Short-term wind power prediction method, computer equipment and storage medium

Abstract

The invention discloses a short-term wind power prediction method, computer equipment and a storage medium, and particularly relates to the field of wind power prediction, and the method comprises the steps: S1, collecting target fan station data, S2, preprocessing the target fan station data, S3, carrying out the analogy of similar fans, S4, fusing a plurality of intelligent models, and S5, outputting a prediction report. According to the short-term wind power prediction method based on multi-model fusion, the computer equipment and the storage medium, target wind field station data is preprocessed, features are further screened, features with small correlation with wind power are removed, and the model prediction efficiency is improved; through the fusion model, the defects of a single model are made up, and the accuracy of wind power prediction is improved; by outputting the prediction report, the output power of the wind power plant in a short time is accurately predicted, a wind power plant manager is helped to make a power generation plan in advance, power generation amount fluctuation caused by wind power fluctuation is reduced, and therefore the operation efficiency of the wind power plant is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power prediction, and more specifically, to a short-term wind power prediction method, a computer device, and a storage medium. Background Art

[0002] With the continuous growth of the global demand for clean energy and the rapid development of technology, the field of renewable energy has attracted much attention. As a clean and sustainable energy form, the scale of development and utilization of wind energy is constantly expanding. However, wind energy itself has the characteristics of volatility, intermittency, and randomness, which make it difficult to stably predict the output power of wind power, posing challenges to the safe and stable operation of the power system.

[0003] Based on various parameters such as rich meteorological information of the wind farm and the operating status of wind turbines, by using physical simulation calculations and scientific statistical methods, the power generation of the wind farm in the future period is predicted, providing a decision-making basis for the power dispatching department, enabling it to more reasonably arrange unit combinations and power generation plans, thus effectively improving the trading efficiency of the power market. At the same time, it provides indispensable data support for the scientific planning and optimal design of the wind farm, strongly promoting the vigorous development of the wind power industry in the direction of high efficiency and sustainability.

[0004] However, in actual use, there are still some drawbacks. For example, the inherent uncertainty and volatility of wind energy result in large prediction errors and it is difficult to meet the accuracy requirements of actual needs; a single prediction model, due to the limitations of its own structure and algorithm, cannot fully adapt to the complex and changeable requirements of the wind power prediction task and it is difficult to achieve a satisfactory accuracy rate; due to the uncertainty of wind power output, the power dispatching department faces great difficulties in arranging unit combinations and power generation plans and it is difficult to ensure the accurate matching of power supply and demand. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides a short-term wind power prediction method, a computer device, and a storage medium, through the following solutions to solve the problems raised in the above background art.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A short-term wind power prediction method, characterized by comprising:

[0008] S1: Collect data of the target wind farm station: In response to the information collection device of the wind farm station, obtain the first fan data group corresponding to the target wind farm station;

[0009] S2: Preprocess the target wind farm station data: Perform preprocessing operations on the first set of wind turbine data. The preprocessing operations are used to obtain the first set of key wind turbine features corresponding to the first set of wind turbine data;

[0010] S3: Analogize similar wind turbines: Obtain an intelligent wind turbine analysis model. According to the intelligent wind turbine analysis model and through the first set of key wind turbine features, obtain the second set of wind turbine data;

[0011] S4: Integrate multiple intelligent models: Obtain an intelligent prediction integration model. According to the intelligent prediction integration model and through the second set of wind turbine data, obtain the third set of wind turbine data;

[0012] S5: Output a prediction report: Based on the first set of wind turbine data, the second set of wind turbine data, and the third set of wind turbine data, obtain a short-term wind power prediction report.

[0013] Preferably, in S1, the first set of wind turbine data includes wind turbine operation data, wind turbine power data, geographical location coordinates of each wind turbine in the wind farm station, spatial layout information between wind turbines, and corresponding historical data processed by spatio-temporal interpolation.

[0014] Preferably, in S2, to obtain the first set of key wind turbine features, specifically includes:

[0015] Based on using the Pearson correlation coefficient as a measurement tool, calculate the degree of linear correlation quantification between variables Specifically expressed as:

[0016]

[0017] Among them, WP represents the wind turbine power, FDS i represents the i-th data in the first set of wind turbine data, i represents the index of the corresponding data in the first set of wind turbine data, cov(WP, FDS i ) is the covariance between WP and FDS i , and σ WP and respectively represent the standard deviations of WP and fds i ;

[0018] When , it indicates a positive correlation between the wind turbine power and the i-th data in the first set of wind turbine data, indicating a negative correlation between the wind turbine power and the i-th data in the first set of wind turbine data.

[0019] Preferably, in S3, pass the first set of key wind turbine features through the target intelligent wind turbine analysis model to obtain a cluster of similar wind turbines composed of similar features. The cluster of similar wind turbines is composed of multiple similar wind turbine features in the same cluster.

[0020] Preferably, in step S3, a plurality of similar fan clusters are obtained by inputting the first key fan feature groups corresponding to a plurality of fans into the target intelligent fan analysis model;

[0021] Based on the fan n and fan m corresponding to the first key fan feature group, calculate the similarity WS between fan n and fan m, which is specifically expressed as:

[0022]

[0023] where FD represents the total number of key features of the first key fan feature group, k represents the index of the key feature of the first key fan feature group, FD nk represents the k-th key feature corresponding to fan n, and FD mk represents the k-th key feature corresponding to fan m.

[0024] Preferably, in step S4, obtaining the third fan data group specifically includes:

[0025] Obtain an intelligent prediction model, which is a prediction model stored in the system operation database;

[0026] Determine the fusion strategy of the intelligent prediction model;

[0027] Input the second fan data group into the intelligent prediction fusion model for prediction, and output the corresponding prediction result, that is, the third fan data group.

[0028] Preferably, in step S4, obtaining the third fan data group specifically includes: The fusion strategy is a strategy of assigning corresponding weights to each prediction model based on the performance of the corresponding historical data verification in the system operation database.

[0029] Preferably, in step S4, obtaining the third fan data group specifically includes:

[0030] Take the second fan data group corresponding to the fan units in the same similar fan cluster as the data set M1. The data set M1 includes m1 features, input it into the GRU model for training and prediction, and the output result gum1 and the data set M1 form the data set M2;

[0031] The data set M2 is input into the LightGBM model for training and prediction, and the prediction result gbm1 and the data set M2 form the data set M3. The data set M3 includes m1 + 2 features;

[0032] The data set M3 is input into the LSTM model for training and prediction, and the prediction result lstm of the LSTM model is used as part of the short-term prediction result;

[0033] The short-term prediction results are generated by weighted fusion of the outputs of the GRU model, the LightGBM model, and the LSTM model.

[0034] To achieve the above object, the present invention provides the following technical solutions: A short-term wind power prediction system based on multi-model fusion, including a system operation database, a system central processing module, and a user information terminal. Implementing the above-mentioned short-term wind power prediction method includes:

[0035] Target wind farm station data acquisition module: Used to respond to the information acquisition device of the wind farm station, obtain the first fan data group corresponding to the target wind farm station, and transmit it to the wind farm station data preprocessing module and the prediction report output module;

[0036] Wind farm station data preprocessing module: Used to perform preprocessing operations on the first fan data group transmitted by the target wind farm station data acquisition module. The preprocessing operation is used to obtain the first key fan feature group corresponding to the first fan data group, and transmit it to the similar fan analogy module;

[0037] Similar fan analogy module: Used to obtain an intelligent fan analysis model, and according to the intelligent fan analysis model and the first key fan feature group transmitted by the wind farm station data preprocessing module, obtain the second fan data group, and transmit it to the intelligent model fusion module and the prediction report output module;

[0038] Intelligent model fusion module: Used to obtain an intelligent prediction fusion model, and according to the intelligent prediction fusion model and the second fan data group transmitted by the similar fan analogy module, obtain the third fan data group, and transmit it to the prediction report output module;

[0039] Prediction report output module: Used to obtain a short-term wind power prediction report based on the first fan data group transmitted by the target wind farm station data acquisition module, the second fan data group transmitted by the similar fan analogy module, and the third fan data group transmitted by the intelligent model fusion module, and output it to the user terminal in a preset manner;

[0040] The system operation database includes all data texts of the short-term wind power prediction system and collects the information texts output by each module in real time. The system central processing module is used to control the information text instructions output by each module in the method, and the user information terminal is an information output device for receiving the short-term wind power prediction system.

[0041] On the other hand, an embodiment of the present application provides a computer device, which includes an input interface and an output interface, and the computer device further includes:

[0042] A processor, adapted to implement one or more instructions; and,

[0043] A computer storage medium stores one or more instructions, and the one or more instructions are adapted to be loaded and executed by the processor to perform the following steps:

[0044] In response to the information collection device of the wind farm station, obtain the first wind turbine data set corresponding to the target wind farm station;

[0045] Perform a preprocessing operation on the first wind turbine data set, and the preprocessing operation is used to obtain the first key wind turbine feature set corresponding to the first wind turbine data set;

[0046] Obtain an intelligent wind turbine analysis model, and according to the intelligent wind turbine analysis model, obtain a second wind turbine data set through the first key wind turbine feature set;

[0047] Obtain an intelligent prediction fusion model, and according to the intelligent prediction fusion model, obtain a third wind turbine data set through the second wind turbine data set;

[0048] Based on the first wind turbine data set, the second wind turbine data set, and the third wind turbine data set, obtain a short-term wind power prediction report.

[0049] On the other hand, an embodiment of the present application provides a computer storage medium, and the computer storage medium stores one or more instructions, and the one or more instructions are adapted to be loaded and executed by the processor to perform the short-term wind power prediction method mentioned above.

[0050] Technical effects and advantages of the present invention:

[0051] 1. By preprocessing the data of the target wind farm station, the present invention further screens features and eliminates features with low correlation with wind power, improving the prediction efficiency of the model;

[0052] 2. By using a fusion model, the present invention makes up for the shortcomings of a single model and improves the accuracy of wind power prediction;

[0053] 3. By outputting a prediction report, the present invention accurately predicts the output power of the wind farm in a short period of time, helps the wind farm manager formulate a power generation plan in advance, reduces the power generation fluctuation caused by the volatility of wind power, and thus improves the operation efficiency of the wind farm. Description of the Drawings

[0054] Figure 1 It is a method step diagram of the present invention.

[0055] Figure 2 It is a method structure schematic diagram of the present invention.

[0056] Figure 3 It is a structure diagram of the multi-model fusion wind power prediction model of the present invention. Detailed Embodiments

[0057] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0058] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to limit the present application. As used in the specification of the present application, the singular forms "a", "an", "the", "above", "said", "this" are also intended to include the plural forms, unless there is a clear indication to the contrary in the context. It should also be understood that the term "and / or" used in the present application refers to and includes any or all possible combinations of one or more of the listed items.

[0059] Hereinafter, the terms "first", "second", and "third" are only used for descriptive purposes and cannot be construed as implying or suggesting relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first", "second", and "third" may explicitly or implicitly include one or more of such features. In the description of the embodiments of the present application, unless otherwise specified, the meaning of "a plurality" is two or more.

[0060] As shown in the attached Figure 1 A short-term wind power prediction method includes S1: collecting target wind farm station data, S2: preprocessing the target wind farm station data, S3: analogizing similar wind turbines, S4: fusing multiple intelligent models, and S5: outputting a prediction report.

[0061] S1: Collecting target wind farm station data: In response to the information collection device of the wind farm station, obtain the first wind turbine data set corresponding to the target wind farm station.

[0062] Specifically, the short-term wind power prediction system responds to the information collection devices at the wind farm station. The information collection devices include, but are not limited to, anemometers, wind vanes, thermometers, barometers, etc. installed at different positions and heights. Based on a preset collection frequency, it collects the corresponding fan data and the corresponding geographical location of the target wind farm station. At the same time, it performs spatio-temporal interpolation processing on the historical data corresponding to the target wind farm station stored in the system operation database of the short-term wind power prediction system. The spatio-temporal interpolation processing is to fill in based on the data collection time point and spatial position through an interpolation algorithm. It integrates and obtains the first fan data group corresponding to the target wind farm station. The first fan data group includes fan operation data, fan power data, the geographical location coordinates of each fan in the wind farm station, the spatial layout information between the fans, and the corresponding historical data processed by spatio-temporal interpolation. Among them, the fan operation data includes the wind speed at which the fan operates, the wind direction at which the fan operates, the temperature during fan operation, and the air pressure during fan operation. The spatial layout information between the fans includes, but is not limited to, the fan spacing, arrangement method, etc.

[0063] In this embodiment, the collection frequency is preset to collect the corresponding fan data of the target wind farm station every 5 minutes. The interpolation algorithms used in the spatio-temporal interpolation processing include, but are not limited to, the Kriging interpolation method, the inverse distance weighted interpolation method, etc.

[0064] S2: Preprocess the data of the target wind farm station: Perform a preprocessing operation on the first fan data group. The preprocessing operation is used to obtain the first key fan feature group corresponding to the first fan data group.

[0065] Specifically, after the short-term wind power prediction system obtains the first fan data group, through the preprocessing operation, it can obtain multiple key features corresponding to the first fan data group, that is, the first key fan feature group. The steps of the preprocessing are as follows:

[0066] A1: Check the continuity of the first fan data group, mark the missing data and duplicate data, and at the same time convert the time to a timestamp and calculate the difference between adjacent time points;

[0067] A2: Check the outliers in the first fan data group and mark the outliers;

[0068] In this embodiment, a density-based clustering method with noise is used. The first fan data group is clustered through a clustering algorithm, and the unclassified data and the classes with less data are determined as outliers. The clustering algorithms used include, but are not limited to, K-means, Gaussian mixture model, and density-based clustering method with noise, etc.;

[0069] A3: Fill in the missing data and abnormal data marked in A1 and A2, and delete the duplicate data;

[0070] In this embodiment, a repair method based on neighboring data is adopted for the marked outliers and missing values, specifically represented as the K-nearest neighbor method. By identifying several data points around the target outlier or missing value, the arithmetic mean of the neighboring data is calculated and used as a replacement value to fill in the abnormal or missing position.

[0071] A4: Normalize the corresponding values of the first fan data group;

[0072] In this embodiment, the maximum-minimum value method is adopted as the normalization method to map the values to the interval (0, 1).

[0073] A5: Perform feature dimensionality reduction on the data processed in A4. The feature dimensionality reduction is to screen out the key features related to the fan power through the feature selection algorithm of the correlation coefficient and discard the irrelevant data, so as to obtain multiple key features corresponding to the first fan data group, that is, the first key fan feature group.

[0074] In a possible implementation manner, obtaining the first key fan feature group includes: calculating the degree of linear correlation quantification between variables based on using the Pearson correlation coefficient as a measurement tool Specifically expressed as:

[0075]

[0076] Among them, WP represents the fan power, FDS i represents the i-th data in the first fan data group, and i represents the index of the corresponding data in the first fan data group. cov(WP, FDS i ) is the covariance of WP and FDS i , and σ WP and respectively represent the standard deviations of WP and fds i ;

[0077] It should be noted that the meaning of the degree of linear correlation quantification ρ WP,FDS between variables is as follows:

[0078] When , it indicates that there is a positive correlation between the fan power and the i-th data in the first fan data group. It indicates that there is a negative correlation between the fan power and the i-th data in the first fan data group; and when , it indicates that there is a certain degree of linear correlation between the fan power and the i-th data in the first fan data group. The closer to 1, the stronger the correlation. The closer to 0, the weaker the correlation.

[0079] In this embodiment, based on Based on the value, the correlation can be divided into three levels. It indicates that the correlation between the fan power and the i-th data in the first fan data group is relatively low. It indicates that the correlation between the fan power and the i-th data in the first fan data group is average. It indicates that the correlation between the fan power and the i-th data in the first fan data group is strong; and the features with relatively low correlation are discarded from the first fan data group, while the features with high correlation are retained.

[0080] S3: Analogize similar fans: Obtain an intelligent fan analysis model, and based on the intelligent fan analysis model, obtain a second fan data group through the first key fan feature group.

[0081] Specifically, the intelligent fan analysis model is a pre-constructed learning model. By inputting the first key fan feature group into the intelligent fan analysis model, the intelligent fan analysis model obtains the second fan data group according to the first key fan feature group. The second fan data group includes a similar fan cluster composed of similar features and corresponding key features.

[0082] In a possible implementation, obtaining the second fan data group includes: obtaining multiple intelligent fan analysis models based on the characteristics of the corresponding wind farm and the characteristics of historical data in the first key fan feature group. The multiple intelligent fan analysis models include principal component analysis, clustering analysis model, support vector machine model, and decision tree model; evaluating the performance of the multiple intelligent fan analysis models through evaluation metrics, and the evaluation metrics include but are not limited to accuracy, recall rate, F1 value, etc., to determine the target intelligent fan analysis model corresponding to the first key fan feature group; inputting the first key fan feature group through the target intelligent fan analysis model to obtain a similar fan cluster composed of similar features, and the similar fan cluster is composed of multiple similar fan features in the same cluster.

[0083] In this embodiment, based on the fan n and fan m corresponding to the first key fan feature group, the similarity WS between fan n and fan m is calculated, which is specifically expressed as:

[0084]

[0085] Among them, FD represents the total number of key features of the first key fan feature group, k represents the index of the key features of the first key fan feature group, FD nk represents the k-th key feature corresponding in fan n, and FD mk represents the k-th key feature corresponding in fan m;

[0086] It should be noted that by inputting the first key fan feature groups corresponding to multiple fans through the target intelligent fan analysis model, multiple similar fan clusters are obtained.

[0087] S4: Integrate multiple intelligent models: Obtain an intelligent prediction fusion model, and based on the intelligent prediction fusion model, obtain a third set of fan data through the second set of fan data.

[0088] Specifically, the intelligent prediction fusion model is a pre-constructed learning model. By inputting the second set of fan data into the intelligent prediction fusion model, the intelligent prediction fusion model obtains the third set of fan data based on the second set of fan data, that is, the short-term prediction result.

[0089] In a possible implementation, obtaining the third set of fan data includes: obtaining multiple intelligent prediction models. The intelligent prediction models are prediction models stored in the system operation database, and the prediction models include, but are not limited to, Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Supervised Learning Ensemble Model (LightGBM), Support Vector Machine (SVM), Random Forest (RF), Gradient Boosting Decision Tree (GBDT), etc.; determining the fusion strategy for the multiple intelligent prediction models. The fusion strategy is a strategy of assigning corresponding weights to each model based on the performance of each prediction model verified by the corresponding historical data in the system operation database. The fusion strategy includes, but is not limited to, weighted average method, voting method, stacking method, etc.; inputting the second set of fan data into the intelligent prediction fusion model for prediction and outputting the corresponding prediction result, that is, the third set of fan data.

[0090] It should be noted that taking the establishment of a fusion model composed of an LSTM model, a LightGBM model, and a GRU model as an example, input the second set of fan data into the fusion model and predict the fan units in the same similar fan cluster. The combined structure of the fusion model of the LSTM model, the LightGBM model, and the GRU model is as Figure 3 shown and consists of three layers in total;

[0091] It should be noted that in the first layer, the second set of fan data corresponding to the fan units in the same similar fan cluster is used as the data set M1. The data set M1 includes m1 features and is directly input into the GRU model for training and prediction. The output result gum1 and the data set M1 form the data set M2; in the second layer, the data set M2 is input into the LightGBM model for training and prediction. The prediction result gbm1 and the data set M2 form the data set M3, and the data set M3 includes m1 + 2 features; in the third layer, the data set M3 is input into the LSTM model for training and prediction. The prediction result lstm of the LSTM model is used as part of the short-term prediction result; the final short-term prediction result is obtained by weighted fusion of the outputs of the GRU model, the LightGBM model, and the LSTM model.

[0092] In this embodiment, the second set of fan data is divided into a training set and a validation set. The training set is allocated at a ratio of 70%, and the validation set is allocated at a ratio of 30%. The training set is used to train each intelligent prediction model, and the validation set is used to evaluate the performance of the intelligent prediction model and adjust the parameters of the intelligent prediction model. Under the weighted average method fusion strategy, the LSTM model outputs the prediction result lstm, the LightGBM model outputs the prediction result gbm1, and the GRU model outputs the prediction result gum1, and the short-term prediction result P is calculated as follows:

[0093] P = 0.3 * gum1 + 0.3 * gbm1 + 0.4 * lstm,

[0094] It should be noted that the weighted average method fusion strategy is used to obtain the weights corresponding to the short-term prediction results of the intelligent prediction model. In the embodiment verification, the accuracy rate of the LSTM model is 70%, the accuracy rate of the LightGBM model is 70%, and the accuracy rate of the GRU model is 93%. Based on the accuracy rate, the weights corresponding to the short-term prediction results are calculated. LSTM weight = 0.70 / (0.70 + 0.70 + 0.93) ≈ 0.3, RF weight = 0.70 / (0.70 + 0.70 + 0.93) ≈ 0.3, SVM weight = 0.93 / (0.70 + 0.70 + 0.93) ≈ 0.4.

[0095] S5: Output the prediction report: Based on the first set of fan data, the second set of fan data, and the third set of fan data, obtain the short-term wind power prediction report.

[0096] Specifically, the historical meteorological data, fan power data, and geographical location information in the first set of fan data are summarized with the feature data obtained after the similar fan analogy processing in the second set of fan data and the prediction results of the fusion model in the third set of fan data, and the correlation relationship between the data is established, so that each data point can correspond to the corresponding fan and time point; generate a short-term wind power prediction report. The structure of the short-term wind power prediction report includes but is not limited to prediction results and analysis, risk assessment and suggestions, and conclusions, etc. The prediction results and analysis are used to present the prediction data, charts, and relevant feature and index analysis of the short-term wind power; the risk assessment and suggestions are used to analyze the uncertain factors and risks existing in the prediction process and put forward corresponding countermeasures; the conclusions are used to summarize the main content and prediction conclusions of the report.

[0097] On the other hand, the embodiment of the present application provides a short-term wind power prediction system based on multi-model fusion, including a system operation database, a system central processing module, and a user information terminal, and further including: a target wind farm station data acquisition module, a wind farm station data preprocessing module, a similar fan analogy module, an intelligent model fusion module, and a prediction report output module.

[0098] Target wind farm station data acquisition module: It is used to respond to the information acquisition device of the wind farm station, obtain the first set of fan data corresponding to the target wind farm station, and transmit it to the wind farm station data preprocessing module and the prediction report output module;

[0099] Wind farm station data preprocessing module: It is used to perform preprocessing operations on the first set of fan data transmitted by the target wind farm station data acquisition module. The preprocessing operations are used to obtain the first set of key fan feature groups corresponding to the first set of fan data, and transmit it to the similar fan analogy module;

[0100] Similar fan analogy module: It is used to obtain the intelligent fan analysis model, and according to the intelligent fan analysis model and the first set of key fan feature groups transmitted by the wind farm station data preprocessing module, obtain the second set of fan data, and transmit it to the intelligent model fusion module and the prediction report output module;

[0101] Intelligent model fusion module: It is used to obtain the intelligent prediction fusion model, and according to the intelligent prediction fusion model and the second set of fan data transmitted by the similar fan analogy module, obtain the third set of fan data, and transmit it to the prediction report output module;

[0102] Prediction report output module: It is used to obtain a short-term wind power prediction report based on the first set of fan data transmitted by the target wind farm station data acquisition module, the second set of fan data transmitted by the similar fan analogy module, and the third set of fan data transmitted by the intelligent model fusion module, and output it to the user terminal in a preset manner;

[0103] The system operation database includes all data texts of the short-term wind power prediction system, and collects the information texts output by each module in real time. The system central processing module is used for the information text instructions output by each module in the central control method, and the user information terminal is an information output device for receiving the short-term wind power prediction system.

[0104] On the other hand, an embodiment of the present application provides a computer device, the computer device includes an input interface and an output interface, and the computer device further includes:

[0105] A processor, adapted to implement one or more instructions; and,

[0106] A computer storage medium, the computer storage medium stores one or more instructions, and the one or more instructions are adapted to be loaded and executed by the processor to perform the following steps:

[0107] Respond to the information acquisition device of the wind farm station, and obtain the first set of fan data corresponding to the target wind farm station;

[0108] Perform a preprocessing operation on the first set of fan data, where the preprocessing operation is used to obtain the first set of key fan features corresponding to the first set of fan data;

[0109] Obtain an intelligent fan analysis model, and based on the intelligent fan analysis model and through the first set of key fan features, obtain a second set of fan data;

[0110] Obtain an intelligent prediction fusion model, and based on the intelligent prediction fusion model and through the second set of fan data, obtain a third set of fan data;

[0111] Based on the first set of fan data, the second set of fan data, and the third set of fan data, obtain a short-term wind power prediction report.

[0112] On the other hand, an embodiment of the present application provides a computer storage medium, which stores one or more instructions, and the one or more instructions are adapted to be loaded and executed by a processor to perform the short-term wind power prediction method mentioned above.

[0113] Secondly: In the attached drawings of the disclosed embodiments of the present invention, only the structures related to the disclosed embodiments are involved, and other structures can refer to the general design. Without conflict, the same embodiment and different embodiments of the present invention can be combined with each other;

[0114] Finally: The above are only the preferred embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A short-term wind power prediction method, characterized in that: include: S1: Collecting target wind turbine station data: Responding to the information collection device of the wind turbine station, obtaining a first wind turbine data group corresponding to the target wind turbine station; S2: preprocessing target wind turbine station data: performing a preprocessing operation on the first wind turbine data group, the preprocessing operation being used to obtain a first key wind turbine feature group corresponding to the first wind turbine data group; S3: Analogy with similar wind turbines: obtaining an intelligent wind turbine analysis model, and obtaining a second wind turbine data group according to the intelligent wind turbine analysis model through the first key wind turbine feature group; S4: integrating multiple intelligent models: obtaining an intelligent prediction fusion model, and obtaining a third wind turbine data group through the second wind turbine data group according to the intelligent prediction fusion model; S5: Outputting a forecast report: Based on the first wind turbine data group, the second wind turbine data group, and the third wind turbine data group, obtaining a short-term wind power forecast report.

2. A short-term wind power prediction method according to claim 1, characterized in that: Said S1, the first wind turbine data group includes wind turbine operation data, wind turbine power data, geographical location coordinates of each wind turbine in the wind turbine plant, spatial layout information between wind turbines, and corresponding historical data processed by spatiotemporal interpolation.

3. A short-term wind power prediction method according to claim 1, characterized in that: The step S2, obtaining a first key wind turbine feature group, specifically includes: Based on the use of Pearson correlation coefficient as a measurement tool, the linear correlation between variables is calculated. Specifically expressed as: Among them, WP represents the fan power, FDS i represents the i-th data in the first wind turbine data group, i represents the index of the corresponding data in the first wind turbine data group, cov(WP, FDS i ) is WP and FDS i The covariance of WP and Represented as WP and fds respectively i The standard deviation of when This indicates that the wind turbine power is positively correlated with the i-th data in the first wind turbine data group. This indicates that there is a negative correlation between the wind turbine power and the i-th data in the first wind turbine data group.

4. A short-term wind power prediction method according to claim 1, characterized in that: In S3, the first key wind turbine feature group is subjected to a target intelligent wind turbine analysis model to obtain a similar wind turbine cluster composed of similar features, wherein the similar wind turbine cluster is a cluster composed of multiple similar wind turbine features.

5. A short-term wind power prediction method according to claim 4, characterized in that: S3, obtaining a plurality of similar wind turbine clusters by passing the first key wind turbine feature groups corresponding to the plurality of wind turbines through the target intelligent wind turbine analysis model; Based on the fans n and the fans m corresponding to the first key fan feature group, the similarity WS between the fans n and the fans m is calculated, which is specifically expressed as: Where FD represents the total number of key features of the first key wind turbine feature group, k represents the index of the key feature of the first key wind turbine feature group, and FD nk It is represented as the kth key feature corresponding to fan n, FD mk It is represented as the corresponding kth key feature in fan m.

6. A short-term wind power prediction method according to claim 1, characterized in that: The step S4, obtaining the third wind turbine data group, specifically includes: Obtaining an intelligent prediction model, where the intelligent prediction model is a prediction model stored in a system operation database; Determine the fusion strategy of intelligent prediction models; The second wind turbine data group is input into the intelligent prediction fusion model for prediction, and the corresponding prediction result, i.e., the third wind turbine data group, is output.

7. A short-term wind power prediction method according to claim 6, characterized in that: The S4, obtaining the third wind turbine data group, specifically includes: the fusion strategy is a strategy of assigning corresponding weights to each model based on the performance of each prediction model verified by corresponding historical data in the system operation database.

8. A short-term wind power prediction method according to claim 6, characterized in that: The step S4, obtaining the third wind turbine data group, specifically includes: The second wind turbine data group corresponding to the wind turbine units of the same similar wind turbine cluster is taken as the data set M1, and the data set M1 includes m1 features, which are input into the GRU model for training and prediction, and the output result gum1 and the data set M1 constitute the data set M2; Dataset M2 is input into the LightGBM model for training prediction. The prediction result gbm1 and data set M2 constitute data set M3, which includes m1+2 features. Dataset M3 is input into the LSTM model for training and prediction, and the prediction result lstm of the LSTM model is used as part of the short-term prediction result; The short-term prediction results are generated by weighted fusion of the outputs of the GRU model, LightGBM model, and LSTM model.

9. A computer device comprising an input interface and an output interface, characterized in that: Also includes: a processor adapted to implement one or more instructions; And, a computer storage medium storing one or more instructions, wherein the one or more instructions are suitable for being loaded by the processor and executing the short-term wind power prediction method according to any one of claims 1-8.

10. A computer storage medium, characterized in that: The computer storage medium stores one or more instructions, and the one or more instructions are suitable for being loaded by a processor and executing the short-term wind power prediction method according to any one of claims 1-8.

Images