Wind profile acquisition method based on multi-source data fusion

Abstract

The invention discloses a wind profile acquisition method based on multi-source data fusion, and relates to the technical field of meteorology, and the method comprises the steps: obtaining horizontal wind detection results reported by a plurality of different horizontal wind detection data sources at each data fusion moment; distance correction is carried out on various horizontal wind detection results at the first K data fusion moments, then fusion is carried out to obtain a fusion wind profile, and the obtained results are used for training to obtain a Kalman filtering model; at a subsequent data fusion moment, the optimal fusion wind profile at the previous data fusion moment is firstly utilized to perform optimal estimation screening on the multi-source data reported at the current moment, and then a result after screening quality control is used as an observation value and is input into a Kalman filtering model to predict and obtain the optimal fusion wind profile at the current moment; the method not only can improve the accuracy of a horizontal wind fusion result, but also is high in calculation speed, simple and efficient in algorithm, high in result reliability and easier to realize engineering, and has openness.

Description

Technical Field

[0001] The present application relates to the field of meteorological technology, and in particular to a method for obtaining wind profiles based on multi-source data fusion. Background Art

[0002] In recent years, climate change has intensified and extreme weather has occurred frequently, posing a huge threat to people's property and life safety. Among them, the forecast and warning of severe convective weather phenomena, including tornadoes, heavy rains, hail, etc., poses a huge challenge to the meteorological department. The horizontal wind field is one of the important indicators of severe convective weather. Therefore, accurately identifying the changes in the horizontal wind field can provide important information for severe convective weather warning and research.

[0003] At present, various meteorological observation and detection equipment such as ground-based automatic weather stations (AWS), laser wind radar (Lidar), wind profile radar (WPR), and sounding data (Sonde) can be used to obtain horizontal wind data. At the same time, multiple global numerical weather forecast models (such as the GFS model) can also provide horizontal wind forecast results. However, the temporal and spatial resolutions of horizontal wind data obtained by the above methods are different: for example, the temporal resolution of ground-based automatic weather stations (AWS) is relatively high, and the detection interval is usually about 2 minutes, but its detection space is relatively limited, and it can only provide single-layer data at a height of 2 meters. For example, the horizontal wind data provided by the GFS model is 0.25°*0.25° grid data in space and the height interval is 50hpa, but the temporal resolution is relatively low, and the detection interval is usually 1 hour. It can be seen that the horizontal wind data obtained by different methods have their own characteristics but also their own limitations. Therefore, the recognition accuracy and precision of horizontal wind still need to be improved. Summary of the invention

[0004] In view of the above problems and technical requirements, this application proposes a method for obtaining wind profiles based on multi-source data fusion. The technical solution of this application is as follows:

[0005] A wind profile acquisition method based on multi-source data fusion, the wind profile acquisition method comprising:

[0006] Obtain horizontal wind detection results reported by multiple different horizontal wind detection data sources between the k-1th data fusion time and the kth data fusion time, where the integer parameter k≥1;

[0007] When the integer parameter k≤K, the horizontal wind detection results reported by various horizontal wind detection data sources are fused after distance correction in the vertical direction to obtain the fused horizontal wind detection results at each data fusion altitude layer, and the fused wind profile at the kth data fusion moment is obtained by connecting the lines; wherein the integer parameter K≥3;

[0008] When the integer parameter k>K, the horizontal wind detection results reported by various horizontal wind detection data sources are distance-corrected in the vertical direction, and the multi-source horizontal wind detection results at each data fusion altitude layer at the k-th data fusion moment are combined and connected to obtain several candidate wind profiles; the fused wind profile at the k-1-th data fusion moment is used as the reference value for optimal estimation screening, and a candidate wind profile is selected from the several candidate wind profiles obtained by combining at the k-th data fusion moment as the wind profile observation value at the k-th data fusion moment; the Kalman filter model is used to obtain the fused wind profile at the k-th data fusion moment based on the wind profile observation value at the k-th data fusion moment; wherein the Kalman filter model is trained using the fused wind profiles at the first K data fusion moments.

[0009] A further technical solution is to select a candidate wind profile from a number of candidate wind profiles as the wind profile observation value at the kth data fusion moment, including:

[0010] For any candidate wind profile, the systematic deviation between the candidate wind profile and the fused wind profile at the k-1th data fusion time is calculated, and the result standard deviation of the candidate wind profile is calculated. The consistency score of the candidate wind profile is calculated based on the systematic deviation and result standard deviation of the candidate wind profile. The smaller the systematic deviation and the result standard deviation, the higher the consistency score.

[0011] The candidate wind profile with the highest consistency score is selected from several candidate wind profiles as the wind profile observation value at the kth data fusion moment.

[0012] A further technical solution is to use the Kalman filter model to obtain the fused wind profile at the kth data fusion moment based on the wind profile observation value at the kth data fusion moment, including:

[0013] For any horizontal wind detection result of the data fusion altitude layer H in the wind profile observation value at the k-th data fusion moment, when the horizontal wind detection result of the data fusion altitude layer H is empty, or the deviation between the horizontal wind detection result of the data fusion altitude layer H and the reference result of the data fusion altitude layer H reaches the deviation threshold, the reference result of the data fusion altitude layer H is updated as the horizontal wind detection result of the data fusion altitude layer H, and the wind profile observation value after the median test at the k-th data fusion moment is obtained;

[0014] The wind profile observation value after the median test at the k-th data fusion moment is input into the Kalman filter model to obtain the fused wind profile at the k-th data fusion moment;

[0015] Among them, the benchmark result of any data fusion altitude layer H is the median of multiple reference horizontal wind results of the data fusion altitude layer H, and the multiple reference horizontal wind results of the data fusion altitude layer H include: the horizontal wind detection results in the wind profile observation values ​​of the data fusion altitude layer H and its upper and lower adjacent data fusion altitude layers at the kth data fusion time, and the horizontal wind detection results in the fused wind profile of the data fusion altitude layer H and its upper and lower adjacent data fusion altitude layers at the k-1th data fusion time.

[0016] Its further technical solution is that the data sources of horizontal wind detection include ground automatic weather stations, laser wind radars, wind profile radars, sounding data and GFS mode, and the time interval between two adjacent data fusion moments is not less than the detection interval of the ground automatic weather stations and the laser wind radars but less than the detection interval of other horizontal wind detection data sources. The horizontal wind detection results reported at each data fusion moment include at least the AWS horizontal wind detection results reported by the ground automatic weather station and the Lidar horizontal wind detection results reported by the laser wind radar.

[0017] A further technical solution is to correct the distance of the horizontal wind detection results reported by various horizontal wind detection data sources in the vertical direction, including:

[0018] The AWS horizontal wind detection result at the kth data fusion moment is used to perform data correction on the Lidar horizontal wind detection result to obtain the corrected Lidar horizontal wind detection result at the kth data fusion moment; the corrected Lidar horizontal wind detection result at the kth data fusion moment is linearly interpolated in the vertical direction to obtain the corrected Lidar horizontal wind detection results at each data fusion altitude layer;

[0019] When the kth data fusion moment also includes horizontal wind detection results reported by other horizontal wind detection data sources, the horizontal wind detection results reported by other horizontal wind detection data sources are linearly interpolated in the vertical direction to obtain the horizontal wind detection results of other horizontal wind detection data sources at each data fusion altitude layer.

[0020] A further technical solution is to obtain the corrected Lidar horizontal wind detection result at the kth data fusion moment, including:

[0021] Substitute the AWS horizontal wind detection result at the kth data fusion moment into the near-surface gale model inversion to obtain the horizontal wind inversion result at any height within the near-surface height range;

[0022] For the Lidar horizontal wind detection result at any height within the near-ground height range at the kth data fusion moment, when the error between the Lidar horizontal wind detection result and the horizontal wind inversion result at the current height is less than the error threshold, the Lidar horizontal wind detection result at the current height is retained, otherwise the Lidar horizontal wind detection result at the current height is replaced by the horizontal wind inversion result at the current height to obtain the corrected Lidar horizontal wind detection result at the kth data fusion moment.

[0023] Its further technical solution is to fuse the fused horizontal wind detection results at each data fusion altitude layer, including:

[0024] When the kth data fusion moment only includes AWS horizontal wind detection results and Lidar horizontal wind detection results, the corrected Lidar horizontal wind detection results at each data fusion altitude layer are used as the fused horizontal wind detection results at each data fusion altitude layer;

[0025] When the kth data fusion moment includes horizontal wind detection results reported by other horizontal wind detection data sources in addition to AWS horizontal wind detection results and Lidar horizontal wind detection results, the corrected Lidar horizontal wind detection results at each data fusion altitude layer and the horizontal wind detection results of other horizontal wind detection data sources are weightedly calculated to obtain the fused horizontal wind detection result at the data fusion altitude layer.

[0026] A further technical solution is to perform weighted calculation on the corrected Lidar horizontal wind detection results at each data fusion altitude layer and the horizontal wind detection results of other horizontal wind detection data sources, including:

[0027] According to the weight combination corresponding to each altitude range, the corrected Lidar horizontal wind detection results at each data fusion altitude within the altitude range and the horizontal wind detection results reported by other horizontal wind detection data sources are weighted averaged; different altitude ranges have different corresponding weight combinations, and the lower the altitude range, the higher the weight of the corrected Lidar horizontal wind detection result.

[0028] A further technical solution is that, when the WPR horizontal wind detection result reported by the wind profile radar is also included at the kth data fusion moment, it is determined that (1) the weight combination corresponding to the altitude range below 3 km includes the weight of the corrected Lidar horizontal wind detection result being 0.6 and the weight of the WPR horizontal wind detection result being 0.4; and (2) the weight combination corresponding to the altitude range from 3 km to 6 km includes the weight of the corrected Lidar horizontal wind detection result being 0.3 and the weight of the WPR horizontal wind detection result being 0.7;

[0029] When the kth data fusion moment also includes the WPR horizontal wind detection result reported by the wind profile radar and the GFS horizontal wind detection result reported by the GFS mode, it is determined that (1) the weight combination corresponding to the altitude range below 3 km includes the weight of the corrected Lidar horizontal wind detection result of 0.5, the weight of the WPR horizontal wind detection result of 0.2, and the weight of the GFS horizontal wind detection result; (2) the weight combination corresponding to the altitude range from 3 km to 6 km includes the weight of the corrected Lidar horizontal wind detection result of 0.2, the weight of the WPR horizontal wind detection result of 0.3. The weight of the wind detection result is 0.4, and the weight of the GFS horizontal wind detection result is 0.4; (3) The weight combination corresponding to the altitude range of 6 km to 10 km includes the weight of the corrected Lidar horizontal wind detection result of 0, the weight of the WPR horizontal wind detection result of 0.3, and the weight of the GFS horizontal wind detection result of 0.7; (4) The weight combination corresponding to the altitude range above 10 km includes the weight of the corrected Lidar horizontal wind detection result of 0, the weight of the WPR horizontal wind detection result of 0, and the weight of the GFS horizontal wind detection result of 1;

[0030] When the kth data fusion moment also includes the WPR horizontal wind detection result reported by the wind profile radar, the GFS horizontal wind detection result reported by the GFS model, and the Sonde horizontal wind detection result reported by the sounding data, it is determined that (1) the weight combination corresponding to the altitude range below 3 km includes the weight of the corrected Lidar horizontal wind detection result of 0.35, the weight of the WPR horizontal wind detection result of 0.15, the weight of the GFS horizontal wind detection result of 0.25, and the weight of the Sonde horizontal wind detection result of 0.25; (2) the weight combination corresponding to the altitude range of 3 km to 6 km includes the weight of the corrected Lidar horizontal wind detection result of 0.2, the weight of the WPR horizontal wind detection result of 0. .25, the weight of the GFS horizontal wind detection result is 0.25, the weight of the Sonde horizontal wind detection result is 0.3; (3) The weight combination corresponding to the altitude range of 6km to 10km includes the weight of the corrected Lidar horizontal wind detection result is 0, the weight of the WPR horizontal wind detection result is 0.2, the weight of the GFS horizontal wind detection result is 0.4, and the weight of the Sonde horizontal wind detection result is 0.4; (4) The weight combination corresponding to the altitude range above 10km includes the weight of the corrected Lidar horizontal wind detection result is 0, the weight of the WPR horizontal wind detection result is 0, the weight of the GFS horizontal wind detection result is 0.5, and the weight of the Sonde horizontal wind detection result is 0.5.

[0031] Its further technical solution is that the near-ground strong wind model is:

[0032] V2=V1(Z2 / Z1) α

[0033] Among them, V2 is the AWS horizontal wind detection result at the kth data fusion moment, Z2 is the equipment installation height of the ground automatic weather station from the ground; V1 is the horizontal wind inversion result at any height Z1; Z0 represents the ground roughness in the area where the ground automatic weather station is located.

[0034] The beneficial technical effects of this application are:

[0035] The present application discloses a method for acquiring wind profiles based on multi-source data fusion. The method fuses horizontal wind detection results from different horizontal wind detection data sources based on Kalman filtering. During the fusion, the optimal fused wind profile at the previous data fusion moment is used to perform optimal estimation screening on the multi-source data arriving at the current moment. The screening quality-controlled results are then input as observation values ​​into the Kalman filter model to obtain the optimal fused wind profile at the next moment. The method can not only improve the accuracy of the horizontal wind fusion results, but also has fast calculation speed, simple and efficient algorithm, high result reliability, easier engineering implementation, and openness. Combined with the wind profile acquisition method of the present application, the organic fusion of wind field data of various height resolutions and time resolutions can be achieved.

[0036] This method combines and connects several candidate wind profiles obtained by combining and connecting the multi-source horizontal wind detection results and compares them with the optimal fused wind profile at the previous data fusion moment to achieve optimal estimation screening, rather than selecting multi-source data by setting a threshold at a single altitude layer, making the final wind profile fusion result more accurate.

[0037] This method uses the near-ground strong wind model to push the ground automatic weather station data up to the horizontal wind result at an altitude of 150m, and corrects the detection data of the laser wind radar at the low level, thus making up for the shortcoming of the laser wind radar having a detection blind spot at low altitude. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a flow chart of a method for obtaining a wind profile according to an embodiment of the present application. DETAILED DESCRIPTION

[0039] The specific implementation of the present application is further described below in conjunction with the accompanying drawings.

[0040] This application discloses a method for obtaining wind profiles based on multi-source data fusion. Please refer to Figure 1 As shown in the flowchart, the wind profile acquisition method comprises the following steps:

[0041] Step 110, obtaining horizontal wind detection results reported by multiple different horizontal wind detection data sources between the k-1th data fusion time and the kth data fusion time, wherein the integer parameter k≥1.

[0042] The k-1th data fusion moment is the previous data fusion moment of the kth data fusion moment. When k=1, the k-1th data fusion moment represents the initial moment.

[0043] Since the time resolutions of different horizontal wind detection data sources are different, the data arrival situations between each two data fusion moments are different. The time interval between two adjacent data fusion moments can be customized, but it is necessary to ensure that the detection interval of at least two horizontal wind detection data sources is not greater than the time interval between two adjacent data fusion moments. This ensures that at least two different horizontal wind detection data sources have data arrival between any two adjacent data fusion moments.

[0044] In one embodiment, the horizontal wind detection data sources used in this application include ground automatic weather stations (AWS), laser wind radar (Lidar), wind profiler radar (WPR), sounding data (Sonde) and GFS (Global Forecast System) mode. The typical time resolution and spatial resolution of these five different horizontal wind detection data sources are shown in the following table:

[0045]

[0046] It can be seen that the ground automatic weather station AWS and the laser wind radar Lidar have the highest time resolution, so the time interval between two adjacent data fusion moments must be greater than or equal to the detection interval of the ground automatic weather station AWS and the laser wind radar Lidar, but in order to maximize the time resolution of the wind profile acquisition method of this application, the time interval of the data fusion moment must also be as small as possible, so in a petrochemical example, the time interval between two adjacent data fusion moments is not less than the detection interval of the ground automatic weather station and the laser wind radar but less than the detection interval of other horizontal wind detection data sources. For example, when the detection intervals of the ground automatic weather station AWS and the laser wind radar Lidar are both 2 minutes, the time interval between two adjacent data fusion moments can be directly selected as 2 minutes.

[0047] On the basis of selecting the above five horizontal wind detection data sources and setting the time interval between two adjacent data fusion moments according to the above principles, based on the relative relationship of the time resolution of various horizontal wind detection data sources, the horizontal wind detection results reported at each data fusion moment include at least the AWS horizontal wind detection results reported by the ground automatic weather station AWS and the Lidar horizontal wind detection results reported by the laser wind radar Lidar. At certain data fusion moments, there will be horizontal wind detection results reported by other horizontal wind detection data sources. The data reporting situations at different data fusion moments are mainly as follows:

[0048] (1) Only the AWS horizontal wind detection results reported by the ground automatic weather station AWS and the Lidar horizontal wind detection results reported by the laser wind radar Lidar are available.

[0049] (2) Includes AWS horizontal wind detection results reported by the ground automatic weather station AWS, Lidar horizontal wind detection results reported by the laser wind radar Lidar, and WPR horizontal wind detection results reported by the wind profile radar WPR.

[0050] (3) Includes AWS horizontal wind detection results reported by the ground automatic weather station AWS, Lidar horizontal wind detection results reported by the laser wind radar Lidar, WPR horizontal wind detection results reported by the wind profiler radar WPR, and GFS horizontal wind detection results reported by the GFS model.

[0051] (4) Includes AWS horizontal wind detection results reported by the ground automatic weather station AWS, Lidar horizontal wind detection results reported by the laser wind radar Lidar, WPR horizontal wind detection results reported by the wind profiler radar WPR, GFS horizontal wind detection results reported by the GFS model, and Sonde horizontal wind detection results reported by the sounding data Sonde.

[0052] After obtaining multiple groups of horizontal wind detection results reported at each data fusion moment, the fused wind profile is obtained according to the following step 120 for the first K data fusion moments, and the fused wind profile is obtained according to the following steps 130 to 160 starting from the K+1th data fusion moment.

[0053] Step 120, when the integer parameter k≤K, the horizontal wind detection results reported by various horizontal wind detection data sources are distance-corrected in the vertical direction and then fused to obtain the fused horizontal wind detection results at each data fusion altitude layer, and the fused wind profile at the kth data fusion moment is obtained by connecting the lines.

[0054] As mentioned above, each data fusion moment must include AWS horizontal wind detection results and Lidar horizontal wind detection results, and may also include other horizontal wind detection data sources. Therefore, when correcting the distance, the data correction of AWS horizontal wind detection results and Lidar horizontal wind detection results must be involved. Since AWS horizontal wind detection results can only provide a single layer at a height of 2 meters in space, it is difficult to directly perform linear interpolation of AWS horizontal wind detection results in the vertical direction. In addition, considering that the laser wind measurement radar Lidar has a detection blind spot at low altitude, in one embodiment, the AWS horizontal wind detection result at the kth data fusion moment is used to perform data correction on the Lidar horizontal wind detection result, and the corrected Lidar horizontal wind detection result at the kth data fusion moment is obtained, including:

[0055] 1. Substitute the AWS horizontal wind detection result at the kth data fusion time into the near-surface strong wind model inversion to obtain the horizontal wind inversion result at any height within the near-surface height range. The near-surface strong wind model is:

[0056] V2=V1(Z2 / Z1) α

[0057] Among them, V2 is the AWS horizontal wind detection result at the kth data fusion moment, Z2 is the equipment installation height of the ground automatic weather station from the ground. V1 is the horizontal wind inversion result at any height Z1. The units of Z1 and Z2 are both meters.

[0058] In the above formula, the power function Z0 represents the ground roughness of the area where the ground automatic weather station is located, in meters. The ground roughness of different areas has its own typical values. The ground roughness of several common areas is shown in the following table:

[0059] City suburbs Farmland grassland Snow surface beach desert <![CDATA[Z0(m)]]> 0.8 0.4 0.05 0.01 0.001 0.001 0.001

[0060] 2. For the Lidar horizontal wind detection result at any height within the near-ground height range at the kth data fusion moment, compare the Lidar horizontal wind detection result at that height with the horizontal wind inversion result at that height:

[0061] When the error between the Lidar horizontal wind detection result and the horizontal wind inversion result at the current height is less than the error threshold, the Lidar horizontal wind detection result at the current height is retained.

[0062] When the error between the Lidar horizontal wind detection result and the horizontal wind inversion result at the current height reaches the error threshold, the Lidar horizontal wind detection result at the current height is replaced by the horizontal wind inversion result at the current height.

[0063] The above error threshold can be customized. After the above processing, the data correction of the result within the near-ground height range of the Lidar horizontal wind detection result at the kth data fusion moment can be completed, thereby obtaining the corrected Lidar horizontal wind detection result at the kth data fusion moment.

[0064] Then, the corrected Lidar horizontal wind detection results at the kth data fusion moment are linearly interpolated in the vertical direction to obtain the corrected Lidar horizontal wind detection results at each data fusion altitude layer, thereby realizing the data fusion and distance correction of these two types of horizontal wind detection results.

[0065] The selection of the data fusion altitude layer of the present application is customized according to the spatial resolution requirements, taking into full consideration the spatial resolution and advantages of each data source. In one embodiment, the setting method of the data fusion altitude layer includes: setting the data fusion altitude layer at intervals of 30 meters in the altitude range below 3km, setting the data fusion altitude layer at intervals of 120 meters in the altitude range of 3km to 6km, setting the data fusion altitude layer at intervals of 240 meters in the altitude range of 6km to 10km, and setting the data fusion altitude layer at intervals of 480 meters in the altitude range above 10km.

[0066] When the kth data fusion moment only includes AWS horizontal wind detection results and Lidar horizontal wind detection results, after obtaining the corrected Lidar horizontal wind detection results at each data fusion altitude layer according to the above method, the corrected Lidar horizontal wind detection results at each data fusion altitude layer are directly used as the fused horizontal wind detection results at each data fusion altitude layer.

[0067] When the kth data fusion moment also includes the horizontal wind detection results reported by other horizontal wind detection data sources, the horizontal wind detection results reported by other horizontal wind detection data sources are linearly interpolated in the vertical direction to obtain the horizontal wind detection results of other horizontal wind detection data sources at each data fusion altitude layer. Then, the corrected Lidar horizontal wind detection results and the horizontal wind detection results of other horizontal wind detection data sources at each data fusion altitude layer are weighted to obtain the fused horizontal wind detection results at the data fusion altitude layer. Among them, since the GFS data is grid data, before fusing it, it is necessary to perform bilinear interpolation in the horizontal direction on the u and v wind data at each altitude layer according to the grid longitude and latitude of the GFS data and the geographic spatial position of the measuring station, and extract the GFS horizontal wind detection results of the GFS at the measuring station position.

[0068] As mentioned above, when the kth data fusion moment also includes the horizontal wind detection results reported by other horizontal wind detection data sources, there are also many different combinations, and the detection characteristics of different horizontal wind detection data sources are different. Then, according to the weight combination corresponding to each altitude layer range, the corrected Lidar horizontal wind detection results at each data fusion altitude layer within the altitude layer range and the horizontal wind detection results reported by other horizontal wind detection data sources are weighted averaged. The weight combinations corresponding to different altitude layer ranges are different, and the lower the altitude layer range, the higher the weight of the corrected Lidar horizontal wind detection result.

[0069] In one embodiment, the weight combinations under different data arrival conditions at the kth data fusion moment are respectively introduced as follows:

[0070] In case (a), when the WPR horizontal wind detection result reported by the wind profile radar is also included at the kth data fusion moment, it is determined that (1) the weight combination corresponding to the altitude range below 3 km includes the weight of the corrected Lidar horizontal wind detection result being 0.6 and the weight of the WPR horizontal wind detection result being 0.4; and (2) the weight combination corresponding to the altitude range from 3 km to 6 km includes the weight of the corrected Lidar horizontal wind detection result being 0.3 and the weight of the WPR horizontal wind detection result being 0.7.

[0071] In case (b), when the kth data fusion moment also includes the WPR horizontal wind detection result reported by the wind profile radar and the GFS horizontal wind detection result reported by the GFS mode, it is determined that (1) the weight combination corresponding to the altitude range below 3 km includes the weight of the corrected Lidar horizontal wind detection result of 0.5, the weight of the WPR horizontal wind detection result of 0.2, and the weight of the GFS horizontal wind detection result; (2) the weight combination corresponding to the altitude range from 3 km to 6 km includes the weight of the corrected Lidar horizontal wind detection result of 0.2, the weight of the WPR horizontal wind detection result of 0.3. The weight of the R horizontal wind detection result is 0.4, and the weight of the GFS horizontal wind detection result is 0.4; (3) The weight combination corresponding to the altitude range of 6km to 10km includes the weight of the corrected Lidar horizontal wind detection result of 0, the weight of the WPR horizontal wind detection result of 0.3, and the weight of the GFS horizontal wind detection result of 0.7; (4) The weight combination corresponding to the altitude range above 10km includes the weight of the corrected Lidar horizontal wind detection result of 0, the weight of the WPR horizontal wind detection result of 0, and the weight of the GFS horizontal wind detection result of 1.

[0072] In case (c), when the kth data fusion moment also includes the WPR horizontal wind detection results reported by the wind profile radar, the GFS horizontal wind detection results reported by the GFS model, and the Sonde horizontal wind detection results reported by the sounding data, it is determined that (1) the weight combination corresponding to the altitude range below 3 km includes the weight of the corrected Lidar horizontal wind detection results of 0.35, the weight of the WPR horizontal wind detection results of 0.15, the weight of the GFS horizontal wind detection results of 0.25, and the weight of the Sonde horizontal wind detection results of 0.25; (2) the weight combination corresponding to the altitude range of 3 km to 6 km includes the weight of the corrected Lidar horizontal wind detection results of 0.2, the weight of the WPR horizontal wind detection results of 0. The weight of the horizontal wind detection result after correction of Lidar is 0, the weight of the horizontal wind detection result after WPR is 0.2, the weight of the horizontal wind detection result after GFS is 0.3, and the weight of the horizontal wind detection result after Sonde is 0.4. (4) The weight combination corresponding to the altitude range above 10 km includes the weight of the horizontal wind detection result after correction of Lidar is 0, the weight of the horizontal wind detection result after WPR is 0, the weight of the horizontal wind detection result after GFS is 0.5, and the weight of the horizontal wind detection result after Sonde is 0.5.

[0073] The summary can be expressed in the following table:

[0074]

[0075]

[0076] After obtaining the fused horizontal wind detection results at each data fusion altitude layer by the above method, the fused horizontal wind detection results at each data fusion altitude layer are connected from bottom to top to obtain the fused wind profile at the kth data fusion moment.

[0077] After obtaining the fused wind profiles of the first K data fusion moments according to this step, the Kalman filter model is trained using the fused wind profiles of the first K data fusion moments. During training, the fused horizontal wind detection results at each data fusion altitude layer are decomposed and entered into the Kalman filter. Taking the horizontal wind east-west component u as an example, the initial value is set to 0, and the u component of K groups of fused wind profiles is sequentially brought into the Kalman filter model. After K groups, the optimal solution gradually converges, and a more stable Kalman filter model can be obtained. According to actual measurements, generally taking K=3 can achieve better convergence, and then the trained Kalman filter model can be used from the K+1th data fusion moment.

[0078] Step 130, when the integer parameter k>K, the horizontal wind detection results reported by various horizontal wind detection data sources are distance-corrected in the vertical direction to obtain the multi-source horizontal wind detection results at each data fusion altitude layer at the kth data fusion time.

[0079] The method for distance correction is similar to the above step 120, and the corrected Lidar horizontal wind detection results at each data fusion altitude layer are obtained by fusing and linearly interpolating. When there are horizontal wind detection results reported by other horizontal wind detection data sources, the horizontal wind detection results of other horizontal wind detection data sources at each data fusion altitude layer are also interpolated. However, different from step 120, when k>K, the fused horizontal wind detection result of each data fusion altitude layer is no longer calculated, but the data at the kth data fusion moment at each data fusion altitude layer is directly aggregated as the multi-source horizontal wind detection result, and the multi-source horizontal wind detection result at each data fusion altitude layer is the corrected Lidar horizontal wind detection result, or the multi-source horizontal wind detection result at each data fusion altitude layer includes the corrected Lidar horizontal wind detection result and the WPR horizontal wind detection result, or the multi-source horizontal wind detection result at each data fusion altitude layer includes the corrected Lidar horizontal wind detection result, the WPR horizontal wind detection result and the GFS horizontal wind detection result, or the multi-source horizontal wind detection result at each data fusion altitude layer includes the corrected Lidar horizontal wind detection result, the WPR horizontal wind detection result, the GFS horizontal wind detection result and the Sonde horizontal wind detection result.

[0080] Step 140, the multi-source horizontal wind detection results at each data fusion altitude at the kth data fusion moment are combined and connected to obtain a number of candidate wind profiles. As described above, according to the different types of reported data, the multi-source horizontal wind detection results at a single data fusion altitude may include 1 to 4 different types of horizontal wind detection results, and when the multi-source horizontal wind detection results at different data fusion altitudes are connected from bottom to top, each data fusion altitude will select a horizontal wind detection result from the multi-source horizontal wind detection results to participate in the connection, and different types of data selected to participate will be connected to obtain different candidate wind profiles, and the horizontal wind detection results selected to participate in the connection at different data fusion altitudes have different permutations and combinations, so a number of different candidate wind profiles can be obtained through different full permutations and combinations.

[0081] Step 150, using the fused wind profile at the k-1th data fusion moment as a reference value to perform optimal estimation screening, and selecting a candidate wind profile from a plurality of candidate wind profiles obtained by combining at the kth data fusion moment as the wind profile observation value at the kth data fusion moment.

[0082] When screening the wind profile observation value at the kth data fusion moment, for any candidate wind profile, calculate the system deviation between the candidate wind profile and the fused wind profile at the k-1th data fusion moment, and calculate the result standard deviation of the candidate wind profile itself. Then, the consistency score of the candidate wind profile is calculated based on the system deviation and result standard deviation of the candidate wind profile. The specific scoring criteria can be customized, but the smaller the system deviation of the self-check between the candidate wind profile and the fused wind profile at the k-1th data fusion moment, or the smaller the result standard deviation of the candidate wind profile itself, the higher the consistency score of the candidate wind profile. Then, select the candidate wind profile with the highest consistency score from several candidate wind profiles as the wind profile observation value at the kth data fusion moment.

[0083] Step 160, using the Kalman filter model to obtain the fused wind profile at the kth data fusion moment based on the wind profile observation value at the kth data fusion moment, including entering the wind profile observation value at the kth data fusion moment into the trained Kalman filter model according to the altitude layer, and solving to obtain the optimal solution of the fused wind profile as the fused wind profile at the kth data fusion moment.

[0084] In one embodiment, in order to further improve the wind profile acquisition precision and accuracy, after obtaining the wind profile observation value at the kth data fusion moment through step 140, it is not directly input into the trained Kalman filter model, but a median test process is further added, including:

[0085] For any horizontal wind detection result of the data fusion altitude layer H in the wind profile observation value at the kth data fusion time, when the horizontal wind detection result of the data fusion altitude layer H is empty, or the deviation between the horizontal wind detection result of the data fusion altitude layer H and the reference result of the data fusion altitude layer H reaches the deviation threshold, the reference result of the data fusion altitude layer H is updated as the horizontal wind detection result of the data fusion altitude layer H, and the wind profile observation value after the median test at the kth data fusion time is obtained. Then the wind profile observation value after the median test at the kth data fusion time is input into the Kalman filter model to obtain the fused wind profile at the kth data fusion time. Among them, the reference result of any data fusion altitude layer H is the median of multiple reference horizontal wind results of the data fusion altitude layer H, and the multiple reference horizontal wind results of the data fusion altitude layer H include: the horizontal wind detection result of the data fusion altitude layer H and its upper and lower adjacent data fusion altitude layers in the wind profile observation value at the kth data fusion time, and the horizontal wind detection result of the data fusion altitude layer H and its upper and lower adjacent data fusion altitude layers in the fused wind profile at the k-1th data fusion time. For example, in one example, assuming that the data fusion altitude layer H = 150 meters, the adjacent data fusion altitude layer above the data fusion altitude layer H is 180 meters, and the adjacent data fusion altitude layer below the data fusion altitude layer H is 120 meters, then the multiple reference horizontal wind results of the data fusion altitude layer H = 150 meters include: the horizontal wind detection results at 120 meters, 150 meters, and 180 meters in the wind profile observation value at the kth data fusion moment, and the horizontal wind detection results at 120 meters, 150 meters, and 180 meters in the fused wind profile at the k-1th data fusion moment.

[0086] The above is only a preferred embodiment of the present application, and the present application is not limited to the above embodiments. It is understood that other improvements and changes directly derived or associated by those skilled in the art without departing from the spirit and concept of the present application should be considered to be included in the protection scope of the present application.

Claims

1. A method for obtaining wind profiles based on multi-source data fusion, characterized in that: The wind profile acquisition method comprises: Obtain horizontal wind detection results reported by multiple different horizontal wind detection data sources between the k-1th data fusion time and the kth data fusion time, where the integer parameter k≥1; When the integer parameter k≤K, the horizontal wind detection results reported by various horizontal wind detection data sources are fused after distance correction in the vertical direction to obtain the fused horizontal wind detection results at each data fusion altitude layer, and the fused wind profile at the kth data fusion moment is obtained by connecting the lines; wherein the integer parameter K≥3; When the integer parameter k>K, the horizontal wind detection results reported by various horizontal wind detection data sources are distance-corrected in the vertical direction, and the multi-source horizontal wind detection results at each data fusion altitude layer at the k-th data fusion moment are combined and connected to obtain several candidate wind profiles; the fused wind profile at the k-1-th data fusion moment is used as the reference value for optimal estimation screening, and a candidate wind profile is selected from the several candidate wind profiles obtained by combining at the k-th data fusion moment as the wind profile observation value at the k-th data fusion moment; the Kalman filter model is used to obtain the fused wind profile at the k-th data fusion moment based on the wind profile observation value at the k-th data fusion moment; wherein the Kalman filter model is trained using the fused wind profiles at the first K data fusion moments.

2. The method for obtaining a wind profile according to claim 1, characterized in that: Selecting a candidate wind profile from a number of candidate wind profiles as the wind profile observation value at the kth data fusion moment includes: For any candidate wind profile, the system deviation between the candidate wind profile and the fused wind profile at the k-1th data fusion moment is calculated, and the result standard deviation of the candidate wind profile is calculated. The consistency score of the candidate wind profile is calculated according to the system deviation and the result standard deviation of the candidate wind profile. The smaller the system deviation and the result standard deviation, the higher the consistency score. The candidate wind profile with the highest consistency score is selected from several candidate wind profiles as the wind profile observation value at the kth data fusion moment.

3. The method for obtaining a wind profile according to claim 1, characterized in that: The method of obtaining the fused wind profile at the kth data fusion moment based on the wind profile observation value at the kth data fusion moment by using the Kalman filter model includes: For any horizontal wind detection result of a data fusion altitude layer H in the wind profile observation value at the k-th data fusion moment, when the horizontal wind detection result of the data fusion altitude layer H is empty, or the deviation between the horizontal wind detection result of the data fusion altitude layer H and the reference result of the data fusion altitude layer H reaches the deviation threshold, the reference result of the data fusion altitude layer H is updated as the horizontal wind detection result of the data fusion altitude layer H, and the wind profile observation value after the median test at the k-th data fusion moment is obtained; The wind profile observation value after the median test at the k-th data fusion moment is input into the Kalman filter model to obtain the fused wind profile at the k-th data fusion moment; Among them, the benchmark result of any data fusion altitude layer H is the median of multiple reference horizontal wind results of the data fusion altitude layer H, and the multiple reference horizontal wind results of the data fusion altitude layer H include: the horizontal wind detection results in the wind profile observation values ​​of the data fusion altitude layer H and its upper and lower adjacent data fusion altitude layers at the kth data fusion moment, and the horizontal wind detection results in the fused wind profile of the data fusion altitude layer H and its upper and lower adjacent data fusion altitude layers at the k-1th data fusion moment.

4. The method for obtaining wind profile according to claim 1, characterized in that: The data sources of horizontal wind detection include ground automatic weather stations, laser wind radars, wind profiler radars, sounding data and GFS mode. The time interval between two adjacent data fusion moments is not less than the detection interval of the ground automatic weather stations and laser wind radars but less than the detection interval of other horizontal wind detection data sources. The horizontal wind detection results reported at each data fusion moment include at least the AWS horizontal wind detection results reported by the ground automatic weather station and the Lidar horizontal wind detection results reported by the laser wind radar.

5. The method for obtaining wind profile according to claim 4, characterized in that: The horizontal wind detection results reported from various horizontal wind detection data sources are corrected in the vertical direction by distance correction, including: The AWS horizontal wind detection result at the kth data fusion moment is used to perform data correction on the Lidar horizontal wind detection result to obtain the corrected Lidar horizontal wind detection result at the kth data fusion moment; the corrected Lidar horizontal wind detection result at the kth data fusion moment is linearly interpolated in the vertical direction to obtain the corrected Lidar horizontal wind detection results at each data fusion altitude layer; When the kth data fusion moment also includes horizontal wind detection results reported by other horizontal wind detection data sources, the horizontal wind detection results reported by other horizontal wind detection data sources are linearly interpolated in the vertical direction to obtain the horizontal wind detection results of other horizontal wind detection data sources at each data fusion altitude layer.

6. The method for obtaining wind profile according to claim 5, characterized in that: The corrected Lidar horizontal wind detection result obtained at the kth data fusion moment includes: Substitute the AWS horizontal wind detection result at the kth data fusion moment into the near-surface gale model inversion to obtain the horizontal wind inversion result at any height within the near-surface height range; For the Lidar horizontal wind detection result at any height within the near-ground height range at the kth data fusion moment, when the error between the Lidar horizontal wind detection result and the horizontal wind inversion result at the current height is less than the error threshold, the Lidar horizontal wind detection result at the current height is retained, otherwise the Lidar horizontal wind detection result at the current height is replaced by the horizontal wind inversion result at the current height to obtain the corrected Lidar horizontal wind detection result at the kth data fusion moment.

7. The method for obtaining wind profile according to claim 5, characterized in that: The fused horizontal wind detection results at each data fusion altitude layer include: When the kth data fusion moment only includes AWS horizontal wind detection results and Lidar horizontal wind detection results, the corrected Lidar horizontal wind detection results at each data fusion altitude layer are used as the fused horizontal wind detection results at each data fusion altitude layer; When the kth data fusion moment includes horizontal wind detection results reported by other horizontal wind detection data sources in addition to AWS horizontal wind detection results and Lidar horizontal wind detection results, the corrected Lidar horizontal wind detection results at each data fusion altitude layer and the horizontal wind detection results of other horizontal wind detection data sources are weightedly calculated to obtain the fused horizontal wind detection result at the data fusion altitude layer.

8. The method for obtaining wind profile according to claim 7, characterized in that: The weighted calculation of the corrected Lidar horizontal wind detection results at each data fusion altitude layer and the horizontal wind detection results from other horizontal wind detection data sources includes: According to the weight combination corresponding to each altitude layer range, the corrected Lidar horizontal wind detection results at each data fusion altitude layer within the altitude layer range and the horizontal wind detection results reported by other horizontal wind detection data sources are weighted averaged; different altitude layer ranges correspond to different weight combinations, and the lower the altitude layer range, the higher the weight of the corrected Lidar horizontal wind detection result.

9. The method for obtaining a wind profile according to claim 8, characterized in that: When the WPR horizontal wind detection result reported by the wind profile radar is also included at the kth data fusion moment, it is determined that (1) the weight combination corresponding to the altitude range below 3 km includes the weight of the corrected Lidar horizontal wind detection result being 0.6 and the weight of the WPR horizontal wind detection result being 0.4; and (2) the weight combination corresponding to the altitude range from 3 km to 6 km includes the weight of the corrected Lidar horizontal wind detection result being 0.3 and the weight of the WPR horizontal wind detection result being 0.7; When the kth data fusion moment also includes the WPR horizontal wind detection result reported by the wind profile radar and the GFS horizontal wind detection result reported by the GFS mode, it is determined that (1) the weight combination corresponding to the altitude range below 3 km includes the weight of the corrected Lidar horizontal wind detection result of 0.5, the weight of the WPR horizontal wind detection result of 0.2, and the weight of the GFS horizontal wind detection result; (2) the weight combination corresponding to the altitude range from 3 km to 6 km includes the weight of the corrected Lidar horizontal wind detection result of 0.2, the weight of the WPR horizontal wind detection result of 0.

3. The weight of the wind detection result is 0.4, and the weight of the GFS horizontal wind detection result is 0.4; (3) The weight combination corresponding to the altitude range of 6 km to 10 km includes the weight of the corrected Lidar horizontal wind detection result of 0, the weight of the WPR horizontal wind detection result of 0.3, and the weight of the GFS horizontal wind detection result of 0.7; (4) The weight combination corresponding to the altitude range above 10 km includes the weight of the corrected Lidar horizontal wind detection result of 0, the weight of the WPR horizontal wind detection result of 0, and the weight of the GFS horizontal wind detection result of 1; When the kth data fusion moment also includes the WPR horizontal wind detection result reported by the wind profile radar, the GFS horizontal wind detection result reported by the GFS model, and the Sonde horizontal wind detection result reported by the sounding data, it is determined that (1) the weight combination corresponding to the altitude range below 3 km includes the weight of the corrected Lidar horizontal wind detection result of 0.35, the weight of the WPR horizontal wind detection result of 0.15, the weight of the GFS horizontal wind detection result of 0.25, and the weight of the Sonde horizontal wind detection result of 0.25; (2) the weight combination corresponding to the altitude range of 3 km to 6 km includes the weight of the corrected Lidar horizontal wind detection result of 0.2, the weight of the WPR horizontal wind detection result of 0. .25, the weight of the GFS horizontal wind detection result is 0.25, the weight of the Sonde horizontal wind detection result is 0.3; (3) The weight combination corresponding to the altitude range of 6km to 10km includes the weight of the corrected Lidar horizontal wind detection result is 0, the weight of the WPR horizontal wind detection result is 0.2, the weight of the GFS horizontal wind detection result is 0.4, and the weight of the Sonde horizontal wind detection result is 0.4; (4) The weight combination corresponding to the altitude range above 10km includes the weight of the corrected Lidar horizontal wind detection result is 0, the weight of the WPR horizontal wind detection result is 0, the weight of the GFS horizontal wind detection result is 0.5, and the weight of the Sonde horizontal wind detection result is 0.

5.

10. The method for obtaining wind profile according to claim 5, characterized in that: The near-surface wind model is: V2=V1(Z2 / Z1) α Among them, V2 is the AWS horizontal wind detection result at the kth data fusion moment, Z2 is the equipment installation height of the ground automatic weather station from the ground; V1 is the horizontal wind inversion result at any height Z1; Z0 represents the ground roughness in the area where the ground automatic weather station is located.

Images