A Wind Power Prediction Method and System Based on Meteorological Data

Abstract

The present application discloses a wind power prediction method and system based on meteorological data, relating to wind power prediction technology, including: pre-training a large model and training an LSTM model; obtaining the geographical location description information of the target wind farm to construct a fine-tuning training set based on the keyword fields; using the fine-tuning training set to fine-tune the pre-trained large model; obtaining the current meteorological prediction data of the area covered by the target wind farm and outputting the wind farm prediction information of multiple distribution locations based on the fine-tuned large model; based on the wind farm prediction information, using the trained LSTM model to perform prediction to obtain the power generation prediction information of multiple distribution locations; calculating the power generation prediction result of the target wind farm based on the power generation prediction information of multiple distribution locations. The present application combines a large model and LSTM, trains and predicts meteorological data through the large model, and uses LSTM to realize the prediction of wind farm power generation data, improving the accuracy of wind power prediction.

Description

Technical Field

[0001] This application relates to the technical field of wind power prediction, and particularly to a wind power prediction method and system based on meteorological data. Background Art

[0002] To meet the energy demand, wind energy has received extensive attention from researchers, and wind power generation technology is also becoming increasingly mature. However, due to the strong randomness and volatility of wind energy, it poses challenges to the safe and stable operation of the power system. High-precision wind power prediction can not only improve the utilization level of wind energy but also effectively alleviate the impact of the uncertainty of wind power on the system.

[0003] With the rapid development of artificial intelligence technology, deep learning models have gradually been introduced into wind power prediction. However, a single model has limitations in wind power prediction and often has preferences in data feature extraction, resulting in incomplete capture of wind power data features. Summary of the Invention

[0004] The embodiments of this application provide a wind power prediction method and system based on meteorological data, which combines a large model and LSTM, trains and predicts meteorological data through the large model, and uses LSTM to realize the prediction of wind farm power generation data, improving the accuracy of wind power prediction.

[0005] The embodiments of this application propose a wind power prediction method based on meteorological data, including the following steps:

[0006] Pre-obtain historical meteorological prediction data and historical actual meteorological data of the areas covered by different wind farms, train a large model, and train an LSTM model according to the historical meteorological prediction data, historical actual meteorological data, and power generation data at multiple distribution locations of different wind farms;

[0007] Obtain the geographical location description information of the target wind farm, and screen out the keyword fields related to wind power generation from the description information to construct a fine-tuning training set based on the keyword fields;

[0008] Use the fine-tuning training set to fine-tune the trained large model;

[0009] Obtain the current meteorological prediction data of the area covered by the target wind farm, and output wind farm prediction information at multiple distribution locations based on the fine-tuned large model;

[0010] Based on the wind farm prediction information, use the trained LSTM model to perform prediction to obtain power generation prediction information at multiple distribution locations;

[0011] Calculate the power generation prediction result of the target wind farm based on the power generation prediction information at multiple distribution locations.

[0012] Optionally, obtain historical meteorological prediction data and historical actual meteorological data for the areas covered by different wind farms in advance, and train the large model, including:

[0013] Use web crawler technology to obtain historical meteorological prediction data for the areas covered by different wind farms from meteorological websites;

[0014] Determine the wind farm center area in the historical meteorological prediction data, and calculate the distance deviation between the wind farm center area and the areas covered by different wind farms;

[0015] Add marks to the historical meteorological prediction data based on the distance deviation, and use the historical meteorological prediction data and the historical actual meteorological data as a set of training data at set time intervals to perform full-parameter training on the large model.

[0016] Optionally, train the LSTM model according to the historical meteorological prediction data, the historical actual meteorological data, and the power generation data of multiple distribution locations of different wind farms, including:

[0017] Obtain the terrain relationship of the different wind farms, and determine the positions of all wind turbines in the different wind farms based on the terrain relationship;

[0018] Determine the terrain characteristics of the positions of each wind turbine;

[0019] According to the positions of all wind turbines in the different wind farms and the terrain characteristics, select wind turbines at multiple positions with the required number of wind turbines to be selected, where the selected wind turbine positions include the position with the largest deviation in the terrain characteristics and multiple positions that are dispersed from the position with the largest deviation;

[0020] Obtain the historical power generation data of the selected wind turbines at multiple positions to train the LSTM model.

[0021] Optionally, obtain the regional location description information of the target wind farm, and screen out the keyword fields related to wind power generation from the description information to construct a fine-tuning training set based on the keyword fields, including:

[0022] Obtain the description information related to the regional location of the target wind farm from the web pages related to the target wind farm and the authoritative geographical platform, and perform word segmentation on the description information;

[0023] Based on the word segmentation results, calculate the edit distance from the pre-configured candidate word segments;

[0024] Determine the keyword fields related to wind power generation from the description information according to the calculation results of the edit distance;

[0025] Construct a fine-tuning training set based on the determined keyword fields related to wind power generation.

[0026] Optionally, it further includes repeating, according to a preset period, obtaining description information related to the geographical location of the target wind farm from relevant web pages of the target wind farm and an authoritative geographical platform, and updating the fine-tuning training set to fine-tune the trained large model in the case where the keyword fields change.

[0027] Optionally, obtain the current meteorological prediction data of the area covered by the target wind farm, and the wind farm prediction information at multiple distribution locations output based on the fine-tuned large model includes:

[0028] Obtain the current meteorological prediction data of the area covered by the target wind farm, and add a mark to the current meteorological prediction data based on the keywords of the target wind farm and input it into the fine-tuned large model, so as to use the fine-tuned large model to output the wind farm prediction information at multiple distribution locations.

[0029] Optionally, calculating the power generation prediction result of the target wind farm based on the power generation prediction information at multiple distribution locations includes:

[0030] Divide all the wind turbine positions of the target wind farm into multiple sub-regions, and any sub-region contains at least two predicted distribution locations;

[0031] Calculate the power generation prediction information of other positions within the sub-region according to the power generation prediction information of the divided sub-region;

[0032] Sum up the power generation prediction information for each to obtain the power generation prediction result of the target wind farm.

[0033] An embodiment of the present application also proposes a wind power prediction system based on meteorological data, including a processor and a memory. A computer program is stored on the memory, and when the computer program is executed by the processor, the steps of the wind power prediction method based on meteorological data as described above are implemented.

[0034] The method of the present application combines a large model and LSTM, trains and predicts meteorological data through the large model, and uses LSTM to realize the prediction of wind farm power generation data, improving the accuracy of wind power prediction.

[0035] The above description is only an overview of the technical solution of the present application. In order to be able to understand the technical means of the present application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of the present application more obvious and understandable, the following specifically illustrates the specific embodiments of the present invention. Description of the Drawings

[0036] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Also, throughout the drawings, the same reference numerals are used to denote the same components. In the drawings:

[0037] Figure 1 This is the basic flowchart of this embodiment. Detailed implementation manners

[0038] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art.

[0039] An embodiment of the present application proposes a wind power prediction method based on meteorological data, as Figure 1 shown, including the following steps:

[0040] In step S101, historical meteorological prediction data and historical actual meteorological data of areas covered by different wind farms are acquired in advance, a large model is trained, and an LSTM model is trained according to the historical meteorological prediction data, historical actual meteorological data, and power generation data of multiple distribution positions of different wind farms.

[0041] In some embodiments, acquiring historical meteorological prediction data and historical actual meteorological data of areas covered by different wind farms in advance and training a large model includes:

[0042] Using web crawler technology to obtain historical meteorological prediction data of areas covered by different wind farms from meteorological websites. For example, historical meteorological prediction data of areas covered by different wind farms can be obtained from authoritative meteorological websites using relevant technologies, or historical meteorological prediction data of areas covered by different wind farms can be obtained according to nearby meteorological monitoring stations and other means.

[0043] Determine the wind field center area in the historical meteorological prediction data, and calculate the distance deviation between the wind field center area and the areas covered by different wind farms. In some examples, the meteorological prediction data is not only for the areas covered by different wind farms, but is for the entire range. In this example, further calculate the distance deviation between the wind field center area and the areas covered by different wind farms. In specific applications, the wind field center area does not only include, for example, the typhoon center, but also includes the starting points of possible wind factors, etc. The method of this application aims at the prediction deviation that may be brought about by the large coverage area of most wind farms, but the meteorological forecast data only predicts large areas. In the embodiments of this application, by calculating the distance deviation, more accurate prediction data that fits the geographical location of the wind turbines can be derived based on the reasoning ability of the large model in subsequent examples. By selecting different wind farms for training, the large model can learn more regional weather information, and through subsequent fine-tuning, it can be adapted to the target wind farm.

[0044] Add a mark to the historical meteorological prediction data based on the distance deviation, and at a set time interval, use the historical meteorological prediction data and the historical actual meteorological data as a set of training data to perform full-parameter training on the large model. In some examples, the time interval can be set according to the location of the actual wind farm, according to the periodicity of, for example, seasonal climate.

[0045] In step S102, obtain the geographical location description information of the target wind farm, and screen out the keyword fields related to wind power generation from the description information, so as to construct a fine-tuning training set based on the keyword fields.

[0046] In step S103, use the fine-tuning training set to fine-tune the trained large model. By fine-tuning the large model, a large model of the target wind farm that is more suitable for the target wind farm can be further obtained.

[0047] In step S104, obtain the current meteorological prediction data of the area covered by the target wind farm, and output wind field prediction information at multiple distribution positions based on the fine-tuned large model. By setting the distribution positions, the large-scale prediction of the wind field can be realized through multiple representative local positions, reducing the computational amount and improving the calculation efficiency.

[0048] In step S105, based on the wind field prediction information, use the trained LSTM model to perform prediction to obtain power generation prediction information at multiple distribution positions.

[0049] In step S106, calculate the power generation prediction result of the target wind farm based on the power generation prediction information at multiple distribution positions.

[0050] The method of the present application combines a large model and LSTM, trains and predicts meteorological data through the large model, and uses LSTM to achieve the prediction of wind farm power generation data, improving the accuracy of wind power prediction.

[0051] In some embodiments, training the LSTM model according to historical meteorological prediction data, historical actual meteorological data, and power generation data at multiple distribution positions of different wind farms includes:

[0052] Obtain the terrain relationship of the different wind farms, and determine the positions of all wind turbines in the different wind farms based on the terrain relationship. For example, the position of any wind farm turbine can be marked in the terrain relationship model.

[0053] Determine the terrain features of the positions of the wind turbines, such as geographical location, slope, etc. These factors are related to power generation on the one hand and are associated factors of the wind farm air duct on the other hand.

[0054] According to the positions of all wind turbines in the different wind farms and the terrain features, select wind turbines at multiple positions with the required number of selected wind turbines to be distributed. The selected wind turbine positions include the position with the largest deviation in the terrain features and multiple positions that are dispersed from the position with the largest deviation. The position with the largest deviation and other distributed positions are selected as the representative positions of the target wind farm for subsequent prediction.

[0055] Obtain the historical power generation data of the wind turbines at the selected multiple positions to train the LSTM model.

[0056] In some embodiments, obtaining the regional location description information of the target wind farm and screening out the keyword fields related to wind power generation from the description information to construct a fine-tuning training set based on the keyword fields includes:

[0057] Obtain the description information related to the regional location of the target wind farm from the relevant web pages of the target wind farm and the authoritative geographical platform, and perform word segmentation on the description information.

[0058] Based on the word segmentation results, calculate the edit distance between the words and the pre-configured candidate words. For example, candidate words related to wind turbines and candidate words related to wind farm positions can be pre-configured. The candidate words can be applicable to different wind farms or for the whole-season climate. The fine-tuned large model is then adapted to the overall situation of the target wind farm.

[0059] According to the calculation results of the edit distance, determine the keyword fields related to wind power generation from the description information. For example, determine the keyword fields related to wind power generation from the description information according to the length of the edit distance.

[0060] Construct a fine-tuning training set based on the determined keyword fields related to wind power generation.

[0061] In some embodiments, it further includes repeating, according to a preset period, obtaining descriptive information related to the geographical location of the target wind farm from relevant web pages of the target wind farm and authoritative geographical platforms, and updating the fine-tuning training set to fine-tune the trained large model when there are changes in the keyword fields. For example, possible terrain changes, seasonal climate changes, etc., can be used to update the trained large model.

[0062] In some embodiments, obtaining the current meteorological prediction data of the area covered by the target wind farm, and the wind farm prediction information of multiple distribution locations output based on the fine-tuned large model includes: obtaining the current meteorological prediction data of the area covered by the target wind farm, and adding markers to the current meteorological prediction data based on the keywords of the target wind farm and inputting it into the fine-tuned large model, so as to use the fine-tuned large model to output the wind farm prediction information of multiple distribution locations. Through the fine-tuned large model, it is possible to predict the wind farm information of multiple distribution locations based on the current meteorological prediction data of the target wind farm, thereby improving the prediction accuracy and being applicable to the characteristics of large-area coverage of wind power.

[0063] In some embodiments, calculating the power generation prediction result of the target wind farm based on the power generation prediction information of multiple distribution locations includes:

[0064] Dividing all the positions of the wind turbines in the target wind farm into multiple sub-regions, and any sub-region contains at least two predicted distribution positions;

[0065] Calculating the power generation prediction information of other positions in the sub-region according to the power generation prediction information of the divided sub-regions. For example, the power generation prediction information of other positions in the sub-region can be calculated by means such as calculating the mean value and fitting.

[0066] Summing up the power generation prediction information of each item to obtain the power generation prediction result of the target wind farm.

[0067] The embodiment of the present application proposes a method for training and predicting meteorological data by combining a large model and LSTM. On the one hand, it uses the inference ability of the large model to improve the accuracy of wind farm prediction based on historical meteorological data. On the other hand, through the memory effect of LSTM, it greatly improves the convenience and accuracy of wind power prediction.

[0068] The embodiment of the present application also proposes a wind power prediction system based on meteorological data, including a processor and a memory. A computer program is stored on the memory, and when the computer program is executed by the processor, it implements the steps of the wind power prediction method based on meteorological data as described above.

[0069] In addition, although exemplary embodiments have been described herein, the scope includes any and all embodiments based on the present disclosure having equivalent elements, modifications, omissions, combinations (e.g., schemes that cross various embodiments), adaptations, or alterations. It is not limited to the examples described in this specification or during the implementation of this application, and the examples will be construed as non-exclusive.

[0070] The above description is intended to be illustrative rather than restrictive. For example, the above examples (or one or more aspects thereof) may be used in combination with each other. For example, other embodiments may be used by those of ordinary skill in the art upon reading the above description.

[0071] The above embodiments are only exemplary embodiments of the present disclosure, and those skilled in the art may make various modifications or equivalent replacements to the present invention within the essence and protection scope of the present disclosure, and such modifications or equivalent replacements should also be regarded as falling within the protection scope of the present invention.

Claims

1. A wind power prediction method based on meteorological data, characterized in that, It includes the following steps: Pre-obtain the historical meteorological prediction data and historical actual meteorological data of the covered areas of different wind farms, train a large model, and train an LSTM model based on the historical meteorological prediction data, historical actual meteorological data, and power generation data at multiple distribution locations of different wind farms; Obtain the geographical location description information of the target wind farm, and screen out the keyword fields related to wind power generation from the description information to construct a fine-tuning training set based on the keyword fields; Use the fine-tuning training set to fine-tune the trained large model; Obtain the current meteorological prediction data of the covered area of the target wind farm, and output the wind farm prediction information of multiple distribution locations based on the fine-tuned large model; Based on the wind farm prediction information, use the trained LSTM model to perform prediction to obtain the power generation prediction information of multiple distribution locations; Calculate the power generation prediction result of the target wind farm based on the power generation prediction information of multiple distribution locations; Pre-obtaining the historical meteorological prediction data and historical actual meteorological data of the covered areas of different wind farms and training a large model includes: Using web crawler technology to obtain the historical meteorological prediction data of the covered areas of different wind farms from meteorological websites; Determine the wind farm central area in the historical meteorological prediction data, and calculate the distance deviation between the wind farm central area and the covered areas of different wind farms; Add marks to the historical meteorological prediction data based on the distance deviation, and use the historical meteorological prediction data and historical actual meteorological data as a set of training data at a set time interval to perform full-parameter training on the large model; Training the LSTM model based on the historical meteorological prediction data, historical actual meteorological data, and power generation data at multiple distribution locations of different wind farms includes: Obtain the terrain relationship of the different wind farms, and determine the positions of all wind turbines in the different wind farms based on the terrain relationship; Determine the terrain characteristics of the positions of each wind turbine; According to the positions of all wind turbines in different wind farms and the terrain characteristics, select wind turbines at multiple positions with the required number of wind turbines to be selected, where the selected wind turbine positions include the position with the largest deviation in the terrain characteristics and multiple positions scattered from the position with the largest deviation; According to the wind turbines at the selected multiple positions, obtain their historical power generation data to train the LSTM model.

2. The wind power prediction method based on meteorological data according to claim 1, wherein, Obtaining the geographical location description information of the target wind farm and screening out the keyword fields related to wind power generation from the description information to construct a fine-tuning training set includes: Obtain the description information related to the geographical location of the target wind farm from the relevant web pages of the target wind farm and authoritative geographical platforms, and perform word segmentation on the description information; Based on the word segmentation results, calculate the edit distance between the candidate word segments pre-configured; According to the calculation result of the edit distance, determine the keyword fields related to wind power generation from the description information; Construct a fine-tuning training set according to the determined keyword fields related to wind power generation.

3. The wind power prediction method based on meteorological data according to claim 2, wherein It further includes repeating, according to a preset period, obtaining description information related to the geographical location of the target wind farm from the relevant web pages of the target wind farm and an authoritative geographical platform, and updating and fine-tuning the training set to fine-tune the trained large model in the case where the keyword fields change.

4. The wind power prediction method based on meteorological data according to claim 2, wherein Obtaining current meteorological prediction data for the area covered by the target wind farm, and outputting wind farm prediction information for multiple distribution locations based on the fine-tuned large model, including: Obtaining current meteorological prediction data for the area covered by the target wind farm, adding a label to the current meteorological prediction data based on the keyword of the target wind farm, and inputting it into the fine-tuned large model, so as to output wind farm prediction information for multiple distribution locations by using the fine-tuned large model.

5. The wind power prediction method based on meteorological data according to claim 2, wherein Calculating the power generation prediction result of the target wind farm based on the power generation prediction information for multiple distribution locations, including: Dividing all the wind turbine locations of the target wind farm into multiple sub-regions, and each sub-region contains at least two predicted distribution locations; Calculating the power generation prediction information for other locations within the sub-region according to the power generation prediction information of the segmented sub-region; Summing up the power generation prediction information for each to obtain the power generation prediction result of the target wind farm.

6. A wind power prediction system based on meteorological data, characterized in that, It includes a processor and a memory, and a computer program is stored on the memory. When the computer program is executed by the processor, the steps of the wind power prediction method based on meteorological data according to any one of claims 1 to 5 are implemented.

Images