Short-time wind speed prediction method, device, equipment and medium

Abstract

The invention discloses a short-time wind speed prediction method, relates to the technical field of wind speed prediction, and is used for solving the problem that short-time wind speed accurate prediction cannot be carried out in the prior art, and the method comprises the following steps: extracting topographic feature information; acquiring an integrated wind speed sensing network, and constructing a three-dimensional wind field model according to the topographic feature information and the wind speed sensing network; performing grid division on the three-dimensional wind field model to obtain a plurality of wind speed distribution areas, and acquiring local wind speed information; inputting the topographic feature information into an ARIMA model for training, fusing the ARIMA model and an LSTM model, and constructing a short-time wind speed prediction model; and inputting the collected local wind speed information into the short-time wind speed prediction model to obtain a short-time wind speed prediction result. The invention further discloses a short-time wind speed prediction device, electronic equipment and a computer storage medium. According to the invention, through three-dimensional modeling and model fusion, an accurate result of short-time wind speed prediction is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of wind speed prediction, and in particular to a short-term wind speed prediction method, device, equipment and medium. Background Art

[0002] The operating efficiency of wind power generation is highly dependent on the accurate prediction of wind speed. Traditional wind speed prediction methods mainly use statistical or physical models. These models are often difficult to accurately predict wind speed changes in complex practical application scenarios, such as wind farms with complex terrain or uneven wind field distribution, especially on a short time scale. In addition, existing wind speed prediction technologies lack consideration of the impact of terrain factors, and lack an effective data integration system to adjust the prediction model in real time to adapt to rapidly changing meteorological conditions. Existing traditional methods have limitations in prediction accuracy and real-time performance, and cannot meet the needs of modern wind power generation, resulting in low utilization of wind power resources and low efficiency of wind turbine operation.

[0003] In summary, the existing wind speed prediction methods for wind farms are difficult to accurately predict short-term wind speeds under complex terrain and changeable meteorological conditions, and their prediction accuracy and real-time performance are insufficient. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, one of the objects of the present invention is to provide a short-term wind speed prediction method, which collects local wind speed information and then obtains a short-term wind speed prediction by fusing an ARIMA model and a LSTM model.

[0005] One of the purposes of the present invention is achieved by the following technical solution: A short-term wind speed prediction method comprises the following steps: Obtain terrain data of the target wind farm and extract terrain feature information; Acquire the positions of wind speed sensors to obtain an integrated wind speed sensor network, and construct a three-dimensional wind field model according to the terrain feature information and the wind speed sensor network; Gridding the three-dimensional wind field model to obtain a number of wind speed distribution areas, and collecting local wind speed information with each distribution area as a collection unit; The terrain feature information is input into the ARIMA model for training, and the trained ARIMA model is integrated with the LSTM model to construct a comprehensive short-term wind speed prediction model; The collected local wind speed information is input into the short-term wind speed prediction model to obtain the short-term wind speed prediction result. Through the multi-factor integrated prediction model and the three-dimensional wind field model, the accuracy and real-time performance of wind speed prediction are significantly improved, and the refined prediction of different areas and heights in the wind farm is realized, thereby optimizing the operating efficiency of wind power generation and reducing the failure rate of wind turbines.

[0006] Furthermore, the terrain data of the target wind farm is obtained, and terrain feature information is extracted, including: Obtaining terrain data of the target wind farm through a geographic information system (GIS), wherein the terrain data includes altitude, slope, terrain undulation, and vegetation coverage; The terrain feature information is obtained by extracting features from the terrain data using a digital elevation model (DEM).

[0007] Furthermore, before obtaining the wind speed sensor position to obtain the integrated wind speed sensor network, the method further includes: Acquire target fan structural parameters, wherein the structural parameters include fan height, blade length, and rotation radius; Acquire terrain features at the target wind turbine location, wherein the terrain features include slope, altitude, and surrounding terrain changes; According to the target wind turbine structural parameters and the terrain characteristics at the location, the wind speed gradient is calculated to obtain the height and location of the wind speed sensor, so as to deploy the wind speed sensor and obtain the corresponding wind speed sensor network after the deployment is completed. By analyzing the wind turbine structure and terrain characteristics, identifying the key height and location, optimizing the sensor deployment, and forming a multi-level wind speed sensor network, accurate wind speed data can be obtained in real time.

[0008] Furthermore, constructing a three-dimensional wind field model according to the terrain feature information and the wind speed sensor network includes: The terrain feature information is spatially registered with the wind speed sensor network, and a three-dimensional wind farm model for simulating the wind speed distribution in different areas and heights in the target wind farm is obtained through advanced computational fluid dynamics (CFD) and data modeling. The three-dimensional wind farm model is constructed to accurately simulate the wind speed distribution in different areas and heights in the wind farm.

[0009] Furthermore, the three-dimensional wind field model is grid-divided to obtain a plurality of wind speed distribution areas, and local wind speed information is collected with each distribution area as a collection unit, including: According to the three-dimensional wind field model, a plurality of local sensing spaces including a plurality of three-dimensional stereo units are preset; Gridding the three-dimensional wind field model using the local sensing space as a unit to obtain the wind speed distribution area; Wind speed sampling is performed in the wind speed distribution area to collect local wind speed information in different grid units. Grid division can finely grasp the dynamic changes of wind speed and direction. Furthermore, the terrain feature information is input into the ARIMA model for training, including: The wind speed sensor network continuously collects real-time high-frequency wind speed observation data and preliminary wind speed distribution of the target wind farm, and obtains meteorological data to obtain a high-frequency wind speed observation data set, a preliminary wind speed distribution set and a meteorological data set; Integrate the high-frequency wind speed observation data set, the meteorological data set and the terrain feature information into a data matrix and input the data into an ARIMA model, use the preliminary wind speed distribution set as the output of the ARIMA model, and train the model; Based on real-time data feedback, the autoregression, difference and moving average parameters of the ARIMA model are adaptively and dynamically adjusted to obtain the adjusted ARIMA model.

[0010] Furthermore, the trained ARIMA model and LSTM model are integrated to construct a comprehensive short-term wind speed prediction model, including: Collect a sample short-time wind speed distribution set, combine it with the preliminary wind speed distribution set, and construct a data time series list, wherein the sample short-time wind speed distribution set and the preliminary wind speed distribution set have the same timestamp; Using the preliminary wind speed distribution set in the data time series list as input data of the LSTM model, using the sample short-time wind speed distribution set as output data of the LSTM model, training the LSTM model, and obtaining a trained LSTM model; The output of the ARIMA model is used as the input of the LSTM model to perform model fusion and obtain a comprehensive short-term wind speed prediction model. The two models are integrated into one prediction model to analyze and fuse linear and nonlinear features, ensuring a more comprehensive analysis of wind speed and more accurate wind speed features.

[0011] A second object of the present invention is to provide a short-term wind speed prediction device, which performs short-term wind speed prediction by fusing two models.

[0012] The second object of the present invention is achieved by adopting the following technical solution: A short-term wind speed prediction device, comprising: A data acquisition module is used to acquire the terrain data of the target wind farm and extract terrain feature information; acquire the location of the wind speed sensor to obtain an integrated wind speed sensor network, and construct a three-dimensional wind farm model based on the terrain feature information and the wind speed sensor network; A data acquisition module, used for gridding the three-dimensional wind field model to obtain a plurality of wind speed distribution areas, and collecting local wind speed information with each distribution area as a collection unit; A model building module is used to input the terrain feature information into an ARIMA model for training, and to fuse the trained ARIMA model with the LSTM model to build a comprehensive short-term wind speed prediction model; The result output module is used to input the collected local wind speed information into the short-term wind speed prediction model to obtain a short-term wind speed prediction result.

[0013] The third object of the present invention is to provide an electronic device for performing one of the objects of the invention, which includes a processor, a storage medium and a computer program, wherein the computer program is stored in the storage medium, and when the computer program is executed by the processor, the above-mentioned short-term wind speed prediction method is implemented.

[0014] A fourth object of the present invention is to provide a computer-readable storage medium storing one of the objects of the invention, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned short-term wind speed prediction method is implemented.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a method for short-term wind speed prediction. By acquiring the position of a wind speed sensor, an integrated wind speed sensor network is obtained, and a three-dimensional wind field model is constructed in combination with terrain feature information, so that the wind field model is accurate to improve the prediction effect. The present invention also performs grid division on the three-dimensional wind field model of the target wind farm to obtain multiple wind speed distribution areas, so as to sample the wind speed with the wind speed distribution area as the analysis target and obtain local wind speed information. By using the terrain feature information, the ARIMA model is adaptively adjusted, and the LSTM model is integrated to construct a short-term wind speed prediction model with good prediction effect. By collecting local wind speed information as the input of the short-term wind speed prediction model, the short-term wind speed is predicted to obtain the field wind speed distribution result. The present method solves the technical problem that the wind farm wind speed prediction fails to effectively cope with the short-term wind speed prediction challenge under complex terrain and changeable meteorological conditions, resulting in insufficient prediction accuracy and real-time performance. The method can improve the accuracy and real-time response capability of short-term wind speed prediction in wind farms with complex terrain, and help wind power generation to perform subsequent wind turbine control strategy adjustment. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flow chart of a short-term wind speed prediction method according to Embodiment 1; Figure 2 is a structural block diagram of a short-term wind speed prediction device according to Embodiment 2; Figure 3 It is a structural block diagram of an electronic device of Embodiment 3. DETAILED DESCRIPTION

[0017] The present invention will be described in more detail below in conjunction with the accompanying drawings. It should be noted that the following description of the present invention with reference to the accompanying drawings is only illustrative and not restrictive. The various embodiments can be combined with each other to form other embodiments not shown in the following description.

[0018] Embodiment 1 Embodiment 1 provides a short-term wind speed prediction method, which aims to predict the short-term wind speed by constructing a three-dimensional wind field model, gridding the wind field model through a preset local perception space, sampling to obtain local wind speed information, and using the local wind speed information as a model input.

[0019] The method of this embodiment is used for short-term wind speed prediction, which is reflected in the processing of time series data and the time scale of prediction during the construction and application of the wind speed prediction model, and specifically includes: 1. High-frequency wind speed observation data collection: Real-time high-frequency wind speed observation data of the target wind farm is continuously collected through the wind speed sensor network. These data are updated frequently and can capture short-term fluctuations and changes in wind speed, reflecting the dynamic changes of wind speed in a relatively short period of time, and providing an immediate data basis for short-term wind speed prediction.

[0020] 2. Collection of sample short-term wind speed distribution sets: Collect sample short-term wind speed distribution sets with medium time resolution to analyze and predict wind speed trends and patterns in the short term, provide an overview of wind speed within a certain time window, and focus on the changing characteristics of wind speed within a shorter time range to meet the data needs of short-term wind speed forecasting.

[0021] 3. Construction of data time series list: Combined with the preliminary wind speed distribution set, a data time series list is constructed. This list records the wind speed data at each time point in detail in chronological order. By orderly organizing the wind speed data in a short period of time, it provides data support for subsequent time series-based model training and prediction, reflecting the attention to the law of short-term wind speed changes.

[0022] 4. LSTM model training and prediction: The data in the data time series table is used for LSTM model training. The LSTM model is good at processing long-term dependencies in time series data. In short-term wind speed prediction, it can learn the dynamic change law of wind speed from historical short-term wind speed data, so as to predict the wind speed in a shorter time in the future. It is particularly suitable for processing nonlinear change patterns of wind speed that may occur in a short period of time.

[0023] 5. Overall application of the short-term wind speed prediction model: The entire short-term wind speed prediction model is based on the processing and analysis of short-time scale data. It takes local wind speed information as input, predicts the short-term wind speed in different areas of the target wind farm, and generates field wind speed distribution results. These results can reflect the changes in wind speed at different heights and positions in the wind farm in a relatively short period of time, so as to timely adjust the wind turbine control strategy, adapt to the changes in short-term wind speed, and improve the operation efficiency and safety of the wind farm in a short period of time.

[0024] Based on the above description, please refer to Figure 1As shown, a short-term wind speed prediction method comprises the following steps: S1. Obtaining terrain data of the target wind farm and extracting terrain feature information; S1 includes: Obtaining terrain data of the target wind farm through a geographic information system (GIS), wherein the terrain data includes altitude, slope, terrain undulation, and vegetation coverage; The terrain feature information is obtained by extracting features from the terrain data using a digital elevation model (DEM).

[0025] A large amount of geographic and spatial data can be processed and analyzed through the Geographic Information System (GIS). Specifically, GIS can be used to accurately draw the terrain map of the wind farm, including but not limited to the terrain height, slope, and terrain undulations. These data are extracted from a multi-source terrain database and field measurement data, ensuring the accuracy and timeliness of the data. Therefore, this embodiment selects GIS to accurately obtain the terrain data of the target wind farm. The terrain data obtained above includes but is not limited to altitude, slope, terrain undulations, and vegetation coverage, which can be selected according to actual conditions and needs. These data provide basic geographical environment information for wind speed prediction, which is helpful for subsequent simulation of wind flow behavior and construction of wind farm models.

[0026] Digital elevation model DEM can provide continuous surface height data, is the important basis of analyzing terrain influence wind speed distribution.For example, DEM can reveal the micro-topography change in wind farm, such as valley and ridge, and these terrain features have remarkable influence on wind flow pattern.These elevation data are analyzed by GIS software, can extract key terrain characteristic parameters such as wind direction blocking coefficient, local terrain slope change rate, these parameters directly affect the local change of wind speed.Digital elevation model DEM can extract key terrain characteristic parameters that affect wind speed distribution, these parameters present embodiment are selected as wind direction blocking coefficient, local terrain slope change rate and valley effect, actual feature extraction can also carry out the selection of different features according to demand, these characteristic parameters can help understand the situation of wind speed under specific terrain conditions, namely above-mentioned terrain characteristic information summarizes the wind speed pattern under specific terrain conditions, for example, wind direction blocking coefficient can indicate how a certain terrain structure affects the speed and direction of path wind flow, and local terrain slope change rate reflects the potential influence of terrain change on local wind speed. Digital elevation model (DEM) processing involves the use of geographic information system (GIS) software to analyze and modify elevation data collected from satellites, aerial surveys, or ground surveys. This process includes data calibration, interpolation, and format conversion to generate a continuous elevation surface. DEM processing enables users to derive terrain features such as slope, direction, and terrain relief from these elevation data to support a variety of geospatial analyses.

[0027] S2, obtaining the position of the wind speed sensor to obtain an integrated wind speed sensor network, and constructing a three-dimensional wind field model according to the terrain feature information and the wind speed sensor network; Before obtaining the wind speed sensor position to obtain the integrated wind speed sensor network in S2, the following steps are also included: Acquire target fan structural parameters, wherein the structural parameters include fan height, blade length, and rotation radius; Acquire terrain features at the target wind turbine location, wherein the terrain features include slope, altitude, and surrounding terrain changes; The wind speed gradient is calculated according to the target wind turbine structural parameters and the terrain characteristics at the location, and the height and location of the wind speed sensor are obtained to deploy the wind speed sensor and obtain the corresponding wind speed sensor network after the deployment is completed.

[0028] For wind farms with complex terrain or uneven wind distribution, a sensor network is built and grid wind speed sampling is performed to obtain real-time wind speed data at different heights and locations above the wind farm. By constructing a three-dimensional wind farm model, refined wind speed predictions for different areas and heights can be achieved.

[0029] Existing wind speed prediction methods are difficult to achieve high-precision real-time prediction under complex terrain and changeable meteorological conditions, and have poor adaptability and insufficient spatial resolution of wind speed prediction, resulting in resource waste and wind turbine failure. Therefore, this embodiment significantly improves the accuracy and real-time performance of wind speed prediction through a multi-factor integrated prediction model and a three-dimensional wind field model, and achieves refined prediction of different areas and heights within a wind farm, thereby optimizing the operating efficiency of wind power generation and reducing the failure rate of wind turbines.

[0030] The specific structure and location information of the target wind turbine is obtained to ensure that the layout of the wind speed sensor can be optimized according to the operating characteristics and geographical environment of the wind turbine. The wind turbine location has a terrain mark, that is, each wind turbine location corresponds to the terrain feature where it is located, such as a ridge, valley or open plain. Because different terrains have a huge impact on wind speed, this step directly affects the energy efficiency and stability of the wind turbine. After obtaining these basic data, the sensors are deployed according to the wind speed gradient. The wind speed gradient refers to the rate of change of wind speed with position in a certain spatial area, which is usually related to terrain changes. For example, on a ridge, the wind speed gradient may be significantly different due to the blocking and guiding effects of the terrain. Based on this concept, multiple sensors are strategically arranged at key locations where wind speed changes are expected to be large, such as above or downwind of the ridge, to capture more accurate wind speed data. Subsequently, these sensors are wirelessly integrated into a network to form a wind speed sensing network, which can not only collect wind speed data from each sensor in real time, but also automatically adjust the data collection frequency or the working state of the sensor according to the preset algorithm to cope with sudden meteorological changes, such as the rapid approach of a storm. In this way, wind speed monitoring of the entire wind farm becomes more efficient and accurate, greatly improving the timeliness and reliability of wind speed data, and providing high-quality input data for subsequent wind speed prediction models.

[0031] Calculating the wind speed gradient means calculating the rate of change of wind speed as the terrain changes. The calculation of wind speed gradient helps identify key heights and locations where wind speed changes significantly, and can provide a basis for selecting the location of wind speed sensors. For example, if a high wind speed gradient is detected near a ridge with obvious terrain obstruction, it indicates that this is an area where wind speed changes dramatically. Specifically, the construction of a simulated wind field can be simulated based on the WAsP model in the prior art. The structural parameters and terrain characteristic parameters of the wind turbine are input into the model to analyze their impact on the local wind field. Parameters such as height and blade size will affect the average wind speed and wind speed distribution felt by the wind turbine. Use the model to simulate wind speed, and output wind speed gradient data and wind field simulation diagrams. The relationship between wind speed and height can be described by the wind speed profile formula in boundary layer meteorology. The most common one is the logarithmic wind speed profile formula, which is expressed as ;in, is the wind speed at height z, is the friction velocity, which represents the dynamic characteristics of air movement in the near-surface layer. is the von Karman constant, which is generally taken as about 0.4. is the roughness length, which indicates the degree of resistance of the ground to the wind flow. and The model can be adjusted based on actual observations. The wind speed gradient can be obtained by taking the derivative of the wind speed formula, that is, calculating the rate at which wind speed changes with height. For logarithmic wind speed profiles , this derivative expression shows the rate of change of wind speed with height. Based on the model output, the wind speed gradient at different heights is analyzed, and the critical height is determined in combination with the wind field simulation diagram. The critical height refers to the location where the wind speed change is significantly different from the wind speed within the conventional stratospheric gradient height range. After determining these critical heights and positions, multiple wind speed sensors are deployed at these locations to form a multi-level wind speed sensing network covering the target wind turbine. This hierarchical and targeted sensor layout strategy allows the system to collect wind speed data from different heights and directions in real time, so that a more accurate and comprehensive wind speed data model can be constructed. The data from these sensors are transmitted to the central processing system in real time for wind speed monitoring and prediction, ensuring that the operation of the wind farm is carried out under optimal wind speed conditions.

[0032] Through the above steps, significant optimization of wind speed monitoring and prediction in wind farms has been achieved, which not only improves the accuracy of wind speed prediction, but also provides solid technical support for risk management and efficiency optimization of wind farms. This sensor network layout scheme based on detailed terrain and structural analysis significantly enhances the adaptability and response efficiency of wind farms to complex environmental conditions.

[0033] This embodiment makes full use of terrain identification and wind speed gradient information through the designed sensor layout and intelligent integrated network, optimizes the wind speed monitoring and data collection process, and provides a solid foundation for achieving high-precision short-term wind speed prediction. This method not only improves the operating efficiency of the wind farm, but also helps to reduce wind turbine failures and maintenance costs caused by wind speed prediction errors.

[0034] In S2, a three-dimensional wind field model is constructed according to the terrain feature information and the wind speed sensor network, including: The terrain feature information is spatially registered with the wind speed sensor network, and a three-dimensional wind field model for simulating wind speed distribution conditions at different areas and heights in a target wind farm is obtained through advanced computational fluid dynamics (CFD) and data modeling.

[0035] S2's three-dimensional wind field model is not only based on terrain feature information, but also integrates wind speed data collected in real time from the wind speed sensor network. The above-mentioned advanced computational fluid dynamics CFD can accurately simulate the changes in wind speed and direction at different heights and locations. For example, the model can show how the wind speeds up on the ridge or slows down in the valley under specific wind speed and terrain conditions, thereby helping to understand the specific impact of different terrains on wind speed distribution.

[0036] Constructing a three-dimensional wind farm model is the core step of realizing the short-term wind speed prediction of the present embodiment, and its process includes detailed terrain data acquisition, feature extraction, and integration with wind speed sensor network data. First, through geographic information system GIS technology, the detailed terrain data of the target wind farm is obtained, including but not limited to altitude, slope, terrain undulation and vegetation coverage. These data provide basic geographical environment information for wind speed prediction and are the key to understanding and simulating wind flow behavior. Subsequently, these terrain data are processed by digital elevation model DEM to extract key terrain characteristic parameters affecting wind speed distribution. These parameters include wind direction blocking coefficient, local terrain slope change rate and valley effect, etc. They are directly related to the behavior of wind speed under specific terrain conditions. These behaviors, namely terrain characteristic information, summarize the wind speed mode under specific terrain conditions. For example, wind direction blocking coefficient can indicate how a certain terrain structure affects the speed and direction of the wind flow passing through, and the local terrain slope change rate reflects the potential impact of terrain change on local wind speed. Among them, digital elevation model DEM processing involves using geographic information system GIS software to automatically analyze and modify elevation data collected from satellites, aerial surveys or ground surveys. This process includes data calibration, interpolation and format conversion to generate a continuous elevation surface. DEM processing enables users to derive terrain features, such as slope, direction and terrain undulations, from these elevation data to support various geospatial analyses. Next, these key terrain feature parameters are spatially aligned with the real-time wind speed data collected by the wind speed sensor network. The spatial alignment refers to aligning the terrain data with the wind speed data in the same geographic coordinate system to ensure that the spatial position of the wind speed data is consistent with its terrain background, thereby improving the accuracy and reliability of the prediction model. Finally, based on the integrated terrain feature information and wind speed data, a detailed three-dimensional wind farm model is constructed. In this embodiment, the model is constructed using advanced computational fluid dynamics CFD and data modeling technology. Terrain and wind speed data are input into the model, and boundary conditions and initial conditions preset by the CFD model are obtained, such as wind speed inlet conditions, ground roughness and atmospheric boundary layer configuration, to simulate and display the wind speed distribution in different areas and at different heights in the target wind farm. This three-dimensional simulation can accurately predict the dynamic changes in wind speed in complex terrain, such as wind speed acceleration in valleys and wind speed reduction on ridges.

[0037] The three-dimensional wind farm model constructed by the above process not only optimizes the technical route of wind speed prediction, but also provides a solid technical foundation for risk management and efficiency optimization of wind farms. This comprehensive model based on GIS terrain data and real-time wind speed monitoring effectively improves the wind farm's response ability to environmental changes and the overall efficiency of wind power utilization.

[0038] S3, gridding the three-dimensional wind field model to obtain a number of wind speed distribution areas, and collecting local wind speed information with each distribution area as a collection unit; S3 specifically includes: According to the three-dimensional wind farm model, a plurality of local perception spaces including a plurality of three-dimensional stereo units are preset; each of the three-dimensional stereo units covers a specific area and height range within the wind farm.

[0039] Gridding the three-dimensional wind field model using the local sensing space as a unit to obtain the wind speed distribution area; Wind speed sampling is performed in the wind speed distribution area to collect local wind speed information in different grid units. Local wind speed information in different grid units is collected to capture the wind speed change characteristics in each area and obtain multiple local wind speed information.

[0040] The above-mentioned local perception space refers to a specific spatial unit area preset in the three-dimensional wind farm model for capturing wind speed changes in detail. Based on these areas, the three-dimensional wind farm model of the target wind farm is divided into several small blocks or grids, each grid representing a specific area in the wind farm. This division allows the model to focus on analyzing local wind speed changes, thereby improving the prediction accuracy and practicality of the entire model. Specifically, based on the three-dimensional wind farm model of the target wind farm, several local perception spaces are set, and these local perception spaces include multiple three-dimensional stereo units, each unit specifically defines a specific area in the wind farm and its corresponding height range. This division allows the model to capture the wind speed and wind direction changes unique to each area in more detail, thereby improving the accuracy and practicality of the prediction. Next, based on these local perception spaces, the entire three-dimensional wind farm model is gridded. This process involves dividing the wind farm into multiple wind speed distribution areas, each of which is marked in detail to facilitate subsequent data collection and analysis. The accuracy of the grid division is the key to ensuring that the model reflects the actual wind speed distribution, which allows the system to accurately locate and predict possible wind speed fluctuations in each area. Subsequently, wind speed sampling is performed in these wind speed distribution areas. Through sensor sampling and wind field model simulation, the system can collect local wind speed information within each area. This data includes not only the size of the wind speed, but also information such as wind direction and frequency of wind speed changes, so that the entire prediction model can capture the subtle wind speed change characteristics in each area. This collection and analysis of local wind speed information provides rich data support for overall wind speed prediction, greatly enhancing the real-time and accuracy of the prediction model.

[0041] Among them, grid division is applied to the three-dimensional wind farm model of the target wind farm. This embodiment uses computer-aided design (CAD) and computational fluid dynamics (CFD) software to convert the complex wind farm model into a structure composed of multiple small grids that is easier to manage and analyze. For example, the wind farm can be divided into different areas according to the height change of the terrain and the expected wind speed gradient. Each area includes several grids, which can represent the wind speed characteristics of its corresponding area. For example, the top of the mountain and the valley may be divided into different grids because the wind speed characteristics of the two are usually significantly different. Then, wind speed sampling is performed with these wind speed distribution areas as the analysis target. In each grid, wind speed data is collected by wind speed sensors arranged at key locations. This local wind speed information is collected in real time and can capture rapid changes in wind speed caused by terrain, weather conditions or other factors. In addition, the collected data is transmitted back to the central processing system in real time, and the data analysis software therein analyzes the data and identifies key wind speed changes that may affect the performance of the wind turbine.

[0042] Through this gridded wind farm model and targeted wind speed sampling strategy, not only can the wind speed be analyzed and predicted at a finer spatial scale, but also the operation of the wind farm can be managed and optimized more effectively. This method provides a technical means to accurately adjust the wind turbine control strategy and optimize the wind farm layout, significantly improving the efficiency of wind power generation and the operational stability of the wind turbine. The above method can realize the accurate monitoring and prediction of the wind speed distribution in the wind farm, providing strong technical support for the operation management and risk assessment of the wind farm. This method not only improves the efficiency of wind power generation, but also helps to optimize the adjustment and maintenance strategy of the wind turbine, thereby maximizing the utilization of wind power generation.

[0043] S4, inputting the terrain feature information into the ARIMA model for training, and fusing the trained ARIMA model with the LSTM model to construct a comprehensive short-term wind speed prediction model; In S4, the terrain feature information is input into the ARIMA model for training, including: The wind speed sensor network continuously collects real-time high-frequency wind speed observation data and preliminary wind speed distribution of the target wind farm, and obtains meteorological data to obtain a high-frequency wind speed observation data set, a preliminary wind speed distribution set and a meteorological data set; Integrate the high-frequency wind speed observation data set, the meteorological data set and the terrain feature information into a data matrix and input the data into an ARIMA model, use the preliminary wind speed distribution set as the output of the ARIMA model, and train the model; Based on real-time data feedback, the autoregression, difference and moving average parameters of the ARIMA model are adaptively and dynamically adjusted to obtain the adjusted ARIMA model.

[0044] The above-mentioned real-time collection of high-frequency wind speed observation data and meteorological data, and combining these data with terrain feature information, can adjust and optimize the short-term wind speed prediction model. Specifically, wind speed data can be collected in real time through a network of high-frequency wind speed sensors installed at key locations in the wind farm. These sensors can provide wind speed measurements multiple times per second, ensuring high frequency and high accuracy of the data, thereby capturing instant changes in wind speed. At the same time, meteorological data including wind direction, temperature, air pressure, etc. are obtained in real time through an interface with the meteorological forecasting system. These data are of great reference value for predicting wind speed because they directly affect changes in wind force and fluctuations in wind speed. In addition, terrain feature information obtained through GIS technology, such as altitude, slope, etc., is also integrated into the prediction model. These terrain data are crucial to understanding the behavior patterns of wind speed under specific geographical conditions.

[0045] Exemplarily, one of the keys to achieve high-precision wind speed forecasting is through the integration of real-time high-frequency wind speed observation data and meteorological data, as well as the adaptive adjustment of the ARIMA model. First, through the wind speed sensor network installed in the target wind farm, the system continuously collects real-time high-frequency wind speed observation data and preliminary wind speed distribution information, which are uploaded in real time through the sensor network to ensure the timeliness and accuracy of the information. Among them, the high-frequency wind speed observation data set mainly includes wind speed, wind direction and other data collected from the wind speed sensor. These data are characterized by high update frequency and can capture short-term fluctuations and changes in wind speed; the preliminary wind speed distribution set is a data set obtained by statistical methods or preliminary model processing based on the observed data, including wind speed distribution at different locations, such as average wind speed, etc., in order to provide a basis for further analysis and model training. At the same time, relevant meteorological data, including wind speed, wind direction, temperature and air pressure, are obtained from the meteorological forecast system to form a comprehensive meteorological data set. These data are crucial to understanding and predicting changes in wind speed because they directly affect the behavior and distribution of wind speed. Next, these high-frequency wind speed observation data sets, meteorological data sets and terrain feature information of the target wind farm are integrated. This step involves data preprocessing, such as data cleaning and normalization, to ensure the quality and applicability of the data. After that, they are integrated into a data matrix based on the time series, and all data variables are integrated into a single data matrix, where each row represents a time point and each column represents a variable, such as wind speed, air pressure and corresponding terrain features at a time point. The integrated data matrix is ​​used as the input of the ARIMA (autoregressive integrated moving average model). The model will try to predict the wind speed based on the autocorrelation in the time series and the association with meteorological and terrain variables, and the preliminary wind speed distribution set composed of the preliminary wind speed distribution information is used as the output of the model, that is, the model will try to predict the wind speed distribution in a certain time period in the future. In this way, the ARIMA model can more accurately learn and simulate the time series characteristics of wind speed data, thereby improving the accuracy of the prediction. Subsequently, the key parameters of the ARIMA model, including autoregressive (AR) parameters, difference (I) operations, and moving average (MA) parameters, are dynamically adjusted based on real-time data feedback, enabling the model to adapt to changing environmental conditions and actual observation data. For example, if the wind speed predicted by the model deviates from the actual observed value, the system will automatically adjust the parameters of the ARIMA model to reduce the prediction error and enhance the stability and reliability of the model.

[0046] Through the above steps, not only the accuracy of wind speed prediction is improved, but also the wind farm's ability to respond to meteorological changes is enhanced. This data-driven model adjustment method provides strong technical support for wind farm operation management and risk assessment, and optimizes the overall efficiency and safety of wind power generation.

[0047] The above-mentioned ARIMA model and LSTM model are integrated to build a comprehensive short-term wind speed prediction model, including: Collect a sample short-time wind speed distribution set, combine it with the preliminary wind speed distribution set, and construct a data time series list, wherein the sample short-time wind speed distribution set and the preliminary wind speed distribution set have the same timestamp; Using the preliminary wind speed distribution set in the data time series list as input data of the LSTM model, using the sample short-time wind speed distribution set as output data of the LSTM model, training the LSTM model, and obtaining a trained LSTM model; The output of the ARIMA model is used as the input of the LSTM model to perform model fusion and obtain a comprehensive short-term wind speed prediction model.

[0048] The ARIMA model is good at processing and predicting linear relationships in data, while LSTM is superior to processing and memorizing long-term dependencies in time series data, and is particularly suitable for the identification and prediction of nonlinear patterns. By fusing these two models together, their respective advantages can be used to provide a more accurate and reliable wind speed forecast. This fusion strategy makes full use of the ability of the LSTM model in processing long-term dependencies in time series data, combined with the advantages of the ARIMA model in processing linear time series data, to build a comprehensive short-term wind speed forecast model. In addition, the model of this embodiment also includes an adaptive adjustment mechanism that can automatically adjust model parameters according to real-time data to respond to rapid changes in environmental and meteorological conditions. For example, if the monitored wind speed suddenly increases, the model will automatically adjust its parameters to more accurately predict the duration and impact of this change. This adaptive adjustment is achieved through a data time series list, in which the data at each time point is used to continuously update and train the model to ensure the real-time and accuracy of the prediction.

[0049] Specifically, first, the system continuously collects sample short-term wind speed distribution sets through the wind speed sensor network. Different from high-frequency wind speed observation data collection, it has medium time resolution and is used to analyze and predict wind speed trends and patterns in the short term, providing an overview of wind speed within a certain time window. Subsequently, combined with the previously obtained preliminary wind speed distribution set, a detailed data time series list is constructed. This list not only records the wind speed data at each time point, but also arranges it in chronological order, providing a basis for subsequent data processing and analysis. Ensure that the data in the two wind speed distribution sets have the same timestamp. If the collection frequency of the two is different, resampling or interpolation may be required to match the time points. Combine the measured data at each time point in the sample short-term wind speed distribution set with the statistical data at the corresponding time point in the preliminary wind speed distribution set. The construction of time series is key to understanding and predicting wind speed trends because it can show the law of wind speed changes over time. Next, the preliminary wind speed distribution sets in these data time series lists are used as input data for the LSTM model, and the sample short-term wind speed distribution set is used as the output data of the model. During the training process of the LSTM model, by continuously adjusting and optimizing the network parameters, the model learns how to capture the dynamic changes of wind speed from historical data. In this process, the powerful memory function of LSTM enables it to effectively process and predict nonlinear patterns in wind speed data, which is particularly important for dynamic and complex wind farm environments. Finally, the trained LSTM model is fused with the previously adjusted and optimized ARIMA model. The specific fusion process refers to the output of the ARIMA model being used as the input of the LSTM model to form a comprehensive short-term wind speed prediction model. This model combines the linear prediction ability of ARIMA and the nonlinear processing ability of LSTM, and can more comprehensively and accurately predict short-term wind speed changes, especially under sudden changes in wind speed or complex meteorological conditions.

[0050] Through the above steps, not only can accurate short-term wind speed prediction be achieved, but the prediction model can also be dynamically adjusted to adapt to complex and changeable environmental conditions, providing strong technical support for the operation and management of wind farms. This method of comprehensive utilization of high-frequency wind speed data, meteorological information and terrain characteristics has significantly improved the performance and application scope of the wind speed prediction model.

[0051] S5. Input the collected local wind speed information into the short-term wind speed prediction model to obtain a short-term wind speed prediction result.

[0052] S5 Specifically, after constructing and optimizing the short-term wind speed prediction model, multiple local wind speed information is used as input data, and the local wind speed information collected from multiple locations and heights is input into the short-term wind speed prediction model to accurately predict the wind speed in a specific area, and obtain detailed field wind speed distribution results from it. These results include not only the wind speed size, but also the specific wind direction and height changes, which provide necessary data support for wind turbine control. That is, the local wind speed data collected from each key grid area is input into the prediction model. These local data reflect the wind speed conditions at different locations and heights in the wind farm, providing the necessary geographical and meteorological context for the prediction model. Then, the prediction model calculates the wind speed distribution of the entire wind farm based on the input local wind speed information, using the algorithm previously constructed by the fusion of ARIMA and LSTM technology. This prediction involves not only the size of the wind speed, but also the direction of the wind speed and possible change trends, which is crucial for the operation and management of the wind farm. For example, the model can predict abnormal changes in wind speed on specific terrain such as valleys or ridges, thereby guiding operation and maintenance personnel to make corresponding adjustment measures in advance. The resulting field wind speed distribution is then used to generate a control parameter deviation data set, which includes the adjustment measures that each wind turbine should take to cope with the predicted wind speed changes. Control parameter deviation refers to the difference between the actual operating parameters of the wind turbine and the ideal state, that is, the difference between the real-time wind speed and the wind turbine's reference wind speed, which is the ideal or common wind speed considered when the wind turbine is designed. Types include pitch angle, speed, and power output. By accurately measuring these deviations, a control parameter deviation data set is formed, which is the key basis for adjusting the wind turbine control strategy. For example, if the predicted wind speed in a certain area exceeds the design standard of the wind turbine, the control strategy may automatically adjust the pitch angle of the wind turbine to reduce the potential damage to the wind speed to the wind turbine. In this way, the operation strategy of each wind turbine can be dynamically adjusted according to the actual wind speed conditions, maximizing the utilization efficiency of wind energy while minimizing the risk of mechanical wear and failure caused by wind speed changes. This comprehensive use of high-precision wind speed prediction models and refined wind turbine control strategies has significantly improved the safety, reliability and economic benefits of wind power generation.

[0053] In summary, the method of this embodiment provides a short-term wind speed prediction method, which can solve the problem that wind farm wind speed prediction cannot effectively cope with short-term wind speed prediction under complex terrain and changeable meteorological conditions. The prediction accuracy and real-time performance are high, achieving the technical effect of improving the accuracy of short-term wind speed prediction and real-time response capability in complex terrain wind farms.

[0054] Embodiment 2 Embodiment 2 discloses a device corresponding to the short-term wind speed prediction method of the above embodiment, which is a virtual device structure of the above embodiment. Please refer to Figure 2 As shown, including: The data acquisition module 210 is used to acquire the terrain data of the target wind farm and extract the terrain feature information; acquire the location of the wind speed sensor to obtain an integrated wind speed sensor network, and construct a three-dimensional wind farm model based on the terrain feature information and the wind speed sensor network; The data acquisition module 220 is used to grid the three-dimensional wind field model to obtain a plurality of wind speed distribution areas, and collect local wind speed information with each distribution area as a collection unit; A model building module 230 is used to input the terrain feature information into an ARIMA model for training, and fuse the trained ARIMA model with the LSTM model to build a comprehensive short-term wind speed prediction model; The result output module 240 is used to input the collected local wind speed information into the short-term wind speed prediction model to obtain a short-term wind speed prediction result.

[0055] Embodiment 3 Figure 3 A schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention is shown in FIG. Figure 3 As shown, the electronic device includes a processor 310, a memory 320, an input device 330 and an output device 340; the number of processors 310 in the computer device can be one or more. Figure 3 A processor 310 is taken as an example; the processor 310, the memory 320, the input device 330 and the output device 340 in the electronic device can be connected via a bus or other means. Figure 3 The example of connecting through bus is taken in the following.

[0056] The memory 320, as a computer-readable storage medium, can be used to store software programs, computer executable programs and modules, such as program instructions / modules corresponding to the short-term wind speed prediction method in the embodiment of the present invention. The processor 310 executes various functional applications and data processing of the electronic device by running the software programs, instructions and modules stored in the memory 320, that is, the short-term wind speed prediction method of the above-mentioned embodiment 1 is implemented.

[0057] The memory 320 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and at least one application required for a function; the data storage area may store data created according to the use of the terminal, etc. In addition, the memory 320 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 320 may further include a memory remotely arranged relative to the processor 310, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0058] The input device 330 may be used to receive input user identity information, wind farm data, etc. The output device 340 may include a display device such as a display screen.

[0059] Embodiment 4 Embodiment 4 of the present invention further provides a storage medium including computer executable instructions, which can be used for a computer to execute a short-term wind speed prediction method, the method comprising: Obtain terrain data of the target wind farm and extract terrain feature information; Acquire the positions of wind speed sensors to obtain an integrated wind speed sensor network, and construct a three-dimensional wind field model according to the terrain feature information and the wind speed sensor network; Gridding the three-dimensional wind field model to obtain a number of wind speed distribution areas, and collecting local wind speed information with each distribution area as a collection unit; The terrain feature information is input into the ARIMA model for training, and the trained ARIMA model is integrated with the LSTM model to construct a comprehensive short-term wind speed prediction model; The collected local wind speed information is input into the short-term wind speed prediction model to obtain a short-term wind speed prediction result.

[0060] Of course, the storage medium containing computer executable instructions provided in an embodiment of the present invention is not limited to the method operations described above, and the computer executable instructions can also execute related operations in the short-term wind speed prediction method provided in any embodiment of the present invention.

[0061] Through the above description of the implementation methods, the technicians in the relevant field can clearly understand that the present invention can be implemented by means of software and necessary general hardware, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for an electronic device (which can be a mobile phone, a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0062] It is worth noting that in the above-mentioned embodiment of the device based on the short-term wind speed prediction method, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0063] For those skilled in the art, various other corresponding changes and deformations can be made according to the technical solutions and concepts described above, and all of these changes and deformations should fall within the protection scope of the claims of the present invention.

Claims

1. A short-term wind speed prediction method, characterized in that: The following steps are involved: Obtain terrain data of the target wind farm and extract terrain feature information; Acquire the positions of wind speed sensors to obtain an integrated wind speed sensor network, and construct a three-dimensional wind field model according to the terrain feature information and the wind speed sensor network; Gridding the three-dimensional wind field model to obtain a number of wind speed distribution areas, and collecting local wind speed information with each distribution area as a collection unit; The terrain feature information is input into the ARIMA model for training, and the trained ARIMA model is integrated with the LSTM model to construct a comprehensive short-term wind speed prediction model; The collected local wind speed information is input into the short-term wind speed prediction model to obtain a short-term wind speed prediction result.

2. The short-term wind speed prediction method according to claim 1, characterized in that: Obtain the terrain data of the target wind farm and extract terrain feature information, including: Obtaining terrain data of the target wind farm through a geographic information system (GIS), wherein the terrain data includes altitude, slope, terrain undulation, and vegetation coverage; The terrain feature information is obtained by extracting features from the terrain data using a digital elevation model (DEM).

3. The short-term wind speed prediction method according to claim 1, characterized in that: Before obtaining the wind speed sensor position to obtain the integrated wind speed sensor network, it also includes: Acquire target fan structural parameters, wherein the structural parameters include fan height, blade length, and rotation radius; Acquire terrain features at the target wind turbine location, wherein the terrain features include slope, altitude, and surrounding terrain changes; The wind speed gradient is calculated according to the target wind turbine structural parameters and the terrain characteristics at the location, and the height and location of the wind speed sensor are obtained to deploy the wind speed sensor and obtain the corresponding wind speed sensor network after the deployment is completed.

4. The short-term wind speed prediction method according to claim 1 or 2, characterized in that: Constructing a three-dimensional wind field model according to the terrain feature information and the wind speed sensor network includes: The terrain feature information is spatially registered with the wind speed sensor network, and a three-dimensional wind field model for simulating wind speed distribution conditions at different areas and heights in a target wind farm is obtained through advanced computational fluid dynamics (CFD) and data modeling.

5. The short-term wind speed prediction method according to claim 1, characterized in that: The three-dimensional wind field model is grid-divided to obtain a plurality of wind speed distribution areas, and local wind speed information is collected with each distribution area as a collection unit, including: According to the three-dimensional wind field model, a plurality of local sensing spaces including a plurality of three-dimensional stereo units are preset; Gridding the three-dimensional wind field model using the local sensing space as a unit to obtain the wind speed distribution area; Wind speed sampling is performed within the wind speed distribution area to collect local wind speed information within different grid units.

6. The short-term wind speed prediction method according to claim 1, characterized in that: The terrain feature information is input into the ARIMA model for training, including: The wind speed sensor network continuously collects real-time high-frequency wind speed observation data and preliminary wind speed distribution of the target wind farm, and obtains meteorological data to obtain a high-frequency wind speed observation data set, a preliminary wind speed distribution set and a meteorological data set; Integrate the high-frequency wind speed observation data set, the meteorological data set and the terrain feature information into a data matrix and input the data into an ARIMA model, use the preliminary wind speed distribution set as the output of the ARIMA model, and train the model; Based on real-time data feedback, the autoregression, difference and moving average parameters of the ARIMA model are adaptively and dynamically adjusted to obtain the adjusted ARIMA model.

7. The short-term wind speed prediction method according to claim 1 or 6, characterized in that: The trained ARIMA model and LSTM model are integrated to build a comprehensive short-term wind speed prediction model, including: Collect a sample short-time wind speed distribution set, and combine it with the preliminary wind speed distribution set to construct a data time series list, wherein the sample short-time wind speed distribution set and the preliminary wind speed distribution set have the same timestamp; Using the preliminary wind speed distribution set in the data time series list as input data of the LSTM model, using the sample short-time wind speed distribution set as output data of the LSTM model, training the LSTM model, and obtaining a trained LSTM model; The output of the ARIMA model is used as the input of the LSTM model to perform model fusion and obtain a comprehensive short-term wind speed prediction model.

8. A short-term wind speed prediction device, characterized in that: It includes: A data acquisition module is used to acquire terrain data of a target wind farm and extract terrain feature information; Acquire the positions of wind speed sensors to obtain an integrated wind speed sensor network, and construct a three-dimensional wind field model according to the terrain feature information and the wind speed sensor network; A data acquisition module, used for gridding the three-dimensional wind field model to obtain a plurality of wind speed distribution areas, and collecting local wind speed information with each distribution area as a collection unit; A model building module is used to input the terrain feature information into an ARIMA model for training, and to fuse the trained ARIMA model with the LSTM model to build a comprehensive short-term wind speed prediction model; The result output module is used to input the collected local wind speed information into the short-term wind speed prediction model to obtain a short-term wind speed prediction result.

9. An electronic device comprising a processor, a storage medium and a computer program, wherein the computer program is stored in the storage medium, wherein: When the computer program is executed by a processor, the short-term wind speed prediction method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the short-term wind speed prediction method according to any one of claims 1 to 7 is implemented.

Images