainstream methods are divided into three types of statistical model methods, physical model methods and machine learning methods which develop rapidly. The traditional statistical model method aims at utilizing a statistical method to establish a mapping relation by acquiring relevant information related to time and space in historical wind power data for prediction; commonly used statistical methods are time series methods, regression fitting methods, gray prediction methods, kalman filtering, and the like. However, the statistical model has higher prediction precision in a stationary time sequence, lower prediction precision for gust mutation conditions and lower long-term prediction accuracy. The essence of the physical model method is that the wind speed of the wind farm needs to be corrected and forecasted by depending on the accuracy of a numerical weather forecasting mode and taking meteorological variables such as wind speed, wind direction and air temperature in NWP (non-Newton-Perot-Watt) as input, and then the predicted value of the wind power is calculated according to the power curve of the wind farm. The core idea of the Machine learning method is to train and learn a large amount of historical actual data through a computer Artificial intelligence algorithm, and capture rules and logic relations implicit in input feature vectors and output data, so that a model can make accurate prediction, such as an Artificial Neural Network (ANN), a Support Vector Machine (SVM), and the like. However, the conventional neural network has weak processing capability for a long-time sequence and is difficult to have higher accuracy. Although the support vector regression algorithm can avoid falling into a local optimal solution, the problem that the convergence speed is slow and the implementation is difficult occurs when a large-scale training sample is processed.
Disclosure of Invention
The invention provides a wind power prediction method and device based on high-dimensional gridding and LightGBM, and provides more accurate wind power prediction aiming at the characteristics of stronger wind power generation volatility and randomness.
In order to achieve the purpose, the technical scheme of the invention is realized as follows:
a wind power prediction method based on high-dimensional gridding and LightGBM comprises the following steps:
s1, extracting meteorological element data of W grid points around the central coordinate position of the wind turbine generator or the wind power plant, and taking each grid point of each meteorological element as a characteristic quantity, wherein the meteorological elements comprise wind speed, wind direction, temperature, humidity and pressure intensity;
s2, linearly interpolating various meteorological elements to obtain time series meteorological data with the same time resolution as the power data to form multi-dimensional meteorological forecast data; wherein the wind direction is obtained by calculating after linear interpolation of the wind speed in the x direction and the wind speed in the y direction respectively;
s3, preprocessing the actual power data of the wind turbine generator or the wind power plant based on a power curve of density distribution to obtain effective actual power data; and taking the effective actual power data and the multidimensional weather forecast data in the corresponding time period as training data, calling a LightGBM algorithm, and establishing a prediction model to obtain a power prediction result.
Further, the specific step of obtaining the power prediction result in step S3 includes:
and taking the obtained effective actual power data and the multi-dimensional weather forecast data in the corresponding time period as training data, calling a python library LightGBM algorithm, automatically carrying out parameter optimization through grid search, and establishing a power prediction model to obtain a power prediction result.
Further, the specific step of obtaining the power prediction result in step S3 includes:
taking actual wind speed data corresponding to the obtained effective actual power data and multi-dimensional weather forecast data of a corresponding time period as training data, calling a python library LightGBM algorithm, automatically performing parameter optimization through grid search, and establishing a wind speed prediction model; and converting the wind speed prediction result of the wind speed prediction model into a power value through the power curve based on the density distribution to obtain a power prediction result.
Preferably, the method for acquiring the power curve based on the density distribution in step S3 includes:
(1) acquiring actual historical operating wind speed and power of a wind turbine generator or a wind power plant, performing primary screening, and removing constant values and abnormal values;
(2) the wind speed is divided at equal intervals by taking the wind speed as an abscissa and the power as an ordinate to obtain N wind speed sections, the power is divided at equal intervals to obtain M power sections, and finally N × M grids are obtained, and the grid is represented by a grid central point coordinate point, namely (V)i,Pj) Wherein, i is 1, 2.., N; j ═ 1, 2, 3,. ·, M;
(3) counting the number Num (i, j) of actual historical power falling on each grid point to obtain the two-dimensional power density number space distribution condition, and taking the median of the non-zero power density numbers of all grids to be recorded as Median;
(4) When the wind speed is less than the threshold value, for the grids corresponding to the maximum power space density in each wind speed section in the area, if the power density corresponding to the grids is the maximum value in the longitudinal direction or the transverse direction, recording the center coordinates of the grid;
(5) when the wind speed is greater than the threshold value, for each wind speed segment V in the areaiCorresponding power segment PiStarting from a high wind power section, namely j is M, M-1, searching a grid point meeting the following conditions, recording the center coordinate of the grid point, and entering the search of the next wind speed section; 1) num (i, j) is equal to the maximum density number of the power segment Is the maximum value of Num (i, j) when i is 1, 2.., N; 2) median M with Num (i, j) greater than density numberedian;
(6) And (5) obtaining a power curve of the wind turbine generator according to the coordinate points obtained in the steps (4) and (5).
Further, the step (6) of generating a power curve according to the coordinate points further includes:
(601) acquired set of coordinate points (V)k,Pk) Wherein k is 1, 2, …, q; at wind speed VkSorting from small to large, removing coordinate points with decreasing power to obtain a new set of coordinates (V)k,Pk) Wherein k is 1, 2, …, p;
(602) from 0 to 30m/s, a wind speed sequence is constructed at intervals of 0.1m/s, and (V)k,Pk) Linearly interpolating to a new wind speed sequence to obtain a power curve B;
(603) setting the power of the power curve B with the wind speed less than K m/s to be zero, wherein K is the cut-in wind speed of the fan;
(604) and taking the maximum power value and the maximum wind speed value in the group of data as rated output power and rated wind speed, wherein the corresponding power value of the wind speed larger than the rated wind speed is equal to the rated output power, and finally obtaining a power curve of the wind turbine generator.
In another aspect of the present invention, there is also provided a wind power prediction apparatus based on high dimensional grid and LightGBM, including:
the extraction unit is used for extracting meteorological element data of W grid points around the central coordinate position of the wind turbine generator or the wind power plant, and each grid point of each meteorological element is used as a characteristic quantity, wherein the meteorological elements comprise wind speed, wind direction, temperature, humidity and pressure intensity;
the interpolation unit is used for linearly interpolating various meteorological elements into time series meteorological data with the same time resolution as the power data to form multi-dimensional meteorological forecast data; wherein the wind direction is obtained by calculating after linear interpolation of the wind speed in the x direction and the wind speed in the y direction respectively;
the prediction unit is used for preprocessing the actual power data of the wind turbine generator or the wind power plant based on a power curve of density distribution to obtain effective actual power data; and taking the effective actual power data and the multidimensional weather forecast data in the corresponding time period as training data, calling a LightGBM algorithm, and establishing a prediction model to obtain a power prediction result.
Furthermore, the prediction unit is provided with a power prediction model module for taking the obtained effective actual power data and the multi-dimensional weather forecast data in the corresponding time period as training data, calling a python library LightGBM algorithm, automatically performing parameter optimization through grid search, and establishing a power prediction model to obtain a power prediction result.
Furthermore, the prediction unit is provided with a wind speed prediction model module for taking the actual wind speed data corresponding to the obtained effective actual power data and the multidimensional meteorological forecast data in the corresponding time period as training data, calling a python library LightGBM algorithm, automatically performing parameter optimization by grid search, and establishing a wind speed prediction model; and converting the wind speed prediction result of the wind speed prediction model into a power value through the power curve based on the density distribution to obtain a power prediction result.
Preferably, the prediction unit is provided with a power curve obtaining subunit, and the power curve obtaining subunit includes:
the preliminary screening module is used for carrying out preliminary screening processing after acquiring actual historical operating wind speed and power of the wind turbine generator or the wind power plant and removing constant values and abnormal values;
the grid division module is used for dividing the wind speed at equal intervals by taking the wind speed as an abscissa and the power as an ordinate to obtain N wind speed sections, dividing the power at equal intervals to obtain M power sections, finally obtaining N x M grids, and representing the grids by the coordinate point of the central point of the grids, namely (V)i,Pj) Wherein, i is 1, 2.., N; j ═ 1, 2, 3,. ·, M;
a density statistic module for counting that the actual historical power falls onThe number Num (i, j) of each grid point obtains the space distribution condition of the two-dimensional power density number, and the median of the non-zero power density numbers of all the grids is taken as Median;
A coordinate acquisition module to: when the wind speed is less than the threshold value, for the grids corresponding to the maximum power space density in each wind speed section in the area, if the power density corresponding to the grids is the maximum value in the longitudinal direction or the transverse direction, recording the center coordinates of the grid; when the wind speed is greater than the threshold value, for each wind speed segment V in the areaiCorresponding power segment PjStarting from a high wind power section, namely j is M, M-1, searching a grid point meeting the following conditions, recording the center coordinate of the grid point, and entering the search of the next wind speed section; 1) num (i, j) is equal to the maximum density number of the power segmentIs the maximum value of Num (i, j) when i is 1, 2.., N; 2) median M with Num (i, j) greater than density numberedian;
And the power curve module is used for obtaining a power curve of the wind turbine generator according to the coordinate points obtained by the coordinate obtaining module.
Still further, the power curve module further comprises:
a sorting submodule for sorting a set of coordinate points (V) acquired by the coordinate acquisition modulek,Pk) Wherein k is 1, 2, …, q; at wind speed VkSorting from small to large, removing coordinate points with decreasing power to obtain a new set of coordinates (V)k,Pk) Wherein k is 1, 2, …, p;
an interpolation submodule for constructing a wind speed sequence from 0 to 30m/s at intervals of 0.1m/s, and for converting (V)k,Pk) Linearly interpolating to a new wind speed sequence to obtain a power curve B;
the zero setting submodule is used for setting the power of the power curve B with the wind speed less than K m/s to be zero, and K is the cut-in wind speed of the fan;
and the curve generation submodule is used for taking the maximum power value and the maximum wind speed value in the group of data as rated output power and rated wind speed, and the corresponding power value of the wind speed larger than the rated wind speed is equal to the rated output power to finally obtain a power curve.
Compared with the prior art, the invention has the following beneficial effects:
(1) according to the invention, high-dimensional input characteristics are constructed through meteorological information around a wind turbine generator or a wind power plant, and the correlation information of the meteorological information and the wind power is mined by utilizing the strong learning capacity of the lightGBM, so that the wind power can be more accurately predicted. Practice shows that a single LightGBM model can achieve better results than a general algorithm;
(2) according to the method, the power curve based on density distribution is used for preprocessing actual power data when a prediction model is constructed, the influence of abnormal data is very small, the accuracy is extremely high, and the samples input by the power model in a training mode are guaranteed to be effective data of the wind power plant;
(3) the method can also carry out wind speed prediction, because high-dimensional meteorological information is used, excellent effects can be obtained, and meanwhile, the wind speed prediction result can also be applied to a power curve based on density distribution to obtain a power prediction result;
(4) the LightGBM algorithm can be replaced by a Catboost or xgboost algorithm, and the same effect can be basically achieved after the parameters are adjusted and optimized.
Drawings
FIG. 1 is a scatter plot of actual data collected for a power curve in an embodiment of the present invention;
FIG. 2 is a schematic diagram of a set of coordinate points of a power curve selected based on a density distribution according to an embodiment of the present invention;
FIG. 3 is a schematic power curve of an embodiment of the present invention;
FIG. 4 is a schematic diagram of the deviation range of the power curve according to the embodiment of the present invention;
FIG. 5 is a diagram illustrating actual power data after preprocessing according to an embodiment of the present invention;
FIG. 6 is a comparison graph of power prediction for an embodiment of the present invention.
Detailed Description
It should be noted that the embodiments and features of the embodiments may be combined with each other without conflict.
In order to make the objects and features of the present invention more comprehensible, embodiments accompanying the present invention are further described below. It is noted that the drawings are in greatly simplified form and employ non-precise ratios for the purpose of facilitating and distinctly aiding in the description of the patented embodiments of the invention.
The design idea of the invention is the preprocessing of the power curve of the data based on density distribution and the power prediction using LightGBM.
Because the electric field operation data are collected from the SCADA system, the electric field operation data are influenced by various factors in the production and collection processes, so that errors are brought, and if more data are introduced into the power model, the prediction accuracy of the wind power is obviously reduced. Therefore, obvious power abnormal points such as power limit and faults are eliminated under the condition that a certain error range needs to be considered before the historical data training model is used. Grid discretization is carried out on the two-dimensional space of the actually measured wind speed and the actually measured power of the wind power plant, the number of power values falling into each grid is counted, power space density number distribution is obtained, a fitting curve of the actually measured wind speed and the actually measured power of the wind power plant is obtained through counting, abnormal, extension and power limiting data are eliminated according to comparison of the actually measured power of the wind power plant and theoretical power converted by the fitting curve, and it is guaranteed that the sample set input by power model training is the effective data of the wind power plant.
According to the LightGBM model, the model effect cannot be influenced due to different dimensions of specific data in the characteristics according to the decision tree principle, so that the characteristics can be automatically pruned and selected without carrying out standardized processing on the characteristics, and compared with other single models, the LightGBM model has strong generalization capability and robustness and is very suitable for short-term wind power prediction.
The LightGBM is a fast, distributed and high-performance gradient lifting framework based on a decision tree algorithm, and for such a tree-based model, the most time-consuming part is to traverse all possible division points and calculate information gain when splitting a feature selection node, so as to find the optimal division point. When the LightGBM algorithm faces massive data and a data set with very high characteristic dimension, the LightGBM algorithm improves the training speed under the condition of not sacrificing precision by optimizing the sampling of sample points during model training and optimizing the characteristic dimension during selecting split points.
The method comprises the following specific steps:
1. performing primary screening on the acquired actual historical operating data, and removing a constant value and an abnormal value; if the data acquisition frequency is 15 minutes and the grid-connected generation capacity of the wind turbine generator or the wind power plant is C, the judgment threshold of the constant value is 24 (the threshold is obtained according to experience and can be set to be an integer larger than 3, but is not too large so as not to play a role in filtering abnormal data); the range of the normal value of the wind speed is [0, 30], and the range of the normal value of the power is [0, C ];
2. the wind speed is divided at equal intervals from 0 to 30m/s by taking the wind speed as an abscissa and the power as an ordinate to obtain N wind speed sections, and the wind speed sections from the low wind speed section to the high wind speed section are numbered 1, 2. Dividing the power from 0 to C at equal intervals to obtain M power sections, numbering from the low power section to the high power section by 1, 2, 3i,Pj) Wherein, i is 1, 2.., N; j ═ 1, 2, 3. Generally, the wind speed interval is 0.02 m/s, the power interval is C/1500, and the interval is a relatively stable value obtained by actual data test experience of more than one hundred wind farms nationwide;
3. counting the number Num (i, j) of actual historical power falling on each grid point to obtain the two-dimensional power density number space distribution condition, and taking the median of the non-zero power density number of all grids to be recorded as Median;
4. When the wind speed is less than VOWhen it is, generally VOThe value is 9-11 m/s (wind speed V)OThe value of (A) is generally close to the rated output power of a fan, the electricity limitation of the wind power plant basically occurs in a high wind speed section, and the wind speedExceeds VOWhen the grid is in a vertical or horizontal state, the power density corresponding to the grid is the maximum value, and the center coordinates of the grid are recorded;
5. when the wind speed is greater than VOFor each wind speed segment V in the regioniCorresponding power segment PjStarting from a high wind power section, namely j is M, M-1, until a grid point meeting the following conditions is searched, stopping the current search immediately once the meeting grid point is found, recording the center coordinate of the grid point, and entering the search of the next wind speed section;
the conditions are as follows: 1) num (i, j) is equal to the maximum density number of the power segmentIs the maximum value of Num (i, j) when i is 1, 2.., N; 2) median M with Num (i, j) greater than density numberedian;MedianThe value obtained in the step 3 is the median of the nonzero power density of all the grid points;
the above condition 1) plays a decisive role; condition 2) is defined for a special case, which can be regarded as special treatment, and it can be said that the condition 2 is satisfied in the case of most electric fields);
6. through two steps 4 and 5, a group of coordinate points (V) can be obtainedk,Pk) Where k is 1, 2, …, q, the following is done for the set of coordinate points: 1) at wind speed VkSorting from small to large, removing coordinate points with decreasing power to obtain a group of new coordinates (V)k,Pk) Wherein k is 1, 2, …, p; 2) from 0 to 30m/s, a wind speed sequence is constructed at intervals of 0.1m/s, and (V)k,Pk) Linearly interpolating to a new wind speed sequence to obtain a power curve B; 3) setting the power value of the power curve B with the wind speed less than K m/s as zero (K is the cut-in wind speed of the fan, the general value range is 2-3.5 m/s, and the value is related to the type of the fan); 4) counting the groupAccording to the maximum power value and the maximum wind speed value as rated output power and rated wind speed, and the corresponding power value of the wind speed larger than the rated wind speed is equal to the rated output power, finally obtaining a power curve of the wind turbine generator based on actual operation data;
7. a group of theoretical generating power P can be obtained by interpolation through the power curve according to the actual wind speedtheroyWhen the deviation between the actual power and the theoretical generated power is within a certain range, the actual power value is considered to be valid and can be used as a model input value, and if the deviation exceeds the range, the actual power value is considered to be abnormal. Typically, the deviation ranges from [ -k 1C, k 2C [ -C]Wherein k1 and k2 are in the range of (0, 0.3)];
8. Acquiring gridding meteorological forecast data, wherein the data can be global meteorological pattern forecast data or mesoscale meteorological forecast model data, extracting the meteorological data of W grid points within about 50km around the central coordinate position of a wind turbine generator or a wind power plant, taking each grid point of each meteorological element as a characteristic quantity, and taking general meteorological elements as wind speed, wind direction, temperature, humidity and pressure intensity, so that the characteristic quantities of the meteorological sources are all 5 x W.
9. Linearly interpolating various meteorological elements to obtain time series multidimensional meteorological forecast data with the same time resolution as the power data, wherein the wind direction is obtained by calculating after linear interpolation of the wind speed in the x direction and the wind speed in the y direction respectively;
10. by the step 7, the actual power after the abnormal data are removed and the multidimensional weather forecast data in the corresponding time period are used as training data, a python library lightGBM algorithm is called, grid search is carried out automatically to carry out parameter optimization, a power prediction model is established, and a power prediction result is obtained;
11. or by the step 7, taking the actual wind speed data corresponding to the actual power without the abnormal data and the multidimensional meteorological forecast data of the time period corresponding to the actual power as training data, calling a python library LightGBM algorithm, automatically performing parameter optimization by grid search, establishing a wind speed prediction model, and converting the wind speed prediction result of the wind speed prediction model into a power value through the power curve obtained in the step 6, so as to obtain a power prediction result.
In the above steps 10 and 11, the LightGBM algorithm may be replaced with the Catboost or xgboost algorithm.
According to the above steps, taking a certain wind power plant as an example, the installed capacity of the wind power plant is 49.5MW, the spatial resolution of meteorological data is 10km × 10km, the number of meteorological grid points in the range of 50km around the wind power plant is 10 × 10, five meteorological elements of wind speed, wind direction, temperature, humidity and pressure are selected, the characteristic dimension obtained by obtaining meteorological forecast data is 500, the group is set as a group B, 5 meteorological elements of a single station of the wind power plant are obtained as a test comparison group a, the single-station meteorological forecast data and high-dimensional meteorological forecast data are respectively input into a lightGBM model, a power prediction model is established, the output of the wind power plant from 11 days 5 and 11 days 2019 to 17 days 5 and 9 is predicted, and the results are shown in the following table and fig. 6.
Date | Accuracy-predicted power A | Precision-predicted power B |
2019-05-11 | 86.23% | 85.82% |
2019-05-12 | 78.11% | 81.49% |
2019-05-13 | 68.51% | 74.84% |
2019-05-14 | 89.61% | 88.82% |
2019-05-15 | 82.35% | 83.40% |
2019-05-16 | 81.06% | 82.89% |
2019-05-17 | 86.12% | 84.49% |
Mean value of | 81.71% | 83.11% |
The method for calculating the wind power prediction accuracy in the table comprises the following steps:
wherein ACC is the daily precision, and n is the number of measurement samples on the day; priIs the actual power; ppiTo a predicted power; cpIs the installed capacity.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (10)
1. A wind power prediction method based on high-dimensional gridding and LightGBM is characterized by comprising the following steps:
s1, extracting meteorological element data of W grid points around the central coordinate position of the wind turbine generator or the wind power plant, and taking each grid point of each meteorological element as a characteristic quantity, wherein the meteorological elements comprise wind speed, wind direction, temperature, humidity and pressure intensity;
s2, linearly interpolating various meteorological elements to obtain time series meteorological data with the same time resolution as the power data to form multi-dimensional meteorological forecast data; wherein the wind direction is obtained by calculating after linear interpolation of the wind speed in the x direction and the wind speed in the y direction respectively;
s3, preprocessing the actual power data of the wind turbine generator or the wind power plant based on a power curve of density distribution to obtain effective actual power data; and taking the effective actual power data and the multidimensional weather forecast data in the corresponding time period as training data, calling a LightGBM algorithm, and establishing a prediction model to obtain a power prediction result.
2. The wind power prediction method based on high-dimensional grid and LightGBM as claimed in claim 1, wherein the specific step of obtaining the power prediction result in step S3 comprises:
and taking the obtained effective actual power data and the multi-dimensional weather forecast data in the corresponding time period as training data, calling a python library LightGBM algorithm, automatically carrying out parameter optimization through grid search, and establishing a power prediction model to obtain a power prediction result.
3. The wind power prediction method based on high-dimensional grid and LightGBM as claimed in claim 1, wherein the specific step of obtaining the power prediction result in step S3 comprises:
taking actual wind speed data corresponding to the obtained effective actual power data and multi-dimensional weather forecast data of a corresponding time period as training data, calling a python library LightGBM algorithm, automatically performing parameter optimization through grid search, and establishing a wind speed prediction model; and converting the wind speed prediction result of the wind speed prediction model into a power value through the power curve based on the density distribution to obtain a power prediction result.
4. The wind power prediction method based on high-dimensional gridding and LightGBM as claimed in claim 1 or 3, wherein the method for obtaining the power curve based on density distribution in step S3 comprises:
(1) acquiring actual historical operating wind speed and power of a wind turbine generator or a wind power plant, performing primary screening, and removing constant values and abnormal values;
(2) the wind speed is divided at equal intervals by taking the wind speed as an abscissa and the power as an ordinate to obtain N wind speed sections, the power is divided at equal intervals to obtain M power sections, and finally N × M grids are obtained, and the grid is represented by a grid central point coordinate point, namely (V)i,Pj) Wherein, i is 1, 2.., N; j ═ 1, 2, 3,. ·, M;
(3) counting the number Num (i, j) of actual historical power falling on each grid point to obtain the two-dimensional power density number space distribution condition, and taking the median of the non-zero power density numbers of all grids to be recorded as Median;
(4) When the wind speed is less than the threshold value, for the grids corresponding to the maximum power space density in each wind speed section in the area, if the power density corresponding to the grids is the maximum value in the longitudinal direction or the transverse direction, recording the center coordinates of the grid;
(5) when the wind speed is greater than the threshold value, for each wind speed segment V in the areaiCorresponding power segment PiStarting from a high wind power section, namely j is M, M-1, searching a grid point meeting the following conditions, recording the center coordinate of the grid point, and entering the search of the next wind speed section; 1) num (i, j) is equal to the maximum density number of the power segment Is the maximum value of Num (i, j) when i is 1, 2.., N; 2) median M with Num (i, j) greater than density numberedian;
(6) And (5) obtaining a power curve of the wind turbine generator according to the coordinate points obtained in the steps (4) and (5).
5. The wind power prediction method based on high-dimensional gridding and LightGBM as claimed in claim 4, wherein the step of generating the power curve according to the coordinate points in step (6) further comprises:
(601) acquired set of coordinate points (V)k,Pk) Wherein k is 1, 2, …, q; at wind speed VkSorting from small to large, removing coordinate points with decreasing power to obtain a new set of coordinates (V)k,Pk) Wherein k is 1, 2, …, p;
(602) from 0 to 30m/s, a wind speed sequence is constructed at intervals of 0.1m/s, and (V)k,Pk) Linearly interpolating to a new wind speed sequence to obtain a power curve B;
(603) setting the power of the power curve B with the wind speed less than K m/s to be zero, wherein K is the cut-in wind speed of the fan;
(604) and taking the maximum power value and the maximum wind speed value in the group of data as rated output power and rated wind speed, wherein the corresponding power value of the wind speed larger than the rated wind speed is equal to the rated output power, and finally obtaining a power curve of the wind turbine generator.
6. A wind power prediction device based on high-dimensional gridding and LightGBM is characterized by comprising:
the extraction unit is used for extracting meteorological element data of W grid points around the central coordinate position of the wind turbine generator or the wind power plant, and each grid point of each meteorological element is used as a characteristic quantity, wherein the meteorological elements comprise wind speed, wind direction, temperature, humidity and pressure intensity;
the interpolation unit is used for linearly interpolating various meteorological elements into time series meteorological data with the same time resolution as the power data to form multi-dimensional meteorological forecast data; wherein the wind direction is obtained by calculating after linear interpolation of the wind speed in the x direction and the wind speed in the y direction respectively;
the prediction unit is used for preprocessing the actual power data of the wind turbine generator or the wind power plant based on a power curve of density distribution to obtain effective actual power data; and taking the effective actual power data and the multidimensional weather forecast data in the corresponding time period as training data, calling a LightGBM algorithm, and establishing a prediction model to obtain a power prediction result.
7. The wind power prediction device based on high-dimensional gridding and LightGBM as claimed in claim 6, wherein the prediction unit is provided with a power prediction model module for using the obtained effective actual power data and the multi-dimensional meteorological forecast data of the corresponding time period as training data, calling a python library LightGBM algorithm, performing grid search for automatic parameter optimization, and establishing a power prediction model to obtain a power prediction result.
8. The wind power prediction device based on high-dimensional gridding and LightGBM as claimed in claim 6, wherein the prediction unit is provided with a wind speed prediction model module for using the actual wind speed data corresponding to the obtained effective actual power data and the multidimensional meteorological forecast data in the corresponding time period as training data, calling a python library LightGBM algorithm, automatically performing parameter optimization by grid search, and establishing a wind speed prediction model; and converting the wind speed prediction result of the wind speed prediction model into a power value through the power curve based on the density distribution to obtain a power prediction result.
9. The wind power prediction device based on high dimensional grid and LightGBM as claimed in claim 6 or 8, wherein the prediction unit is provided with a power curve obtaining subunit, and the power curve obtaining subunit comprises:
the preliminary screening module is used for carrying out preliminary screening processing after acquiring actual historical operating wind speed and power of the wind turbine generator or the wind power plant and removing constant values and abnormal values;
the grid division module is used for dividing the wind speed at equal intervals by taking the wind speed as an abscissa and the power as an ordinate to obtain N wind speed sections, dividing the power at equal intervals to obtain M power sections, finally obtaining N x M grids, and representing the grids by the coordinate point of the central point of the grids, namely (V)i,Pj) Wherein, i is 1, 2.., N; j ═ 1, 2, 3,. ·, M;
a density statistic module for counting the number Num (i, j) of actual historical power falling on each grid point to obtain the two-dimensional power density number space distribution condition, and taking the median of the non-zero power density numbers of all grids as Median;
A coordinate acquisition module to: when the wind speed is less than the threshold value, for the grids corresponding to the maximum power space density in each wind speed section in the area, if the power density corresponding to the grids is the maximum value in the longitudinal direction or the transverse direction, recording the center coordinates of the grid; when the wind speed is greater than the threshold value, for each wind speed segment V in the areaiCorresponding power segment PjStarting from a high wind power section, namely j is M, M-1, searching a grid point meeting the following conditions, recording the center coordinate of the grid point, and entering the search of the next wind speed section; 1) num (i, j) is equal to the maximum density number of the power segment Is the maximum value of Num (i, j) when i is 1, 2.., N; 2) median M with Num (i, j) greater than density numberedian;
And the power curve module is used for obtaining a power curve of the wind turbine generator according to the coordinate points obtained by the coordinate obtaining module.
10. The wind power prediction device based on high dimensional grid and LightGBM of claim 9, wherein the power curve module further comprises:
a sorting submodule for sorting a set of coordinate points (V) acquired by the coordinate acquisition modulek,Pk) Wherein k is 1, 2, …, q; at wind speed VkSorting from small to large, removing coordinate points with decreasing power to obtain a new set of coordinates (V)k,Pk) Wherein k is 1, 2, …, p;
an interpolation submodule for constructing a wind speed sequence from 0 to 30m/s at intervals of 0.1m/s, and for converting (V)k,Pk) Linearly interpolating to a new wind speed sequence to obtain a power curve B;
the zero setting submodule is used for setting the power of the power curve B with the wind speed less than K m/s to be zero, and K is the cut-in wind speed of the fan;
and the curve generation submodule is used for taking the maximum power value and the maximum wind speed value in the group of data as rated output power and rated wind speed, and the corresponding power value of the wind speed larger than the rated wind speed is equal to the rated output power to finally obtain a power curve.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010332939.3A CN113554203B (en) | 2020-04-24 | 2020-04-24 | Wind power prediction method and device based on high-dimensional meshing and LightGBM |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010332939.3A CN113554203B (en) | 2020-04-24 | 2020-04-24 | Wind power prediction method and device based on high-dimensional meshing and LightGBM |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113554203A true CN113554203A (en) | 2021-10-26 |
CN113554203B CN113554203B (en) | 2023-12-26 |
Family
ID=78101274
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010332939.3A Active CN113554203B (en) | 2020-04-24 | 2020-04-24 | Wind power prediction method and device based on high-dimensional meshing and LightGBM |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113554203B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115189356A (en) * | 2022-09-13 | 2022-10-14 | 国能日新科技股份有限公司 | Wind and light power prediction correction method and device based on atmospheric circulation stable state |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150204922A1 (en) * | 2012-08-07 | 2015-07-23 | Korea Institute Of Energy Research | Method for Predicting Wind Power Density |
CN106355512A (en) * | 2016-08-26 | 2017-01-25 | 华北电力大学 | Power curve fitting method of wind turbine generator based on optimizing maximum value of probability density |
CN106548253A (en) * | 2016-11-08 | 2017-03-29 | 中国地质大学(武汉) | Method based on the wind power prediction of nonparametric probability |
CN107507097A (en) * | 2017-07-03 | 2017-12-22 | 上海电力学院 | A kind of short-term wind power prediction method |
CN109242143A (en) * | 2018-07-31 | 2019-01-18 | 中国电力科学研究院有限公司 | A kind of neural network wind power forecasting method and system |
CN110288136A (en) * | 2019-06-11 | 2019-09-27 | 上海电力学院 | Wind power multi-step Predictive Model method for building up |
CN110705772A (en) * | 2019-09-26 | 2020-01-17 | 国家电网公司华北分部 | Regional power grid wind power generation power prediction optimization method and device |
-
2020
- 2020-04-24 CN CN202010332939.3A patent/CN113554203B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150204922A1 (en) * | 2012-08-07 | 2015-07-23 | Korea Institute Of Energy Research | Method for Predicting Wind Power Density |
CN106355512A (en) * | 2016-08-26 | 2017-01-25 | 华北电力大学 | Power curve fitting method of wind turbine generator based on optimizing maximum value of probability density |
CN106548253A (en) * | 2016-11-08 | 2017-03-29 | 中国地质大学(武汉) | Method based on the wind power prediction of nonparametric probability |
CN107507097A (en) * | 2017-07-03 | 2017-12-22 | 上海电力学院 | A kind of short-term wind power prediction method |
CN109242143A (en) * | 2018-07-31 | 2019-01-18 | 中国电力科学研究院有限公司 | A kind of neural network wind power forecasting method and system |
CN110288136A (en) * | 2019-06-11 | 2019-09-27 | 上海电力学院 | Wind power multi-step Predictive Model method for building up |
CN110705772A (en) * | 2019-09-26 | 2020-01-17 | 国家电网公司华北分部 | Regional power grid wind power generation power prediction optimization method and device |
Non-Patent Citations (5)
Title |
---|
JU Y等: "A model combining convolutional neural network and LightGBM algorithm for ultra-short-term wind power forecasting", IEEE ACCESS, vol. 7, pages 28309 - 28318, XP011715192, DOI: 10.1109/ACCESS.2019.2901920 * |
WANYING ZHANG等: "A multi-step probability density prediction model based on gaussian approximation of quantiles for offshore wind power", RENEWABLE ENERGY, vol. 202, pages 992 - 1011, XP087233178, DOI: 10.1016/j.renene.2022.11.111 * |
WU YX等: "Data-driven wind speed forecasting using deep feature extraction and LSTM", IET RENEWABLE POWER GENERATION, vol. 13, no. 12, pages 2062 - 2069, XP006086855, DOI: 10.1049/iet-rpg.2018.5917 * |
程启明;陈路;程尹曼;张强;高杰;: "基于EEMD和LS-SVM模型的风电功率短期预测方法", 电力自动化设备, no. 05 * |
齐霞;安磊;张妍;李芬花;张浩;阎洁;韩爽;李莉;: "风电机组功率曲线建模方法对比研究", 可再生能源, no. 04 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115189356A (en) * | 2022-09-13 | 2022-10-14 | 国能日新科技股份有限公司 | Wind and light power prediction correction method and device based on atmospheric circulation stable state |
CN115189356B (en) * | 2022-09-13 | 2022-12-02 | 国能日新科技股份有限公司 | Wind and light power prediction correction method and device based on atmospheric circulation stable state |
Also Published As
Publication number | Publication date |
---|---|
CN113554203B (en) | 2023-12-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106875033B (en) | Wind power cluster power prediction method based on dynamic self-adaption | |
Yang et al. | Day-ahead forecasting of photovoltaic output power with similar cloud space fusion based on incomplete historical data mining | |
CN102236795B (en) | Method for forecasting wind speed in wind power station | |
CN109063276B (en) | Wind power plant dynamic equivalent modeling method suitable for long-time domain wind speed fluctuation | |
CN110766200A (en) | Method for predicting generating power of wind turbine generator based on K-means mean clustering | |
CN111475909B (en) | Wind turbine generator output correlation mapping modeling method based on long-term and short-term memory network | |
CN110601250B (en) | Method for evaluating stability of static voltage of access point based on new energy power station | |
CN105512766A (en) | Wind power plant power predication method | |
CN112580874A (en) | Short-term wind power prediction method based on random forest algorithm and TCN | |
CN113344288B (en) | Cascade hydropower station group water level prediction method and device and computer readable storage medium | |
CN112186761B (en) | Wind power scene generation method and system based on probability distribution | |
CN110674864A (en) | Wind power abnormal data identification method with synchronous phasor measurement device | |
CN112651576A (en) | Long-term wind power prediction method and device | |
CN110334847A (en) | Based on the wind power prediction method for improving K-means cluster and support vector machines | |
CN114819374A (en) | Regional new energy ultra-short term power prediction method and system | |
CN115358606A (en) | Method and system for evaluating energy efficiency of in-service wind power plant under gentle terrain | |
CN113505909B (en) | Error compensation method for short-term wind power trend prediction | |
CN113554203B (en) | Wind power prediction method and device based on high-dimensional meshing and LightGBM | |
CN110852492A (en) | Photovoltaic power ultra-short-term prediction method for finding similarity based on Mahalanobis distance | |
CN117200241A (en) | New forms of energy intelligent voltage regulation system that is incorporated into power networks | |
CN111460360B (en) | Power curve fitting data preprocessing method and device based on density distribution | |
CN106251238B (en) | Wind power plant modeling sequence discretization step length selection and model error analysis method | |
CN113887839A (en) | Clustering-based wind power plant ultra-short-term wind power prediction method and system | |
CN114239920A (en) | Short-term wind power prediction method | |
CN111539577A (en) | Short-term wind power prediction method based on wind speed change rate and Gaussian process regression |
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 |