A comprehensive quantitative evaluation method for meteorological data based on cloud model

Abstract

A comprehensive quantitative evaluation method for meteorological data based on cloud model, with the following specific steps: Step 1: Data normalization processing, this technology divides meteorological data indicators into two attribute types: real number type and fuzzy type, and divides each indicator into positive and negative attributes; Step 2: Data cloud processing, the attribute indicator data are unified using cloud model C i (E x , E n , H e ) represents; Step 3: Establish a judgment matrix and use the hierarchical analysis method to determine the weight w of the attribute cloud model i (E x , E n , H e ); Step 4: Weight the multi-source attribute cloud to generate the cloud model C to be evaluated z (E x , E n , H e ); Step 5: Construct a subordinate cloud scale and calculate the similarity between the cloud to be evaluated and a similar base cloud; Step 6: Based on the calculated similarity between the evaluated cloud and a similar base cloud, qualitatively and quantitatively evaluate multi-source meteorological data. This invention proposes a quantitative meteorological data evaluation technique based on cloud model theory, characterized by: comprehensive consideration of multi-source data attribute indicators; comprehensive consideration of data characteristics such as real and fuzzy attributes; and the use of a unified cloud model representation for objective qualitative and quantitative analysis of data characteristics.

Description

Technical Field

[0001] The present invention belongs to the field of multi-source meteorological data application, and in particular relates to a comprehensive quantitative evaluation method for meteorological data based on a cloud model. Background Art

[0002] Meteorological monitoring, forecasting, and early warning are among the most important services provided by meteorological departments. The quality of meteorological data directly impacts meteorological operations, and in turn, the disaster prevention and mitigation efforts of governments at all levels. Currently, a variety of approaches exist for monitoring and forecasting meteorological elements, including satellite monitoring, weather station monitoring, numerical forecasting, and artificial intelligence forecasting. However, each method for acquiring meteorological elements has its own advantages and disadvantages. For example, satellite data allows for large-scale monitoring, but is also prone to outliers. Weather station data offers high accuracy but struggles with large-scale macro-monitoring. Numerical forecasting, based on physical analysis, allows for short-term prediction of meteorological elements, but suffers from low data resolution and high computational complexity. Furthermore, different meteorological observation purposes require different meteorological data metrics. Therefore, effectively and quantitatively evaluating meteorological data, while comprehensively considering the advantages and disadvantages of multiple sources for different observation purposes, remains a key and challenging issue in the meteorological field.

[0003] Currently, linear weighting is often used to evaluate data. This involves constructing a data evaluation index system, assigning weights to each index, and finally conducting a fusion evaluation of the data. However, the determination of index weights is highly subjective, and the problem of unifying qualitative and quantitative indicators has rarely been solved. Alternatively, methods based on absolute mathematical statistical analysis are divorced from the inherent attributes of meteorological data and fail to effectively reflect the ambiguity of data information. Existing methods have not yet provided effective comprehensive evaluation conclusions that combine qualitative and quantitative methods. This technology addresses the problem of unified representation of multi-source data by first introducing cloud model theory, unifying different attribute indicators using a cloud model. This cloud model is then used to assign weights to the attribute indicators, completing qualitative and quantitative evaluations of meteorological data for a given observation target. Summary of the Invention

[0004] In order to solve the above problems, a comprehensive quantitative evaluation method for meteorological data based on cloud model is proposed. It comprehensively considers the attribute indicators of multi-source data; comprehensively considers data characteristics such as real number and fuzzy attributes; uses cloud model for unified representation, and objectively analyzes data characteristics qualitatively and quantitatively.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] A comprehensive quantitative evaluation method for meteorological data based on cloud models, the specific steps are as follows:

[0007] Step 1: Data normalization. This technology divides meteorological data indicators into two attribute types: real number type and fuzzy type, and divides each indicator into positive and negative attributes;

[0008] Step 2: Cloud processing of real number data, uniformly representing each attribute indicator data with the cloud model Ci (Ex, En, He);

[0009] The formula for cloudification of real number data is:

[0010]

[0011] E n =(π / 2) 1 / 2 ×B,H e =|S 2 -E n 2 | 1 / 2 (2)

[0012] Step 3: Establish a judgment matrix and use the hierarchical analysis method to determine the weight wi (Ex, En, He) of the attribute cloud model;

[0013] The calculation formula of the weight cloud model is:

[0014]

[0015] in Represents U i The expectation, entropy and super entropy in the cloud model weights corresponding to the attribute cloud, and Represents the cloud scale C in the judgment matrix ij After consistency test, the cloud model weight matrix W is finally obtained. i (E x , e n , H e );

[0016] Step 4: Weight the multi-source attribute clouds to generate the cloud model C to be evaluated z (E x , E n , H e );

[0017] The formula for generating the medium-level evaluation cloud is:

[0018]

[0019] For property cloud C i (Ex i ,En i , Hei ) and the corresponding weight cloud W i (Ex zi ,En zi , He zi ) multiplication operation adopts formula (5), for the attribute cloud C zi (Ex i ,En i , He i ) and attribute cloud C zj (Ex j ,En j , He j ) The addition operation of two clouds uses formula (6)

[0020] W i ×C i =C(Ex wi ×Ex i ,En wi ×En i , He wi ×He i ) (5)

[0021] C zi +C zj =C(Ex zi +Ex zj ,En zi +En zj , He zi +He zj ) (6)

[0022] Step 5: Construct a cloud scale to calculate the similarity between the cloud to be evaluated and the similar basic cloud;

[0023] Step 6: Qualitatively and quantitatively evaluate multi-source meteorological data based on the similarity calculation between the evaluation cloud and similar basic clouds.

[0024] As a further improvement of the present invention, the calculation formula for the intersection area in step 5 is:

[0025] a. For two intersection points p1 and p2, let p1 < p2. The intersection calculation formula is:

[0026]

[0027] Then S ∩ It consists of three parts:

[0028]

[0029] b. There is a single intersection point p. The intersection calculation formula is:

[0030]

[0031] Then S ∩ It consists of two parts,

[0032]

[0033] c. If there are clouds that overlap or contain each other, the smaller of the two clouds is integrated to obtain the solution. If the clouds overlap, choose either of them.

[0034]

[0035] Where φ(x) is the standard normal distribution function.

[0036] The similarity calculation formula is as follows:

[0037] D i =S ∩i / S z

[0038] (12).

[0039] As a further improvement of the present invention, the quantitative evaluation calculation formula in step 6 is:

[0040]

[0041] Where D1 and D2 are the cloud similarities of two basic clouds close to the cloud to be evaluated Cz.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] A comprehensive quantitative evaluation method for meteorological data based on a cloud model has the following specific steps: Step 1: Data normalization. This technology divides meteorological data indicators into two attribute types: real and fuzzy, and divides each indicator into positive and negative attributes; Step 2: Data cloudification. Each attribute indicator data is uniformly represented by the cloud model Ci (Ex, En, He); Step 3: Establishing a judgment matrix and using the analytic hierarchy process to determine the weights wi (Ex, En, He) of the attribute cloud model; Step 4: Weighting the multi-source attribute clouds to generate the cloud model to be evaluated Cz (Ex, En, He); Step 5: Constructing a subordinate cloud scale and calculating the similarity between the cloud to be evaluated and a similar base cloud; Step 6: Based on the similarity calculation between the evaluation cloud and a similar base cloud, qualitatively and quantitatively evaluating multi-source meteorological data. The present invention proposes a quantitative evaluation technology for meteorological data based on cloud model theory, characterized by: comprehensively considering multi-source data attribute indicators; comprehensively considering data characteristics such as real and fuzzy attributes; using the cloud model for unified representation, and objectively analyzing data characteristics qualitatively and quantitatively. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 : Technical architecture and execution flow chart;

[0045] Figure 2 ; ERA5 wind field color map based on K-means;

[0046] Figure 3 ; ERA5 wind field clustering color map based on K-means;

[0047] Figure 4 ;ERA5 data anomaly rate attribute cloud;

[0048] Figure 5 ;ERA5 data accuracy attribute cloud;

[0049] Figure 6 ;ERA5 data attribute cloud and basic evaluation cloud;

[0050] Figure 7 ; The intersection points and intersection areas of ERA5 data attribute cloud and basic cloud;

[0051] Figure 8 ; The intersection points and intersection areas of ERA5 data attribute cloud and “middle” basic cloud. DETAILED DESCRIPTION

[0052] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, specifically:

[0053] Step 1: Determine the meteorological data requirement index U based on the purpose and needs of meteorological observation i , where i is the number of indicators. The data corresponding to each indicator is divided into two types: real number type and fuzzy type, and positive and negative attributes are distinguished at the same time. The larger the expected value of the positive attribute, the better the indicator; the smaller the expected value of the negative attribute, the better the indicator. Use p i It represents real number attribute data, and the five levels of "excellent, good, medium, poor, and very poor" represent fuzzy attribute data.

[0054] For real number attribute data, normalize each data item as follows:

[0055]

[0056] Step 2: Process the data of various evaluation indicators in the cloud and use the cloud model C i (E x , E n , H e ) is used to evaluate the uncertainty of each attribute. For real-number attribute data, the reverse cloud algorithm is used to calculate the n sampling samples p in each attribute cloud model data. i (i=1,2,3,…,n), where the expected value (Ex ), Entropy (E n ) and super entropy (H e ) is calculated as follows:

[0057]

[0058] E n =(π / 2) 1 / 2 ×B,H e =|S 2 -E n 2 | 1 / 2 (3)

[0059] For fuzzy attribute data, we use five fuzzy evaluation values ("excellent, good, fair, poor, and very poor") and, based on expert experience and a model-driven approach based on the golden ratio, develop a fuzzy evaluation attribute cloud model, as shown in Table 1. "Excellent" is described as a semi-rising cloud, "very poor" as a semi-falling cloud, and the other fuzzy evaluation values as a complete cloud.

[0060] Table 1 Fuzzy evaluation attribute cloud model

[0061] Fuzzy evaluation value Property Cloud Model excellent C(1.0000，0.1309，0.0262) good C(0.7000，0.0809，0.0162) middle C(0.5000，0.050，0.01000) Difference C(0.3000，0.0809，0.00162) Special C(0.0000，0.01309，0.0262)

[0062] Step 3: Use the scaling method in Table 2 to construct the pairwise comparison judgment matrix X as follows:

[0063]

[0064] Among them, C ij Indicates the index U i with U j Judgment cloud scale.

[0065] The judgment matrix X is determined by the hierarchical analysis method. i The cloud model weight corresponding to the attribute cloud is calculated by the AHP method as follows:

[0066]

[0067] in Represents U i The expectation, entropy and super entropy in the cloud model weights corresponding to the attribute cloud, and Represents the cloud scale C in the judgment matrix ij After consistency test, the cloud model weight matrix W is finally obtained. i (E x , e n , H e ).

[0068] Table 2 Cloud model scale and meaning

[0069] <![CDATA[Scale C(E x , e n , H e )]]> meaning C1(1，0，0) Comparing two by two is equally important C2(3，0.33，0.05) Comparing the two, the former is slightly more important than the latter C3(5，0.33，0.05) Comparing the two, the former is obviously more important than the latter C4(7，0.33，0.05) Comparing the two, the former is more important than the latter. C5(9，0.33，0.05) Comparing the two, the former is extremely important than the latter C6(1 / 3，0.33 / 9，0.55 / 9) Comparing the two, the former is slightly less important than the latter C7(1 / 5，0.33 / 25，0.05 / 25) Comparing the two, the former is obviously less important than the latter C8(1 / 7，0.33 / 49，0.05 / 49) Comparing the two, the former is not as strong as the latter. C5(9，0.33 / 81，0.05 / 81) Comparing the two, the former is not as important as the latter

[0070] The consistency check process is as follows:

[0071] 1) For the judgment matrix X, there is

[0072]

[0073] Among them C Exij Indicates C ij Expectations of Ex.

[0074] For the judgment matrix W, we have

[0075] W Ex =[W Ex1 , W Ex2 ,…,W Exn ] T (7)

[0076] Where W Exi Indicates U i The expected Ex in the cloud model weight corresponding to the attribute cloud.

[0077] 2) Calculate the eigenvector λ using the following formula:

[0078]

[0079] Where n is the number of indicators.

[0080] 3) Consistency test, the calculation formula is as follows:

[0081]

[0082] If CR < 0.1, the consistency test is passed. Where n is the number of indicators, RI is the number of indicators, and the test table is searched according to the number of indicators, as shown in Table 3.

[0083] Table 3 RI values of consistency test

[0084] Order 1 2 3 4 5 6 7 8 … RI 0 0 0.52 0.89 1.12 1.26 1.36 1.41 …

[0085] Step 4: Perform one-dimensional linear weighted calculation on each indicator attribute cloud and generate the object cloud to be evaluated using formula (10).

[0086]

[0087] For property cloud C i (Ex i ,En i , He i) and the corresponding weight cloud W i (Ex zi ,En zi , He zi ) multiplication operation adopts formula (11), for the attribute cloud C zi (Ex i ,En i , He i ) and attribute cloud C zj (Ex j ,En j , He j ) The addition operation of two clouds uses formula (12)

[0088] W i ×C i =C(Ex wi ×Ex i ,En wi ×En i , He Wi ×He i ) (11)

[0089] C zi +C zj =C(Ex zi +Ex zj ,En zi +En zj , He zi +He zj ) (12)

[0090] Step 5: Construct a subordinate cloud scale, which is divided into five basic clouds: "excellent, good, medium, poor, and very poor". Each level of the cloud scale still uses the cloud model representation method in Table 2. Based on the cloud scale, find the two basic clouds C1 and C2 that are closest to the cloud to be evaluated Cz. The similarity D between Cz and Ci can be expressed as:

[0091] D i =S ∩i / S2 (13)

[0093] In formula (13): S ∩i For Cloud C i and C z The intersection area, S z For Cloud C z The area, S ∩i and S z Both are calculated using truncated entropy and the integral of the cloud expectation curve. This technology elevates cloud model similarity calculation to a higher level, placing greater emphasis on the uncertainty of the evaluation process. This reflects the interdependent nature of cloud models, resulting in reasonable and stable evaluation conclusions.

[0094] Step 7: Select the basic cloud scale with the largest similarity as the qualitative evaluation of the cloud to be evaluated. Consider the “3E n The influence of the rule and super entropy is quantitatively evaluated using formula (14):

[0095]

[0096] Where D1 and D2 are the cloud similarities of the two basic clouds C1 and C2 that are close to the cloud to be evaluated Cz. Specific embodiments

[0098] This paper takes Jiangsu Province, China as the research area and power generation forecast as the purpose, and conducts a quantitative evaluation of the 2022 ERA5 reanalysis wind speed product based on a cloud model.

[0099] Step 1: Indicator construction and calculation. The attribute indicators are measured using six indicators: data accuracy, data anomaly rate, data spatial resolution, data temporal resolution, data processing simplicity, and data acquisition difficulty. These indicators cover four items of real data, two items of fuzzy data, and three items each of positive and negative attribute data. The specific description and calculation of attribute data are as follows:

[0100] Data accuracy: Based on the wind speed measured at the meteorological observation station, after linear grid interpolation, the Pearson coefficient is used to measure the correlation of each grid point of the data, which is a positive attribute. Two one-dimensional random variables x = (x1, x2, ..., x n ) and y=(y1,y2,…,y n The calculation formula of the correlation coefficient between ) is:

[0101]

[0102] Data anomaly rate: Based on the K-means outlier detection method, the number of abnormal data is calculated and the data anomaly rate is obtained, which is a negative attribute. The specific method is as follows:

[0103] Step 1: Randomly determine k sample points as the initial cluster centers.

[0104] Step 2: Traverse all samples in the data set in order, calculate the distance to all cluster centers, select the cluster center that is closest to the sample point, and then add the point to the cluster to which this cluster center belongs.

[0105] Step 3: Calculate the latest centroid of each cluster. The centroid is the average value of all points in the updated cluster and is also the latest cluster center.

[0106] Step 4: The algorithm ends when the number of iterations reaches the upper limit or the centroid of each cluster no longer changes. Otherwise, the algorithm will repeat steps 2 and 3.

[0107] Step 5: Calculate the Euclidean distance d(x) between each data point and the cluster center i ,m) and the average Euclidean distance d avg , if d is greater than d avg , then record the data point and write down the d(x i ,m) and d avg , is compared with 1.67 times the standard deviation s of the distance from all data points in the cluster to the cluster center. This comparison result is used as the criterion for judging whether the data point is normal. Data that is smaller than the latter is normal data, otherwise it is regarded as abnormal data, and the coordinates of the data point are output. The clustering effect is as follows Figure 2 、 3 shown.

[0108] Among them, the K-means algorithm Euclidean distance calculation formula is as follows:

[0109]

[0110] The error square sum function is another important basic concept of the algorithm, which refers to the distance from each sample point r to each cluster center x. i The sum of squared distances, m i It represents the mean of the family and is calculated as follows:

[0111]

[0112] Data spatial resolution: The number of data pixels within a 1°×1° area. A greater number indicates higher spatial resolution, a higher data evaluation, and a positive attribute. For wind forecasting, a spatial resolution of 4000m is required, meaning 625 pixels are contained within a 1°×1° area.

[0113] Data temporal resolution: The number of times data is updated in a day. A higher temporal resolution indicates a higher data evaluation and is considered a positive attribute. For wind farm forecasts, data needs to be updated every 15 minutes, or 96 times a day.

[0114] Data processing difficulty: Taking into account the processing processes such as data reading, calibration, and interpolation, the processing difficulty is fuzzy evaluated. When it is extremely poor, the data processing is the easiest, the data evaluation is the highest, and it is a negative attribute.

[0115] Data acquisition difficulty: Taking into account the data retrieval, downloading, storage and other acquisition processes, the difficulty is fuzzy evaluated. When it is extremely poor, the data acquisition difficulty is the lowest, the data evaluation is the highest, and it is a negative attribute.

[0116] Step 2: Using formulas (2) and (3), all attribute data are unified using the cloud model C i (E x , E n , He ) indicates that the ERA 5 wind farm product evaluation index table and attribute cloud model are established. Among them, the attribute clouds of U1 and U2 are as follows Figure 4 、 5 shown.

[0117] Table 4 ERA 5 wind farm product evaluation index table

[0118] Attribute indicators Attribute Type Positive / negative attributes data Normalization Property Cloud Model <![CDATA[Data accuracy (U1)]]> Real number type just 0.3548 0.3583 (0.3583，0.1939，0.0432) <![CDATA[Data anomaly rate (U2)]]> Real number type burden 0.008 0.5 (0.5000，0.0043，0.0002) <![CDATA[Data space resolution (U3)]]> Real number type just 16 0.0256 C(0.0256，0.0000，0.0000) <![CDATA[Data time resolution (U4)]]> Real number type just 24 0.25 C(0.2500，1.0000，1.0000) <![CDATA[Data processing difficulty (U5)]]> Fuzzy burden Special - C(1.0000，0.1309，0.0262) <![CDATA[Difficulty of data acquisition (U6)]]> Fuzzy burden Difference - C(0.7000，0.0809，0.0162)

[0119] Step 3: Establish the judgment matrix and use formula (4) to obtain the weight cloud matrix, as shown in Table 5.

[0120] Table 5 ERA 5 indicator comparison judgment matrix and weight cloud parameters

[0121]

[0122] The CR calculated by formula (5) is less than 0.1, passing the consistency test.

[0123] Step 4: Generate the object cloud Cz(0.3890, 0.0877, 0.025) using formula (6).

[0124] Step 5: Create a basic cloud ruler, such as Figure 6 The two basic clouds closest to the cloud to be measured, Cz, are "poor" and "medium". The intersection area and the evaluation cloud area are calculated as follows:

[0125] For the ‘bad’ base cloud intersection area, there is only one intersection point in the interval y∈[0,1], and the intersection point is

[0126]

[0127] in They are respectively the object cloud C to be evaluated z The expectation and entropy of are the base cloud expectation and entropy of the 'difference' respectively.

[0128] The calculated value p = 0.342, then the intersection area consists of two parts, S1 and S2, as shown in the following example: Figure 7 、 8 shown.

[0129]

[0130] Φ(x) is the probability distribution function of the standard normal distribution, which can be obtained by looking up the table. ∩ It is 0.2198.

[0131] For the 'medium' basic cloud intersection area, there are two intersection points in the interval y∈[0,1], and the intersection points are

[0132]

[0133] in They are respectively the object cloud C to be evaluated z The expectation and entropy of are the base cloud expectation and entropy of 'medium' respectively.

[0134] The calculated values p1 = 0.4597, p2 = 0.6472, and the intersection area is composed of three parts: S1, S2, and S3. Figure 7 、 8 shown.

[0135]

[0136] Φ(x) is the probability distribution function of the standard normal distribution, which can be obtained by looking up the table. ∩ 0.0633

[0137] For evaluating cloud area S z , the calculation formula is as follows:

[0138]

[0139] Obtain S z =0.2198.

[0140] Finally, according to formula (9), the similarities between the cloud to be evaluated and the 'poor' basic cloud and the 'medium' basic cloud are 0.573 and 0.288 respectively.

[0141] Step 6: Based on the similarity, the ERA5 reanalysis data is qualitatively evaluated as "poor" for the forecast demand of wind power generation. Finally, according to formula (10), the quantitative evaluation of the ERA5 reanalysis data for the forecast demand of wind power generation is 0.502.

[0142] The above description is merely a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any modification or equivalent variation based on the technical essence of the present invention shall still fall within the scope of protection claimed by the present invention.

Claims

1. A comprehensive quantitative evaluation method for meteorological data based on a cloud model, comprising the following steps, characterized in that: Step 1: Data normalization. This technology divides meteorological data indicators into two attribute types: real number type and fuzzy type, and divides each indicator into positive and negative attributes; Step 2: Cloud processing of real number data, uniformly representing each attribute indicator data with the cloud model Ci (Ex, En, He); The formula for cloudification of real number data is: （1） （2） Step 3: Establish a judgment matrix and use the hierarchical analysis method to determine the weight wi (Ex, En, He) of the attribute cloud model; The calculation formula of the weight cloud model is: （3） in 、 、 Represents U i The expectation, entropy and super entropy in the cloud model weights corresponding to the attribute cloud, 、 and Represents the cloud scale C in the judgment matrix ij After consistency test, the cloud model weight matrix is finally obtained. ( , , ); Step 4: Weight the multi-source attribute clouds to generate the cloud model C to be evaluated z (E x , E n , H e ); The formula for generating the medium-level evaluation cloud is: （4） For Property Cloud ) and the corresponding weight cloud ) multiplication operation uses formula (5), for the attribute cloud ) and attribute cloud ) The addition operation of two clouds uses formula (6) （5） （6） Step 5: Construct a cloud scale to calculate the similarity between the cloud to be evaluated and the similar basic cloud; Step 6: Qualitatively and quantitatively evaluate multi-source meteorological data based on the similarity calculation between the evaluation cloud and similar basic clouds.

2. A cloud model-based comprehensive quantitative evaluation method for meteorological data according to claim 1, characterized in that: The formula for calculating the intersection area in step 5 is: a. Double intersection ,set up , the intersection calculation formula is: （7） but It consists of three parts:

3. （8） b. There is a single intersection point p. The intersection calculation formula is: （9） but It consists of two parts, （10） c. If there are clouds that overlap or contain each other, the smaller of the two clouds is integrated to obtain the solution. If the clouds overlap, choose either of them. （11） Where φ(x) is the standard normal distribution function; The similarity calculation formula is as follows: （12）。 4. The method for comprehensive quantitative evaluation of meteorological data based on cloud model according to claim 1, characterized in that: The quantitative evaluation calculation formula in step six is: （13） in and is the cloud similarity of the two basic clouds Cz that are close to the cloud to be evaluated.

Images