Sun and rain period classification method based on microwave attenuation data multi-time scale combination discrimination

Abstract

The invention provides a sunny and rainy period classification method based on microwave signal multi-time scale combination discrimination, and relates to the technical field of rainfall monitoring, and the method comprises the steps: obtaining microwave attenuation signal intensity data in a target drainage basin and rainfall data of a ground station; preprocessing the data; selecting and counting characteristic values of the microwave attenuation signals under different time scales to form a characteristic matrix, and taking the characteristic matrix as a data set of a machine learning model; first 75% of the data set is selected as a training set, and machine learning models of different time scales are trained respectively; optimal parameters are determined through Bayesian optimization; the last 25% of the data set is selected as a verification set, and the machine learning models of different time scales are verified; the verified model is driven by the actually measured data of the target drainage basin to obtain discrimination results of different time scales, and the rainfall period is discriminated through a weight factor scoring method, so that the feature information of multiple time scales can be fused, the limitation of single time scale discrimination is made up, and the reliability of rainfall period discrimination is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of precipitation monitoring, and particularly relates to a method for classifying sunny and rainy periods by combining microwave signals at multiple time scales for discrimination. Background Art

[0002] In order to effectively improve the early warning level of precipitation monitoring, it is necessary to improve the accuracy and real-time performance of precipitation monitoring. Currently, the commonly used precipitation monitoring means include the following three: rain gauges, weather radars, and rain-measuring satellites. Although these methods have been widely used in the water conservancy and meteorology industries, they have significant limitations such as being prone to missing heavy rain centers, having limited monitoring effects on the transformation from "clouds in the air" to actual "rain on the ground", low inversion accuracy, and poor real-time performance.

[0003] The method of classifying sunny and rainy periods by using the rain attenuation effect of microwave links is an important part of microwave link precipitation monitoring and is also the basis for inverting the rainfall intensity along the path. However, most of the existing methods for classifying sunny and rainy periods based on the rain attenuation effect of microwave links study classification problems at a single time scale. They are often limited by the selection of the representative time scale, do not consider that the characteristic information of data is different at different time scales, and do not fully combine the multi-scale characteristic advantages of data, and there are deficiencies in combining multi-scale characteristics.

[0004] Therefore, it is necessary to provide a method for classifying sunny and rainy periods by combining microwave signals at multiple time scales for discrimination to solve the above technical problems. Summary of the Invention

[0005] To solve the above technical problems, the present invention provides a method for classifying sunny and rainy periods by combining microwave signals at multiple time scales for discrimination, which is used to solve the problem that most of the existing methods for classifying sunny and rainy periods based on the rain attenuation effect of microwave links study classification problems at a single time scale, are limited by the selection of the representative time scale, and cannot further explore the advantages of microwave attenuation data at multiple time resolutions, and thus cannot achieve the classification of sunny and rainy periods with multi-scale feature combination.

[0006] The method for classifying sunny and rainy periods by combining microwave signals at multiple time scales for discrimination provided by the present invention includes: Obtaining microwave attenuation signal data of a target basin and rainfall data of a ground station; Preprocessing the microwave attenuation signal data; Selecting and statistically analyzing the characteristic values of the preprocessed microwave attenuation signal data at different time scales, determining the corresponding characteristic matrix, and combining the rainfall data to construct an input data set for a machine learning model; Dividing the input data set into a training set and a validation set, and training machine learning models at different time scales based on the training set; Determine the optimal parameters of machine learning models at different time scales through the Bayesian optimization algorithm; Verify the effects of machine learning models at different time scales based on the validation set; Based on the microwave attenuation signal data, determine the rainfall period model discrimination results of machine learning models at different time scales, and use the weight factor scoring method to determine the final rainfall period discrimination results according to the rainfall period model discrimination results.

[0007] Preferably, based on the microwave signal transceiver device, obtain the microwave attenuation signal data of the target basin at 15-minute and 1-hour resolutions for each microwave link.

[0008] Preferably, the rainfall data of the ground station includes the measured rainfall data of rain gauges, meteorological stations, and hydrological stations.

[0009] Preferably, the preprocessing of the microwave attenuation signal data includes removing the microwave attenuation signal data corresponding to abnormal microwave links and filtering clutter from the microwave attenuation signal data corresponding to effective microwave links.

[0010] Preferably, select and statistically analyze the eigenvalues of the preprocessed microwave attenuation signal data at different time scales, determine the corresponding feature matrix, and combine the rainfall data to construct the input dataset of the machine learning model, specifically including: Select and statistically analyze the eigenvalues of the preprocessed microwave attenuation signal data at different time scales to construct the corresponding feature matrix , where , , , , , respectively represent the microwave link length, microwave attenuation signal frequency, microwave attenuation signal mean, microwave attenuation signal median, microwave attenuation signal maximum, and microwave attenuation signal minimum; k represents the sequence number of the microwave attenuation signal data, ; Determine the label value according to the rainfall data , construct the input dataset of the machine learning model according to the feature matrix and the label value , and determine the datasets Q 1 、Q 2 corresponding to 15-minute and 1-hour resolutions, where the label value takes the value of 0, indicating the non-rainfall period, and the label value takes the value of 1, indicating the rainfall period.

[0011] Preferably, based on the dataset Q 1For the corresponding training set, set the parameters of the support vector machine model and select the Gaussian radial basis kernel function as the kernel function of the support vector machine model. The calculation formula of the Gaussian radial basis kernel function is as follows: In the formula, represents the Gaussian radial basis kernel function; represents the feature matrix and the label value of the Euclidean distance; represents the width of the Gaussian radial basis kernel function, ; represents the exponential function with the natural constant e as the base; Based on the dataset Q 2 corresponding training set, set the parameters of the class boosting model and create a data pool.

[0012] Preferably, the parameters of the class boosting model include a loss function, a decision tree depth, a learning rate, a subsample ratio for each iteration of training, a feature sampling ratio for each level, a maximum number of iterations, and the number of rounds for early stopping.

[0013] Preferably, the optimal parameters of the machine learning model at different time scales are determined by the Bayesian optimization algorithm, specifically including: Obtain the measured rainfall period results and the rainfall period prediction results of the machine learning model, and set the confusion matrix corresponding to the machine learning model, including true positive (TP), true negative (TN), false positive (FP), and false negative (FN). Among them, true positive (TP) means that both the measured rainfall period results and the rainfall period prediction results of the machine learning model are wet periods; true negative (TN) means that both the measured rainfall period results and the rainfall period prediction results of the machine learning model are dry periods; false positive (FP) means that the measured rainfall period result is a dry period and the rainfall period prediction result of the machine learning model is a wet period; false negative (FN) means that the measured rainfall period result is a wet period and the rainfall period prediction result of the machine learning model is a dry period; Calculate the precision corresponding to the machine learning model according to the confusion matrix, and use the precision as the objective function. The calculation formula of the precision is as follows: In the formula, ACC represents precision; TP represents true positive; TN represents true negative; FP represents false positive; FN represents false negative; Set the objective space of the Bayesian optimization algorithm, select a uniform distribution as the prior distribution of the Bayesian optimization algorithm, maximize the precision, and determine the optimal parameters of the machine learning model.

[0014] Preferably, based on the microwave attenuation signal data, the rainfall period model discrimination results of machine learning models at different time scales are determined, and the final rainfall period discrimination result is determined by using the weight factor scoring method, which specifically includes: Input the microwave attenuation signal data into the verified support vector machine model, and the rainfall period model discrimination result at a 15-minute resolution is output. Then input the microwave attenuation signal data into the verified class boosting model, and the rainfall period model discrimination result at a 1-hour resolution is output; Set every 1 hour as the statistical period. The rainfall period model discrimination results every 15 minutes output by the support vector machine model are 、 、 、 The rainfall period model discrimination results every 1 hour output by the class boosting model are Calculate the total score f by using the weight factor scoring method, and the corresponding calculation formula is as follows: In the formula, f represents the total score; represents the weight score; represents the rainfall period model discrimination result. Among them, when the rainfall period model discrimination result is non-rainfall period, takes the value of 0, and when the rainfall period model discrimination result is rainfall period, takes the value of 1; If then determine that the final rainfall period discrimination result is rainfall period; otherwise, determine that the final rainfall period discrimination result is non-rainfall period.

[0015] Preferably, the weight scores 、 、 、 、 take the values of 0.1, 0.15, 0.2, 0.25, and 0.3 respectively.

[0016] Compared with the related technology, a sunny-rainy period classification method for microwave signal multi-time scale combination discrimination provided by the present invention has the following beneficial effects: The present invention obtains the microwave attenuation signal data of the target basin and the rainfall data of the ground station; preprocesses the microwave attenuation signal data; selects and counts the eigenvalues of the preprocessed microwave attenuation signal data at different time scales, determines the corresponding feature matrix, and constructs the input data set of the machine learning model in combination with the rainfall data; divides the input data set into a training set and a validation set, and trains the machine learning models at different time scales based on the training set; determines the optimal parameters of the machine learning models at different time scales through the Bayesian optimization algorithm; verifies the effects of the machine learning models at different time scales based on the validation set; determines the rainfall period model discrimination results of the machine learning models at different time scales based on the microwave attenuation signal data, and uses the weight factor scoring method to determine the final rainfall period discrimination result according to the rainfall period model discrimination results, so as to further explore the advantages of microwave attenuation data at multiple time resolutions, fuse the feature information of multiple time scales, make up for the limitations of single-time-scale discrimination, improve the reliability of rainfall period discrimination, and thus realize the classification of sunny and rainy periods with multi-scale feature combination, solve the problem that most existing methods study classification problems at a single time scale and are limited by the selection of representative time scales. At the same time, the method of the present invention is widely applied to the field of rainfall inversion by microwave signals. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a flowchart of a sunny and rainy period classification method for multi-time-scale combination discrimination of microwave signals according to the present invention; Figure 2 is a distributed scatter plot of the verification results of the present invention; Figure 3 is a sunny and rainy period discrimination result diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

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

[0019] As Figure 1 shown, a sunny and rainy period classification method for multi-time-scale combination discrimination of microwave signals includes: S1. Obtain the microwave attenuation signal data of the target basin and the rainfall data of the ground station; S2. Preprocess the microwave attenuation signal data; S3. Select and count the eigenvalues of the preprocessed microwave attenuation signal data at different time scales, determine the corresponding feature matrix, and construct the input data set of the machine learning model in combination with the rainfall data; S4. Divide the input data set into a training set and a validation set, and train the machine learning models at different time scales based on the training set; S5. Determine the optimal parameters of machine learning models at different time scales through the Bayesian optimization algorithm; S6. Verify the effects of machine learning models at different time scales based on the validation set; S7. Based on the microwave attenuation signal data, determine the rainfall period model discrimination results of machine learning models at different time scales, and use the weight factor scoring method to determine the final rainfall period discrimination result according to the rainfall period model discrimination results.

[0020] It should be noted that the research area of the present invention is located in a near-dam basin in a mountainous area of Yichang City, belonging to the mountainous landform of the Three Gorges of the Yangtze River. The elevation difference in this basin is large. In the upper and middle reaches, there are rolling mountains and many river valleys, with the risk of falling rocks; in the lower reaches, the river channel widens to dozens of meters, mainly consisting of open piedmont plains, where there are many villages and farmlands. This basin belongs to the subtropical continental monsoon climate. Since the establishment of the representative hydrological station in this basin, there have been mainly two major floods, and the peak flood flow at the control section has reached as high as 732 m 3 / s, and this basin faces the risk of sudden local rainstorm floods.

[0021] In the specific implementation process, based on the microwave signal transceiver device, obtain the microwave attenuation signal data of the target basin at 15-minute and 1-hour resolutions for each microwave link.

[0022] In practical applications, based on 50 microwave signal transceiver devices deployed in this basin, the microwave attenuation signal data of each microwave link at 15-minute and 1-hour resolutions can be obtained.

[0023] The rainfall data of the ground station includes the measured rainfall data of rain gauges, meteorological stations, and hydrological stations.

[0024] In practical applications, the rainfall data of the ground station specifically includes the measured rainfall data of 18 newly built meteorological stations and 10 existing hydrological stations in this basin.

[0025] The preprocessing of the microwave attenuation signal data includes removing the microwave attenuation signal data corresponding to abnormal microwave links and filtering clutter from the microwave attenuation signal data corresponding to effective microwave links.

[0026] Among them, the preprocessing process of the microwave attenuation signal data specifically includes: removing abnormal microwave links with obvious abnormal fluctuations in the signal, excessive extreme values, and large-area data missing, and filtering clutter from effective microwave links.

[0027] The selection and statistics of the eigenvalue of the preprocessed microwave attenuation signal data at different time scales, determining the corresponding feature matrix, and combining the rainfall data to construct the input data set of the machine learning model specifically include: Select and count the eigenvalues of the preprocessed microwave attenuation signal data at different time scales, and construct the corresponding feature matrix , where , , , , , respectively represent the microwave link length, microwave attenuation signal frequency, microwave attenuation signal mean, microwave attenuation signal median, microwave attenuation signal maximum, and microwave attenuation signal minimum; k represents the sequence number of the microwave attenuation signal data, , n = 100; Determine the label value according to the rainfall data , and construct the input data set of the machine learning model according to the feature matrix and the label value , and determine the data sets Q 1 , Q 2 corresponding to 15-minute and 1-hour resolutions, where the label value takes the value of 0, indicating the non-rainy period, and the label value takes the value of 1, indicating the rainy period.

[0028] Based on the data set Q 1 corresponding training set, set the parameters of the support vector machine model, and select the Gaussian radial basis kernel function as the kernel function of the support vector machine model. The calculation formula of the Gaussian radial basis kernel function is as follows: In the formula, represents the Gaussian radial basis kernel function; represents the feature matrix and the label value 's Euclidean distance; represents the width of the Gaussian radial basis kernel function, ; represents the exponential function with the natural constant e as the base; Based on the data set Q 2 corresponding training set, set the parameters of the class boosting model, and create a data pool.

[0029] It should be noted that when training the support vector machine (SVM) model, the first 75% of the data in the data set Q1 can be used as the training set, set the parameters of the support vector machine model, and select the Gaussian radial basis kernel function as the kernel function of the model, so as to distinguish the rainy period.

[0030] When training a CatBoost model, the first 75% of the data in dataset Q2 can be used as the training set, that is, the data from 0:00 on May 1st to 24:00 on July 31st. Then, set the parameters of the CatBoost model, create a Pool class, and further identify the rainfall period.

[0031] The parameters of the CatBoost model include loss function, decision tree depth, learning rate, subsample ratio for each iteration of training, feature sampling ratio for each level, maximum number of iterations, and number of rounds for early stopping.

[0032] Specifically, the parameters of the CatBoost model are as follows: loss function (loss_function) with a corresponding value of Logloss, decision tree depth (depth) with a corresponding value of 8, learning rate (learning_rate) with a corresponding value of 0.04, subsample ratio for each iteration of training (subsample) with a corresponding value of 0.8, feature sampling ratio for each level (colsample_bylevel) with a corresponding value of 0.8, maximum number of iterations (iterations) with a corresponding value of 1500, and number of rounds for early stopping (early_stopping_rounds) with a corresponding value of 20.

[0033] Determining the optimal parameters of machine learning models at different time scales through the Bayesian optimization algorithm specifically includes: As shown in Table 1, obtain the measured rainfall period results and the rainfall period prediction results of the machine learning model, and set the confusion matrix corresponding to the machine learning model, including true positive (TP), true negative (TN), false positive (FP), and false negative (FN). Among them, true positive (TP) means that both the measured rainfall period result and the rainfall period prediction result of the machine learning model are wet periods; true negative (TN) means that both the measured rainfall period result and the rainfall period prediction result of the machine learning model are dry periods; false positive (FP) means that the measured rainfall period result is a dry period and the rainfall period prediction result of the machine learning model is a wet period; false negative (FN) means that the measured rainfall period result is a wet period and the rainfall period prediction result of the machine learning model is a dry period. Table 1 Table of Confusion Matrix Elements

[0034] Calculate the precision corresponding to the machine learning model according to the confusion matrix, and use the precision as the objective function. The formula for calculating the precision is as follows: In the formula, ACC represents precision; TP represents true positive; TN represents true negative; FP represents false positive; FN represents false negative. Set the objective space of the Bayesian optimization algorithm, select the uniform distribution as the prior distribution of the Bayesian optimization algorithm, maximize the precision rate, and determine the optimal parameters of the machine learning model.

[0035] In practical applications, select the last 25% of the data in the input dataset as the validation set, that is, the data from 0:00 on August 1st to 24:00 on August 31st, and verify the effects of machine learning models with different time scales based on the validation set. The distributed dot plot of the verification results is as Figure 2 shown.

[0036] Based on the microwave attenuation signal data, determine the rainfall period model discrimination results of machine learning models with different time scales, and use the weight factor scoring method to determine the final rainfall period discrimination result according to the rainfall period model discrimination results, specifically including: Input the microwave attenuation signal data into the verified support vector machine model, and output the rainfall period model discrimination result at a 15-minute resolution. Then input the microwave attenuation signal data into the verified class boosting model, and output the rainfall period model discrimination result at a 1-hour resolution; Set every 1 hour as the statistical period. The rainfall period model discrimination results for every 15 minutes output by the support vector machine model are , , , , and the rainfall period model discrimination results for every 1 hour output by the class boosting model are , and use the weight factor scoring method to calculate the total score f. The corresponding calculation formula is as follows: In the formula, f represents the total score; represents the weight score; represents the rainfall period model discrimination result. Among them, when the rainfall period model discrimination result is a non-rainfall period, takes the value of 0, and when the rainfall period model discrimination result is a rainfall period, takes the value of 1; If , then determine that the final rainfall period discrimination result is a rainfall period; otherwise, determine that the final rainfall period discrimination result is a non-rainfall period. The corresponding discrimination result is as Figure 3 shown.

[0037] The weight scores , , , , take the values of 0.1, 0.15, 0.2, 0.25, and 0.3 respectively.

[0038] Through the introduction of the above embodiments, the present invention provides a method for classifying sunny and rainy periods by combining microwave signal discrimination on multiple time scales. The method includes obtaining microwave attenuation signal data of the target basin and rainfall data of the ground station; preprocessing the microwave attenuation signal data; selecting and statistically analyzing the eigenvalues of the preprocessed microwave attenuation signal data at different time scales to determine the corresponding feature matrix, and combining the rainfall data to construct the input data set of the machine learning model; dividing the input data set into a training set and a validation set, and training machine learning models at different time scales based on the training set respectively; determining the optimal parameters of the machine learning models at different time scales through the Bayesian optimization algorithm; verifying the effects of the machine learning models at different time scales based on the validation set respectively; determining the rainfall period model discrimination results of the machine learning models at different time scales based on the microwave attenuation signal data, and using the weighted factor scoring method to determine the final rainfall period discrimination result according to the rainfall period model discrimination results. Thus, the advantages of microwave attenuation data at multiple time resolutions can be further explored, the feature information of multiple time scales can be fused, the limitations of single-time-scale discrimination can be compensated, and the reliability of rainfall period discrimination can be improved. Therefore, the classification of sunny and rainy periods with multi-scale feature combination can be realized, solving the problem that most existing methods study classification problems on a single time scale and are limited by the selection of representative time scales. At the same time, the method of the present invention is widely applied to the field of rainfall inversion by microwave signals.

[0039] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for realizing the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0040] Those of ordinary skill in the art will appreciate that all or part of the steps in the various methods of the above embodiments can be accomplished by instructing relevant hardware through a program, which can be stored in a computer-readable storage medium. The storage medium includes read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, magnetic disk storage, tape storage, or any other medium that can be used to carry or store data and is computer-readable.

[0041] It should also be noted that the term "comprising", "including" or any other variation thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the element.

Claims

1. A method for classifying sunny and rainy seasons by combining microwave signals at multiple time scales, characterized in that: include: Obtain microwave attenuation signal data of the target basin and rainfall data from the ground station; Preprocessing the microwave attenuation signal data; Select and count the eigenvalues ​​of the preprocessed microwave attenuation signal data at different time scales, determine the corresponding feature matrix, and construct an input data set of the machine learning model in combination with the rainfall data; Dividing the input data set into a training set and a validation set, and training machine learning models of different time scales based on the training sets; Determine the optimal parameters of machine learning models at different time scales through Bayesian optimization algorithms; Verifying the effects of machine learning models at different time scales based on the verification set; Based on the microwave attenuation signal data, the rainfall period model discrimination results of the machine learning models at different time scales are determined, and the final rainfall period discrimination result is determined according to the rainfall period model discrimination results using a weight factor scoring method.

2. The method for classifying sunny and rainy seasons by combining microwave signals at multiple time scales according to claim 1 is characterized in that: Based on the microwave signal transceiver, the microwave attenuation signal data of the target watershed at a resolution of 15 minutes and 1 hour is obtained for each microwave link.

3. The method for classifying sunny and rainy seasons by combining microwave signals at multiple time scales according to claim 1 is characterized in that: The rainfall data of the ground station includes rainfall measured data of the rain gauge station, meteorological station and hydrological station.

4. The method for classifying sunny and rainy seasons by combining microwave signals at multiple time scales according to claim 1 is characterized in that: The preprocessing of the microwave attenuation signal data includes removing the microwave attenuation signal data corresponding to abnormal microwave links and performing clutter filtering on the microwave attenuation signal data corresponding to valid microwave links.

5. The method for classifying sunny and rainy seasons by combining microwave signals at multiple time scales according to claim 1 is characterized in that: The selecting and counting the eigenvalues ​​of the preprocessed microwave attenuation signal data at different time scales, determining the corresponding feature matrix, and combining the rainfall data to construct an input data set for the machine learning model specifically includes: Select and count the eigenvalues ​​of the preprocessed microwave attenuation signal data at different time scales to construct the corresponding feature matrix ,in, , , , , , They represent the microwave link length, microwave attenuation signal frequency, microwave attenuation signal mean, microwave attenuation signal median, microwave attenuation signal maximum, and microwave attenuation signal minimum respectively; k represents the sequence number of the microwave attenuation signal data, ; Determine the tag value according to the rainfall data , construct the input data set of the machine learning model according to the feature matrix and the label value , and determine the data sets Q1 and Q2 corresponding to the 15-minute and 1-hour resolutions, where the label value The value is 0, indicating a non-rainy period. The value is 1, indicating the rainfall period.

6. The method for classifying sunny and rainy seasons by combining microwave signals at multiple time scales according to claim 5 is characterized in that: Based on the training set corresponding to the data set Q1, the parameters of the support vector machine model are set, and the Gaussian radial basis kernel function is selected as the kernel function of the support vector machine model. The calculation formula of the Gaussian radial basis kernel function is as follows: In the formula, represents the Gaussian radial basis kernel function; Represents the feature matrix With label value The Euclidean distance of represents the width of the Gaussian radial basis kernel function, ; It represents the exponential function with the natural constant e as the base; Based on the training set corresponding to the data set Q2, the parameters of the category boosting model are set, and a data pool is created.

7. The method for classifying sunny and rainy seasons by combining microwave signals at multiple time scales according to claim 6 is characterized in that: The parameters of the category boosting model include loss function, decision tree depth, learning rate, subsample ratio for each iteration training, feature sampling ratio for each level, maximum number of iterations, and number of early stopping rounds.

8. The method for classifying sunny and rainy seasons by combining microwave signals at multiple time scales according to claim 1 is characterized in that: The method of determining the optimal parameters of the machine learning model at different time scales by using the Bayesian optimization algorithm specifically includes: Obtain the measured rainfall period result and the rainfall period prediction result of the machine learning model, and set the confusion matrix corresponding to the machine learning model, including true positive examples TP, true negative examples TN, false positive examples FP and false negative examples FN, wherein the true positive example TP indicates that the measured rainfall period result and the rainfall period prediction result of the machine learning model are both wet periods; the true negative example TN indicates that the measured rainfall period result and the rainfall period prediction result of the machine learning model are both dry periods; the false positive example FP indicates that the measured rainfall period result is a dry period, and the rainfall period prediction result of the machine learning model is a wet period; the false negative example FN indicates that the measured rainfall period result is a wet period, and the rainfall period prediction result of the machine learning model is a dry period; The accuracy rate corresponding to the machine learning model is calculated according to the confusion matrix, and the accuracy rate is used as the objective function. The calculation formula of the accuracy rate is as follows: Where ACC is the accuracy; TP is the true positive; TN is the true negative; FP is the false positive; FN is the false negative. The target space of the Bayesian optimization algorithm is set, and uniform distribution is selected as the prior distribution of the Bayesian optimization algorithm to maximize the accuracy and determine the optimal parameters of the machine learning model.

9. A method for classifying sunny and rainy seasons by combining microwave signals at multiple time scales according to claim 1 or 6, characterized in that: The method of determining the rainfall period model discrimination results of the machine learning models of different time scales based on the microwave attenuation signal data, and determining the final rainfall period discrimination result according to the rainfall period model discrimination results by using a weight factor scoring method, specifically includes: Input the microwave attenuation signal data into the verified support vector machine model, and output the rainfall period model discrimination result at a resolution of 15 minutes, and input the microwave attenuation signal data into the verified category promotion model, and output the rainfall period model discrimination result at a resolution of 1 hour; Set every hour as the statistical period, and the model discrimination result of every 15-minute rainfall period output by the support vector machine model is: , , , The classification improvement model outputs the model discrimination result of each hour of rainfall period as follows: , and the total score f is calculated using the weighted factor scoring method. The corresponding calculation formula is as follows: In the formula, f represents the total score; Indicates weight score; represents the rainfall period model discrimination result, where the rainfall period model discrimination result is the non-rainfall period, The value is 0, and the rainfall period model judgment result is the rainfall period. The value is 1; like , then the final rainfall period determination result is determined to be a rainfall period; otherwise, the final rainfall period determination result is determined to be a non-rainfall period.

10. The method for classifying sunny and rainy seasons by combining microwave signals at multiple time scales according to claim 9 is characterized in that: The weighted , , , , The values ​​are 0.1, 0.15, 0.2, 0.25, and 0.3 respectively.

Images