Dry-wet period discrimination method based on comprehensive voting

Abstract

The invention provides a dry-wet period discrimination method based on comprehensive voting, and relates to the technical field of surface meteorological monitoring, and the method comprises the steps: obtaining the data of each microwave link through a microwave base station in a research region: selecting clustering features through a feature selection method; dry and wet period discrimination is carried out based on a K-Means method; dry and wet period discrimination is carried out based on an FCM method; dry and wet period discrimination is carried out based on a sliding standard deviation method; and the results obtained by the three dry-wet period discrimination methods are integrated in a voting mode. According to the method, dry-wet period judgment results of K-Means, FCM and a sliding standard deviation method are combined through a comprehensive voting method, compared with dry-wet period judgment by adopting a single sliding standard deviation method, the sensitivity of a microwave link in a rainfall capturing process can be remarkably improved, accurate judgment of the dry-wet period is realized, and the accuracy of the dry-wet period judgment is improved. Therefore, a solution thought and method can be provided for dry-wet period discrimination based on microwave attenuation characteristics, and research on rainfall monitoring of a wireless microwave link is promoted.

Description

Technical Field

[0001] The invention relates to the technical field of ground meteorological monitoring, and in particular to a dry and wet period discrimination method based on comprehensive voting. Background Art

[0002] Using microwave links in the field of wireless communications to monitor precipitation is one of the rapidly developing atmospheric environment monitoring technologies in recent years. Since rainfall discrimination is the prerequisite for ultra-high frequency microwave rainfall inversion calculation, before using wireless microwave data to invert and calculate the rainfall process, it is necessary to first distinguish the dry and wet periods, and then the calculated rainfall process can be corrected based on the dry and wet period discrimination results, so that the rainfall inversion process is more consistent with the actual situation.

[0003] Compared with other meteorological elements, the dispersion of microwave signal attenuation during rainfall is more significant, and the sliding standard deviation method is usually used to distinguish between dry and wet periods. However, this method is highly random and often uses repeated thresholds, resulting in a low degree of data fitting, which may not ensure the sensitivity of the microwave link to capture the rainfall process, affecting the accurate distinction of dry and wet periods.

[0004] Therefore, it is necessary to provide a dry and wet period discrimination method based on comprehensive voting to solve the above technical problems. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a dry and wet period discrimination method based on comprehensive voting, which is used to solve the problem that the existing single sliding standard deviation method cannot ensure the sensitivity of the microwave link to capture the rainfall process, thus affecting the accurate discrimination of the dry and wet periods.

[0006] The present invention provides a method for distinguishing dry and wet periods based on comprehensive voting, comprising: Obtain microwave attenuation data of each microwave link through a microwave base station; Based on the feature selection algorithm, the microwave attenuation data of a single link is clustered and selected to obtain the microwave attenuation characteristics of the single link; The microwave attenuation characteristics of each link are clustered based on the K-means clustering algorithm to obtain the first dry and wet period discrimination result; The microwave attenuation characteristics of each link are clustered based on the fuzzy C-means clustering algorithm to obtain the second dry and wet period discrimination results; The microwave attenuation characteristics of each link are clustered based on the sliding standard deviation method to obtain the discrimination results of the third dry and wet period; According to the first dry and wet period discrimination results, the second dry and wet period discrimination results and the third dry and wet period discrimination results, the comprehensive voting method is used to determine the target dry and wet period discrimination results.

[0007] Preferably, the clustering feature selection of the microwave attenuation data of a single link based on a feature selection algorithm to obtain the microwave attenuation feature of the single link specifically includes: Based on the preset attenuation time, select the maximum microwave attenuation rate A in the microwave attenuation data of a single link max , standard deviation of microwave attenuation rate S and maximum microwave attenuation rate change rate V max As clustering features, the microwave attenuation characteristics of a single link are obtained; Among them, the standard deviation S of the microwave attenuation rate within the preset attenuation time and the qth microwave attenuation rate change rate The calculation formula is as follows: Where, S represents the standard deviation of microwave attenuation rate within the preset attenuation time; k represents the number of microwave attenuation data within the preset attenuation time; It indicates the qth microwave attenuation rate within the preset attenuation time, in dB / km; Indicates the average microwave attenuation rate within the preset attenuation time, in dB / km; represents the q-1th microwave attenuation rate within the preset attenuation time, in dB / km; T represents the receiving time interval of microwave attenuation data, in seconds; represents the rate of change of the qth microwave attenuation rate; ∑ represents the summation symbol; Select the maximum value of all microwave attenuation rate change rates as the maximum microwave attenuation rate change rate V max ; The expression of the microwave attenuation characteristic FE of a single link within the preset attenuation time is as follows: Where FE represents the microwave attenuation characteristics of a single link within the preset attenuation time; It represents the maximum microwave attenuation rate within the preset attenuation time; S represents the standard deviation of the microwave attenuation rate within the preset attenuation time; Indicates the maximum microwave attenuation rate change rate within the preset attenuation time.

[0008] Preferably, clustering the microwave attenuation characteristics of each link based on the K-means clustering algorithm to obtain the first dry-wet period discrimination result specifically includes: Assume that the data set D = {FE1, FE2, ..., FE i , …, FE n}, where n represents the number of links; FE i represents the i-th microwave attenuation feature in the data set D; In the data set D, the microwave attenuation characteristics of K links are selected as the initial clustering centers; Calculate the Euclidean distance between other microwave attenuation features in the data set D and each initial cluster center as the similarity index, and cluster each microwave attenuation feature in the data set D into the category cluster corresponding to the initial cluster center with the highest similarity index; For each category cluster, the average value of all microwave attenuation features within the category cluster is calculated as the new cluster center; Repeat the above cluster center updating process until the updated cluster center remains unchanged, that is, the criterion function of the K-means clustering algorithm converges, and the first dry and wet period discrimination result is obtained.

[0009] Preferably, the calculation formula of the Euclidean distance between other microwave attenuation features in the data set D and each initial cluster center is as follows: Where D represents the data set composed of microwave attenuation characteristics of all links; Represents the value of the pth attribute of other microwave attenuation features X in the data set D; Represents the value of the pth attribute of the initial cluster center Y; Among them, the dimensions of other microwave attenuation features X and initial clustering centers Y in the data set D are both n.

[0010] Preferably, the expression of the criterion function of the K-means clustering algorithm is as follows: In the formula, Z represents the criterion function of the K-means clustering algorithm; R represents the total number of cluster categories, that is, the total number of category clusters; x represents the rth category cluster Microwave attenuation characteristics in ; Represents the rth category cluster The cluster center of ; ∑ represents the summation symbol.

[0011] Preferably, clustering the microwave attenuation characteristics of each link based on the fuzzy C-means clustering algorithm to obtain the second dry-wet period discrimination result specifically includes: The microwave attenuation characteristics of all links in the data set D each include d eigenvalues, namely FE i ={FE i1 , FE i2 , …, FE ij , …, FE id}, where FE ij Represents the i-th microwave attenuation feature FE in the data set D i The j-th eigenvalue of ; Initialize the membership matrix U, divide the data set D into c clusters, and calculate the corresponding cluster centers as follows: In the formula, represents the jth cluster center; represents the i-th microwave attenuation feature in the data set D; represents the i-th microwave attenuation characteristic FE i With the jth cluster center The membership degree of ; n represents the number of links; m represents the fuzzy index, and m>1; ∑ represents the summation symbol; Update the membership matrix U, and the corresponding calculation formula is as follows: In the formula, represents the updated i-th microwave attenuation feature FE i With the jth cluster center The membership degree of ; c represents the number of clusters; represents the i-th microwave attenuation characteristic FE i With the jth cluster center distance; represents the i-th microwave attenuation characteristic FE i With the e-th cluster center distance; m represents the fuzzy index, and m>1; ∑ represents the summation symbol; The objective function and constraints of the fuzzy C-means clustering algorithm are calculated as follows: In the formula, represents the objective function of the fuzzy C-means clustering algorithm; U represents the membership matrix; V represents the cluster center matrix; n represents the number of links; m represents the fuzzy index, and m>1; represents the i-th microwave attenuation characteristic FE i With the jth cluster center The degree of membership; represents the i-th microwave attenuation characteristic FE i With the jth cluster center The distance; ∑ represents the summation symbol; If the objective function of the fuzzy C-means clustering algorithm is less than the preset termination threshold, the iteration process of the fuzzy C-means clustering algorithm is stopped and the second dry-wet period discrimination result is output.

[0012] Preferably, clustering the microwave attenuation characteristics of each link based on the sliding standard deviation method to obtain the third dry and wet period discrimination result specifically includes: Setting a sliding time window , where t represents the current time; s represents the duration, and s>0; The mean and standard deviation of the microwave attenuation value of a single link within the sliding time window are calculated as follows: In the formula, Represents the mean value of the microwave attenuation value of a single link within the sliding time window; represents the standard deviation of the microwave attenuation value of a single link within the sliding time window; N represents the total number of moments within the sliding time window; represents a sliding time window; represents the microwave attenuation value of a single link at the gth moment in the sliding time window; ∑ represents the summation symbol; The microwave attenuation threshold is determined based on the percentile of the dry period attenuation of a single link. The microwave attenuation threshold is expressed as follows: In the formula, represents the microwave attenuation threshold; represents the 99% quantile of the dry period decay time series; represents the standard deviation of the microwave attenuation value of a single link in the sliding time window; t represents the current time; D represents the dry period; Based on the decision rule, the standard deviation of the microwave attenuation value of a single link in the sliding time window is compared with the microwave attenuation threshold to obtain the discrimination result of the third dry and wet period. The expression of the decision rule is as follows: Where, P represents the third dry and wet period discrimination result; W represents the wet period; D represents the dry period; Represents the standard deviation of the microwave attenuation value of a single link within the sliding time window; Indicates the microwave attenuation threshold.

[0013] Preferably, according to the third dry-wet period discrimination result, the microwave attenuation value of a single link at the g-th moment in the sliding time window is decomposed to obtain: In the formula, Indicates the microwave attenuation value of a single link at the gth moment in the sliding time window; Indicates the basic attenuation value of a single link at the gth moment in the sliding time window; It represents the attenuation value caused by the rainfall event at the gth moment in the sliding time window of a single link; P represents the third dry and wet period discrimination result; W represents the wet period; and D represents the dry period.

[0014] Preferably, the method of determining the target dry and wet period discrimination result by using a comprehensive voting method according to the first dry and wet period discrimination result, the second dry and wet period discrimination result and the third dry and wet period discrimination result specifically includes: If at least two of the first dry-wet period discrimination result, the second dry-wet period discrimination result and the third dry-wet period discrimination result are wet periods, the target dry-wet period discrimination result is determined to be a wet period; otherwise, the target dry-wet period discrimination result is determined to be a dry period.

[0015] Compared with the related art, the dry and wet period discrimination method based on comprehensive voting provided by the present invention has the following beneficial effects: The present invention obtains microwave attenuation data of each microwave link through a microwave base station; clusters the microwave attenuation data of a single link based on a feature selection algorithm to obtain the microwave attenuation characteristics of the single link; clusters the microwave attenuation characteristics of each link based on a K-means clustering algorithm to obtain a first dry and wet period discrimination result; clusters the microwave attenuation characteristics of each link based on a fuzzy C-means clustering algorithm to obtain a second dry and wet period discrimination result; clusters the microwave attenuation characteristics of each link based on a sliding standard deviation method to obtain a third dry and wet period discrimination result; and adopts a comprehensive voting method to determine a target dry and wet period discrimination result according to the first dry and wet period discrimination result, the second dry and wet period discrimination result and the third dry and wet period discrimination result, thereby improving the sensitivity of the microwave link in capturing the rainfall process and realizing accurate discrimination of the dry and wet periods.

[0016] The present invention combines the dry and wet period discrimination results of the K-means clustering algorithm, the fuzzy C-means clustering algorithm and the sliding standard deviation method through a comprehensive voting method. Compared with the use of a single sliding standard deviation method for dry and wet period discrimination, the sensitivity of the microwave link in capturing the rainfall process can be significantly improved, and the accurate discrimination of the dry and wet periods can be achieved. Therefore, a solution and method can be provided for the dry and wet period discrimination based on microwave attenuation characteristics, and the research on rainfall monitoring of wireless microwave links is promoted. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow chart of a method for distinguishing dry and wet periods based on comprehensive voting of the present invention; Figure 2 This is a diagram of the total attenuation process of microwave signals in each link from March 1 to 20, 2021 of the present invention. DETAILED DESCRIPTION

[0018] The present invention will be further described below in conjunction with the accompanying drawings and implementation modes.

[0019] Embodiment 1

[0020] like Figure 1 As shown, a method for distinguishing dry and wet periods based on comprehensive voting includes: S1, obtaining microwave attenuation data of each microwave link through a microwave base station; S2, clustering features are selected for the microwave attenuation data of a single link based on a feature selection algorithm to obtain microwave attenuation features of the single link; S3, clustering the microwave attenuation characteristics of each link based on the K-means clustering algorithm to obtain the first dry and wet period discrimination result; S4, clustering the microwave attenuation characteristics of each link based on the fuzzy C-means clustering algorithm to obtain the second dry and wet period discrimination result; S5, clustering the microwave attenuation characteristics of each link based on the sliding standard deviation method to obtain the third dry and wet period discrimination results; S6, according to the first dry and wet period discrimination result, the second dry and wet period discrimination result and the third dry and wet period discrimination result, a comprehensive voting method is used to determine the target dry and wet period discrimination result.

[0021] In the specific implementation process, the signal recorded by the microwave base station in the basin near the dam of a reservoir was processed from March 1 to 20, 2021, and the microwave signal attenuation process of 7 links in the study area was obtained. During the study period, the signal values ​​of the transmitting and receiving ends of the microwave link were recorded in real time every 30 seconds and transmitted to the microwave attenuation signal database. Then, the size of the transmitted microwave signal can be subtracted from the size of the received microwave signal to obtain the microwave signal attenuation in dB.

[0022] like Figure 2 As shown in the figure, the change process of microwave signal attenuation of 7 microwave links during the study period. Among them, the attenuation process of the 7 links is relatively similar as a whole. The total attenuation is in a relatively stable range for most of the time, and the fluctuation range is not large. When rainfall occurs, the attenuation of the link signal will fluctuate significantly, and the time when the microwave signals of each link fluctuate is coincident. This fluctuation indicates that the microwave link can capture the occurrence of rainfall, and the impact of rainfall on the microwave signal is very significant. At the same time, during the study period, the microwave signal attenuation process of the 7 links has differences in details. Therefore, the wireless microwave link network can effectively capture the phenomenon of uneven spatial distribution of rainfall, and compared with the rain gauge, the selected 7 links can more finely reflect the spatial distribution of rainfall.

[0023] In a specific implementation process, the clustering feature selection is performed on the microwave attenuation data of a single link based on the feature selection algorithm to obtain the microwave attenuation feature of the single link, specifically including: Based on the preset attenuation time, select the maximum microwave attenuation rate A in the microwave attenuation data of a single link max , standard deviation of microwave attenuation rate S and maximum microwave attenuation rate change rate V max As clustering features, the microwave attenuation characteristics of a single link are obtained; Among them, the standard deviation S of the microwave attenuation rate within the preset attenuation time and the qth microwave attenuation rate change rate The calculation formula is as follows: Where, S represents the standard deviation of microwave attenuation rate within the preset attenuation time; k represents the number of microwave attenuation data within the preset attenuation time; It indicates the qth microwave attenuation rate within the preset attenuation time, in dB / km; Indicates the average microwave attenuation rate within the preset attenuation time, in dB / km; represents the q-1th microwave attenuation rate within the preset attenuation time, in dB / km; T represents the receiving time interval of microwave attenuation data, in seconds; represents the rate of change of the qth microwave attenuation rate; ∑ represents the summation symbol; Select the maximum value of all microwave attenuation rate change rates as the maximum microwave attenuation rate change rate V max ; The expression of the microwave attenuation characteristic FE of a single link within the preset attenuation time is as follows: Where FE represents the microwave attenuation characteristics of a single link within the preset attenuation time; It represents the maximum microwave attenuation rate within the preset attenuation time; S represents the standard deviation of the microwave attenuation rate within the preset attenuation time; Indicates the maximum microwave attenuation rate change rate within the preset attenuation time.

[0024] The microwave attenuation characteristics of each link are clustered based on the K-means clustering algorithm to obtain the first dry and wet period discrimination result, which specifically includes: Assume that the data set D = {FE1, FE2, ..., FE i , …, FE n}, where n represents the number of links; FE i represents the i-th microwave attenuation feature in the data set D; In the data set D, the microwave attenuation characteristics of K links are selected as the initial clustering centers; Calculate the Euclidean distance between other microwave attenuation features in the data set D and each initial cluster center as the similarity index, and cluster each microwave attenuation feature in the data set D into the category cluster corresponding to the initial cluster center with the highest similarity index; For each category cluster, the average value of all microwave attenuation features within the category cluster is calculated as the new cluster center; Repeat the above cluster center updating process until the updated cluster center remains unchanged, that is, the criterion function of the K-means clustering algorithm converges, and the first dry and wet period discrimination result is obtained.

[0025] The calculation formula of the Euclidean distance between other microwave attenuation features in the data set D and each initial cluster center is as follows: Where D represents the data set composed of microwave attenuation characteristics of all links; Represents the value of the pth attribute of other microwave attenuation features X in the data set D; Represents the value of the pth attribute of the initial cluster center Y; Among them, the dimensions of other microwave attenuation features X and initial clustering centers Y in the data set D are both n.

[0026] The expression of the criterion function of the K-means clustering algorithm is as follows: In the formula, Z represents the criterion function of the K-means clustering algorithm; R represents the total number of cluster categories, that is, the total number of category clusters; x represents the rth category cluster Microwave attenuation characteristics in ; Represents the rth category cluster The cluster center of ; ∑ represents the summation symbol.

[0027] It can be understood that the K-means clustering (K-Means) algorithm can be used to cluster the microwave attenuation characteristics of each link in the study period, and then distinguish the dry and wet process. As shown in Table 1, it shows the dry and wet period discrimination results obtained using the K-means clustering algorithm.

[0028] Table 1 The results of dry and wet period discrimination from March 1 to 20 based on K-Means method

[0029] In Table 1, TA represents the total number of hours with no rainfall both in actual measurement and judged by the model; TB represents the total number of hours with no rainfall in actual measurement but judged by the model as rainfall; TC represents the total number of hours with rainfall in actual measurement but judged by the model as no rainfall; TD represents the total number of hours with rainfall both in actual measurement and judged by the model, that is, the number of hours with rainfall successfully captured.

[0030] When evaluating the accuracy of model discrimination, the accuracy rate RI, capture rate CA and false alarm rate FA are used to evaluate the excellence of the model in distinguishing dry and wet periods. The calculation formulas of the accuracy rate RI, capture rate CA and false alarm rate FA are as follows:

[0031] In the formula, RI represents the accuracy rate, that is, the proportion of hours with successful dry and wet period discrimination in the total time. The larger the proportion, the better the model recognition result. CA represents the capture rate, that is, the proportion of hours successfully captured by the microwave link in the total hours with actual rainfall. The closer this proportion is to 1, the better the performance of the dry and wet period discrimination model. FA represents the false alarm rate, that is, the proportion of hours with no actual rainfall but judged as rainfall by the model in the total hours judged as rainfall by the model. The closer this proportion is to 0, the better the performance of the dry and wet period discrimination model.

[0032] The results in Table 1 show that from March 1 to 20, in the first dry and wet period discrimination results obtained by the K-means clustering algorithm, among the 7 links, except for link 6 and link 7, the clustering accuracy of the other 5 links reached 0.7 or above, among which link 1, link 2 and link 4 had an accuracy greater than or equal to 0.8, indicating that the proportion of hours for which the dry and wet period discrimination results during this period were consistent with the measured results was relatively high, and the overall result was acceptable.

[0033] In addition, during this period, except for link 2, which has a capture rate of 0.67 for the number of hours when rainfall occurred, the capture rates of the remaining 6 links all reached above 0.7, indicating that the K-Means method has a strong ability to identify the occurrence of rainfall, that is, if rainfall occurs within a certain hour, this method can effectively identify the occurrence of rainfall. However, the K-Means method has a minimum false alarm rate of 0.53 and a maximum of 0.72, which means that this method clusters more dry periods into wet periods.

[0034] The method of clustering the microwave attenuation characteristics of each link based on the fuzzy C-means clustering algorithm to obtain the second dry-wet period discrimination result specifically includes: The microwave attenuation characteristics of all links in the data set D each include d eigenvalues, namely FE i ={FE i1 , FE i2 , …, FE ij , …, FE id}, where FE ij Represents the i-th microwave attenuation feature FE in the data set D i The j-th eigenvalue of ; Initialize the membership matrix U, divide the data set D into c clusters, and calculate the corresponding cluster centers as follows: In the formula, represents the jth cluster center; represents the i-th microwave attenuation feature in the data set D; represents the i-th microwave attenuation characteristic FE iWith the jth cluster center The membership degree of ; n represents the number of links; m represents the fuzzy index, and m>1; ∑ represents the summation symbol; Update the membership matrix U, and the corresponding calculation formula is as follows: In the formula, represents the updated i-th microwave attenuation feature FE i With the jth cluster center The membership degree of ; c represents the number of clusters; represents the i-th microwave attenuation characteristic FE i With the jth cluster center distance; represents the i-th microwave attenuation characteristic FE i With the e-th cluster center distance; m represents the fuzzy index, and m>1; ∑ represents the summation symbol; The objective function and constraints of the fuzzy C-means clustering algorithm are calculated as follows: In the formula, represents the objective function of the fuzzy C-means clustering algorithm; U represents the membership matrix; V represents the cluster center matrix; n represents the number of links; m represents the fuzzy index, and m>1; represents the i-th microwave attenuation characteristic FE i With the jth cluster center The degree of membership; represents the i-th microwave attenuation characteristic FE i With the jth cluster center The distance; ∑ represents the summation symbol; If the objective function of the fuzzy C-means clustering algorithm is less than the preset termination threshold, the iteration process of the fuzzy C-means clustering algorithm is stopped and the second dry-wet period discrimination result is output.

[0035] In practical applications, the fuzzy C-means clustering (FCM) algorithm can be used to cluster the microwave attenuation characteristics of each link during the study period, and then distinguish the dry and wet processes. As shown in Table 2, it shows the dry and wet period discrimination results obtained using the FCM method.

[0036] Table 2 The dry and wet period discrimination results from March 1 to 20 based on the FCM method

[0037] The results in Table 2 show that from March 1 to 20, for the dry and wet season discrimination results obtained by clustering using the FCM method, among the 7 links, except for link 6 with a clustering accuracy of 0.64, which was lower than that of the K-Means clustering method, the accuracy of the other 6 links reached 0.85 or above, which was higher than that of the K-Means clustering method. This shows that in terms of overall accuracy, the dry and wet season discrimination results of this method are better.

[0038] For the rainfall capture rate, except for link 2 and link 5, which are 0.53 and 0.64 respectively, the rainfall capture rates of the other five links are all above 0.7. Although the capture rate results are not as good as those of the K-Means method, they are still acceptable.

[0039] From the perspective of false alarm rate, the minimum false alarm rate during this period was 0.44 and the maximum reached 0.75, but it has been improved compared with the K-Means method.

[0040] The microwave attenuation characteristics of each link are clustered based on the sliding standard deviation method to obtain the third dry and wet period discrimination result, which specifically includes: Setting a sliding time window , where t represents the current time; s represents the duration, and s>0; The mean and standard deviation of the microwave attenuation value of a single link within the sliding time window are calculated as follows: In the formula, Represents the mean value of the microwave attenuation value of a single link within the sliding time window; represents the standard deviation of the microwave attenuation value of a single link within the sliding time window; N represents the total number of moments within the sliding time window; represents a sliding time window; represents the microwave attenuation value of a single link at the gth moment in the sliding time window; ∑ represents the summation symbol; The microwave attenuation threshold is determined based on the percentile of the dry period attenuation of a single link. The microwave attenuation threshold is expressed as follows: In the formula, represents the microwave attenuation threshold; represents the 99% quantile of the dry period decay time series; represents the standard deviation of the microwave attenuation value of a single link in the sliding time window; t represents the current time; D represents the dry period; Based on the decision rule, the standard deviation of the microwave attenuation value of a single link in the sliding time window is compared with the microwave attenuation threshold to obtain the discrimination result of the third dry and wet period. The expression of the decision rule is as follows: Where, P represents the third dry and wet period discrimination result; W represents the wet period; D represents the dry period; Represents the standard deviation of the microwave attenuation value of a single link within the sliding time window; Indicates the microwave attenuation threshold.

[0041] According to the third dry-wet period discrimination result, the microwave attenuation value of a single link at the g-th moment in the sliding time window is decomposed to obtain: In the formula, Indicates the microwave attenuation value of a single link at the gth moment in the sliding time window; Indicates the basic attenuation value of a single link at the gth moment in the sliding time window; It represents the attenuation value caused by the rainfall event at the gth moment in the sliding time window of a single link; P represents the third dry and wet period discrimination result; W represents the wet period; and D represents the dry period.

[0042] It should be noted that the sliding standard deviation method can be used to distinguish between dry and wet. For the microwave attenuation threshold, it can be determined according to the percentile of the dry period attenuation of the selected microwave link, and the corresponding calculation formula is: , where D represents the dry period, Represents the 99% quantile of the time series. The thresholds used by the 7 links selected by the present invention are shown in Table 3.

[0043] Table 3 Dry and wet period discrimination thresholds for each link

[0044] During the study period, the standard deviation of the microwave attenuation series was solved using the sliding standard deviation method and compared with the set microwave attenuation threshold to identify the dry and wet periods. The dry and wet period identification results are shown in Table 4.

[0045] Table 4 The dry and wet period discrimination results from March 1 to 20 based on the sliding standard deviation method

[0046] In Table 4, from March 1 to 20, the accuracy of the seven links exceeded 0.7, and most of them were above 0.85, which is an excellent accuracy indicator.

[0047] Regarding the rainfall capture rate, the results of this method are insufficient, with the indicators of three links being lower than 0.6 and only two links reaching 0.7.

[0048] Compared with the K-Means method and the FCM method, the sliding standard deviation method has better control over the null alarm rate. Except for link 6, the null alarm rates of the other six links are all less than 0.3. Therefore, the null alarm rate of this method is within a relatively ideal range.

[0049] The method of determining the target dry and wet period discrimination result by using a comprehensive voting method according to the first dry and wet period discrimination result, the second dry and wet period discrimination result and the third dry and wet period discrimination result specifically includes: If at least two of the first dry-wet period discrimination result, the second dry-wet period discrimination result and the third dry-wet period discrimination result are wet periods, the target dry-wet period discrimination result is determined to be a wet period; otherwise, the target dry-wet period discrimination result is determined to be a dry period.

[0050] In practical applications, the results obtained by the three methods for distinguishing dry and wet periods can be comprehensively considered by voting. Specifically, if more than or equal to two of the methods determine that rainfall occurs within a certain hour, then rainfall occurs within this hour and it is a wet period, otherwise it is a dry period. The dry and wet period distinction results obtained by the comprehensive voting method are shown in Table 5.

[0051] Table 5 The results of dry and wet period discrimination from March 1 to 20 based on the comprehensive voting method

[0052] In Table 5, by voting on the dry and wet period discrimination results from March 1 to 20, the final dry and wet discrimination results can be obtained, and their accuracy is not much different from the results obtained by the three methods.

[0053] For the rainfall capture rate, the comprehensive voting method has an overall decrease compared with the K-Means and FCM methods, but the decrease is not large and is within an acceptable range. However, compared with the sliding standard deviation method, among the 7 links, except for link 4 and link 6, which decreased by 0.05 and 0.07, the other 5 links increased, with the maximum increase of 0.5 and the minimum increase of 0.34. This shows that in March, the comprehensive voting of the three methods to determine the dry and wet periods has improved the rainfall capture rate compared with the traditional method of using only the sliding standard deviation. This reflects to a certain extent the improvement of the wireless microwave link's ability to capture rainfall.

[0054] For the false alarm rate, in the dry-wet discrimination results determined by comprehensive voting using the three methods, except for link 6, the false alarm rates of the other six links have significantly decreased compared with the K-Means and FCM methods. Although the false alarm rate has increased slightly compared with the sliding standard deviation method, it is still within an acceptable range.

[0055] In summary, compared with the traditional method of using only the sliding standard deviation, the use of the comprehensive voting method to distinguish dry and wet periods has a slightly higher false alarm rate, but the sensitivity of the microwave link to capture rainfall has increased significantly. Therefore, this method is more applicable to distinguish dry and wet periods than the traditional method.

[0056] The present invention uses a dry and wet period discrimination method based on comprehensive voting to obtain microwave attenuation data of each microwave link through a microwave base station; clustering feature selection is performed on the microwave attenuation data of a single link based on a feature selection algorithm to obtain the microwave attenuation feature of the single link; the microwave attenuation features of each link are clustered based on a K-means clustering algorithm to obtain a first dry and wet period discrimination result; the microwave attenuation features of each link are clustered based on a fuzzy C-means clustering algorithm to obtain a second dry and wet period discrimination result; the microwave attenuation features of each link are clustered based on a sliding standard deviation method to obtain a third dry and wet period discrimination result; according to the first dry and wet period discrimination result, the second dry and wet period discrimination result and the third dry and wet period discrimination result, a comprehensive voting method is used to determine a target dry and wet period discrimination result, thereby improving the sensitivity of the microwave link in capturing the rainfall process and realizing accurate discrimination of the dry and wet period.

[0057] The present invention combines the dry and wet period discrimination results of the K-means clustering algorithm, the fuzzy C-means clustering algorithm and the sliding standard deviation method through a comprehensive voting method. Compared with the use of a single sliding standard deviation method for dry and wet period discrimination, the sensitivity of the microwave link in capturing the rainfall process can be significantly improved, and the accurate discrimination of the dry and wet periods can be achieved. Therefore, a solution and method can be provided for the dry and wet period discrimination based on microwave attenuation characteristics, and the research on rainfall monitoring of wireless microwave links is promoted.

[0058] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0059] Those skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0060] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

Claims

1. A method for distinguishing dry and wet periods based on comprehensive voting, characterized in that: include: Obtain microwave attenuation data of each microwave link through a microwave base station; Based on the feature selection algorithm, the microwave attenuation data of a single link is clustered and selected to obtain the microwave attenuation characteristics of the single link; The microwave attenuation characteristics of each link are clustered based on the K-means clustering algorithm to obtain the first dry and wet period discrimination result; The microwave attenuation characteristics of each link are clustered based on the fuzzy C-means clustering algorithm to obtain the second dry and wet period discrimination results; The microwave attenuation characteristics of each link are clustered based on the sliding standard deviation method to obtain the discrimination results of the third dry and wet period; According to the first dry and wet period discrimination results, the second dry and wet period discrimination results and the third dry and wet period discrimination results, the comprehensive voting method is used to determine the target dry and wet period discrimination results.

2. A method for distinguishing dry and wet periods based on comprehensive voting according to claim 1, characterized in that: The clustering feature selection of the microwave attenuation data of a single link based on the feature selection algorithm to obtain the microwave attenuation feature of the single link specifically includes: Based on the preset attenuation time, select the maximum microwave attenuation rate A in the microwave attenuation data of a single link max , standard deviation of microwave attenuation rate S and maximum microwave attenuation rate change rate V max As clustering features, the microwave attenuation characteristics of a single link are obtained; Among them, the standard deviation S of the microwave attenuation rate within the preset attenuation time and the qth microwave attenuation rate change rate The calculation formula is as follows: Where, S represents the standard deviation of microwave attenuation rate within the preset attenuation time; k represents the number of microwave attenuation data within the preset attenuation time; It indicates the qth microwave attenuation rate within the preset attenuation time, in dB / km; Indicates the average microwave attenuation rate within the preset attenuation time, in dB / km; represents the q-1th microwave attenuation rate within the preset attenuation time, in dB / km; T represents the receiving time interval of microwave attenuation data, in seconds; represents the rate of change of the qth microwave attenuation rate; ∑ represents the summation symbol; Select the maximum value of all microwave attenuation rate change rates as the maximum microwave attenuation rate change rate V max ; The expression of the microwave attenuation characteristic FE of a single link within the preset attenuation time is as follows: Where FE represents the microwave attenuation characteristics of a single link within the preset attenuation time; It represents the maximum microwave attenuation rate within the preset attenuation time; S represents the standard deviation of the microwave attenuation rate within the preset attenuation time; Indicates the maximum microwave attenuation rate change rate within the preset attenuation time.

3. The method for distinguishing dry and wet periods based on comprehensive voting according to claim 1 is characterized in that: The microwave attenuation characteristics of each link are clustered based on the K-means clustering algorithm to obtain the first dry and wet period discrimination result, which specifically includes: Assume that the data set D = {FE1, FE2, ..., FE i , …, FE n }, where n represents the number of links; FE i represents the i-th microwave attenuation feature in the data set D; In the data set D, the microwave attenuation characteristics of K links are selected as the initial clustering centers; Calculate the Euclidean distance between other microwave attenuation features in the data set D and each initial cluster center as the similarity index, and cluster each microwave attenuation feature in the data set D into the category cluster corresponding to the initial cluster center with the highest similarity index; For each category cluster, the average value of all microwave attenuation features within the category cluster is calculated as the new cluster center; Repeat the above cluster center updating process until the updated cluster center remains unchanged, that is, the criterion function of the K-means clustering algorithm converges, and the first dry and wet period discrimination result is obtained.

4. A method for distinguishing dry and wet periods based on comprehensive voting according to claim 3, characterized in that: The calculation formula of the Euclidean distance between other microwave attenuation features in the data set D and each initial cluster center is as follows: Where D represents the data set composed of microwave attenuation characteristics of all links; Represents the value of the pth attribute of other microwave attenuation features X in the data set D; Represents the value of the pth attribute of the initial cluster center Y; Among them, the dimensions of other microwave attenuation features X and initial clustering centers Y in the data set D are both n.

5. The method for distinguishing dry and wet periods based on comprehensive voting according to claim 3 is characterized in that: The expression of the criterion function of the K-means clustering algorithm is as follows: In the formula, Z represents the criterion function of the K-means clustering algorithm; R represents the total number of cluster categories, that is, the total number of category clusters; x represents the rth category cluster Microwave attenuation characteristics in ; Represents the rth category cluster The cluster center of ; ∑ represents the summation symbol.

6. A method for distinguishing dry and wet periods based on comprehensive voting according to claim 1 or 3, characterized in that: The method of clustering the microwave attenuation characteristics of each link based on the fuzzy C-means clustering algorithm to obtain the second dry-wet period discrimination result specifically includes: The microwave attenuation characteristics of all links in the data set D each include d eigenvalues, namely FE i ={FE i1 , FE i2 , …, FE ij , …, FE id }, where FE ij Represents the i-th microwave attenuation feature FE in the data set D i The j-th eigenvalue of ; Initialize the membership matrix U, divide the data set D into c clusters, and calculate the corresponding cluster centers as follows: In the formula, represents the jth cluster center; represents the i-th microwave attenuation feature in the data set D; represents the i-th microwave attenuation characteristic FE i With the jth cluster center The membership degree of ; n represents the number of links; m represents the fuzzy index, and m>1; ∑ represents the summation symbol; Update the membership matrix U, and the corresponding calculation formula is as follows: In the formula, represents the updated i-th microwave attenuation feature FE i With the jth cluster center The membership degree of ; c represents the number of clusters; represents the i-th microwave attenuation characteristic FE i With the jth cluster center distance; represents the i-th microwave attenuation characteristic FE i With the e-th cluster center distance; m represents the fuzzy index, and m>1; ∑ represents the summation symbol; The objective function and constraints of the fuzzy C-means clustering algorithm are calculated as follows: In the formula, represents the objective function of the fuzzy C-means clustering algorithm; U represents the membership matrix; V represents the cluster center matrix; n represents the number of links; m represents the fuzzy index, and m>1; represents the i-th microwave attenuation characteristic FE i With the jth cluster center The degree of membership; represents the i-th microwave attenuation characteristic FE i With the jth cluster center The distance; ∑ represents the summation symbol; If the objective function of the fuzzy C-means clustering algorithm is less than the preset termination threshold, the iteration process of the fuzzy C-means clustering algorithm is stopped and the second dry-wet period discrimination result is output.

7. The method for distinguishing dry and wet periods based on comprehensive voting according to claim 1 is characterized in that: The microwave attenuation characteristics of each link are clustered based on the sliding standard deviation method to obtain the third dry and wet period discrimination result, which specifically includes: Setting a sliding time window , where t represents the current time; s represents the duration, and s>0; The mean and standard deviation of the microwave attenuation value of a single link within the sliding time window are calculated as follows: In the formula, Represents the mean value of the microwave attenuation value of a single link within the sliding time window; represents the standard deviation of the microwave attenuation value of a single link within the sliding time window; N represents the total number of moments within the sliding time window; represents a sliding time window; represents the microwave attenuation value of a single link at the gth moment in the sliding time window; ∑ represents the summation symbol; The microwave attenuation threshold is determined based on the percentile of the dry period attenuation of a single link. The microwave attenuation threshold is expressed as follows: In the formula, represents the microwave attenuation threshold; represents the 99% quantile of the dry period decay time series; represents the standard deviation of the microwave attenuation value of a single link in the sliding time window; t represents the current time; D represents the dry period; Based on the decision rule, the standard deviation of the microwave attenuation value of a single link in the sliding time window is compared with the microwave attenuation threshold to obtain the discrimination result of the third dry and wet period. The expression of the decision rule is as follows: Where, P represents the third dry and wet period discrimination result; W represents the wet period; D represents the dry period; Represents the standard deviation of the microwave attenuation value of a single link within the sliding time window; Indicates the microwave attenuation threshold.

8. The method for distinguishing dry and wet periods based on comprehensive voting according to claim 7 is characterized in that: According to the third dry-wet period discrimination result, the microwave attenuation value of a single link at the g-th moment in the sliding time window is decomposed to obtain: In the formula, Indicates the microwave attenuation value of a single link at the gth moment in the sliding time window; Indicates the basic attenuation value of a single link at the gth moment in the sliding time window; It represents the attenuation value caused by the rainfall event at the gth moment in the sliding time window of a single link; P represents the third dry and wet period discrimination result; W represents the wet period; and D represents the dry period.

9. The method for distinguishing dry and wet periods based on comprehensive voting according to claim 1 is characterized in that: The method of determining the target dry and wet period discrimination result by using a comprehensive voting method according to the first dry and wet period discrimination result, the second dry and wet period discrimination result and the third dry and wet period discrimination result specifically includes: If at least two of the first dry-wet period discrimination result, the second dry-wet period discrimination result and the third dry-wet period discrimination result are wet periods, the target dry-wet period discrimination result is determined to be a wet period; otherwise, the target dry-wet period discrimination result is determined to be a dry period.

Images