A nearest neighbor weighted interpolation method for meteorological elements based on terrain importance

Abstract

The present invention provides a nearest neighbor weighted interpolation method for meteorological elements based on terrain importance, including: constructing a terrain element importance evaluation model; obtaining a terrain element importance evaluation result of each original meteorological grid; obtaining the neighboring original meteorological grid of the target interpolation grid based on the closest spatial distance measurement basis; obtaining the original meteorological grid closest to the target interpolation grid in each terrain element dimension among the neighboring original meteorological grids, and weighting the target meteorological element value of the obtained original meteorological grid using the terrain element importance of the target interpolation grid to obtain the target meteorological element value of the target interpolation grid. The terrain importance quantitative evaluation model and the nearest neighbor weighted interpolation model of the present invention quantitatively evaluate the importance of terrain elements to target meteorological elements, quantify the correlation between terrain elements and target meteorological elements, realize the accurate interpolation of the target meteorological elements at the target interpolation position, and improve the accuracy of the interpolation result.

Description

Technical Field

[0001] The invention belongs to the technical field of meteorological data processing, and in particular relates to a nearest neighbor weighted interpolation method of meteorological elements based on terrain importance. Background Art

[0002] As the demand for the development of new energy power stations becomes more urgent, the scale of new energy is also growing rapidly. In the process of new energy construction and management, various basic data such as meteorology and topography are important bases for supporting the site selection and type selection of power stations and the calculation of power generation capacity. The unified management of the above data has a great promoting effect on the planning, design and operation of new energy power stations. Among the above data, meteorological data covers a wide area, a long time, a large amount of data, and the spatial scales of different data sources are not unified. Interpolating meteorological data into a unified spatial scale can achieve standard unified management of multi-source data and effectively improve data retrieval and calculation efficiency.

[0003] At present, meteorological data interpolation mainly includes nearest neighbor interpolation, linear interpolation, inverse distance weighting, Kriging interpolation and other methods. Among the above methods, the nearest neighbor interpolation mainly measures the spatial distance between the target interpolation object and the surrounding original data, and takes the closest data as the interpolation result; linear interpolation converts the value of the adjacent point to the interpolation object through the linear proportion of the spatial distance; inverse distance weighting takes the inverse ratio of the distance between the target interpolation object and its surrounding points as the weight, and takes the value of the surrounding points as the interpolation result after weighting; Kriging interpolation considers the spatial relationship between the points in the surrounding area of ​​the target object and the random error, and takes the optimal coefficient of the solution as the weight to obtain the interpolation result. The above interpolation methods are more based on the spatial distance as the weight basis of interpolation, and consider fewer factors, resulting in the need to improve the accuracy of meteorological data interpolation results. Summary of the invention

[0004] In view of the defects of the prior art, the present invention provides a nearest neighbor weighted interpolation method of meteorological elements based on terrain importance, which can effectively solve the above problems.

[0005] The technical solution adopted by the present invention is as follows:

[0006] The present invention provides a nearest neighbor weighted interpolation method of meteorological elements based on terrain importance, comprising the following steps:

[0007] Step S1, obtaining multiple original meteorological grids in the study area according to the grid form of the original meteorological data and the distribution in the study area; gridding the study area according to the spatial scale requirements of the target meteorological elements to be interpolated to form multiple target interpolation grids;

[0008] Step S2, determine the target meteorological element Y that needs to be interpolated for the target interpolation grid; according to the target meteorological element Y, determine the terrain meteorological comprehensive feature X = [G, U] closely related to the target meteorological element Y; wherein the terrain meteorological comprehensive feature X is the splicing feature of the terrain element feature G and the basic meteorological element feature U; the terrain element feature G = [g1, g2, ..., g n ],g1,g2,...,g n , respectively represent the first terrain element, the second terrain element, ..., the nth terrain element; basic meteorological element characteristics U = [u1, u2, ..., u m ],u1,u2,...,u m , respectively represent the first basic meteorological element, the second basic meteorological element, …, the mth basic meteorological element;

[0009] Step S3, constructing a terrain element importance evaluation model;

[0010] The terrain element importance evaluation model includes a clustering model and a random forest importance evaluation model; the terrain meteorological comprehensive characteristic values ​​and target meteorological element values ​​of all original meteorological grids in the study area form a training sample set; the terrain element importance evaluation model is trained using the training sample set to obtain a trained terrain element importance evaluation model; according to the trained terrain element importance evaluation model, the terrain element importance evaluation result of each original meteorological grid is obtained, which is expressed as: W i =(w i,1 ,w i,2 ,...,w i,n ), where W i represents the evaluation result of the importance of terrain elements of the original meteorological grid i, i = 1, 2, ..., N, N represents the total number of original meteorological grids; w i,1 ,w i,2 ,...,w i,n They represent the importance of the first terrain element of the original meteorological grid i to the target meteorological element Y, the importance of the second terrain element to the target meteorological element Y, …, the importance of the nth terrain element to the target meteorological element Y;

[0011] Step S4: for each target interpolation grid j in the study area, the target meteorological element value Y is obtained by the following method: j :

[0012] Step S4.1, obtaining a number of neighboring original meteorological grids of the target interpolation grid j based on the closest spatial distance metric;

[0013] Step S4.2, among several adjacent original meteorological grids, taking the first type of terrain element closest as the metric basis, obtain the first original meteorological grid I1 that is closest to the target interpolation grid j in the first type of terrain element dimension; taking the second type of terrain element closest as the metric basis, obtain the second original meteorological grid I2 that is closest to the target interpolation grid j in the second type of terrain element dimension; and so on, taking the nth type of terrain element closest as the metric basis, obtain the nth original meteorological grid I2 that is closest to the target interpolation grid j in the nth type of terrain element dimension. n ;

[0014] Step S4.3, determine the original meteorological grid where the center point of the target interpolation grid j is located, denoted as the original meteorological grid l, l = 1, 2, ..., N, and obtain the terrain element importance evaluation result W of the original meteorological grid l obtained in step S3 l =(w l,1 ,w l,2 ,...,w l,n ), as the terrain element importance of the target interpolation grid j;

[0015] Step S4.4, use the following formula to obtain the target meteorological element value Y of the target interpolation grid j: j :

[0016] Y j =Y(I1)w l,1 +Y(I2)w l,2 +...+Y(I n ) l,n

[0017] Where: Y(I1),Y(I2),...,Y(I n ), representing the first original meteorological grid I1, the second original meteorological grid I2, ..., the nth original meteorological grid I n The target meteorological element value.

[0018] Preferably, in step S1, the spatial resolutions of the original meteorological grid and the target interpolation grid are the same or different.

[0019] Preferably, in step S2, the target meteorological element Y is irradiance; the terrain element feature G includes three terrain elements, namely slope, aspect and altitude; the basic meteorological element feature G includes three basic meteorological elements, namely precipitation, air pressure and cloud cover.

[0020] Preferably, step S3 specifically comprises:

[0021] Step S3.1, assuming that the total number of original meteorological grids in the study area is N; for each original meteorological grid i, i = 1, 2, ..., N, obtain its terrain meteorological comprehensive characteristic value Xi and target meteorological element value Y i , forming training samples S i =(X i ,Y i ); Therefore, for N original meteorological grids, a training sample set S = [S1, S2, ..., S N ];

[0022] Step S3.2: Set the training sample set S = [S1, S2, ..., S N ] is input into the clustering model; the clustering model uses the similarity of the terrain and meteorological comprehensive features X between the training samples as the clustering feature, clusters the N training samples, and forms K sample subsets {C1, C2, ..., C K};

[0023] Step S3.3, each sample subset C k Input into the random forest importance evaluation model for training, k = 1, 2, ..., K. After the training is completed, each sample subset C is obtained k Evaluation results of terrain element importance W k =(w k,1 ,w k,2 ,...,w k,n ), where w k,1 ,w k,2 ,...,w k,n , represents the sample subset C k The importance of the first terrain element to the target meteorological element Y, the importance of the second terrain element to the target meteorological element Y, …, the importance of the nth terrain element to the target meteorological element Y;

[0024] Each sample subset C k Evaluation results of terrain element importance W k =(w k,1 ,w k,2 ,...,w k,n ), as the sample subset C k The importance evaluation results of terrain elements in each original meteorological grid.

[0025] Preferably, in step S3.1, the terrain meteorological comprehensive characteristic value X of the original meteorological grid i i The characteristic value of the terrain element G i =[g i,1 ,g i,2 ,...,g i,n ], for the fth terrain element value g i,f , f=1,2,...,n, and is obtained by the following method:

[0026] Obtain a topographic grid within the study area;

[0027] Filter terrain grids that overlap or intersect with the original meteorological grid i; take the average value of the f-th terrain element value of all the filtered terrain grids as the f-th terrain element value g of the original meteorological grid i. i,f , the formula is as follows:

[0028]

[0029] Where: num represents the number of terrain grids that overlap or intersect with the original meteorological grid i; Represents the fth terrain element value of the yth terrain grid that overlaps or intersects with the original meteorological grid i.

[0030] Preferably, step S3.2 is specifically:

[0031] Step S3.2.1, determine the number of sample subsets K according to the number of original meteorological grids N in the study area, the formula is:

[0032] K=min{N / 100,2 n}

[0033] Where: n is the number of terrain feature classes;

[0034] Step S3.2.2, training sample set S = [S1, S2, ..., S N ] as the initial subset C of clustering;

[0035] Step S3.2.3, for each training sample S i The comprehensive characteristic value of terrain and meteorology X i =(g i,1 ,g i,2 ,...,g i,n ,u i,1 ,u i,2 ,...,u i,m ) to obtain the standardized terrain and meteorological comprehensive characteristic value The standardization method is:

[0036] X i =(g i,1 ,g i,2 ,...,g i,n ,u i,1 ,u i,2 ,...,u i,m ), which is n+m-dimensional data; the standardization method of each dimensional data is the same;

[0037] For u i,m , using the formula To standardize;

[0038] Where: u i,m is the training sample S i The value of the mth basic meteorological element before standardization; is the training sample S i The standardized value of the mth basic meteorological element; m,min is the minimum value of the mth basic meteorological element of N training samples; u m,max is the maximum value of the mth basic meteorological element of N training samples;

[0039] Step S3.2.4, using the average distance indicator d between training samples avg Measure the distance between each sample subset, merge the sample subsets with the closest distance one by one; repeat the sample subset merging step until the number of sample subsets is equal to K, and generate a total of K types of sample subsets {C1, C2, ..., C K};

[0040] Among them: The calculation formula for the distance between sample subsets is:

[0041]

[0042]

[0043] in:

[0044] C k ,C h for any two different sample subsets;

[0045] d avg (C k ,C h ) is the sample subset C k ,C h The distance between

[0046] s a is the sample subset C k Samples in b is the sample subset C h Samples in

[0047] dist(s a ,s b ) represents sample s a and samples b The distance between

[0048] ‖s a -s b ‖2 represents sample s a and samples b The distance between

[0049] Representative samples a The standardized data of each dimension;

[0050] Representative samples b The standardized data of each dimension.

[0051] Preferably, step S3.3 is specifically:

[0052] Step S3.3.1, for each sample subset C k , both construct random forest importance evaluation models and execute steps S3.3.2 to S3.3.5;

[0053] Step S3.3.2, use self-sampling to select the sample subset C k The samples in are sampled to generate the samples required by each decision tree in the random forest; each sample has n terrain elements and m basic meteorological elements, so it has n+m features;

[0054] Step S3.3.3, in the process of generating each decision tree, randomly select n types of features from the n+m features of the sample as the input of the decision tree, and use the feature that reduces the root mean square error the most before and after the decision tree splits as the feature for splitting the decision tree;

[0055] Step S3.3.4, repeating steps S3.3.2 and S3.3.3 until a set number z of decision trees are generated;

[0056] Step S3.3.5, count the terrain feature g1, g2, ..., g in each decision tree n The contribution values ​​are e1, e2, ..., e n ; Among them, for the characteristic value e of the f-th terrain element f , f=1,2,...,n, calculated using the following formula:

[0057]

[0058] Where: Δrmse f is the root mean square error reduction contributed by the f-th terrain element before and after splitting in each tree building process, e f is the sum of the root mean square error reduction contributed by the f-th terrain element in the random forest model; d is the number of splits in which each terrain element feature participates in the tree building process;

[0059] Use the following formula, with contribution values ​​e1, e2, ..., e nAs n types of terrain features g1, g2, ..., g n The degree of influence on the target meteorological element Y is denoted as W k =(w k,1 ,w k,2 ,...,w k,n ):

[0060]

[0061] Among them, w k,f is the fth terrain feature g f The degree of influence w on the target meteorological element Y k,f .

[0062] Preferably, in step S4.1, the nine neighboring original meteorological grids of the target interpolation grid j are obtained by:

[0063] The coordinates of the center point of the target interpolation grid j are determined, and the original meteorological grid where the target interpolation grid j is located is determined according to the coordinates of the center point; the original meteorological grid and the 8 surrounding original meteorological grids are used as the 9 neighboring original meteorological grids of the target interpolation grid j.

[0064] The nearest neighbor weighted interpolation method of meteorological elements based on terrain importance provided by the present invention has the following advantages:

[0065] The present invention provides a nearest neighbor weighted interpolation method for meteorological elements based on terrain importance. The method proposes a terrain importance quantitative evaluation model and a nearest neighbor weighted interpolation model, which can quantitatively evaluate the importance of different terrain elements to target meteorological elements, quantify the correlation between terrain elements and target meteorological elements, and combine the spatial relationship of the target meteorological elements themselves to achieve accurate interpolation of target meteorological elements at the target interpolation position. The present invention can effectively improve the accuracy of meteorological data interpolation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 A schematic flow chart of a nearest neighbor weighted interpolation method for meteorological elements based on terrain importance provided by the present invention;

[0067] Figure 2 It is a schematic diagram of a method for acquiring terrain data of a meteorological grid according to an embodiment of the present invention;

[0068] Figure 3 Schematic diagram of terrain element importance evaluation carried out in an embodiment of the present invention;

[0069] Figure 4 It is a schematic diagram of the nearest original meteorological grid for determining a target interpolation grid according to an embodiment of the present invention. DETAILED DESCRIPTION

[0070] In order to make the technical problems, technical solutions and beneficial effects solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0071] The present invention provides a nearest neighbor weighted interpolation method for meteorological elements based on terrain importance. The method proposes a terrain importance quantitative evaluation model and a nearest neighbor weighted interpolation model, which can quantitatively evaluate the importance of different terrain elements to target meteorological elements, quantify the correlation between terrain elements and target meteorological elements, and combine the spatial relationship of the target meteorological elements themselves to achieve accurate interpolation of target meteorological elements at the target interpolation position. The present invention can effectively improve the accuracy of meteorological data interpolation results.

[0072] See also Figure 1 The present invention provides a nearest neighbor weighted interpolation method of meteorological elements based on terrain importance, comprising the following steps:

[0073] Step S1, obtaining multiple original meteorological grids in the study area according to the grid form of the original meteorological data and the distribution in the study area; gridding the study area according to the spatial scale requirements of the target meteorological elements to be interpolated to form multiple target interpolation grids;

[0074] In this step, the spatial resolutions of the original meteorological grid and the target interpolation grid are the same or different. For example, the original meteorological grid data is ERA5, the spatial resolution is 0.25°×0.25°, and the spatial resolution of the target interpolation grid is 30km×30km. The purpose of the present invention is to interpolate the meteorological data of the original meteorological grid of 0.25° into the meteorological data of the target interpolation grid of 30km.

[0075] Step S2, determine the target meteorological element Y that needs to be interpolated for the target interpolation grid; according to the target meteorological element Y, determine the terrain meteorological comprehensive feature X = [G, U] closely related to the target meteorological element Y; wherein the terrain meteorological comprehensive feature X is the splicing feature of the terrain element feature G and the basic meteorological element feature U; the terrain element feature G = [g1, g2, ..., g n ],g1,g2,...,g n , respectively represent the first terrain element, the second terrain element, ..., the nth terrain element; basic meteorological element characteristics U = [u1, u2, ..., u m ],u1,u2,...,u m , respectively represent the first basic meteorological element, the second basic meteorological element, …, the mth basic meteorological element;

[0076] For example, the target meteorological element Y is irradiance; the terrain element feature G includes three terrain elements, namely slope, aspect and altitude; the basic meteorological element feature G includes three basic meteorological elements, namely precipitation, air pressure and cloud cover. Therefore, in the present invention, for the irradiance that needs to be interpolated, not only the terrain elements such as slope, aspect and altitude that affect the irradiance are considered, but also the basic meteorological elements such as precipitation, air pressure and cloud cover that are closely related to the irradiance are considered. By comprehensively considering the associated terrain elements and basic meteorological elements, the accuracy of the irradiance that needs to be interpolated can be improved.

[0077] Step S3, constructing a terrain element importance evaluation model;

[0078] The terrain element importance evaluation model includes a clustering model and a random forest importance evaluation model; the terrain meteorological comprehensive characteristic values ​​and target meteorological element values ​​of all original meteorological grids in the study area form a training sample set; the terrain element importance evaluation model is trained using the training sample set to obtain a trained terrain element importance evaluation model; according to the trained terrain element importance evaluation model, the terrain element importance evaluation result of each original meteorological grid is obtained, which is expressed as: W i =(w i,1 ,w i,2 ,...,w i,n ), where W i represents the evaluation result of the importance of terrain elements of the original meteorological grid i, i = 1, 2, ..., N, N represents the total number of original meteorological grids; w i,1 ,w i,2 ,...,w i,n They represent the importance of the first terrain element of the original meteorological grid i to the target meteorological element Y, the importance of the second terrain element to the target meteorological element Y, …, the importance of the nth terrain element to the target meteorological element Y;

[0079] See also Figure 3 , step S3 is specifically as follows:

[0080] Step S3.1, assuming that the total number of original meteorological grids in the study area is N; for each original meteorological grid i, i = 1, 2, ..., N, obtain its terrain meteorological comprehensive characteristic value X i and target meteorological element value Y i , forming training samples S i =(X i ,Y i ); Therefore, for N original meteorological grids, a training sample set S = [S1, S2, ..., S N ];

[0081] In this step, the terrain meteorological comprehensive characteristic value X of the original meteorological grid ii The characteristic value of the terrain element G i =[g i,1 ,g i,2 ,...,g i,n ], for the fth terrain element value g i,f , f=1,2,...,n, and is obtained by the following method:

[0082] In the study area, multiple terrain grids are formed according to the small-scale gridding method;

[0083] Filter terrain grids that overlap or intersect with the original meteorological grid i; take the average value of the f-th terrain element value of all the filtered terrain grids as the f-th terrain element value g of the original meteorological grid i. i,f , the formula is as follows:

[0084]

[0085] Where: num represents the number of terrain grids that overlap or intersect with the original meteorological grid i; Represents the fth terrain element value of the yth terrain grid that overlaps or intersects with the original meteorological grid i.

[0086] For example, Figure 2 As shown, the study area has an original meteorological grid with a spatial resolution of 0.25°×0.25° and a terrain grid with a spatial resolution of 30m×30m. Therefore, the original meteorological grid and the terrain grid have different spatial resolutions. In the present invention, various terrain element values ​​of the original meteorological grid can be obtained through the spatial relationship between the original meteorological grid and the terrain grid.

[0087] The values ​​of the basic meteorological elements and target meteorological elements of the original meteorological grid i can be directly obtained. When interpolation is required for each period, the values ​​of the basic meteorological elements and target meteorological elements are taken as the average values ​​of the corresponding period.

[0088] Step S3.2: Set the training sample set S = [S1, S2, ..., S N ] is input into the clustering model; the clustering model uses the similarity of the terrain and meteorological comprehensive features X between the training samples as the clustering feature, clusters the N training samples, and forms K sample subsets {C1, C2, ..., C K};

[0089] Step S3.2 is specifically as follows:

[0090] Step S3.2.1, determine the number of sample subsets K according to the number of original meteorological grids N in the study area, the formula is:

[0091] K=min{N / 100,2 n}

[0092] Where: n is the number of terrain element classes; the meaning of this formula is: take N / 100 and 2 n The smaller value of is taken as the number of sample subsets K.

[0093] Step S3.2.2, training sample set S = [S1, S2, ..., S N ] as the initial subset C of clustering;

[0094] Step S3.2.3, for each training sample S i The comprehensive characteristic value of terrain and meteorology X i =(g i,1 ,g i,2 ,...,g i,n ,u i,1 ,u i,2 ,...,u i,m ) to obtain the standardized terrain and meteorological comprehensive characteristic value The standardization method is:

[0095] X i =(g i,1 ,g i,2 ,...,g i,n ,u i,1 ,u i,2 ,...,u i,m ), which is n+m-dimensional data; the standardization method of each dimensional data is the same;

[0096] For u i,m , using the formula To standardize;

[0097] Where: u i,m is the training sample S i The value of the mth basic meteorological element before standardization; is the training sample S i The standardized value of the mth basic meteorological element; m,min is the minimum value of the mth basic meteorological element of N training samples; u m,max is the maximum value of the mth basic meteorological element of N training samples;

[0098] Step S3.2.4, using the average distance indicator d between training samples avg Measure the distance between each sample subset, merge the sample subsets with the closest distance one by one; repeat the sample subset merging step until the number of sample subsets is equal to K, and generate a total of K types of sample subsets {C1, C2, ..., C K};

[0099] Among them: The calculation formula for the distance between sample subsets is:

[0100]

[0101] in:

[0102] C k ,C h for any two different sample subsets;

[0103] d avg (C k ,C h ) is the sample subset C k ,C h The distance between

[0104] s a is the sample subset C k Samples in b is the sample subset C h Samples in

[0105] dist(s a ,s b ) represents sample s a and samples b The distance between

[0106] ‖s a -s b ‖2 represents sample s a and samples b The distance between

[0107] Representative samples a The standardized data of each dimension;

[0108] Representative samples b The standardized data of each dimension.

[0109] Step S3.3, each sample subset C k Input into the random forest importance evaluation model for training, k = 1, 2, ..., K. After the training is completed, each sample subset C is obtained k Evaluation results of terrain element importance W k =(w k,1 ,w k,2 ,...,w k,n ), where w k,1 ,w k,2 ,...,w k,n , represents the sample subset C kThe importance of the first terrain element to the target meteorological element Y, the importance of the second terrain element to the target meteorological element Y, …, the importance of the nth terrain element to the target meteorological element Y;

[0110] Each sample subset C k Evaluation results of terrain element importance W k =(w k,1 ,w k,2 ,...,w k,n ), as the sample subset C k The importance evaluation results of terrain elements in each original meteorological grid.

[0111] Step S3.3 is specifically as follows:

[0112] Step S3.3.1, for each sample subset C k , both construct random forest importance evaluation models and execute steps S3.3.2 to S3.3.5;

[0113] Step S3.3.2, use self-sampling to select the sample subset C k The samples in are sampled to generate the samples required by each decision tree in the random forest; each sample has n terrain elements and m basic meteorological elements, so it has n+m features;

[0114] Step S3.3.3, in the process of generating each decision tree, randomly select n types of features from the n+m features of the sample as the input of the decision tree, and use the feature that reduces the root mean square error the most before and after the decision tree splits as the feature for splitting the decision tree;

[0115] Step S3.3.4, repeating steps S3.3.2 and S3.3.3 until a set number z of decision trees are generated;

[0116] Step S3.3.5, count the terrain feature g1, g2, ..., g in each decision tree n The contribution values ​​are e1, e2, ..., e n ; Among them, for the characteristic value e of the f-th terrain element f , f=1,2,...,n, calculated using the following formula:

[0117]

[0118] Where: Δrmse f is the root mean square error reduction contributed by the f-th terrain element before and after splitting in each tree building process, e fis the sum of the root mean square error reduction contributed by the f-th terrain element in the random forest model; d is the number of splits in which each terrain element feature participates in the tree building process;

[0119] Use the following formula, with contribution values ​​e1, e2, ..., e n As n types of terrain features g1, g2, ..., g n The degree of influence on the target meteorological element Y is denoted as W k =(w k,1 ,w k,2 ,...,w k,n ):

[0120]

[0121] Among them, w k,f is the fth terrain feature g f The degree of influence w on the target meteorological element Y k,f .

[0122] Step S4: for each target interpolation grid j in the study area, the target meteorological element value Y is obtained by the following method: j :

[0123] Step S4.1, obtaining a number of neighboring original meteorological grids of the target interpolation grid j based on the closest spatial distance metric;

[0124] In practical applications, the nine neighboring original meteorological grids of the target interpolation grid j are obtained by:

[0125] The coordinates of the center point of the target interpolation grid j are determined, and the original meteorological grid where the target interpolation grid j is located is determined according to the coordinates of the center point; the original meteorological grid and the 8 surrounding original meteorological grids are used as the 9 neighboring original meteorological grids of the target interpolation grid j.

[0126] Step S4.2, among several adjacent original meteorological grids, taking the first type of terrain element closest as the metric basis, obtain the first original meteorological grid I1 that is closest to the target interpolation grid j in the first type of terrain element dimension; taking the second type of terrain element closest as the metric basis, obtain the second original meteorological grid I2 that is closest to the target interpolation grid j in the second type of terrain element dimension; and so on, taking the nth type of terrain element closest as the metric basis, obtain the nth original meteorological grid I2 that is closest to the target interpolation grid j in the nth type of terrain element dimension. n ;

[0127] Step S4.3, determine the original meteorological grid where the center point of the target interpolation grid j is located, denoted as the original meteorological grid l, l = 1, 2, ..., N, and obtain the terrain element importance evaluation result W of the original meteorological grid l obtained in step S3 l =(w l,1 ,w l,2 ,...,w l,n ), as the terrain element importance of the target interpolation grid j;

[0128] Step S4.4, use the following formula to obtain the target meteorological element value Y of the target interpolation grid j: j :

[0129] Y j =Y(I1)w l,1 +Y(I2)w l,2 +...+Y(I n ) l,n

[0130] Where: Y(I1),Y(I2),...,Y(I n ), representing the first original meteorological grid I1, the second original meteorological grid I2, ..., the nth original meteorological grid I n The target meteorological element value.

[0131] For example, refer to Figure 4 , the dotted box represents a target interpolation grid, taking the target meteorological element as irradiance, the terrain meteorological comprehensive characteristics including three terrain elements: slope, aspect, and altitude, and three basic meteorological elements: precipitation, air pressure, and cloud cover.

[0132] The terrain element importance evaluation model constructed in step S3 is used to obtain the terrain element importance evaluation result w of each original meteorological grid. k =(w slp , w asp , w alt ), where w slp , w asp , w alt They represent the importance of slope, aspect and altitude to irradiance respectively.

[0133] exist Figure 4 In the metric, the closest spatial distance is used to obtain the 9 neighboring original meteorological grids of the target interpolation grid, which are: original meteorological grids 1 to 9;

[0134] Among the original meteorological grids 1 to 9, the original meteorological grids closest to the slope, aspect and altitude of the target interpolation grid are obtained respectively, assuming that they are: original meteorological grids 1, 3 and 8, that is, original meteorological grid 1 is the grid closest to the slope; original meteorological grid 3 is the grid closest to the aspect; original meteorological grid 8 is the grid closest to the altitude;

[0135] exist Figure 4 In the figure, the center point of the target interpolation grid is located in the original meteorological grid 5. Therefore, the evaluation result of the terrain element importance of the original meteorological grid 5 is used as the terrain element importance of the target interpolation grid, expressed as: k =(w slp , w asp , w alt );

[0136] Therefore, the irradiance of the target interpolation grid is obtained using the following formula:

[0137] Irradiance of target interpolation grid = irradiance of original meteorological grid 1 * slope importance w of target interpolation grid slp + irradiance of original meteorological grid 3 * aspect importance w of target interpolation grid asp + irradiance of original meteorological grid 8 * altitude importance w of target interpolation grid alt .

[0138] The invention provides a nearest weighted interpolation method for meteorological elements based on terrain importance. Firstly, meteorological and terrain data in a study area are extracted, and a terrain element importance evaluation model based on clustering and random forest is used to quantitatively evaluate the importance of different terrain data to target meteorological elements. Then, the nearest original meteorological grid of the target interpolation grid under different categories of terrain data is determined based on the spatial distance and terrain difference between grids. Finally, the weighted average value of the target interpolation grid under different categories of terrain data is calculated based on the target meteorological data of the nearest original meteorological grid selected as the weight and the terrain importance evaluation result.

[0139] The present invention provides a nearest neighbor weighted interpolation method for meteorological elements based on terrain importance. The method proposes a terrain importance quantitative evaluation model and a nearest neighbor weighted interpolation model, which can quantitatively evaluate the importance of different terrain elements to target meteorological elements, quantify the correlation between terrain elements and target meteorological elements, and combine the spatial relationship of the target meteorological elements themselves to achieve accurate interpolation of target meteorological elements at the target interpolation position. The present invention can effectively improve the accuracy of meteorological data interpolation results.

[0140] By using the difference between terrain elements as a distance judgment index, the present invention can more accurately reflect the impact of terrain elements compared to spatial distance; by constructing an importance evaluation model based on clustering and random forest, it can more accurately reflect and quantitatively derive the degree of influence of terrain elements on target meteorological elements. This method retains the computational efficiency of the nearest neighbor interpolation algorithm, while taking into account the impact of terrain elements on target meteorological elements, and can further improve the reliability of meteorological data interpolation results.

[0141] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be considered as the scope of protection of the present invention.

Claims

1. A nearest neighbor weighted interpolation method for meteorological elements based on terrain importance, characterized in that: The following steps are involved: Step S1, obtaining multiple original meteorological grids in the study area according to the grid form of the original meteorological data and the distribution in the study area; gridding the study area according to the spatial scale requirements of the target meteorological elements to be interpolated to form multiple target interpolation grids; Step S2, determine the target meteorological element Y that needs to be interpolated for the target interpolation grid; according to the target meteorological element Y, determine the terrain meteorological comprehensive feature X = [G, U] closely related to the target meteorological element Y; wherein the terrain meteorological comprehensive feature X is the splicing feature of the terrain element feature G and the basic meteorological element feature U; the terrain element feature G = [g1, g2, ..., g n ],g1,g2,...,g n , respectively represent the first terrain element, the second terrain element, ..., the nth terrain element; basic meteorological element characteristics U = [u1, u2, ..., u m ],u1,u2,...,u m , respectively represent the first basic meteorological element, the second basic meteorological element, …, the mth basic meteorological element; Step S3, constructing a terrain element importance evaluation model; The terrain element importance evaluation model includes a clustering model and a random forest importance evaluation model; the terrain meteorological comprehensive characteristic values ​​and target meteorological element values ​​of all original meteorological grids in the study area form a training sample set; the terrain element importance evaluation model is trained using the training sample set to obtain a trained terrain element importance evaluation model; according to the trained terrain element importance evaluation model, the terrain element importance evaluation result of each original meteorological grid is obtained, which is expressed as: W i =(w i,1 ,w i,2 ,...,w i,n ), where W i represents the evaluation result of the importance of terrain elements of the original meteorological grid i, i = 1, 2, ..., N, N represents the total number of original meteorological grids; w i,1 ,w i,2 ,...,w i,n They represent the importance of the first terrain element of the original meteorological grid i to the target meteorological element Y, the importance of the second terrain element to the target meteorological element Y, …, the importance of the nth terrain element to the target meteorological element Y; Step S4: for each target interpolation grid j in the study area, the target meteorological element value Y is obtained by the following method: j : Step S4.1, obtaining a number of neighboring original meteorological grids of the target interpolation grid j based on the closest spatial distance metric; Step S4.2, among several adjacent original meteorological grids, taking the first type of terrain element closest as the metric basis, obtain the first original meteorological grid I1 that is closest to the target interpolation grid j in the first type of terrain element dimension; taking the second type of terrain element closest as the metric basis, obtain the second original meteorological grid I2 that is closest to the target interpolation grid j in the second type of terrain element dimension; and so on, taking the nth type of terrain element closest as the metric basis, obtain the nth original meteorological grid I2 that is closest to the target interpolation grid j in the nth type of terrain element dimension. n ; Step S4.3, determine the original meteorological grid where the center point of the target interpolation grid j is located, denoted as the original meteorological grid l, l = 1, 2, ..., N, and obtain the terrain element importance evaluation result W of the original meteorological grid l obtained in step S3 l =(w l,1 ,w l,2 ,...,w l,n ), as the terrain element importance of the target interpolation grid j; Step S4.4, use the following formula to obtain the target meteorological element value Y of the target interpolation grid j: j : Y j =Y(I1)w l,1 +Y(I2)w l,2 +...+Y(I n )w l,n Where: Y(I1),Y(I2),...,Y(I n ), representing the first original meteorological grid I1, the second original meteorological grid I2, ..., the nth original meteorological grid I n The target meteorological element value.

2. The method of nearest neighbor weighted interpolation of meteorological elements based on terrain importance according to claim 1, characterized in that: In step S1, the spatial resolution of the original meteorological grid and the target interpolation grid are the same.

3. The method of nearest neighbor weighted interpolation of meteorological elements based on terrain importance according to claim 1, characterized in that: In step S1, the spatial resolutions of the original meteorological grid and the target interpolation grid are different.

4. The method of nearest neighbor weighted interpolation of meteorological elements based on terrain importance according to claim 1, characterized in that: In step S2, the target meteorological element Y is irradiance; the terrain element feature G includes three terrain elements, namely slope, aspect and altitude; the basic meteorological element feature G includes three basic meteorological elements, namely precipitation, air pressure and cloud cover.

5. The method of nearest neighbor weighted interpolation of meteorological elements based on terrain importance according to claim 1, characterized in that: Step S3 is specifically as follows: Step S3.1, assuming that the total number of original meteorological grids in the study area is N; for each original meteorological grid i, i = 1, 2, ..., N, obtain its terrain meteorological comprehensive characteristic value X i and target meteorological element value Y i , forming training samples S i =(X i ,Y i ); Therefore, for N original meteorological grids, a training sample set S = [S1, S2, ..., S N ]; Step S3.2: Set the training sample set S = [S1, S2, ..., S N ] is input into the clustering model; the clustering model uses the similarity of the terrain and meteorological comprehensive characteristics X between the training samples as the clustering feature, clusters the N training samples, and forms K sample subsets {C1, C2, ..., C K }; Step S3.3, each sample subset C k Input into the random forest importance evaluation model for training, k = 1, 2, ..., K. After the training is completed, each sample subset C is obtained k Evaluation results of terrain element importance W k =(w k,1 ,w k,2 ,...,w k,n ), where w k,1 ,w k,2 ,...,w k,n , represents the sample subset C k The importance of the first terrain element to the target meteorological element Y, the importance of the second terrain element to the target meteorological element Y, …, the importance of the nth terrain element to the target meteorological element Y; Each sample subset C k Evaluation results of terrain element importance W k =(w k,1 ,w k,2 ,...,w k,n ), as the sample subset C k The importance evaluation results of terrain elements in each original meteorological grid.

6. The method of nearest neighbor weighted interpolation of meteorological elements based on terrain importance according to claim 5, characterized in that: In step S3.1, the terrain meteorological comprehensive characteristic value X of the original meteorological grid i i The characteristic value of the terrain element G i =[g i,1 ,g i,2 ,...,g i,n ], for the fth terrain element value g i,f , f=1,2,...,n, and is obtained by the following method: Obtain a topographic grid within the study area; Filter terrain grids that overlap or intersect with the original meteorological grid i; take the average value of the f-th terrain element value of all the filtered terrain grids as the f-th terrain element value g of the original meteorological grid i. i,f , the formula is as follows: Where: num represents the number of terrain grids that overlap or intersect with the original meteorological grid i; Represents the fth terrain element value of the yth terrain grid that overlaps or intersects with the original meteorological grid i.

7. The method of nearest neighbor weighted interpolation of meteorological elements based on terrain importance according to claim 5, characterized in that: Step S3.2 is specifically as follows: Step S3.2.1, determine the number of sample subsets K according to the number of original meteorological grids N in the study area, the formula is: K=min{N / 100.2 n } Where: n is the number of terrain feature classes; Step S3.2.2, training sample set S = [S1, S2, ..., S N ] as the initial subset C of clustering; Step S3.2.3, for each training sample S i The comprehensive characteristic value of terrain and meteorology X i =(g i,1 ,g i,2 ,...,g i,n ,u i,1 ,u i,2 ,...,u i,m ) to obtain the standardized terrain and meteorological comprehensive characteristic value The standardization method is: X i =(g i,1 ,g i,2 ,...,g i,n ,u i,1 ,u i,2 ,...,u i,m ), which is n+m-dimensional data; the standardization method of each dimensional data is the same; For u i,m , using the formula To standardize; Where: u i,m is the training sample S i The value of the mth basic meteorological element before standardization; is the training sample S i The standardized value of the mth basic meteorological element; m,min is the minimum value of the mth basic meteorological element of N training samples; u m,max is the maximum value of the mth basic meteorological element of N training samples; Step S3.2.4, using the average distance indicator d between training samples avg Measure the distance between each sample subset, merge the sample subsets with the closest distance one by one; repeat the sample subset merging step until the number of sample subsets is equal to K, and generate a total of K types of sample subsets {C1, C2, ..., C K }; Among them: The calculation formula for the distance between sample subsets is: in: C k , C h for any two different sample subsets; d avg (C k , C h ) is the sample subset C k , C h The distance between s a is the sample subset C k Samples in b is the sample subset C h Samples in dist(s a ,s b ) represents sample s a and samples b The distance between ||s a -s b ||2 represents sample s a and samples b The distance between Representative samples a The standardized data of each dimension; Representative samples b The standardized data of each dimension.

8. The method of nearest neighbor weighted interpolation of meteorological elements based on terrain importance according to claim 5, characterized in that: Step S3.3 is specifically as follows: Step S3.3.1, for each sample subset C k , both construct random forest importance evaluation models and execute steps S3.3.2 to S3.3.5; Step S3.3.2, use self-sampling to select the sample subset C k The samples in are sampled to generate the samples required by each decision tree in the random forest; each sample has n terrain elements and m basic meteorological elements, so it has n+m features; Step S3.3.3, in the process of generating each decision tree, randomly select n types of features from the n+m features of the sample as the input of the decision tree, and use the feature that reduces the root mean square error the most before and after the decision tree splits as the feature for splitting the decision tree; Step S3.3.4, repeating steps S3.3.2 and S3.3.3 until a set number z of decision trees are generated; Step S3.3.5, count the terrain feature g1, g2, ..., g in each decision tree n The contribution values ​​are e1, e2, ..., e n ; Among them, for the characteristic value e of the f-th terrain element f , f=1,2,...,n, calculated using the following formula: Where: Δrmse f is the root mean square error reduction contributed by the f-th terrain element before and after splitting in each tree building process, e f is the sum of the root mean square error reduction contributed by the f-th terrain element in the random forest model; d is the number of splits in which each terrain element feature participates in the tree building process; Use the following formula, with contribution values ​​e1, e2, ..., e n As n types of terrain features g1, g2, ..., g n The degree of influence on the target meteorological element Y is denoted as W k =(w k,1 ,w k,2 ,...,w k,n ): Among them, w k,f is the fth terrain feature g f The degree of influence w on the target meteorological element Y k,f .

9. The method of nearest neighbor weighted interpolation of meteorological elements based on terrain importance according to claim 1, characterized in that: In step S4.1, the nine original meteorological grids adjacent to the target interpolation grid j are obtained by: The coordinates of the center point of the target interpolation grid j are determined, and the original meteorological grid where the target interpolation grid j is located is determined according to the coordinates of the center point; the original meteorological grid and the 8 surrounding original meteorological grids are used as the 9 neighboring original meteorological grids of the target interpolation grid j.

Images