Method, apparatus, device and medium for calculating fusion wind speed based on transfer learning

Abstract

The present disclosure relates to a method, device, equipment and medium for calculating fused wind speed based on transfer learning. Among them, the method for calculating fused wind speed based on transfer learning includes: determining a reference anemometer tower according to wind speed correlation among multiple anemometer towers; obtaining first true observed wind speed data and first simulated wind speed data of the reference anemometer tower, and obtaining second simulated wind speed data of a target point; performing sequential wind speed calculation on the first true observed wind speed data, the first simulated wind speed data and the second simulated wind speed data to obtain first sequential wind speed data of the target point; and performing model calculation on the first sequential wind speed data through a transfer learning neural network model to obtain first fused sequential wind speed data of the target point. According to the embodiments of the present disclosure, the method of performing fused calculation through multiple data can effectively improve the accuracy of wind speed calculation.

Description

Technical Field

[0001] The present disclosure relates to the technical field of wind speed data simulation, and particularly to a fusion wind speed calculation method, device, equipment and medium based on transfer learning. Background Art

[0002] The wind speed at each point in a wind farm, especially at the point where a wind turbine is to be built, is an important parameter for measuring the accuracy of a wind resource assessment software.

[0003] In related technologies, the computational fluid dynamics (CFD) method is usually used to calculate the spatial steady-state flow field distribution under different incoming wind direction conditions, and then based on the time series / frequency domain wind measurement data of a reference anemometer tower, the annual time series / frequency domain wind speed distribution at each point in the wind farm is calculated. In this process, due to the influence of factors such as terrain processing, turbulence model, boundary conditions, and numerical solution, the wind speed error of CFD numerical simulation is inevitably caused, resulting in a low accuracy of the calculated wind speed. Summary of the Invention

[0004] To solve the above technical problems or at least partially solve the above technical problems, the present disclosure provides a fusion wind speed calculation method, device, equipment and medium based on transfer learning.

[0005] In a first aspect, the present disclosure provides a fusion wind speed calculation method based on transfer learning, including:

[0006] Determining a reference anemometer tower according to the wind speed correlation among multiple anemometer towers;

[0007] Obtaining the first true observed wind speed data and the first simulated wind speed data of the reference anemometer tower, and obtaining the second simulated wind speed data of the target point;

[0008] Performing time series wind speed calculation on the first true observed wind speed data, the first simulated wind speed data and the second simulated wind speed data to obtain the first time series wind speed data of the target point;

[0009] Performing model calculation on the first time series wind speed data through a transfer learning neural network model to obtain the first fusion time series wind speed data of the target point.

[0010] In a second aspect, the present disclosure provides a fusion wind speed calculation device based on transfer learning, including:

[0011] A data determination module for determining a reference anemometer tower according to the wind speed correlation among multiple anemometer towers;

[0012] A first acquisition module for obtaining the first true observed wind speed data and the first simulated wind speed data of the reference anemometer tower, and obtaining the second simulated wind speed data of the target point;

[0013] The first calculation module is configured to perform sequential wind speed calculation on the first real observed wind speed data, the first simulated wind speed data, and the second simulated wind speed data to obtain the first sequential wind speed data of the target point.

[0014] The second calculation module is configured to perform model calculation on the first sequential wind speed data through a transfer learning neural network model to obtain the first fused sequential wind speed data of the target point.

[0015] In a third aspect, the present disclosure provides a fused wind speed calculation device based on transfer learning, including:

[0016] A processor;

[0017] A memory for storing executable instructions;

[0018] Wherein, the processor is configured to read the executable instructions from the memory and execute the executable instructions to implement the fused wind speed calculation method based on transfer learning in the first aspect.

[0019] In a fourth aspect, the present disclosure provides a computer-readable storage medium storing a computer program, which when executed by a processor, causes the processor to implement the fused wind speed calculation method based on transfer learning in the first aspect.

[0020] The technical solutions provided in the embodiments of the present disclosure have the following advantages compared with the prior art:

[0021] The fused wind speed calculation method, device, equipment, and medium based on transfer learning in the embodiments of the present disclosure can determine a reference anemometer tower according to the wind speed correlation among multiple anemometer towers, then obtain the first real observed wind speed data and the first simulated wind speed data of the reference anemometer tower, and obtain the second simulated wind speed data of the target point. Then, perform sequential wind speed calculation on the first real observed wind speed data, the first simulated wind speed data, and the second simulated wind speed data to obtain the first sequential wind speed data of the target point. Finally, perform model calculation on the first sequential wind speed data through a transfer learning neural network model to obtain the first fused sequential wind speed data of the target point. Thus, the first fused sequential wind speed data of the target point can be obtained through the fusion calculation of the first real observed wind speed data, the first simulated wind speed data, the second simulated wind speed data of the target point, and the transfer learning neural network model. Therefore, the wind speed calculation accuracy can be effectively improved through the method of fusion calculation with multiple data. Description of the Drawings

[0022] In combination with the accompanying drawings and with reference to the following specific embodiments, the above and other features, advantages, and aspects of the various embodiments of the present disclosure will become more apparent. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the original elements and elements are not necessarily drawn to scale.

[0023] Figure 1 It is a schematic flowchart of a method for calculating fused wind speed based on transfer learning provided by an embodiment of the present disclosure;

[0024] Figure 2 It is a heat map of the wind speed correlation provided by an embodiment of the present disclosure;

[0025] Figure 3 It is a schematic structural diagram of a device for calculating fused wind speed based on transfer learning provided by an embodiment of the present disclosure;

[0026] Figure 4 It is a schematic structural diagram of a device for calculating fused wind speed based on transfer learning provided by an embodiment of the present disclosure. Specific Embodiments

[0027] The embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although some embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. On the contrary, these embodiments are provided to more thoroughly and completely understand the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not used to limit the protection scope of the present disclosure.

[0028] It should be understood that the various steps recited in the method embodiments of the present disclosure can be executed in a different order and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this regard.

[0029] The term "including" and its variations used herein are open-ended, that is, "including but not limited to". The term "based on" is "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.

[0030] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules, or units, and are not used to limit the order of the functions executed by these devices, modules, or units or their interdependent relationships.

[0031] It should be noted that the modifiers "one" and "multiple" mentioned in this disclosure are illustrative rather than restrictive. Those skilled in the art should understand that unless clearly specified otherwise in the context, it should be understood as "one or more".

[0032] The names of the messages or information exchanged between multiple devices in the embodiments of this disclosure are only for illustrative purposes and are not used to limit the scope of these messages or information.

[0033] The wind speed at each point in the wind farm, especially at the points where wind turbines are planned to be built, is an important parameter for measuring the accuracy of wind resource assessment software. In wind resource assessment software, the CFD method is usually used to calculate the spatial steady-state flow field distribution under different incoming wind direction conditions, and then based on the time series / frequency domain wind measurement data of the reference meteorological tower, the annual time series / frequency domain wind speed distribution at each point in the wind farm is calculated. During this process, due to the influence of factors such as terrain processing, turbulence model, boundary conditions, and numerical solution, it is inevitable to cause wind speed errors in the CFD numerical simulation, resulting in relatively low wind speed calculation accuracy. If there is more than one meteorological tower in the wind farm, the wind measurement data of these towers can be fused with the wind speed calculated by the CFD numerical simulation, etc., to effectively improve the wind speed calculation accuracy of the wind resource assessment software.

[0034] In view of the actual situation that the number of meteorological towers for the wind farm to be developed and constructed is limited and the wind speed distribution at each potential turbine point cannot be obtained through measurement, based on the true observed wind speeds of several meteorological towers, the whole-field CFD numerical simulation wind speeds, and the mesoscale simulation data of the wind farm, an intelligent algorithm is used to realize the fusion of wind speed data, and the time series / frequency domain wind speeds at several points are extrapolated to the time series / frequency domain wind speeds at multiple target points in the wind farm, so as to improve the wind speed calculation accuracy of the wind resource assessment software and provide basic data for subsequent calculations such as the annual power generation of the target layout points, serving the micro-siting work of the wind farm.

[0035] Academia and industry at home and abroad mainly focus on research on high-precision mesoscale numerical models and assimilation methods based on a large amount of local observational data. The Technical University of Denmark has proposed a method for offshore wind resource assessment by fusing lidar and mesoscale data, proving that multi-source data fusion can improve the assessment accuracy.

[0036] At present, there is no research on fusing the true observed wind speeds of meteorological towers with the whole-field CFD numerical simulation wind speeds and the mesoscale climate data of the wind farm to study the wind speed data at the target points of the wind farm to be developed and constructed. In view of the relatively scarce resources of true observed data for the wind farm to be developed and constructed and the inability of existing technologies to perform high-precision resource assessment, etc., the research on wind speed data fusion based on multi-source data for transfer learning is a current frontier hot topic in this field.

[0037] To solve the above problems, embodiments of the present disclosure provide a method, apparatus, device, and medium for calculating fused wind speed based on transfer learning.

[0038] First, the method for calculating fused wind speed based on transfer learning provided by the embodiments of the present disclosure will be described below in conjunction with Figure 1-2 the following content.

[0039] In the embodiments of the present disclosure, the method for calculating fused wind speed based on transfer learning can be executed by an electronic device. Among them, the electronic device may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Tablet Computers), PMPs (Portable Multimedia Players), vehicle terminals (such as vehicle navigation terminals), wearable devices, etc., and fixed terminals such as digital TVs, desktop computers, smart home devices, etc.

[0040] Figure 1 FIG. shows a schematic flowchart of a method for calculating fused wind speed based on transfer learning provided by an embodiment of the present disclosure.

[0041] As Figure 1 shown, the method for calculating fused wind speed based on transfer learning may include the following steps.

[0042] S110. Determine a reference anemometer tower among multiple anemometer towers according to the wind speed correlation.

[0043] In the embodiments of the present disclosure, the electronic device may determine a reference anemometer tower among multiple anemometer towers through the wind speed correlation.

[0044] Optionally, the anemometer tower may be a device for measuring the wind speed at a specific point.

[0045] Optionally, the wind speed correlation may be used to represent the correlation of wind speeds between each anemometer tower.

[0046] Optionally, the reference anemometer tower may be an anemometer tower used as a calculation reference for other anemometer towers.

[0047] Specifically, the electronic device may determine the wind speed correlation between each anemometer tower, thereby determining a reference anemometer tower from multiple anemometer towers.

[0048] Optionally, S110 may specifically include: obtaining the time-series wind speed sequences of multiple anemometer towers at a first preset height, and calculating the wind speed correlation of multiple anemometer towers; determining the anemometer tower with the largest wind speed correlation as the reference anemometer tower.

[0049] In the embodiments of the present disclosure, the electronic device may obtain the time-series wind speed sequences of multiple anemometer towers at a first preset height, and calculate the wind speed correlation of multiple anemometer towers.

[0050] Optionally, the first preset height may be a preset height. For example, the first preset height may be the hub height of a wind turbine, such as 90m, 100m, etc., which is not limited here.

[0051] Optionally, the time series wind speed sequence may be a time series wind speed data sequence of multiple anemometer towers within a preset time. Among them, the preset time may be greater than or equal to 1 year. For example, the time series wind speed sequence may include all time series wind speed data obtained from multiple anemometer towers at a data interval of 10 minutes within 1 year.

[0052] Specifically, the electronic device may obtain the time series wind speed sequences corresponding to multiple anemometer towers within a preset time, and calculate the wind speed correlation between each anemometer tower through the time series wind speed sequence. Among them, the method for calculating the wind speed correlation may be a known calculation method, which is not limited here.

[0053] Furthermore, after obtaining the wind speed correlations of multiple anemometer towers, the electronic device may determine the anemometer tower with the largest wind speed correlation as the reference anemometer tower.

[0054] In the embodiments of the present disclosure, after calculating the wind speed correlations between each anemometer tower, the electronic device may determine the anemometer tower with the largest wind speed correlation with other anemometer towers, and use this anemometer tower as the reference anemometer tower.

[0055] Figure 2 Shows a thermal diagram of a wind speed correlation provided by an embodiment of the present disclosure.

[0056] As Figure 2 shown, it includes the wind speed correlations between 5 anemometer towers. The darker the color, the greater the wind speed correlation, and the larger the numerical identifier, the greater the wind speed correlation. Among them, the electronic device may determine an anemometer tower with the largest wind speed correlation with other anemometer towers, such as anemometer tower #2, and use this anemometer tower #2 as the reference anemometer tower.

[0057] Thus, the reference anemometer tower can be determined through the wind speed correlation, so that in subsequent calculation processes, calculations can be performed based on this reference anemometer tower, which can improve the calculation accuracy.

[0058] S120. Obtain the first true observed wind speed data and the first simulated wind speed data of the reference anemometer tower, and obtain the second simulated wind speed data of the target point.

[0059] In the embodiments of the present disclosure, after determining the reference anemometer tower, the electronic device may obtain the first true observed wind speed data and the first simulated wind speed data of this reference anemometer tower, and obtain the second simulated wind speed data of the target point.

[0060] Optionally, the first true observed wind speed data may be the wind speed truly measured at the reference anemometer tower.

[0061] Optionally, the first simulated wind speed data may be the simulated wind speed calculated by the CFD method at the reference anemometer tower.

[0062] Optionally, the target point may be the point where the wind speed needs to be calculated.

[0063] Optionally, the second simulated wind speed data may be the simulated wind speed calculated by the CFD method at the target point.

[0064] Specifically, after determining the reference anemometer tower, the electronic device may obtain the first true observed wind speed data and the first simulated wind speed data of the reference anemometer tower, and obtain the second simulated wind speed data of the target point.

[0065] S130. Perform a temporal wind speed calculation on the first true observed wind speed data, the first simulated wind speed data, and the second simulated wind speed data to obtain the first temporal wind speed data of the target point.

[0066] In the embodiments of the present disclosure, after the electronic device obtains the first true observed wind speed data, the first simulated wind speed data, and the second simulated wind speed data, it may perform a temporal wind speed calculation on the first true observed wind speed data, the first simulated wind speed data, and the second simulated wind speed data to obtain the first temporal wind speed data of the target point.

[0067] Optionally, S130 may specifically include: substituting the first true observed wind speed data, the first simulated wind speed data, and the second simulated wind speed data into the temporal wind speed calculation formula to obtain the first temporal wind speed data of the target point.

[0068] Optionally, the temporal wind speed calculation may be performed by the temporal wind speed calculation formula.

[0069] Optionally, the temporal wind speed calculation formula may be:

[0070]

[0071] Where, taking the point where the reference anemometer tower is located as A and the target point as B, V A may be used to represent the first true observed wind speed data of the reference anemometer tower, that is, point A, V CFDA may be used to represent the first simulated wind speed data of the reference anemometer tower, that is, point A, V CFDB may be used to represent the second simulated wind speed data of the target point, that is, point B, V CFD may be used to represent the first temporal wind speed data of the target point, that is, point B.

[0072] Optionally, the first time-series wind speed data may be the preliminary time-series wind speed data at the target location.

[0073] Specifically, the electronic device may perform time-series wind speed calculation on the first real observed wind speed data, the first simulated wind speed data, and the second simulated wind speed data through the time-series wind speed calculation formula, so as to obtain the first time-series wind speed data at the target location, that is, to obtain the preliminary time-series wind speed data.

[0074] For example, the electronic device may perform time-series wind speed calculation on the first real observed wind speed data V at point A where the reference anemometer tower is located through the time-series wind speed calculation formula. A and the first simulated wind speed data V CFDA , as well as the second simulated wind speed data V at the target location B CFDB to perform time-series wind speed calculation, so as to obtain the first time-series wind speed data V at the target location CFD .

[0075] S140. Perform model calculation on the first time-series wind speed data through the transfer learning neural network model to obtain the first fused time-series wind speed data at the target location.

[0076] In the embodiments of the present disclosure, after obtaining the first time-series wind speed data, the electronic device may perform model calculation on the first time-series wind speed data through the transfer learning neural network model to obtain the first fused time-series wind speed data at the target location.

[0077] Optionally, the transfer learning neural network model may be a pre-trained model.

[0078] Optionally, the model calculation may be to perform calculation on the first time-series wind speed data through the transfer learning neural network model.

[0079] Optionally, the first time-series wind speed data may be the final time-series wind speed data at the target location.

[0080] Specifically, the electronic device may perform model calculation on the first time-series wind speed data through the transfer learning neural network model, that is, further fuse and correct the preliminary time-series wind speed data, so as to obtain the first fused time-series wind speed data at the target location, that is, to obtain the final time-series wind speed data.

[0081] Optionally, S140 may specifically include: inputting the first time-series wind speed data into the transfer learning neural network model, so that the transfer learning neural network model performs model calculation on the first time-series wind speed data to obtain the first fused time-series wind speed data at the target location.

[0082] In an embodiment of the present disclosure, after obtaining the first time-series wind speed data, the electronic device may input the first time-series wind speed data into a transfer learning neural network model. The transfer learning neural network model may receive and perform model calculations on the first time-series wind speed data to obtain the first fused time-series wind speed data, and output the first fused time-series wind speed data. The electronic device may obtain the first fused time-series wind speed data of the target point.

[0083] Thus, in an embodiment of the present disclosure, a reference anemometer tower can be determined based on wind speed correlation among multiple anemometer towers. Then, the first true observed wind speed data and the first simulated wind speed data of the reference anemometer tower, as well as the second simulated wind speed data of the target point, are obtained. Then, time-series wind speed calculations are performed on the first true observed wind speed data, the first simulated wind speed data, and the second simulated wind speed data to obtain the first time-series wind speed data of the target point. Finally, the transfer learning neural network model performs model calculations on the first time-series wind speed data to obtain the first fused time-series wind speed data of the target point. Thus, the first fused time-series wind speed data of the target point can be obtained through fusion calculations using the first true observed wind speed data, the first simulated wind speed data, the second simulated wind speed data of the target point, and the transfer learning neural network model. Therefore, the wind speed calculation accuracy can be effectively improved by the method of performing fusion calculations using multiple data.

[0084] Optionally, before S120, the fusion wind speed calculation method based on transfer learning may further include: obtaining a simulated wind speed data table for each point through computational fluid dynamics methods. The simulated wind speed data table includes time-series wind direction label features, and the time-series wind direction label features are mesoscale wind direction data at a second preset height. The simulated wind speed data table includes the first simulated wind speed data and the second simulated wind speed data.

[0085] In an embodiment of the present disclosure, the electronic device may obtain a simulated wind speed data table for each point through computational fluid dynamics methods.

[0086] Optionally, the computational fluid dynamics (CFD) method may be used to calculate the model wind speed data for each point.

[0087] Optionally, the simulated wind speed data table may include the wind speed data at a first preset height under different wind directions for each point. Among them, the simulated wind speed data table may include time-series wind direction label features.

[0088] Optionally, the time-series wind direction label features may be mesoscale wind direction data at a second preset height for each point.

[0089] Optionally, the second preset height can be a predefined height. For example, the second preset height can be 1500m, 1600m, etc., which is not limited here.

[0090] Optionally, the mesoscale wind direction data can be data used to characterize the wind direction at the second preset height. For example, the mesoscale wind direction data can be data used to characterize wind directions such as southeast wind, northeast wind, etc. at the second preset height, which is not limited here.

[0091] Specifically, the electronic device can obtain the wind speed data at the first preset height under different wind directions at each point through the computational fluid dynamics (CFD) method, so as to obtain a simulated wind speed data table.

[0092] For example, the electronic device can calculate the numerical simulation of the wind farm flow field under 16 wind direction sectors through the CFD method, and obtain the model wind speed data at the first preset height of each point (including the target point) in the wind farm under 16 wind direction sectors under neutral conditions, forming a simulated wind speed data table. Among them, each anemometer tower can correspond to a point.

[0093] Optionally, due to the influence of factors such as terrain and landform, there are differences in the wind directions at each point in the wind farm. The wind direction of a certain anemometer tower cannot be used as the inlet wind direction boundary condition for the CFD simulation of the wind farm flow field. Therefore, it is necessary to obtain the mesoscale wind direction data at the second preset height, such as 1500m, as the time-series wind direction label feature in the simulated wind speed data table.

[0094] Optionally, the simulated wind speed data table can include the first simulated wind speed data at the point where the reference anemometer tower is located and the second simulated wind speed data at the target point.

[0095] Thus, in the embodiments of the present disclosure, a simulated wind speed data table can be obtained through the CFD method, so as to provide corresponding data for subsequent calculations and effectively improve the accuracy of wind speed calculation.

[0096] Optionally, before S140, the fusion wind speed calculation method based on transfer learning may further include: training the neural network model to be trained according to the first true observed wind speed data and the first simulated wind speed data of the reference anemometer tower, and the second true observed wind speed data and the third simulated wind speed data of the fusion anemometer tower, to obtain a trained neural network model. The fusion anemometer tower is an anemometer tower other than the reference anemometer tower among multiple anemometer towers;

[0097] Perform transfer learning processing on the trained neural network model to obtain a transfer learning neural network model.

[0098] In an embodiment of the present disclosure, the electronic device may train a neural network model to be trained based on the first true observed wind speed data and the first simulated wind speed data of the reference anemometer tower, and the second true observed wind speed data and the third simulated wind speed data of the integrated anemometer tower, to obtain a trained neural network model.

[0099] Optionally, the integrated anemometer tower may be an anemometer tower other than the reference anemometer tower among multiple anemometer towers.

[0100] For example, there are 5 anemometer towers. After the electronic device determines that anemometer tower #2 is the reference anemometer tower, it can determine that #1, #3, #4, and #5 can be integrated anemometer towers.

[0101] Optionally, the second true observed wind speed data may be the wind speed truly measured at the integrated anemometer tower.

[0102] Optionally, the third simulated wind speed data may be the simulated wind speed calculated at the integrated anemometer tower by the CFD method.

[0103] Specifically, the electronic device may obtain the first true observed wind speed data of the reference anemometer tower and the second true observed wind speed data of the integrated anemometer tower through measurement, then obtain the first simulated wind speed data in the simulated wind speed data table and the third simulated wind speed data of the integrated anemometer tower in the simulated wind speed data table, and train the neural network model to be trained with the above data to obtain a trained neural network model, that is, obtain the parameters in the trained neural network model.

[0104] Further, after the electronic device obtains the trained neural network model, it may perform transfer learning processing on the trained neural network model to obtain a transfer learning neural network model.

[0105] Optionally, the transfer learning processing may be to transfer the trained model parameters to a new model to help train the new model.

[0106] Specifically, after the electronic device obtains the trained neural network model, it may transfer the parameters in the trained neural network model to a new model for model training to obtain a transfer learning neural network model.

[0107] For example, the electronic device may perform transfer learning processing on the trained neural network models of the reference anemometer tower and the integrated anemometer tower to obtain a transfer learning neural network model corresponding to the target location. Among them, the electronic device uses an Artificial Neural Network (ANN) as the kernel network model for transfer learning, and applies the Back Propagation (BP) algorithm to train the model to establish a relatively reliable transfer learning neural network model.

[0108] Thus, in the embodiments of the present disclosure, the electronic device can obtain a relatively accurate transfer learning neural network model through transfer learning processing, thereby improving the accuracy of subsequent wind speed calculation.

[0109] Optionally, after performing transfer learning processing on the trained neural network model to obtain a transfer learning neural network model, the fusion wind speed calculation method based on transfer learning may further include: performing error analysis calculation on the transfer learning neural network model through the second time-series wind speed data and the second fused time-series wind speed data of the test wind measurement tower to obtain an error result, and the error analysis calculation is to calculate the root mean square error of the second time-series wind speed data, the second fused time-series wind speed data, and the third true observed wind speed data of the test wind measurement tower.

[0110] In the embodiments of the present disclosure, the electronic device can perform error analysis calculation on the transfer learning neural network model through the second time-series wind speed data and the second fused time-series wind speed data of the test wind measurement tower to obtain an error result.

[0111] Optionally, the test wind measurement tower can be any one of multiple fused wind measurement towers.

[0112] For example, there are 5 wind measurement towers. After the electronic device determines that the wind measurement tower #2 is the reference wind measurement tower, it can determine that #1, #3, #4, and #5 can be fused wind measurement towers, and then it can determine that the wind measurement tower #1 is the test wind measurement tower.

[0113] Optionally, the second time-series wind speed data can be the preliminary time-series wind speed data at the test wind measurement tower. Among them, the second time-series wind speed data can be obtained by calculating the time-series wind speed according to the time-series wind speed calculation formula. The specific implementation manner refers to the above, and will not be elaborated here.

[0114] Optionally, the second fused time-series wind speed data can be the final time-series wind speed data at the test wind measurement tower. Among them, the second fused time-series wind speed data can be obtained by performing model calculation through the transfer learning neural network model. The specific implementation manner refers to the above, and will not be elaborated here.

[0115] Optionally, the error analysis calculation is to calculate the root mean square error of the second time-series wind speed data, the second fused time-series wind speed data, and the third true observed wind speed data of the test wind measurement tower.

[0116] Optionally, the third true observed wind speed data can be the wind speed truly measured at the test wind measurement tower.

[0117] Optionally, the root mean square error can be the square root of the ratio of the sum of the squares of the deviations of the second time-series wind speed data and the second fused time-series wind speed data from the third true observed wind speed data to the number of observations n.

[0118] Optionally, the error analysis calculation can be performed through an error analysis calculation formula.

[0119] Optionally, the error analysis calculation formula can be:

[0120]

[0121] where \(i = 1, 2, 3,\cdots, N\), \(N\) is the length of the selected time series (resolution is 10 min), \(V\) model can represent the second time series wind speed data or the second fused time series wind speed data, and \(V\) obs can represent the third true observed wind speed data.

[0122] Specifically, the electronic device can perform error analysis calculations on the second time series wind speed data and the third true observed wind speed data, and the second fused time series wind speed data and the third true observed wind speed data respectively according to the error analysis calculation formula, so as to obtain the corresponding error results, that is, the corresponding root mean square errors.

[0123] For example, there are 5 wind measurement towers. Wind measurement tower #2 is the reference wind measurement tower, wind measurement tower #1 is the test wind measurement tower, and #3, #4, and #5 can be fused wind measurement towers. The electronic device can perform error analysis calculations on the test wind measurement tower #1 through the error analysis calculation formula to obtain the corresponding root mean square errors. For example, the root mean square error between the second time series wind speed data and the third true observed wind speed data is 1.03 m / s, and the root mean square error between the second fused time series wind speed data and the third true observed wind speed data is 0.81 m / s. Thus, it can be seen that the root mean square error of the second fused time series wind speed data calculated by the model through the transfer learning neural network model is smaller than that of the second time series wind speed data, that is, the second fused time series wind speed data is more accurate.

[0124] Therefore, in the embodiments of the present disclosure, through the method of fusing multiple data for calculation, the accuracy of wind speed calculation can be effectively improved.

[0125] The embodiments of the present disclosure also provide a fused wind speed calculation device based on transfer learning. The following will be described in conjunction with Figure 3 for illustration.

[0126] In the embodiments of the present disclosure, the fused wind speed calculation device based on transfer learning can be an electronic device. Among them, the electronic device can include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs, PADs, PMPs, vehicle terminals (such as vehicle navigation terminals), wearable devices, etc. and fixed terminals such as digital TVs, desktop computers, smart home devices, etc.

[0127] Figure 3The figure shows a schematic structural diagram of a fusion wind speed calculation device based on transfer learning provided by an embodiment of the present disclosure.

[0128] As Figure 3 shown, the fusion wind speed calculation device 300 based on transfer learning may include a data determination module 310, a first acquisition module 320, a first calculation module 330, and a second calculation module 340.

[0129] The data determination module 310 may be configured to determine a reference anemometer tower according to wind speed correlation among multiple anemometer towers.

[0130] The first acquisition module 320 may be configured to acquire first true observed wind speed data and first simulated wind speed data of the reference anemometer tower, and acquire second simulated wind speed data of a target location.

[0131] The first calculation module 330 may be configured to perform time-series wind speed calculation on the first true observed wind speed data, the first simulated wind speed data, and the second simulated wind speed data to obtain first time-series wind speed data of the target location.

[0132] The second calculation module 340 may be configured to perform model calculation on the first time-series wind speed data through a transfer learning neural network model to obtain first fused time-series wind speed data of the target location.

[0133] In the embodiment of the present disclosure, a reference anemometer tower can be determined according to wind speed correlation among multiple anemometer towers, then the first true observed wind speed data and the first simulated wind speed data of the reference anemometer tower are acquired, and the second simulated wind speed data of the target location is acquired. Then, time-series wind speed calculation is performed on the first true observed wind speed data, the first simulated wind speed data, and the second simulated wind speed data to obtain first time-series wind speed data of the target location. Finally, model calculation is performed on the first time-series wind speed data through a transfer learning neural network model to obtain first fused time-series wind speed data of the target location. Thus, first fused time-series wind speed data of the target location can be obtained through fusion calculation of the first true observed wind speed data, the first simulated wind speed data, the second simulated wind speed data of the target location, and the transfer learning neural network model, and the wind speed calculation accuracy can be effectively improved through the method of fusion calculation with multiple data.

[0134] In some embodiments of the present disclosure, the data determination module 310 may specifically include a data acquisition unit and a data determination unit.

[0135] The data acquisition unit may be configured to acquire time-series wind speed sequences of multiple anemometer towers at a first preset height and calculate the wind speed correlation among the multiple anemometer towers.

[0136] The data determination unit may be configured to determine the anemometer tower with the largest wind speed correlation as the reference anemometer tower.

[0137] In some embodiments of the present disclosure, the fusion wind speed calculation device 300 based on transfer learning may further include a second acquisition module.

[0138] The second acquisition module may be configured to, before acquiring the first true observed wind speed data and the first simulated wind speed data of the reference anemometer tower and the second simulated wind speed data of the target point, obtain a simulated wind speed data table for each point through computational fluid dynamics method. The simulated wind speed data table includes temporal wind direction label features, where the temporal wind direction label features are mesoscale wind direction data at a second preset height. The simulated wind speed data table includes the first simulated wind speed data and the second simulated wind speed data.

[0139] In some embodiments of the present disclosure, the first calculation module 330 may specifically include a first calculation unit.

[0140] The first calculation unit may be configured to substitute the first true observed wind speed data, the first simulated wind speed data, and the second simulated wind speed data into the temporal wind speed calculation formula to obtain the first temporal wind speed data of the target point.

[0141] In some embodiments of the present disclosure, the fusion wind speed calculation device 300 based on transfer learning may further include a model training module and a model processing module.

[0142] The model training module may be configured to, before performing model calculation on the temporal wind speed data through a transfer learning neural network model to obtain the first fused temporal wind speed data of the target point, perform model training on the neural network model to be trained according to the first true observed wind speed data and the first simulated wind speed data of the reference anemometer tower, and the second true observed wind speed data and the third simulated wind speed data of the fused anemometer tower, to obtain a trained neural network model. The fused anemometer tower is an anemometer tower other than the reference anemometer tower among multiple anemometer towers.

[0143] The model processing module may be configured to perform transfer learning processing on the trained neural network model to obtain a transfer learning neural network model.

[0144] In some embodiments of the present disclosure, the fusion wind speed calculation device 300 based on transfer learning may further include a third calculation module.

[0145] The third calculation module may be configured to, after performing transfer learning processing on the trained neural network model to obtain a transfer learning neural network model, perform error analysis calculation on the transfer learning neural network model through the second temporal wind speed data and the second fused temporal wind speed data of the test anemometer tower to obtain an error result. The error analysis calculation is to calculate the root mean square error of the second temporal wind speed data, the second fused temporal wind speed data, and the third true observed wind speed data of the test anemometer tower.

[0146] In some embodiments of the present disclosure, the second calculation module 340 may specifically include a second calculation unit.

[0147] The second calculation unit may be configured to input the first time-series wind speed data into the transfer learning neural network model, so that the transfer learning neural network model performs model calculations on the first time-series wind speed data to obtain the first fused time-series wind speed data of the target point.

[0148] It should be noted that Figure 3 the fusion wind speed calculation device 300 based on transfer learning shown can execute Figures 1 to 2 each step in the method embodiments shown, and implement Figures 1 to 2 each process and effect in the method embodiments shown, which will not be elaborated here.

[0149] Embodiments of the present disclosure further provide an electronic device, which may include a processor and a memory, and the memory may be used to store executable instructions. Among them, the processor may be configured to read the executable instructions from the memory and execute the executable instructions to implement the fusion wind speed calculation method based on transfer learning in the above embodiments.

[0150] Figure 4 Fig. shows a schematic structural diagram of a fusion wind speed calculation device based on transfer learning provided by an embodiment of the present disclosure.

[0151] In some embodiments of the present disclosure, Figure 4 the fusion wind speed calculation device based on transfer learning shown may be an electronic device for a user to perform wind speed calculation operations. Among them, the electronic device may include, but is not limited to, mobile terminals such as laptop computers, PDAs (Personal Digital Assistants), PADs (Tablet Computers), etc. and fixed terminals such as digital TVs, desktop computers, etc.

[0152] As Figure 4 shown, the fusion wind speed calculation device based on transfer learning may include a processor 401 and a memory 402 storing computer program instructions.

[0153] Specifically, the above-mentioned processor 401 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0154] The memory 402 may include a mass memory for information or instructions. By way of example and not limitation, the memory 402 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 402 may include removable or non-removable (or fixed) media. Where appropriate, the memory 402 may be internal or external to the integrated gateway device. In a particular embodiment, the memory 402 is a non-volatile solid-state memory. In a particular embodiment, the memory 402 includes a read-only memory (ROM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically alterable ROM (EAROM), or a flash memory, or a combination of two or more of these.

[0155] The processor 401 reads and executes the computer program instructions stored in the memory 402 to perform the steps of the method for calculating the fusion wind speed based on transfer learning provided by the embodiments of the present disclosure.

[0156] In one example, the device for calculating the fusion wind speed based on transfer learning may further include a transceiver 403 and a bus 404. Among them, as Figure 4 shown, the processor 401, the memory 402, and the transceiver 403 are connected through the bus 404 and complete communication with each other.

[0157] The bus 404 includes hardware, software, or both. By way of example and not limitation, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side BUS (FSB), a Hyper Transport (HT) interconnect, an Industrial Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a MicroChannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses or a combination of two or more of these. Where appropriate, the bus 404 may include one or more buses. Although embodiments of the present application describe and illustrate specific buses, the present application contemplates any suitable bus or interconnect.

[0158] Embodiments of the present disclosure also provide a computer-readable storage medium that may store a computer program, which when executed by a processor, causes the processor to implement the method for calculating a fusion wind speed based on transfer learning provided by the embodiments of the present disclosure.

[0159] The above storage medium may include, for example, a memory 402 storing computer program instructions, and the above instructions may be executed by a processor 401 of a device for calculating a fusion wind speed based on transfer learning to complete the method for calculating a fusion wind speed based on transfer learning provided by the embodiments of the present disclosure. Optionally, the storage medium may be a non-transitory computer-readable storage medium. For example, the non-transitory computer-readable storage medium may be a ROM, a Random Access Memory (RAM), a Compact Disc ROM (CD-ROM), a magnetic tape, a floppy disk, and an optical data storage device, etc.

[0160] It should be noted that in this document, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising" is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or elements inherent to such process, method, article or device.

[0161] The above are only specific embodiments of the present disclosure, enabling those skilled in the art to understand or implement the present disclosure. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present disclosure. Therefore, the present disclosure will not be limited to the embodiments described herein, but rather to the broadest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for calculating the fusion wind speed based on transfer learning, characterized in that Including: Determine a reference anemometer tower according to wind speed correlation among multiple anemometer towers; Obtain the first true observed wind speed data and the first simulated wind speed data of the reference anemometer tower, and obtain the second simulated wind speed data of the target point; Perform time-series wind speed calculation on the first true observed wind speed data, the first simulated wind speed data and the second simulated wind speed data to obtain the first time-series wind speed data of the target point; Perform model calculation on the first time-series wind speed data through a transfer learning neural network model to obtain the first fused time-series wind speed data of the target point; Wherein, before performing model calculation on the time-series wind speed data through the transfer learning neural network model to obtain the first fused time-series wind speed data of the target point, the method further includes: Perform model training on the neural network model to be trained according to the first true observed wind speed data and the first simulated wind speed data of the reference anemometer tower, and the second true observed wind speed data and the third simulated wind speed data of the fused anemometer tower, to obtain a trained neural network model, where the fused anemometer tower is the anemometer tower other than the reference anemometer tower among the multiple anemometer towers; Perform transfer learning processing on the trained neural network model to obtain the transfer learning neural network model.

2. The method according to claim 1, characterized in that, The determining a reference anemometer tower according to wind speed correlation among multiple anemometer towers includes: Obtain the time-series wind speed sequences of the multiple anemometer towers at a first preset height, and calculate the wind speed correlation of the multiple anemometer towers; Determine the anemometer tower with the largest wind speed correlation as the reference anemometer tower.

3. The method according to claim 1, wherein Before obtaining the first true observed wind speed data and the first simulated wind speed data of the reference anemometer tower, and obtaining the second simulated wind speed data of the target point, the method further includes: Obtain the simulated wind speed data tables of each point through the computational fluid dynamics method, where the simulated wind speed data tables include time-series wind direction label features, the time-series wind direction label features are mesoscale wind direction data at a second preset height, and the simulated wind speed data tables include the first simulated wind speed data and the second simulated wind speed data.

4. The method according to claim 1, wherein The performing time-series wind speed calculation on the first true observed wind speed data, the first simulated wind speed data and the second simulated wind speed data to obtain the first time-series wind speed data of the target point includes: Substitute the first true observed wind speed data, the first simulated wind speed data and the second simulated wind speed data into the time-series wind speed calculation formula to obtain the first time-series wind speed data of the target point.

5. The method according to claim 1, wherein After performing transfer learning processing on the trained neural network model to obtain the transfer learning neural network model, the method further includes: Perform error analysis calculation on the transfer learning neural network model through the second time-series wind speed data and the second fused time-series wind speed data of the test anemometer tower to obtain an error result, where the error analysis calculation is to calculate the root mean square error of the second time-series wind speed data and the second fused time-series wind speed data and the third true observed wind speed data of the test anemometer tower.

6. The method according to claim 1, wherein Performing model calculation on the first time-series wind speed data by the transfer learning neural network model to obtain the first fused time-series wind speed data of the target point includes: Inputting the first time-series wind speed data into the transfer learning neural network model, so that the transfer learning neural network model performs model calculation on the first time-series wind speed data to obtain the first fused time-series wind speed data of the target point.

7. A fusion wind speed calculation device based on transfer learning, characterized in that, Including: A data determination module, configured to determine a reference anemometer tower according to wind speed correlation among multiple anemometer towers; A first acquisition module, configured to acquire the first true observed wind speed data and the first simulated wind speed data of the reference anemometer tower, and acquire the second simulated wind speed data of the target point; A first calculation module, configured to perform time-series wind speed calculation on the first true observed wind speed data, the first simulated wind speed data and the second simulated wind speed data to obtain the first time-series wind speed data of the target point; A second calculation module, configured to perform model calculation on the first time-series wind speed data by the transfer learning neural network model to obtain the first fused time-series wind speed data of the target point; Wherein, the device further includes: A model training module, configured to, before performing model calculation on the time-series wind speed data by the transfer learning neural network model to obtain the first fused time-series wind speed data of the target point, perform model training on a neural network model to be trained according to the first true observed wind speed data and the first simulated wind speed data of the reference anemometer tower, and the second true observed wind speed data and the third simulated wind speed data of a fused anemometer tower, to obtain a trained neural network model, where the fused anemometer tower is an anemometer tower other than the reference anemometer tower among the multiple anemometer towers; A model processing module, configured to perform transfer learning processing on the trained neural network model to obtain the transfer learning neural network model.

8. A fusion wind speed calculation device based on transfer learning, characterized in that, Including: A processor; A memory, configured to store executable instructions; Wherein, the processor is configured to read the executable instructions from the memory and execute the executable instructions to implement the transfer learning-based fused wind speed calculation method according to any one of claims 1-6 above.

9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, causes the processor to implement the transfer learning-based fused wind speed calculation method according to any one of claims 1-6 above.

Images