A quality control method for wind data in L-band high-altitude meteorological sounding

Abstract

The application relates to a quality control method for wind data in L-band high-altitude meteorological detection, which comprises the following steps: acquiring wind data, wherein the wind data comprises time data, height data, elevation angle data, slant range data and azimuth angle data; performing quality inspection on the height data, marking the height data as abnormal or normal; repairing the height data marked as abnormal; dividing the elevation angle data, the slant range data and the azimuth angle data into several segments according to the height, respectively determining the quality inspection standards of the segment data; performing quality inspection on the elevation angle data, the slant range data and the azimuth angle data according to the quality inspection standards, marking the elevation angle data, the slant range data and the azimuth angle data as abnormal or normal; repairing the elevation angle data, the slant range data and the azimuth angle data marked as abnormal; and outputting the quality inspection result and the repair result of the wind data.

Description

TECHNICAL FIELD

[0001] The application relates to a quality control method for wind data in L-band high-altitude meteorological sounding, and belongs to the field of high-altitude wind measurement. BACKGROUND

[0002] At present, the temperature, pressure, humidity and other data in the L-band second-level sounding data mainly rely on manual screening, and the second-level wind data is still in the stage of manually marking the sounding termination layer, and has not been deeply screened. In the face of a large amount of L-band second-level wind data, if the manual screening is also used like the temperature, pressure and humidity data, not only a large amount of manpower and time cost will be consumed, but also the manual screening of the wind data has subjectivity and inconsistency, different personnel may make different judgments on the same data, which may affect the reliability and consistency of the data. In summary, in order to efficiently and accurately process a large amount of L-band second-level wind data, an automatic quality control method is needed to realize rapid, accurate and unified data judgment and processing.

[0003] CN115525637A discloses a wind vector observation data quality control and processing method, system and equipment: the wind vector observation data is decomposed into vectors, the vector decomposition results of the wind vector observation data are used for data validity judgment, the singular values of the U component and the V component after vector decomposition are obtained by using the interpolation method or the statistical method to obtain the quality controlled vector decomposition results, and the quality controlled vector decomposition results are combined according to the parallelogram rule to obtain the quality controlled wind vector observation data. The patent performs quality control on the wind data based on the fixed point position (the data acquisition device is a wind measuring instrument), rather than the wind data obtained based on the radar detection of the moving point position (the data acquisition device is a radar detection sounding balloon). Overall, their application scenarios are different, the input of the method is different, and the method logic is also different. SUMMARY

[0004] In order to overcome the problems in the prior art, the application designs a quality control method for wind data in L-band high-altitude meteorological sounding, which considers that the wind data has different variation laws with the increase of the sounding height, and uses the height data to segmentally detect the wind data, so that the wind data quality control precision can be effectively improved.

[0005] In order to achieve the above purpose, the application adopts the following technical scheme:

[0006] A quality control method for wind data in L-band high-altitude meteorological sounding, comprising the following steps:

[0007] Obtaining wind data, including time data, height data, elevation angle data, slant range data and azimuth angle data;

[0008] The height data is inspected, and the height data is marked as abnormal or normal; the height data marked as abnormal is repaired;

[0009] The elevation data, slant range data and azimuth data are divided into several segments according to the height, and the inspection standards of the segment data are determined respectively; the elevation data, slant range data and azimuth data are inspected according to the inspection standards, and the elevation data, slant range data and azimuth data are marked as abnormal or normal; the elevation data, slant range data and azimuth data marked as abnormal are repaired;

[0010] The inspection results of the wind data and the repair results are output.

[0011] Further, the height data is inspected, and the height data is marked as abnormal or normal; the height data marked as abnormal is repaired;

[0012] The height data is divided into several segments; the median of each segment is calculated, and the upper and lower limits of each segment are set according to the median of each segment and the empirical offset value; the height data to be inspected is compared with the upper and lower limits of the segment to which it belongs, and the height data is marked as abnormal or normal according to the comparison result.

[0013] Further, the height data is inspected, and the height data is marked as abnormal or normal; the height data marked as abnormal is repaired;

[0014] The height data is detected by bidirectional longest monotone subsequence, and the longest positive sequence and the longest reverse sequence are obtained; the height data belonging to the longest positive sequence or the longest reverse sequence is marked as normal, and the height data not belonging to the positive sequence and not belonging to the reverse sequence is marked as abnormal.

[0015] Further, the height data is inspected, and the height data is marked as abnormal or normal; the height data marked as abnormal is repaired;

[0016] The height data is divided into several clusters by using DBSCAN algorithm; the height data not belonging to any cluster is marked as abnormal, and the height data belonging to any cluster is marked as normal.

[0017] Further, the height data is inspected, and the height data is marked as abnormal or normal; the height data marked as abnormal is repaired;

[0018] The height data is fitted to obtain a height fitting curve; the residual of the fitting value and the detection value of each height data is calculated; the reference interval is determined according to the distribution characteristics of the residual, and the residual is compared with the reference interval, and the height data corresponding to the residual is marked as abnormal or normal according to the comparison result.

[0019] Further, the height data marked as abnormal is repaired, and the height data is marked as abnormal or normal; the height data marked as abnormal is repaired;

[0020] The height data is locally weighted fitted to obtain a height fitting curve; the value of the height data marked as abnormal is replaced by the fitting value.

[0021] Further, the quality inspection on the elevation data and the slant range data comprises the following steps:

[0022] The distribution characteristics of the change rate of the elevation data and the slant range data in each segment are calculated; according to the distribution characteristics, the reference interval of the change rate of the elevation data and the slant range data in each segment is determined as the quality inspection standard; if the data change rate is not in the reference interval of the segment to which it belongs, the data is marked as abnormal; otherwise, it is marked as normal.

[0023] Further, the quality inspection on the elevation data and the slant range data comprises the following steps:

[0024] The elevation data and the slant range data are fitted respectively to obtain fitting curves; the residual error of the fitting value and the detected value of the elevation data and the slant range data and the distribution characteristics of the residual error are calculated; the reference interval is determined according to the distribution characteristics of the residual error; if the residual error is not in the reference interval, the elevation data and the slant range data corresponding to the residual error are marked as abnormal; otherwise, they are marked as normal.

[0025] Further, the quality inspection on the azimuth data comprises the following steps:

[0026] The vector wind data is calculated according to the azimuth data; the distribution characteristics of the change rate of the vector wind data in each segment are calculated; according to the distribution characteristics, the reference interval of the change rate of the vector wind data in each segment is determined as the quality inspection standard; if the change rate of the vector wind data is not in the reference interval of the segment to which it belongs, the azimuth data corresponding to the vector wind data is marked as abnormal; otherwise, it is marked as normal.

[0027] Further, the repair on the elevation data, the slant range data and the azimuth data is as follows:

[0028] The elevation data and the slant range data are locally weighted fitted respectively to obtain the fitting curves of the elevation data and the slant range data; the values of the elevation data and the slant range data marked as abnormal are replaced by the fitting values;

[0029] The vector wind data is calculated according to the azimuth data; the vector wind data is locally weighted fitted to obtain the fitting curve of the vector wind; the residual error of the fitting value and the detected value of each vector wind data is calculated; the reference interval is determined according to the distribution characteristics of the residual error; according to the reference interval, each vector wind data is marked as abnormal or normal; the value of the vector wind data marked as abnormal is replaced by the fitting value, and the azimuth data is back calculated according to the fitting value of the vector wind data; the value of the azimuth data is replaced by the back calculated value.

[0030] Compared with the prior art, the present application has the following characteristics and beneficial effects:

[0031] The present application considers that the elevation angle data, slant range data and azimuth angle data have different change rules with the increase of the sounding height, and uses the height data to segmentally check the elevation angle data, slant range data and azimuth angle data, so that the quality control precision of the elevation angle data, slant range data and azimuth angle data can be effectively improved.

[0032] The present application carries out data repair on abnormal height data, and carries out segmental quality inspection on the elevation angle data, slant range data and azimuth angle data according to the repaired height data, so that the quality control precision of the elevation angle data, slant range data and azimuth angle data can be further improved.

[0033] The present application provides various height data quality inspection methods: first, the median of each segment of height data is calculated, and the upper and lower bounds are obtained based on the manual experience offset of the median, so that most of the abnormal data that violate the common sense can be screened out, and the subsequent screening program of the height data is reduced. Then, the bidirectional longest monotone subsequence algorithm is used to determine the change trend of the height data, and the abnormal data is screened out according to the change trend; the DBSCAN clustering algorithm is further used to screen out the abnormal data of the drift type; and the residual error of the height data is calculated by fitting the curve, and whether the height data is abnormal is judged according to the residual error and the height data is repaired by using the fitting value. The present application automatically screens abnormal height data from different angles, which can effectively improve the reliability of the height data, and further improve the quality control effect of the subsequent elevation angle data, slant range data and azimuth angle data.

[0034] The present application uses the height to segmentally calculate the change rate of the elevation angle data, slant range data and azimuth angle data, and judges whether the elevation angle data, slant range data and azimuth angle data are abnormal according to the change rate; in addition, the vector wind data is calculated by using the azimuth angle data, and whether the azimuth angle data is abnormal is judged according to the change rate of the vector wind data, so that the abnormal value of the elevation angle data, slant range data and azimuth angle data is automatically screened from multiple angles, and the reliability of the data can be effectively improved.

[0035] The present application uses the azimuth angle data to calculate the vector wind data and the vector wind fitting curve, and uses the vector wind fitting value to screen and repair abnormal azimuth angle data, so that the accuracy of the azimuth angle data can be effectively improved. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 is the flow chart of the present application;

[0037] Figure 2 is the experience threshold detection schematic diagram of height data;

[0038] Figure 3 is the azimuth angle continuous processing. DETAILED DESCRIPTION

[0039] The present application will be described in more detail below in combination with embodiments.

[0040] AsFigure 1 As shown in the method for quality control of wind data in L-band high-altitude meteorological sounding, the method comprises the following steps:

[0041] The sounding system performs high-altitude atmospheric sounding according to a sampling period, and obtains wind data by continuously tracking a sounding instrument (such as a sounding balloon) by radar. The observation data obtained in each sounding process includes time data, height data, elevation angle data, slant range data, and azimuth angle data. The height, elevation angle, slant range, and azimuth angle at the same time constitute a group of data.

[0042] Each height data is subjected to at least one quality inspection, and if the result of any quality inspection is abnormal, the height data is marked as abnormal. The height data marked as abnormal is repaired.

[0043] Similarly, the elevation angle data, slant range data, and azimuth angle data are subjected to at least one quality inspection, and if the result of the quality inspection of any one of the three attributes of the elevation angle, slant range, and azimuth angle at the same time is abnormal, the group of data is considered abnormal. The height data, elevation angle data, slant range data, and azimuth angle data marked as abnormal are repaired. In particular, in this embodiment, the height data, elevation angle data, slant range data, and azimuth angle data are divided into several segments according to the height data after quality inspection, and the quality inspection standards of the height data, elevation angle data, slant range data, and azimuth angle data of each segment are determined; the height data, elevation angle data, slant range data, and azimuth angle data are subjected to quality inspection according to the quality inspection standards, and the height data, elevation angle data, slant range data, and azimuth angle data are marked as abnormal or normal.

[0044] The quality inspection result and the repair result of the observation data are output.

[0045] In one embodiment, the quality inspection of the height data comprises the following steps: dividing the height data into several segments according to time; calculating the median of each segment, and setting the upper and lower limits of each segment according to the median of each segment and an empirical offset value; comparing the height data to be inspected with the upper and lower limits of the segment to which it belongs, and marking the height data as normal if it is within the upper and lower limits; otherwise, marking it as abnormal.

[0046] In one embodiment, the quality inspection of the height data comprises the following steps: performing bidirectional longest monotone subsequence detection on the height data to obtain a longest positive sequence and a longest reverse sequence; marking the height data belonging to the longest positive sequence or the longest reverse sequence as normal, and marking the height data not belonging to the longest positive sequence and not belonging to the longest reverse sequence as abnormal.

[0047] In one embodiment, the quality inspection of the height data comprises the following steps: dividing the height data into several clusters using the DBSCAN algorithm; marking the height data not belonging to any of the clusters as abnormal, and marking the height data belonging to any of the clusters as normal.

[0048] In one embodiment, the quality inspection of the height data comprises the following steps: fitting the height data to obtain a height fitting curve; calculating the residual of each height data between the fitting value and the detected value; determining a reference interval according to at least one distribution characteristic of the residual, which includes but is not limited to mean, variance, standard deviation, median, etc. The following embodiments are the same as this and will not be repeated. If the residual is not in the reference interval, the height data corresponding to the residual is marked as abnormal; otherwise, it is marked as normal.

[0049] In one embodiment, the height data marked as abnormal is repaired, comprising the following steps:

[0050] The height data is locally weighted fitted by using the LOESS algorithm to obtain a height fitting curve; and the value of the height data marked as abnormal is replaced by the fitting value.

[0051] In one embodiment, the quality inspection of the elevation data, slant range data and azimuth data comprises the following steps: dividing the elevation data, slant range data and azimuth data into multiple segments according to the height; calculating the rate of change of the elevation data, slant range data and azimuth data; calculating the distribution characteristic of the rate of change of the elevation data, slant range data and azimuth data in each segment, and determining the reference interval of the rate of change of the elevation data, slant range data and azimuth data in each segment according to the distribution characteristic; if the rate of change of certain elevation data, slant range data and azimuth data is not in the reference interval of the segment to which it belongs, the elevation data, slant range data and azimuth data are marked as abnormal; otherwise, they are marked as normal. Before the segmentation, the azimuth data is pre-processed for continuity.

[0052] In one embodiment, the quality inspection of the elevation data and slant range data comprises the following steps: fitting the elevation data and slant range data to obtain a fitting curve of the elevation data and slant range data; calculating the residual of each elevation data and slant range data between the fitting value and the detected value; determining a reference interval according to the distribution characteristic of the residual; if the residual is not in the reference interval, the elevation data and slant range data corresponding to the residual are marked as abnormal; otherwise, they are marked as normal.

[0053] In one embodiment, the quality inspection of the azimuth data comprises the following steps:

[0054] The vector wind data (zonal wind, meridional wind) is calculated according to the azimuth data; the vector wind data is divided into multiple segments according to the height; the rate of change of each vector wind data is calculated; the distribution characteristic of the rate of change of the vector wind data in each segment is calculated; the reference interval of the rate of change of the vector wind data in each segment is determined according to the distribution characteristic; if the rate of change of certain vector wind data is not in the reference interval of the segment to which it belongs, the vector wind data or the azimuth data corresponding to the vector wind data is marked as abnormal; otherwise, it is marked as normal.

[0055] In the prior art, after receiving the high-altitude atmospheric sounding data, the technician sequentially checks the data and marks the data as reliable data or suspicious data. For suspicious data, there is sufficient evidence to prove that it is error data, which is directly excluded; or special marking is performed to prompt the technician to further analyze and process, and after confirming that it is error data, it is deleted, lacking a repair method for abnormal values. The embodiment provides a wind measurement data repair method, including the following steps:

[0056] The LOESS algorithm is used for local weighted fitting of the elevation angle data and the slant range data to obtain a fitting curve; and the values of the elevation angle data and the slant range data marked as abnormal are replaced by the fitting values.

[0057] The LOESS algorithm is used for local weighted fitting of the vector wind data (meridional wind and latitudinal wind) to obtain a vector wind fitting curve; the residuals of the fitting values and the detection values of each vector wind data are calculated; a reference interval is determined according to the distribution characteristics of the residuals; each vector wind data is marked as abnormal or normal according to the reference interval. The values of the vector wind data marked as abnormal are replaced by the fitting values, and the azimuth angle data is inversely deduced according to the vector wind data fitting values; and the values of the azimuth angle data are replaced by the inversely deduced values.

[0058] In one embodiment, a quality control method for wind measurement data in L-band high-altitude meteorological sounding includes the following steps:

[0059] The L-band sounding system measures a set of observation data every second, including height, elevation angle, slant range and azimuth angle, and records and saves them in the form of [time T, height H, elevation angle E, slant range S, azimuth angle A]. The height, elevation angle, slant range and azimuth angle are backed up to obtain backup arrays H recr , E recr , S recr , A recr , which are used to store the repaired data. Label arrays Tag H , Tag E , Tag S , Tag A are set to store the abnormality determination results of the height, elevation angle, slant range and azimuth angle data, and an array Tag is set to store the quality control result label of each set of data, wherein 1 is normal and 0 is abnormal, and the initial values of all Tag arrays are 1.

[0060] QC1 stage:

[0061] The height data H in the historical data set is divided into a segment every 2 minutes according to time, and the median of the segmented median is obtained by counting. The segmented median offset set sequence is

[0062] U={mk +(k+6.0)(rk*1.4)|k≥0,k∈Z}

[0063] L={m k +(k+4.5)(rk*1.2)|k≥0,k∈Z}

[0064] Among them, U and L are the median set sequences of each segment of the upper and lower bounds of the empirical threshold, k is the segment number, m k is the median of the kth segment, r is an empirical constant, which is 220 in this embodiment, and Z represents an integer; the obtained median set sequence is subjected to row linear interpolation, and the interpolation formula is:

[0065]

[0066] Among them, y(x) is a linear interpolation function, where x is the time of the point to be interpolated, (x a ,y a ) and (x b ,y b ) are the coordinates of the two median points before and after the interpolation point. After interpolation, the upper and lower bound curves of the height data empirical threshold Thold are obtained up (T i ) and Thold low (T i ),like Figure 2 As shown. Label the height data according to the following formula:

[0067]

[0068] Where H[i] is the value of the i-th point in the height data, and T[i] is the value of the time at the i-th point.

[0069] According to the results of the above steps, the bidirectional longest monotone subsequence detection is performed on the height data, that is, the longest monotone increasing subsequence is used in the positive sequence to obtain the sequence set Set LIS , use the longest monotonically decreasing subsequence in reverse order to get the sequence set Set LDS , take the union of the two sequence sets to get the set Set. Tag the data according to the following formula:

[0070]

[0071] The sequence of retained data sets obtained by the above steps is recorded as D, and the DBSCAN algorithm is used to cluster D. Initially, all objects in the given data set D are marked as unvisited. DBSCAN randomly selects an unvisited object p, marks p as visited, and checks whether the ε-neighborhood of p contains at least MinPts objects. If not, p is marked as a noise point, otherwise a new cluster C is created for p, and all objects in the ε-neighborhood of p are placed in the candidate set N. DBSCAN iteratively adds objects in N that do not belong to other clusters to C. In this process, the corresponding object p marked as unvisited in N is ′ , DBSCAN marks it as visited and checks its ε-neighborhood. If p's ε-neighborhood contains at least MinPts objects, then all objects in p's ε-neighborhood are added to N. DBSCAN continues to add objects to C until C cannot be expanded, that is, until N is empty. At this point, cluster C is generated. In order to find the next cluster, DBSCAN randomly selects an unvisited object from the remaining objects. The clustering process continues until all objects are visited. The final output is Cs, which is the union of all clusters. Tag the data according to the following formula:

[0072]

[0073] Among them, ε-neighborhood: The ε-neighborhood of an object is the space with the object as the center and ε as the radius. Core object: The user specifies a parameter MinPts to specify the density threshold of the dense area. If an ε-neighborhood contains at least MinPts objects, then the object is called a core object. Direct density reachability: For a specified object set K, there are objects p and q. If object p is in the neighborhood of object q and q is a core object, then object p is said to be directly density reachable from object q. Density reachability: Suppose there is an object chain p1, p2, ..., p n And p1=q,p n =p. If for p i (1≤i≤n), there is p i+1 It is from p i If p is directly density-reachable with respect to ε and MinPts, then p is said to be density-reachable from object q with respect to ε and MinPts. Density connected: For a given set of objects K, if there exists an object O such that p and q are density-reachable from O with respect to ε and MinPts, then objects p and q are density-connected with respect to ε and MinPts. Noise: For an object p, if its ε-neighborhood contains fewer than MinPts objects, then the object is called a noise point.

[0074] Let the reserved data set sequence obtained in the above steps be D H, use the Loess fitting algorithm to check for missed judgments and recover from false judgments on the height data. Set the parameters: r represents the number of selected nearest neighbors, that is, the fitting window size; t represents the number of rounds of fitting required, and the present invention sets t = 2; W(x) is a weight function, including the following properties: W(x)>0, |x|<1; W(-x)=W(x); when x≥0, W(x) is non-increasing; W(x)=0, |x|≥1. Calculate the initial weight of each point in the current window

[0075]

[0076] Among them, h i is x i The distance to its rth nearest neighbor, x i is the current point, x k Represents sequence D H The horizontal coordinate value of the kth point in w, that is, time, k∈{1, 2, ..., n}; k (x i ) represents x i is the weight of other points in the window of the current point.

[0077] The fitting curve of the current window is calculated by the least squares method. The formula is as follows:

[0078]

[0079] Among them, y k is the dependent variable; β0,…,β d represents the coefficient of the fitting curve; d represents the order of the selected model; n represents the sequence D H The number of data points in .

[0080] Then calculate the fitted value

[0081]

[0082] in, Represents sequence D H The ordinate fitting value of the kth point in ; represents the estimated value of the j-th coefficient in the model.

[0083] Let the quadratic weight function be B,

[0084]

[0085] Calculate the residual e for each point i ,

[0086]

[0087] Let s be {|e iThe median of {H[i], 1≤i≤n} is defined as the robust weight

[0088]

[0089] The new fitting value is calculated by the least square fitting d-degree polynomial with weight δ k w k (x i ) of the fitting residual.

[0090] The process of residual calculation, robust weight definition and new fitting value calculation is repeated t-1 times, and the final fitting curve is denoted as LH.

[0091] The fitting residual H diff is calculated, and the mean μ and the standard deviation σ are counted. According to the 3sigma principle of Gaussian distribution, the height data is marked with Tag:

[0092]

[0093] According to the Tag marking, the height data is repaired with Loess fitting value:

[0094] H recr [i]=LH[i],ifTag H [i]=1

[0095] For the elevation data E and the slant range data S, according to the repaired height H recr , it is segmented (such as segmented every 1km), and the elevation change rate E cr and the slant range S cr change rate of all points are counted, and the mean μ E , μ S and the standard deviation σ E , σ S of each segment change rate are calculated, and according to the 3sigma principle of Gaussian distribution, the elevation and slant range data are marked:

[0096]

[0097] The reserved data set sequences obtained by the above steps are denoted as D E and D S , and the Loess fitting algorithm is used to check the missing judgment and recover the misjudgment of the elevation data E and the slant range data S respectively. The process is the same as the missing judgment check and misjudgment recovery of the height data H, and the repaired results E recr and S recr are obtained.

[0098] In the QC2 stage:

[0099] The azimuth angle is processed continuously to solve the jump of azimuth angle data from 360° to 0°, such as Figure 3 shown.

[0100] For the azimuth data A, according to the repaired height H recr Divide it into segments and calculate the azimuth change rate A of all points cr , calculate the mean μ of the rate of change A and standard deviation σ A , the azimuth data are marked according to the 3sigma principle of Gaussian distribution:

[0101]

[0102] According to the restored elevation data E recr With slope distance data S recr And the azimuth data A is used to calculate the vector wind data, that is, the meridional wind (U wind) W U and zonal wind (V wind) W V .

[0103]

[0104] According to the height H after repair recr Divide it into segments and calculate the meridional wind change rate W of all points Ucr and zonal wind rate of change Calculate the mean μ of the rate of change U 、μ V and standard deviation σ U , σ V , the azimuth angle data is marked according to the 3sigma principle of Gaussian distribution,

[0105]

[0106] The reserved dataset sequences obtained in the above steps are respectively D A , and use the Loess fitting algorithm to calculate the meridional wind W U With the zonal wind W V Perform missed detection and false positive recovery. The process is the same as the missed detection and false positive recovery of height data H, and the repair result U is obtained. recr and V recr The azimuth repair data A_recr is obtained by reverse calculation using the formula.

[0107] Tag=Tag E * Tags S * Tags H * Tags A

[0108] Finally, output {T, E, S, H, A, Tag} to get the quality control result of the wind data; output {T, E recr ,S recr ,H recr ,A recr ,Tag E ,Tag S ,Tag H ,Tag A} to get the wind data repair result.

[0109] In this embodiment, two stages of QC1 and QC2 are included. In the QC1 stage, for the height data, first, two threshold curves are obtained based on manual experience statistics, and most of the data that violates the common sense is first removed through the two threshold curves to reduce interference for subsequent judgment; second, the bidirectional longest monotone subsequence algorithm is used to determine the trend of the height data, and the sounding termination layer can be determined at the same time; then the DBSCAN clustering algorithm is used to filter out the abnormal data of the drift type; then the Loess algorithm is used to check the missed judgment and recover the misjudgment of the height data. For the slope data, the height after repair is used to segment the slope data, and the change rate of each point is counted, and the data exceeding 3 times the standard deviation of the change rate of each segment is marked as abnormal according to the 3sigma principle of Gaussian distribution, the elevation angle and range data are calculated according to the repaired height data and the geometric height calculation formula for abnormal detection, and finally the Loess algorithm is used for abnormal detection and repair of the elevation angle and range. In the QC2 stage, first, the azimuth angle is processed continuously to solve the jump of the azimuth angle data from 360° to 0°, and the change rate of the azimuth angle is counted according to the height, and the data exceeding 3 times the standard deviation of the change rate of each segment is marked as abnormal according to the 3sigma principle of Gaussian distribution, and then the vector wind data is calculated according to the repaired elevation angle and range and azimuth angle, the height data is used to segment and count the change rate of the vector wind data, and the abnormality is marked according to the 3sigma principle, and finally the Loess algorithm is used for abnormal detection and repair of the vector wind data, and finally the label of the source azimuth angle data and the repaired azimuth angle data are obtained. Compared with the traditional manual auditing method, the method proposed in this paper can not only greatly reduce the consumption of human resources, but also help to unify the quality control standard of sounding data, and effectively improve the reliability and accuracy of the quality control of sounding data.

[0110] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit the protection scope of the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should analyze that the technical solutions of the present application can be modified or replaced by equivalents without departing from the essence and scope of the technical solutions of the present application.

Claims

1. A quality control method for wind measurement data in L-band high-altitude meteorological detection, characterized in that: The following steps are involved: Obtain wind measurement data, including time data, altitude data, elevation data, slant range data and azimuth data; Perform quality inspection on the height data and mark it as abnormal or normal; repair the height data marked as abnormal; Dividing the elevation data, slant range data, and azimuth data into several segments according to their height, and determining quality inspection standards for the data of each segment; performing quality inspection on the elevation data, slant range data, and azimuth data according to the quality inspection standards, and marking the elevation data, slant range data, and azimuth data as abnormal or normal; Repair the elevation data, slant range data and azimuth data marked as abnormal; Outputting the quality inspection results and repair results of the wind measurement data; The quality inspection of the height data is as follows: Divide the height data into several segments; calculate the median of each segment, and set the upper and lower limits of each segment based on the median and empirical offset value; compare the height data to be inspected with the upper and lower limits of the segment to which it belongs, and mark the height data as abnormal or normal based on the comparison results; Perform bidirectional longest monotonic subsequence detection on the height data to obtain the longest positive sequence and the longest reverse sequence; mark the height data belonging to the longest positive sequence or the longest reverse sequence as normal, and mark the height data that does not belong to the longest positive sequence or the longest reverse sequence as abnormal; The DBSCAN algorithm is used to divide the height data into several clusters; the height data that does not belong to any cluster is marked as abnormal, and the height data that belongs to any cluster is marked as normal; the height data is fitted to obtain a height fitting curve; the residual of the fitting value and the detection value of each height data is calculated; a reference interval is determined based on the distribution characteristics of the residual, and the residual is compared with the reference interval. Based on the comparison result, the height data corresponding to the residual is marked as abnormal or normal; wherein, the height data marked as abnormal is repaired as follows: Perform local weighted fitting on the height data to obtain a height fitting curve; replace the values ​​of the height data marked as abnormal with the fitting values.

2. The quality control method for wind measurement data in L-band high-altitude meteorological detection according to claim 1, characterized in that: The quality inspection of the elevation angle data and the slant range data comprises the following steps: Calculate the distribution characteristics of the rate of change of the elevation angle data and slant range data in each segment; based on the distribution characteristics, determine the reference interval of the rate of change of the elevation angle data and slant range data in each segment as the quality inspection standard; if the data change rate is not within the reference interval of the segment to which it belongs, mark the data as abnormal; otherwise, mark it as normal.

3. The quality control method for wind measurement data in L-band high-altitude meteorological detection according to claim 1, characterized in that: It also includes the following quality inspection steps for elevation angle data and slant range data: Fit the elevation angle data and slant range data respectively to obtain fitting curves; calculate the residuals between the fitting values ​​and the detection values ​​of the elevation angle data and slant range data, as well as the distribution characteristics of the residuals; determine the reference interval based on the distribution characteristics of the residuals; If the residual is not within the reference interval, the elevation angle data and slant range data corresponding to the residual are marked as abnormal; otherwise, they are marked as normal.

4. The quality control method for wind measurement data in L-band high-altitude meteorological detection according to claim 1, characterized in that: The azimuth data is quality inspected, comprising the following steps: calculating vector wind data based on the azimuth data; calculating the distribution characteristics of the rate of change of the vector wind data in each segment; determining, based on the distribution characteristics, a reference interval of the rate of change of the vector wind data in each segment as a quality inspection standard; if the rate of change of the vector wind data is not within the reference interval of the segment to which it belongs, marking the azimuth data corresponding to the vector wind data as abnormal; otherwise, marking it as normal.

5. The quality control method for wind measurement data in L-band high-altitude meteorological detection according to claim 1 is characterized in that: The elevation data, slant range data and azimuth data are repaired as follows: Perform local weighted fitting on the elevation data and slant range data to obtain fitting curves for the elevation data and slant range data; replace the elevation data and slant range data values ​​marked as abnormal with the fitting values; calculate the vector wind data based on the azimuth data; perform local weighted fitting on the vector wind data to obtain a vector wind fitting curve; calculate the residual between the fitting value and the detection value of each vector wind data; and determine the reference interval based on the distribution characteristics of the residuals; According to the reference interval, each vector wind data is marked as abnormal or normal; the value of the vector wind data marked as abnormal is replaced with the fitting value, and the azimuth data is inferred based on the fitting value of the vector wind data; the value of the azimuth data is replaced with the inferred value.

Images