A clustering-based ultra-short-term wind power prediction method and system for wind farms

Abstract

The present invention discloses a clustering-based ultra-short-term wind power prediction method and system for a wind farm, wherein the method comprises: obtaining basic data and real-time measurement data of the wind farm; clustering all wind turbines in the wind farm using a Gaussian mixture model according to the basic data to obtain multiple wind turbine clusters; equating each wind turbine cluster to an equivalent wind turbine, establishing multiple equivalent wind turbine prediction models; establishing a combined prediction model of the wind farm according to the multiple equivalent wind turbine prediction models; and obtaining the ultra-short-term wind power real-time prediction value of the wind farm according to the real-time measurement data and the combined prediction model. The present invention clusters all wind turbines in the wind farm and establishes prediction models respectively, thereby reducing the number of ultra-short-term wind power value prediction models to be constructed and improving the accuracy of ultra-short-term wind power prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of wind farm wind power generation, and in particular to a clustering-based ultra-short-term wind power prediction method and system for a wind farm. Background Art

[0002] As the most economical and mature renewable energy source, wind power generation has been highly valued by countries around the world. In recent years, the centralized development and utilization of large-scale wind farms in my country has become the mainstream direction of wind power generation. However, the grid-connected operation of large-scale wind power has brought certain difficulties to the dispatching of power systems, which has led to the phenomenon of wind power abandonment. The main reason for wind power abandonment is that the wind power prediction accuracy of wind farms is not enough, which makes traditional thermal power units unable to meet the uncertainty brought by wind power generation.

[0003] At present, the wind power prediction methods of wind farms mainly include numerical weather forecasting (NWP), physical forecasting and statistical forecasting methods. Among them, the numerical weather forecasting method is based on the actual situation of the atmosphere, and uses numerical calculation methods to describe the evolution of the weather, so as to predict the future weather and then predict the output of wind power; the physical forecasting method uses meteorological parameters, geographical information of wind farms and other data to calculate the wind direction and wind speed at the hub position of the wind turbine, and then calculates the wind power output power in combination with the change of the wind farm output power curve with wind speed, and then obtains the predicted value of power; the statistical forecasting method is to construct a mapping relationship between the wind farm measurement data, numerical meteorological forecasting model and other input values ​​and wind power, so as to predict wind power; with the continuous expansion of the scale of wind farms, the prediction accuracy of the above three prediction models is not high, and it is difficult to meet the needs of the actual grid-connected operation of wind farms. Therefore, high-precision wind power prediction technology for wind farms is urgently needed. Summary of the invention

[0004] The purpose of the present invention is to provide a clustering-based ultra-short-term wind power prediction method and system for a wind farm, which clusters all wind turbines in the wind farm and establishes prediction models for each of them, thereby reducing the number of ultra-short-term wind power value prediction models to be constructed and improving the accuracy of ultra-short-term wind power prediction.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A clustering-based ultra-short-term wind power prediction method for a wind farm, comprising:

[0007] Obtaining basic data and real-time measurement data of the wind farm; the basic data includes wind turbine parameter data, wind turbine wind power history data, climate background data, and topographic data; the real-time measurement data includes weather forecast data, weather real-time data, and wind tower real-time data;

[0008] According to the basic data, clustering is performed on all wind turbines in the wind farm using a Gaussian mixture model to obtain a plurality of wind turbine clusters;

[0009] Each wind turbine cluster is equal to an equivalent wind turbine, and multiple equivalent wind turbine prediction models are established;

[0010] Establishing a combined prediction model of the wind farm based on a plurality of equivalent wind turbine prediction models;

[0011] An ultra-short-term wind power real-time forecast value of the wind farm is obtained according to the real-time measurement data and the combined forecasting model.

[0012] Optionally, establishing a combined prediction model of a wind farm based on a plurality of equivalent wind turbine prediction models specifically includes:

[0013] Obtain historical measurement data of wind farms;

[0014] Determining the weight of each equivalent wind turbine prediction model based on the historical measurement data;

[0015] The combined prediction model is established according to the weight and a plurality of equivalent wind turbine prediction models.

[0016] Optionally, the equivalent wind turbine prediction model is one of a BP prediction model, a LSSVM prediction model, a GA-LSSVM prediction model, an ARIMA prediction model or a KALMAN filtering prediction model.

[0017] Optionally, determining the weight of each equivalent wind turbine prediction model according to the historical measurement data specifically includes:

[0018] Inputting the historical measurement data into each of the equivalent wind turbine prediction models respectively to obtain historical wind power prediction values ​​of multiple equivalent wind turbines;

[0019] Calculate the variance between the historical wind power prediction value and the historical wind power value of each equivalent wind turbine group respectively;

[0020] According to the multiple variances, using the formula and formula Determine the weight of each equivalent wind turbine prediction model;

[0021] Where σ is the sum of the variances of the historical wind power prediction values ​​and the historical wind power values ​​of multiple equivalent wind turbines, n is the total number of equivalent wind turbine prediction models, k is i is the weight coefficient of the prediction model of the i-th equivalent wind turbine group, i=1,...n; ξ iis the variance between the historical wind power prediction value and the historical wind power value of the i-th equivalent wind turbine; e i is the prediction error of the prediction model of the i-th equivalent wind turbine; ξ j is the variance between the historical wind power prediction value and the historical wind power value of the j-th equivalent wind turbine, i = 1, ... n and i ≠ j; e j is the prediction error of the prediction model of the j-th equivalent wind turbine; λ is the Lagrangian constant.

[0022] Optionally, the combined prediction model is:

[0023] Among them, P is the real-time forecast value of the ultra-short-term wind power of the wind farm, P i is the real-time predicted value of wind power of the ith equivalent wind turbine prediction model; the real-time predicted value of wind power is obtained by inputting the real-time measurement data into the equivalent wind turbine prediction model.

[0024] A clustering-based ultra-short-term wind power prediction system for a wind farm, comprising:

[0025] The data acquisition module is used to acquire basic data and real-time measurement data of the wind farm; the basic data includes wind turbine parameter data, wind turbine wind power history data, climate background data, and topographic data; the real-time measurement data includes weather forecast data, weather real-time data, and wind tower real-time data;

[0026] A wind turbine cluster determination module, configured to perform clustering processing on all wind turbines in the wind farm using a Gaussian mixture model according to the basic data to obtain a plurality of wind turbine clusters;

[0027] An equivalent wind turbine prediction model establishment module is used to treat each wind turbine cluster as an equivalent wind turbine and establish multiple equivalent wind turbine prediction models;

[0028] A combined prediction model establishment module, used to establish a combined prediction model of a wind farm based on a plurality of equivalent wind turbine prediction models;

[0029] The ultra-short-term wind power real-time prediction value determination module is used to obtain the ultra-short-term wind power real-time prediction value of the wind farm according to the real-time measurement data and the combined prediction model.

[0030] Optionally, the combined prediction model establishment module specifically includes:

[0031] A historical measurement data acquisition unit, used for acquiring historical measurement data of the wind farm;

[0032] A weight determination unit, used to determine the weight of each equivalent wind turbine prediction model according to the historical measurement data;

[0033] The combined prediction model establishing unit is used to establish the combined prediction model according to the weight and a plurality of equivalent wind turbine prediction models.

[0034] Optionally, the equivalent wind turbine prediction model is one of a BP prediction model, a LSSVM prediction model, a GA-LSSVM prediction model, an ARIMA prediction model or a KALMAN filtering prediction model.

[0035] Optionally, the weight determination unit specifically includes:

[0036] A wind power historical prediction value determination subunit, used for inputting the historical measurement data into each of the equivalent wind turbine prediction models to obtain wind power historical prediction values ​​of multiple equivalent wind turbines;

[0037] A calculation subunit, used for respectively calculating the variance between the historical wind power prediction value and the historical wind power value of each equivalent wind turbine group;

[0038] The weight determination subunit is used to determine the weight according to the plurality of variances using the formula and formula Determine the weight of each equivalent wind turbine prediction model;

[0039] Where σ is the sum of the variances of the historical wind power prediction values ​​and the historical wind power values ​​of multiple equivalent wind turbines, n is the total number of equivalent wind turbine prediction models, k is i is the weight coefficient of the prediction model of the i-th equivalent wind turbine group, i=1,...n; ξ i is the variance between the historical wind power prediction value and the historical wind power value of the i-th equivalent wind turbine; e i is the prediction error of the prediction model of the i-th equivalent wind turbine; ξ j is the variance between the historical wind power prediction value and the historical wind power value of the j-th equivalent wind turbine, i = 1, ... n and i ≠ j; e j is the prediction error of the prediction model of the j-th equivalent wind turbine; λ is the Lagrangian constant.

[0040] Optionally, the combined prediction model is:

[0041] Among them, P is the real-time forecast value of the ultra-short-term wind power of the wind farm, P i is the real-time predicted value of wind power of the ith equivalent wind turbine prediction model; the real-time predicted value of wind power is obtained by inputting the real-time measurement data into the equivalent wind turbine prediction model.

[0042] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0043] The present invention discloses a clustering-based ultra-short-term wind power prediction method and system for a wind farm, wherein the method comprises: obtaining basic data and real-time measurement data of the wind farm; clustering all wind turbines in the wind farm using a Gaussian mixture model according to the basic data to obtain multiple wind turbine clusters; equating each wind turbine cluster to an equivalent wind turbine, establishing multiple equivalent wind turbine prediction models; establishing a combined prediction model of the wind farm according to the multiple equivalent wind turbine prediction models; and obtaining the ultra-short-term wind power real-time prediction value of the wind farm according to the real-time measurement data and the combined prediction model. The present invention clusters all wind turbines in the wind farm and establishes prediction models respectively, thereby reducing the number of ultra-short-term wind power value prediction models to be constructed and improving the accuracy of ultra-short-term wind power prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0045] Figure 1 This is a flow chart of a method for predicting ultra-short-term wind power in a wind farm based on clustering in an embodiment of the present invention;

[0046] Figure 2 This is a schematic diagram of the structure of a wind speed and direction measuring device for a wind farm according to an embodiment of the present invention;

[0047] Figure 3 This is a schematic diagram of the structure of a wind farm field data acquisition unit in an embodiment of the present invention;

[0048] Figure 4 is the ultra-short-term wind power prediction result in the embodiment of the present invention;

[0049] Figure 5 The structure of the ultra-short-term wind power prediction system for a wind farm based on clustering in an embodiment of the present invention is schematically shown. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] The purpose of the present invention is to provide a clustering-based ultra-short-term wind power prediction method and system for a wind farm, which clusters all wind turbines in the wind farm and establishes prediction models for each of them, thereby reducing the number of ultra-short-term wind power value prediction models to be constructed and improving the accuracy of ultra-short-term wind power prediction.

[0052] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0053] Figure 1 FIG. 1 is a flow chart of a method for predicting ultra-short-term wind power in a wind farm based on clustering in an embodiment of the present invention. Figure 1 The present invention provides a clustering-based ultra-short-term wind power prediction method for a wind farm, comprising:

[0054] Step 101: Obtain basic data and real-time measurement data of the wind farm; the basic data includes wind turbine parameter data, wind turbine wind power history data, climate background data, and topographic data; the real-time measurement data includes weather forecast data, weather real-time data, and wind tower real-time data;

[0055] Step 102: clustering all wind turbines in the wind farm using a Gaussian mixture model based on the basic data to obtain multiple wind turbine clusters;

[0056] Step 103: each wind turbine cluster is equalized to an equalized wind turbine, and multiple equalized wind turbine prediction models are established;

[0057] Step 104: establishing a combined prediction model of the wind farm according to a plurality of equivalent wind turbine prediction models;

[0058] Step 105: Obtain the ultra-short-term wind power real-time forecast value of the wind farm according to the real-time measurement data and the combined forecast model.

[0059] Step 104 specifically includes:

[0060] Obtain historical measurement data of wind farms;

[0061] Determine the weight of each equivalent wind turbine prediction model based on historical measurement data;

[0062] A combined prediction model is established based on weights and multiple equivalent wind turbine prediction models.

[0063] The equivalent wind turbine prediction model is one of a BP prediction model, a LSSVM prediction model, a GA-LSSVM prediction model, an ARIMA prediction model or a KALMAN filter prediction model.

[0064] Specifically, based on historical measurement data, the weight of each equivalent wind turbine prediction model is determined, including:

[0065] Inputting the historical measurement data into the prediction model of each equivalent wind turbine group respectively, and obtaining the historical wind power prediction values ​​of multiple equivalent wind turbine groups;

[0066] Calculate the variance between the historical wind power prediction value and the historical wind power value of each equivalent wind turbine group respectively;

[0067] Based on multiple variances, using the formula and formula Determine the weight of each equivalent wind turbine prediction model;

[0068] Where σ is the sum of the variances of the historical wind power prediction values ​​and the historical wind power values ​​of multiple equivalent wind turbines, n is the total number of equivalent wind turbine prediction models, k is i is the weight coefficient of the prediction model of the i-th equivalent wind turbine group, i=1,...n; ξ i is the variance between the historical wind power prediction value and the historical wind power value of the i-th equivalent wind turbine; e i is the prediction error of the prediction model of the i-th equivalent wind turbine; ξ j is the variance between the historical wind power prediction value and the historical wind power value of the j-th equivalent wind turbine, i = 1, ... n and i ≠ j; e j is the prediction error of the prediction model of the j-th equivalent wind turbine; λ is the Lagrangian constant.

[0069] The combined prediction model is:

[0070] Among them, P is the real-time forecast value of the ultra-short-term wind power of the wind farm, P i is the real-time predicted value of wind power of the ith equivalent wind turbine prediction model; the real-time predicted value of wind power is obtained by inputting real-time measurement data into the equivalent wind turbine prediction model.

[0071] Specifically, according to the measured data samples of wind power of all wind turbines in the wind farm, a method based on Gaussian mixture model is used to cluster all wind turbines in the wind farm, and a large number of wind turbines are grouped into several clusters, which effectively avoids the problem of difficulty in selecting the number of clustering in K-means method. According to the characteristics of equivalent units, the wind turbine groups in each cluster are equivalent to one wind turbine. By equating a large number of units in the wind farm to several wind turbines, the wind power prediction of the wind farm is simplified to only predicting equivalent units, which greatly improves the calculation efficiency. A combined prediction model is used to predict equivalent wind turbines. The combined prediction model is to construct multiple prediction models according to the method of linear superposition. The mathematical model of the combined prediction model is:

[0072]

[0073] Where n is the total number of combined prediction models, k1, k2, ..., k i , …, k n is the weight coefficient corresponding to the prediction model, P1, P2, ..., P i ,…,P n is the output predicted power value corresponding to the prediction model.

[0074] Among them, the weight coefficient in the wind power prediction process is calculated by minimizing the sum of variances, specifically:

[0075] Among them, e1, e2, …, e n is the prediction result error of each prediction model, ξ1, ξ2, …ξ n is the variance. The Lagrange method is used to find the minimum value of the sum of the predicted variances, and the weight coefficient k in the combined model can be further found. i . The Lagrangian function is:

[0076] In order to improve the accuracy of clustering of wind turbines in wind farms, the present invention has a preferred technical solution that the historical data used in clustering include wind speed, air pressure, humidity, temperature and power of wind turbines, wherein the data sampling period is 15 minutes. According to the data characteristics provided by the wind farm and the meteorological department, a prediction scheme combining multiple prediction models is proposed, and the weights of each model are corrected in real time using the measured data until the prediction accuracy reaches the best, and the prediction results for the next 4 hours are output. In order to reduce costs and improve the reliability and real-time performance of data transmission in the wind farm, the wireless communication network RoLa is used to send data to the wind farm server. Since the meteorological department is far away from the wind farm, the wireless network 5G is used to send meteorological data to the wind farm server. In order to improve the accuracy of ultra-short-term wind power prediction of wind farms, historical meteorological data and wind farm data of the past 5 years are used. At the same time, the wind power prediction system stores a combination of 5 ultra-short-term wind power prediction models. According to the real-time analysis of data characteristics, the best combination prediction scheme is selected, that is, the weight coefficient of the combination model is dynamically adjusted in real time.

[0077] Figure 2 Schematic diagram of the structure of a wind speed and direction measuring device for a wind farm according to an embodiment of the present invention; Figure 2As shown, an ultra-short-term wind power prediction method and process provided by the present invention are first read in the wind farm historical data and measured data, specifically including wind turbine parameter data, wind turbine wind power historical data, climate background data, topography data, meteorological forecast data, meteorological real-time data, wind tower real-time data and wind turbine wind power real-time data. According to the read data, all wind turbines in the wind farm are clustered and grouped by a method based on a Gaussian mixture model, and all wind turbines in the wind farm are grouped into n clusters. Then, wind power is predicted according to the equivalent groups after grouping. In order to improve the speed of wind power prediction, each cluster of groups is equivalent to one group for prediction, that is, n equivalent groups in the wind farm are predicted. Finally, the weight coefficient of each equivalent group in the combined prediction model is dynamically adjusted according to the measured data, and the calculation of the wind power prediction value is completed. The wind power prediction of the entire wind farm is realized by superimposing the wind power prediction results of n equivalent groups. The weight coefficient of each equivalent unit is calculated by minimizing the sum of the variances of the prediction errors. The weight coefficient of the combined prediction model is continuously corrected using measured data and historical data, and the prediction accuracy is improved through dynamic real-time optimization and correction.

[0078] Figure 3 Schematic diagram of the structure of a wind farm field data acquisition unit in an embodiment of the present invention; Figure 3 The present invention includes a data module, a data processing module and a combined prediction model module, wherein the wind turbine parameter data, wind turbine wind power historical data, climate background data, and topographic data in the data module are directly stored in the prediction system, while the meteorological forecast data, meteorological real-time data, wind tower real-time data, and wind turbine wind power real-time data are transmitted to the wind power prediction system through a wireless communication network. The data processing module cleans the historical data and measured data, removes bad data, improves the fidelity of the data, and clusters the wind turbines in the wind farm in combination with the processed data. Equivalence processing is performed on the clustered wind turbine clusters, and each cluster of wind turbines is equivalent to a unit. Then, wind power prediction is performed for each equivalent unit, and the weight coefficient of the combined prediction model is corrected in real time in combination with the measured data characteristics, and the wind power prediction value of the equivalent unit is output in real time. The wind power prediction values ​​of all equivalent units are superimposed to obtain the ultra-short-term wind power prediction result of the entire wind farm and output it.

[0079] The present invention uses the wind power sequence collected in March of each year in the past five years as experimental sample data of the wind farm, with a data sampling period of 15 minutes, and completes the prediction through data preprocessing, clustering, and weight coefficient allocation. Among them, 16 output wind power sampling points in the last four hours of March of the fifth year of each equivalent wind turbine are selected as test data, and the remaining data are experimental data for wind power prediction. The predicted value is compared with the test data to verify the effectiveness of the wind power prediction of the present invention. The prediction results are as follows Figure 4As shown, the present invention not only provides the ultra-short-term wind power forecast value in the form of a curve, but also can provide it in the form of a report or a bar graph.

[0080] The advantages and beneficial effects of the present invention are:

[0081] 1. The present invention uses a large amount of data, including forecast data, measured data, historical data and inherent attribute data, such as topographic data and wind turbine parameter data. By analyzing a large amount of data, the present invention provides the best model and method for wind power prediction.

[0082] 2. The present invention adopts a variety of combined prediction models, and dynamically allocates weight coefficients of the combined prediction models according to different weather conditions, thereby effectively improving the accuracy of wind power prediction.

[0083] 3. The present invention adopts a wireless communication network to transmit data, thereby reducing the cost of the wind power prediction system of the wind farm.

[0084] Figure 5 FIG. 1 is a schematic diagram of the structure of an ultra-short-term wind power prediction system for a wind farm based on clustering in an embodiment of the present invention. Figure 5 The present invention also provides a clustering-based ultra-short-term wind power prediction system for a wind farm, comprising:

[0085] The data acquisition module 501 is used to acquire basic data and real-time measurement data of the wind farm; the basic data includes wind turbine parameter data, wind turbine wind power history data, climate background data, and topographic data; the real-time measurement data includes weather forecast data, weather real-time data, and wind tower real-time data;

[0086] The wind turbine cluster determination module 502 is used to perform clustering processing on all wind turbines in the wind farm using a Gaussian mixture model according to basic data to obtain multiple wind turbine clusters;

[0087] An equivalent wind turbine prediction model establishment module 503 is used to treat each wind turbine cluster as an equivalent wind turbine and establish multiple equivalent wind turbine prediction models;

[0088] A combined prediction model building module 504 is used to build a combined prediction model of a wind farm based on a plurality of equivalent wind turbine prediction models;

[0089] The ultra-short-term wind power real-time prediction value determination module 505 is used to obtain the ultra-short-term wind power real-time prediction value of the wind farm according to the real-time measurement data and the combined prediction model.

[0090] Combined prediction model building module, including:

[0091] A historical measurement data acquisition unit, used for acquiring historical measurement data of the wind farm;

[0092] A weight determination unit, used to determine the weight of each equivalent wind turbine prediction model based on historical measurement data;

[0093] The combined prediction model building unit is used to build a combined prediction model according to the weights and multiple equivalent wind turbine prediction models.

[0094] The equivalent wind turbine prediction model is one of a BP prediction model, a LSSVM prediction model, a GA-LSSVM prediction model, an ARIMA prediction model or a KALMAN filter prediction model.

[0095] The weight determination unit specifically includes:

[0096] A wind power historical prediction value determination subunit is used to input the historical measurement data into each equivalent wind turbine prediction model to obtain wind power historical prediction values ​​of multiple equivalent wind turbines;

[0097] A calculation subunit, used for respectively calculating the variance between the historical wind power prediction value and the historical wind power value of each equivalent wind turbine generator set;

[0098] The weight determination subunit is used to determine the weight according to multiple variances using the formula and formula Determine the weight of each equivalent wind turbine prediction model;

[0099] Where σ is the sum of the variances of the historical wind power prediction values ​​and the historical wind power values ​​of multiple equivalent wind turbines, n is the total number of equivalent wind turbine prediction models, k is i is the weight coefficient of the prediction model of the i-th equivalent wind turbine group, i=1,...n; ξ i is the variance between the historical wind power prediction value and the historical wind power value of the i-th equivalent wind turbine; e i is the prediction error of the prediction model of the i-th equivalent wind turbine; ξ j is the variance between the historical wind power prediction value and the historical wind power value of the j-th equivalent wind turbine, i = 1, ... n and i ≠ j; e j is the prediction error of the prediction model of the j-th equivalent wind turbine; λ is the Lagrangian constant.

[0100] The combined prediction model is:

[0101] Among them, P is the real-time forecast value of the ultra-short-term wind power of the wind farm, P i is the real-time predicted value of wind power of the ith equivalent wind turbine prediction model; the real-time predicted value of wind power is obtained by inputting real-time measurement data into the equivalent wind turbine prediction model.

[0102] The clustering-based ultra-short-term wind power prediction system for wind farms provided by the present invention also includes a data preprocessing module. In order to prevent loss or measurement errors in data transmission, the real-time data provided by the wind farm and the meteorological department are centrally processed, bad data are eliminated, and the authenticity of the data is ensured.

[0103] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0104] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A clustering-based ultra-short-term wind power prediction method for wind farms, characterized in that: The method comprises: Obtaining basic data and real-time measurement data of the wind farm; the basic data includes wind turbine parameter data, wind turbine wind power history data, climate background data, and topographic data; the real-time measurement data includes weather forecast data, weather real-time data, and wind tower real-time data; According to the basic data, clustering is performed on all wind turbines in the wind farm using a Gaussian mixture model to obtain a plurality of wind turbine clusters; Each wind turbine cluster is equal to an equivalent wind turbine, and multiple equivalent wind turbine prediction models are established; Establishing a combined prediction model of the wind farm based on a plurality of equivalent wind turbine prediction models; Obtaining an ultra-short-term wind power real-time forecast value of the wind farm according to the real-time measurement data and the combined forecast model; According to the multiple equivalent wind turbine prediction models, a combined prediction model of the wind farm is established, which specifically includes: Obtain historical measurement data of wind farms; Determining the weight of each equivalent wind turbine prediction model based on the historical measurement data; Establishing the combined prediction model according to the weight and a plurality of equivalent wind turbine prediction models; According to the historical measurement data, the weight of each equivalent wind turbine prediction model is determined, specifically including: Inputting the historical measurement data into each of the equivalent wind turbine prediction models respectively to obtain historical wind power prediction values ​​of multiple equivalent wind turbines; Calculate the variance between the historical wind power prediction value and the historical wind power value of each equivalent wind turbine group respectively; According to the multiple variances, using the formula and formula Determine the weight of each equivalent wind turbine prediction model; Where σ is the sum of the variances of the historical wind power prediction values ​​and the historical wind power values ​​of multiple equivalent wind turbines, n is the total number of equivalent wind turbine prediction models, k is i is the weight coefficient of the prediction model of the i-th equivalent wind turbine group, i=1,...n; ξ i is the variance between the historical wind power prediction value and the historical wind power value of the i-th equivalent wind turbine; e i is the prediction error of the prediction model of the i-th equivalent wind turbine group; i = 1, ... n and i ≠ j; e j is the prediction error of the prediction model of the j-th equivalent wind turbine; λ is the Lagrangian constant.

2. The clustering-based ultra-short-term wind power prediction method for a wind farm according to claim 1, characterized in that: The equivalent wind turbine prediction model is one of a BP prediction model, a LSSVM prediction model, a GA-LSSVM prediction model, an ARIMA prediction model or a KALMAN filtering prediction model.

3. The clustering-based ultra-short-term wind power prediction method for a wind farm according to claim 2, characterized in that: The combined prediction model is: Among them, P is the real-time forecast value of the ultra-short-term wind power of the wind farm, P i is the real-time predicted value of wind power of the ith equivalent wind turbine prediction model; the real-time predicted value of wind power is obtained by inputting the real-time measurement data into the equivalent wind turbine prediction model.

4. A clustering-based ultra-short-term wind power prediction system for wind farms, characterized in that: The system comprises: The data acquisition module is used to acquire basic data and real-time measurement data of the wind farm; the basic data includes wind turbine parameter data, wind turbine wind power history data, climate background data, and topographic data; the real-time measurement data includes weather forecast data, weather real-time data, and wind tower real-time data; A wind turbine cluster determination module, configured to perform clustering processing on all wind turbines in the wind farm using a Gaussian mixture model according to the basic data to obtain a plurality of wind turbine clusters; An equivalent wind turbine prediction model establishment module is used to treat each wind turbine cluster as an equivalent wind turbine and establish multiple equivalent wind turbine prediction models; A combined prediction model establishment module, used to establish a combined prediction model of a wind farm based on a plurality of equivalent wind turbine prediction models; An ultra-short-term wind power real-time prediction value determination module is used to obtain an ultra-short-term wind power real-time prediction value of a wind farm according to the real-time measurement data and the combined prediction model; The combined prediction model establishment module specifically includes: A historical measurement data acquisition unit, used for acquiring historical measurement data of the wind farm; A weight determination unit, used to determine the weight of each equivalent wind turbine prediction model according to the historical measurement data; A combined prediction model establishing unit, used to establish the combined prediction model according to the weight and a plurality of equivalent wind turbine prediction models; The weight determination unit specifically includes: A wind power historical prediction value determination subunit, used for inputting the historical measurement data into each of the equivalent wind turbine prediction models to obtain wind power historical prediction values ​​of multiple equivalent wind turbines; A calculation subunit, used for respectively calculating the variance between the historical wind power prediction value and the historical wind power value of each equivalent wind turbine generator set; The weight determination subunit is used to determine the weight according to the plurality of variances using the formula and formula Determine the weight of each equivalent wind turbine prediction model; Where σ is the sum of the variances of the historical wind power prediction values ​​and the historical wind power values ​​of multiple equivalent wind turbines, n is the total number of equivalent wind turbine prediction models, k is i is the weight coefficient of the prediction model of the i-th equivalent wind turbine group, i=1,...n; ξ i is the variance between the historical wind power prediction value and the historical wind power value of the i-th equivalent wind turbine; e i is the prediction error of the prediction model of the i-th equivalent wind turbine group; i = 1, ... n and i ≠ j; e j is the prediction error of the prediction model of the j-th equivalent wind turbine; λ is the Lagrangian constant.

5. The clustering-based ultra-short-term wind power prediction system for wind farms according to claim 4, characterized in that: The equivalent wind turbine prediction model is one of a BP prediction model, a LSSVM prediction model, a GA-LSSVM prediction model, an ARIMA prediction model or a KALMAN filtering prediction model.

6. The clustering-based ultra-short-term wind power prediction system for wind farms according to claim 5, characterized in that: The combined prediction model is: Among them, P is the real-time forecast value of the ultra-short-term wind power of the wind farm, P i is the real-time predicted value of wind power of the ith equivalent wind turbine prediction model; the real-time predicted value of wind power is obtained by inputting the real-time measurement data into the equivalent wind turbine prediction model.

Images