Time-space intelligent area rainfall qualitative forecasting method based on Beidou water vapor

Abstract

The invention discloses a time-space intelligent area rainfall qualitative forecasting method based on Beidou water vapor. The method comprises the following steps: 1, acquiring atmospheric precipitable water and meteorological historical data inverted by Beidou; 2, constructing and training a BPNN rainfall model; 3, obtaining a rainfall-free threshold optimization value based on the trained BPNN rainfall model; 4, obtaining a rainfall grade threshold optimization value based on the trained BPNN rainfall model; and 5, qualitatively forecasting subsequent rainfall and rainfall grades. The method is simple in steps and reasonable in design, firstly, the BPNN rainfall model is trained through the atmosphere precipitable water amount and meteorological historical data inverted by Beidou, the rainfall-free threshold value optimization value and the rainfall grade threshold value optimization value are obtained through the trained BPNN rainfall model, intelligentization of subsequent area rainfall qualitative forecasting is achieved, rainfall qualitative forecasting in future hours is achieved, and the method is suitable for large-scale popularization and application. And important data support is provided for water conservancy disaster monitoring and water resource management.

Description

Technical Field

[0001] The invention belongs to the technical field of qualitative rainfall forecasting, and in particular relates to a spatial-temporal intelligent surface rainfall qualitative forecasting method based on Beidou water vapor. Background Art

[0002] The water vapor detection technology of the BeiDou navigation satellite system is not affected by weather and has the advantages of high precision, low cost, and continuous operation. It can provide high-precision, high-resolution multi-dimensional water vapor information. Therefore, the forecast of atmospheric precipitable water volume PWV based on BeiDou inversion came into being.

[0003] Existing rainfall forecasting methods using PWV mainly include two types: traditional statistical methods based on the least squares fitting principle and neural network methods. PWV short-term rainfall forecast based on traditional statistical methods is mainly based on long-term historical data. It fits long-term rainfall, PWV and meteorological factors through the least squares principle to achieve short-term rainfall forecast and warning. The nonlinear relationship between rainfall, PWV and meteorological data obtained based on the neural network method can also achieve accurate prediction and warning of short-term rainfall, but neither method can achieve the acquisition of rainfall level threshold.

[0004] Short-term rainfall often constitutes a part of heavy rain, which can easily cause rainstorm and flood disasters. However, the standards for short-term rainfall vary from region to region due to different climatic characteristics. Therefore, it is necessary to formulate scientific and reasonable standards for short-term imminent rainfall levels to make the rainfall, that is, surface rainfall level in the study area, more accurate.

[0005] Therefore, a reasonably designed spatiotemporal intelligent surface rainfall qualitative forecasting method based on Beidou water vapor is needed. First, the atmospheric precipitable water inverted by Beidou and historical meteorological data are used to train the BPNN rainfall model, and the trained BPNN rainfall model is used to obtain the optimized value of the no-rainfall threshold and the optimized value of the rainfall level threshold, so as to realize the intelligence of the subsequent surface rainfall qualitative forecast and the qualitative forecast of future hourly rainfall, providing important data support for water disaster monitoring and water resources management. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a spatiotemporal intelligent surface rainfall qualitative forecasting method based on Beidou water vapor in view of the deficiencies in the above-mentioned prior art. The method has simple steps and reasonable design. The BPNN rainfall model is first trained using the atmospheric precipitable water inverted by Beidou and historical meteorological data, and the trained BPNN rainfall model is used to obtain the optimized value of the no-rainfall threshold and the optimized value of the rainfall level threshold, so as to realize the intelligence of the subsequent surface rainfall qualitative forecast and the qualitative forecast of future hourly rainfall, thereby providing important data support for water disaster monitoring and water resources management.

[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is: a spatial-temporal intelligent surface rainfall qualitative forecasting method based on Beidou water vapor, characterized in that the method comprises the following steps:

[0008] Step 1: Obtain atmospheric precipitable water and historical meteorological data inverted by Beidou:

[0009] Obtain the s-hour mean atmospheric precipitable water, s-hour mean temperature, s-hour mean air pressure, s-hour mean relative humidity and s-hour cumulative rainfall corresponding to each annual accumulation time in the area to be studied to form a training set; where s is a positive integer;

[0010] Step 2: Construction and training of BPNN rainfall model:

[0011] Step 201, constructing N BPNN rainfall models; wherein N is a positive integer;

[0012] Step 202: Take the average atmospheric precipitable water of an annual accumulation time and the corresponding s-hour average temperature, s-hour average air pressure, s-hour average relative humidity and s-hour cumulative rainfall value in the training set as input, and the s-hour cumulative rainfall value of the subsequent annual accumulation time as output, and input them into the BPNN rainfall model for training until N trained BPNN rainfall models are obtained;

[0013] Step 3: Obtain the optimized value of the no-rainfall threshold based on the trained BPNN rainfall model:

[0014] Step 301, input an annual accumulation time in the test set and the corresponding s-hour atmospheric precipitable water mean, s-hour temperature mean, s-hour air pressure mean, s-hour relative humidity mean and s-hour rainfall accumulation value into the first trained BPNN rainfall model, and output the predicted value of the s-hour rainfall accumulation value in the subsequent annual accumulation time;

[0015] Step 302, obtaining the true value of the accumulated rainfall corresponding to the subsequent annual accumulation time in step 301 from the test set as the true value of the accumulated rainfall for s hours in the subsequent annual accumulation time;

[0016] Step 303: Set the no-rainfall threshold value in the range of 0.1 mm to 4.0 mm, and compare the predicted value of the cumulative value of rainfall in s hours in the subsequent year and the true value of the cumulative value of rainfall in s hours in the subsequent year with the no-rainfall threshold value in accordance with a step length of 0.1 mm, and obtain the average value of the short-term rainfall forecast score corresponding to each no-rainfall threshold value under N trained BPNN rainfall models;

[0017] Step 304, selecting the no-rainfall threshold corresponding to the maximum short-term rainfall forecast score mean value from step 303 as the no-rainfall threshold optimization value for the area to be studied;

[0018] Step 4: Obtain the optimized value of rainfall level threshold based on the trained BPNN rainfall model:

[0019] Step 401: The optimized value of the no-rainfall threshold is used as the lower limit threshold of light rain, and the upper limit threshold of light rain is between the optimized value of the no-rainfall threshold + 0.1 mm and H E mm interval, with a step length of 0.1 mm, the predicted value of the cumulative rainfall in the next s hours of the annual accumulation and the true value of the cumulative rainfall in the next s hours of the annual accumulation are respectively compared with the upper limit threshold of light rain to judge the rainfall level, and the average score of the short-term forecast rainfall level corresponding to each upper limit threshold of light rain under N trained BPNN rainfall models is obtained; where H E Set a value for the upper threshold of light rain;

[0020] Step 402: Based on the mean values ​​of the short-term forecast rainfall grade scores corresponding to the light rain upper limit thresholds, a slope change point analysis method is used to obtain an optimized value of the light rain upper limit threshold for the area to be studied;

[0021] Step 403: The optimized value of the upper threshold of light rain is used as the lower threshold of moderate rain, and the upper threshold of moderate rain is within the range of the optimized value of the upper threshold of light rain + 0.1 mm to H′ E In the interval, according to the method from step 401 to step 402, the optimized value of the upper threshold of moderate rain in the area to be studied is obtained; wherein, H′ E Set a value for the upper moderate rain threshold;

[0022] Step 5: Qualitative forecast of subsequent rainfall and rainfall level:

[0023] Step 501: input the s-hour mean atmospheric precipitable water, s-hour mean temperature, s-hour mean air pressure, s-hour mean relative humidity and s-hour cumulative rainfall corresponding to the current annual accumulation time into N trained BPNN rainfall models, and output the predicted values ​​of the cumulative rainfall corresponding to N subsequent annual accumulation times;

[0024] Step 502: average the predicted values ​​of the cumulative rainfall corresponding to N subsequent years to obtain the average of the predicted values ​​of the cumulative rainfall during the subsequent years, and record the average of the predicted values ​​of the cumulative rainfall during the subsequent years as the surface rainfall in the area to be studied;

[0025] Step 503: Compare the surface rainfall in the area to be studied with the optimized value of the no-rainfall threshold, (the optimized value of the lower threshold of light rain to the upper threshold of light rain), and (the optimized value of the lower threshold of moderate rain to the upper threshold of moderate rain) to obtain a qualitative forecast of subsequent rainfall and rainfall level.

[0026] The above-mentioned spatial-temporal intelligent surface rainfall qualitative forecasting method based on Beidou water vapor is characterized by: step one, the specific process is as follows:

[0027] Step 101, using a computer to obtain the atmospheric precipitable water PWV at each measurement time of the Beidou inversion in the area to be studied according to the formula PWV=Π·ZWD; wherein ZWD represents the zenith wet delay, and Π represents the atmospheric water vapor conversion coefficient;

[0028] The atmospheric precipitable water PWV inverted by Beidou at each measurement time in the study area within one hour is averaged to obtain the 1-hour atmospheric precipitable water mean;

[0029] Step 102: Use a temperature and humidity sensor to monitor the area to be studied, obtain the temperature, air pressure, and relative humidity of the area to be studied at each sampling time within one hour, and perform mean processing on the temperature, air pressure, and relative humidity of the area to be studied at each sampling time within one hour to obtain a 1-hour temperature mean, a 1-hour air pressure mean, and a 1-hour relative humidity mean;

[0030] Step 103: Use a rainfall sensor to monitor the area to be studied, obtain the rainfall at each sampling time in the area to be studied within one hour, and accumulate and sum the rainfall at each sampling time in the area to be studied within one hour to obtain a 1-hour rainfall cumulative value;

[0031] Step 104, repeating steps 101 to 103 multiple times, obtaining the average value of atmospheric precipitable water for each hour, the average value of temperature for each hour, the average value of air pressure for each hour, the average value of relative humidity for each hour, and the cumulative value of rainfall for each hour;

[0032] Step 105: According to Get the annual cumulative time T corresponding to the jth time point j ; Where t(j) represents the cumulative day corresponding to the date at the jth time point; j is a natural number, and the value of j ranges from 0 to 23;

[0033] according to Get the annual cumulative time T corresponding to the j-(s-1)th time point j-(s-1) ;

[0034] The annual accumulation time T j-(s-1) To T j The corresponding 1-hour temperature averages are then processed to obtain the annual cumulative temperature T j The corresponding s hourly mean temperature; the annual accumulation time T j-(s-1) To T j The corresponding 1-hour pressure averages are then averaged to obtain the annual cumulative time T j The corresponding s-hour average pressure; the annual accumulation time Tj-(s-1) To T j The corresponding 1-hour relative humidity averages are then averaged to obtain the annual cumulative time T j The corresponding s hour relative humidity average; the annual accumulation time T j-(s-1) To T j The corresponding 1-hour rainfall accumulation values ​​are then accumulated and summed to obtain the annual accumulation time T j The corresponding s-hour rainfall accumulation value; the annual accumulation time T j-(s-1) To T j The corresponding hourly atmospheric precipitation water averages are then processed to obtain the annual cumulative time T j The corresponding s hourly average of atmospheric precipitable water constitutes the historical data set.

[0035] The above-mentioned spatial-temporal intelligent surface rainfall qualitative forecasting method based on Beidou water vapor is characterized in that: the acquisition of the zenith wet delay ZWD and the atmospheric water vapor conversion coefficient Π in step 101 is specifically carried out as follows:

[0036] Step A, using GAMIT / GLOBK software to solve the Beidou station data by computer to obtain the zenith total delay ZTD;

[0037] Step B: Using the Saastamoinen model using a computer The zenith dry delay ZHD is obtained; where P represents the surface pressure at the Beidou station location, represents the latitude of the Beidou station, and H represents the geodetic height of the Beidou station;

[0038] Step C, using a computer to obtain the zenith wet delay ZWD according to the formula ZWD=ZTD-ZHD;

[0039] Step D: Use a computer to calculate the The atmospheric water vapor conversion coefficient Π is obtained; where ρ w represents the density of liquid water, R v is the water vapor gas constant, T m represents the weighted average temperature of the atmosphere, k2′ represents the first constant of the atmospheric refractive index, and k3 represents the second constant of the atmospheric refractive index.

[0040] The above-mentioned spatial-temporal intelligent surface rainfall qualitative forecasting method based on Beidou water vapor is characterized in that: step 303, the specific process is as follows:

[0041] Step 3031, the rain state corresponding to the predicted value of the accumulated rainfall value for s hours in the subsequent year in step 301 is recorded as y; the rain state corresponding to the true value of the accumulated rainfall value for s hours in the subsequent year in step 302 is recorded as x;

[0042] The first no-rainfall threshold is set to l1 and is 0.1 mm. If the predicted value of the cumulative rainfall in the next s hours of the year is greater than or equal to l1, then y = 1, indicating rain; otherwise y = 0, indicating no rain; if the true value of the cumulative rainfall in the next s hours of the year is greater than or equal to l1, then x = 1, indicating rain; otherwise x = 0, indicating no rain;

[0043] Step 3032: If x=0 and y=0, the number of correct rainless forecasts Num TN Add 1; if x=0 and y=1, then the number of empty reports Num FP Add 1; if x=1 and y=0, then the number of missed reports Num FN Add 1; if x=1 and y=1, then the number of correct rainfall forecasts is Num TP Add 1; where Num TN 、Num FP 、Num FN 、Num TP The initial value of is zero;

[0044] Step 3033: Follow the method from step 3031 to step 3032 until the data in the test set is processed and the final correct number of rainless forecasts Num is obtained. TN 、Number of empty reports FP 、Num number of missed reports FN 、Num number of correct rainfall forecasts TP ;

[0045] Step 3034: According to Obtain the short-term rainfall forecast score TS(1) corresponding to the first trained BPNN rainfall model under the first no-rainfall threshold l1;

[0046] Step 3035, repeating steps 3031 to 3034 multiple times, inputting the test set into the nth trained BPNN rainfall model, and obtaining the short-term rainfall forecast score TS(n) corresponding to the nth trained BPNN rainfall model under the first no-rainfall threshold l1;

[0047] Step 3036: According to the formula Get the mean score of the short-term rainfall forecast under the first no-rainfall threshold l1 Wherein, n is a positive integer, and 1≤n≤N;

[0048] Step 3037: Set the no-rainfall threshold of the i-th node to l i , and l i =0.1+(i-1)×0.1, according to the method from step 3031 to step 3036, the i-th no-rainfall threshold l is obtained. iMean value of short-term and imminent rainfall forecast scores under where i is a positive integer;

[0049] Step 3038: Repeat step 3037 multiple times, iterate the non-rainfall threshold by a step size of 0.1 mm until the non-rainfall threshold l of the I-th one I is 4.0 mm, then obtain I mean values of short-term and imminent rainfall forecast scores; where I are all positive integers, and 1 ≤ i ≤ I;

[0050] Step 304 is specifically as follows:

[0051] Select the non-rainfall threshold corresponding to the maximum mean value of short-term and imminent rainfall forecast scores from the I mean values of short-term and imminent rainfall forecast scores and denote it as the optimized non-rainfall threshold value.

[0052] The above-mentioned spatio-temporal intelligent areal rainfall qualitative forecasting method based on Beidou water vapor is characterized in that: Step 401, the specific process is as follows:

[0053] Step 4011: Set the upper limit threshold H1 of the first light rain as the optimized non-rainfall threshold value + 0.1 mm, compare the predicted value of the cumulative s-hour rainfall in the subsequent annual accumulation time in step 301 with the lower limit threshold H1 of the first light rain, and obtain the rainfall level corresponding to the predicted value of the cumulative s-hour rainfall in the subsequent annual accumulation time and denote it as yd;

[0054] Compare the true value of the cumulative rainfall in the subsequent annual accumulation time in step 302 with the lower limit threshold H1 of the first light rain, and obtain the rainfall level corresponding to the true value of the cumulative rainfall in the subsequent annual accumulation time and denote it as xd; where the rainfall levels are 1, 2, 3, 1 is light rain, 2 is moderate rain, and 3 is heavy rain;

[0055] Step 4012: If xd < yd, then the number of times Num OT of overestimated rainfall levels is incremented by 1; if xd > yd, then the number of times Num UT of underestimated rainfall levels is incremented by 1; if xd = yd, then the number of times Num TT of correctly predicted rainfall levels is incremented by 1; where Num OT 、Num UT 、Num TT all have an initial value of zero;

[0056] Step 4013: According to the method of step 4011 to step 4012, until the data in the test set is processed, obtain the final number of times Num OT of overestimated rainfall levels, the number of times Num UT of underestimated rainfall levels, and the number of times Num TT of correctly predicted rainfall levels;

[0057] Step 4014: Obtain the short-term forecast rainfall grade score TS′(1) corresponding to the first trained BPNN rainfall model under the first light rain upper limit threshold H1;

[0058] Step 4015, repeating steps 4011 to 4014 multiple times, inputting the test set into the nth trained BPNN rainfall model, and obtaining the short-term forecast rainfall grade score TS′(n) corresponding to the nth trained BPNN rainfall model under the first light rain upper limit threshold H1;

[0059] Step 4016: According to the formula Get the mean score of the short-term forecast level of light rain under the first light rain upper limit threshold H1

[0060] Step 4017: Set the upper limit threshold of the e-th light rain to H e , and H e =H1+(e-1)×0.1, according to the method of step 4011 to step 4016, the upper limit threshold value H of the e-th light rain is obtained. e The average score of the short-term forecast level light rain under

[0061] Step 4018: Repeat step 4017 multiple times, and iterate the upper limit threshold of light rain with a step length of 0.1 mm until the upper limit threshold of light rain of the Eth time is H. E , then we get E short-term forecast level light rain score means; where e and E are both positive integers, and 1≤e≤E.

[0062] The above-mentioned spatial-temporal intelligent surface rainfall qualitative forecasting method based on Beidou water vapor is characterized in that: step 402, the specific process is as follows:

[0063] Step 4021: Using a computer, take E light rain upper limit thresholds as abscissas, take E short-term forecast level light rain score means corresponding to the E light rain upper limit thresholds as ordinates, and draw a scatter plot of the short-term forecast level light rain score means;

[0064] Step 4022: Using a computer, record the centers of two adjacent light rain upper limit thresholds as exploration points on the horizontal axis to obtain E exploration points; among them, the e-th exploration point x e The upper limit threshold H of the light rain at the e-1th e-1 and the e-th light rain upper limit threshold H e the center of

[0065] Step 4023: Use a computer to obtain the e-th exploration point x e Perform linear regression on the first k light rain upper thresholds to obtain the front slope of the front linear regression equation Use a computer to take the e-th exploration point x e The k light rain upper limit thresholds are linearly regressed to obtain the posterior slope of the posterior linear regression equation. k is a positive integer, and the value of k is 2, 3 or 4; e is a positive integer greater than 4;

[0066] Step 4024: When k is 2, 3 or 4, the computer is used to calculate Get the e-th exploration point x e Weighted average slope when sliding forward window size is k and the e-th exploration point x e Weighted average slope when sliding backward with window size k

[0067] Step 4025: According to the method of step 4024, when e is greater than k+1, the e-1th exploration point x is obtained. e-1 Weighted average slope when sliding forward window size is k and the e-1th exploration point x e-1 Weighted average slope when sliding backward with window size k

[0068] Step 4026: Using a computer according to Get the e-th exploration point x e Second order difference value Δ 2 S(x e );

[0069] Step 4027, repeating steps 4021 to 4026 multiple times until the value of e is Ek, completing the second-order difference value before and after the Ek-th exploration point, and obtaining the second-order difference value before and after each exploration point;

[0070] Step 4028: Obtain the maximum second-order difference value of the exploration point from the second-order difference values ​​of each exploration point, and record the two exploration points corresponding to the maximum second-order difference value of the exploration point as the previous exploration point x e′-1 and the later exploration point x e′ ;

[0071] Step 4029: Using a computer according to Get the optimized value x of the upper threshold of light rain * ; Among them, Δ 2 S(x e′ ) represents the later explored point x e′ The corresponding second-order difference value before and after, Δ 2 S(x e′-1 ) represents the previous exploration point x e′-1 The corresponding second-order difference value before and after, Δ 2S(x e′+1 ) represents the later explored point x e′ A detection point x behind e′+1 The corresponding second-order difference values ​​before and after.

[0072] Compared with the prior art, the present invention has the following advantages:

[0073] 1. The method of the present invention has simple steps and reasonable design, which can solve the problem of current surface rainfall qualitative forecast and ensure the accuracy and reliability of rainfall forecast results.

[0074] 2. The present invention uses the atmospheric precipitable water inverted by Beidou and meteorological historical data to train the constructed N BPNN rainfall models respectively, and obtains N trained BPNN rainfall models; on the one hand, considering the randomness of the BPNN model grid initialization, therefore, N BPNN rainfall models are established for random training to obtain the predicted value of the cumulative rainfall value, thereby improving the accuracy of the subsequent no-rainfall threshold optimization value and the rainfall level threshold optimization value; on the other hand, the rainfall forecast results obtained based on the N trained BPNN rainfall models are aggregated to improve the accuracy of future rainfall forecasts.

[0075] 3. When the present invention obtains the optimized value of the no-rainfall threshold based on the trained BPNN rainfall model, the no-rainfall threshold is set in the range of 0.1mm to 4mm, and the mean short-term rainfall forecast score corresponding to each no-rainfall threshold under N trained BPNN rainfall models is obtained according to a step size of 0.1mm; then, the no-rainfall threshold corresponding to the largest mean short-term rainfall forecast score is selected as the optimized value of the no-rainfall threshold for the area to be studied.

[0076] 4. When the present invention obtains the optimized value of the rainfall level threshold based on the trained BPNN rainfall model, the slope change point analysis method is adopted to obtain the optimized value of the upper threshold of light rain and the upper threshold of moderate rain in the area to be studied, so as to effectively adapt to the changing trend of rainfall.

[0077] In summary, the method of the present invention has simple steps and reasonable design. First, the atmospheric precipitable water inverted by Beidou and historical meteorological data are used to train the BPNN rainfall model, and the trained BPNN rainfall model is used to obtain the optimized value of the no-rainfall threshold and the optimized value of the rainfall level threshold, so as to realize the intelligentization of the subsequent qualitative forecast of surface rainfall and the qualitative forecast of future hourly rainfall, providing important data support for water disaster monitoring and water resources management.

[0078] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 The figure is a flowchart of the method of the present invention. DETAILED DESCRIPTION

[0080] like Figure 1 As shown, the spatial-temporal intelligent surface rainfall qualitative forecasting method based on Beidou water vapor of the present invention comprises the following steps:

[0081] Step 1: Obtain atmospheric precipitable water and historical meteorological data inverted by Beidou:

[0082] Obtain the s-hour mean atmospheric precipitable water, s-hour mean temperature, s-hour mean air pressure, s-hour mean relative humidity and s-hour cumulative rainfall corresponding to each annual accumulation time in the area to be studied to form a training set; where s is a positive integer;

[0083] Step 2: Construction and training of BPNN rainfall model:

[0084] Step 201: construct N BPNN rainfall models; wherein each of the N BPNN rainfall models includes an input layer, a hidden layer and an output layer, and the number of neurons in the input layer is N. in , and N in =6, the number of neurons in the hidden layer is N hid , and 2×N in +1, the number of neurons in the output layer is 1; N is a positive integer;

[0085] Step 202: Take the average atmospheric precipitable water of an annual accumulation time and the corresponding s-hour average temperature, s-hour average air pressure, s-hour average relative humidity and s-hour cumulative rainfall value in the training set as input, and the s-hour cumulative rainfall value of the subsequent annual accumulation time as output, and input them into the BPNN rainfall model for training until N trained BPNN rainfall models are obtained;

[0086] Step 3: Obtain the optimized value of the no-rainfall threshold based on the trained BPNN rainfall model:

[0087] Step 301, input an annual accumulation time in the test set and the corresponding s-hour atmospheric precipitable water mean, s-hour temperature mean, s-hour air pressure mean, s-hour relative humidity mean and s-hour rainfall accumulation value into the first trained BPNN rainfall model, and output the predicted value of the s-hour rainfall accumulation value in the subsequent annual accumulation time;

[0088] Step 302, obtaining the true value of the accumulated rainfall corresponding to the subsequent annual accumulation time in step 301 from the test set as the true value of the accumulated rainfall for s hours in the subsequent annual accumulation time;

[0089] Step 303: Set the no-rainfall threshold value in the range of 0.1 mm to 4.0 mm, and compare the predicted value of the cumulative value of rainfall in s hours in the subsequent year and the true value of the cumulative value of rainfall in s hours in the subsequent year with the no-rainfall threshold value in accordance with a step length of 0.1 mm, and obtain the average value of the short-term rainfall forecast score corresponding to each no-rainfall threshold value under N trained BPNN rainfall models;

[0090] Step 304, selecting the no-rainfall threshold corresponding to the maximum short-term rainfall forecast score mean value from step 303 as the no-rainfall threshold optimization value for the area to be studied;

[0091] Step 4: Obtain the optimized value of rainfall level threshold based on the trained BPNN rainfall model:

[0092] Step 401: The optimized value of the no-rainfall threshold is used as the lower limit threshold of light rain, and the upper limit threshold of light rain is between the optimized value of the no-rainfall threshold + 0.1 mm and H E mm interval, with a step length of 0.1 mm, the predicted value of the cumulative rainfall in the next s hours of the annual accumulation and the true value of the cumulative rainfall in the next s hours of the annual accumulation are respectively compared with the upper limit threshold of light rain to judge the rainfall level, and the average score of the short-term forecast rainfall level corresponding to each upper limit threshold of light rain under N trained BPNN rainfall models is obtained; where H E Set a value for the upper threshold of light rain;

[0093] Step 402: Based on the mean values ​​of the short-term forecast rainfall grade scores corresponding to the light rain upper limit thresholds, a slope change point analysis method is used to obtain an optimized value of the light rain upper limit threshold for the area to be studied;

[0094] Step 403: The optimized value of the upper threshold of light rain is used as the lower threshold of moderate rain, and the upper threshold of moderate rain is within the range of the optimized value of the upper threshold of light rain + 0.1 mm to H′ E In the interval, according to the method from step 401 to step 402, the optimized value of the upper threshold of moderate rain in the area to be studied is obtained; wherein, H′ E Set a value for the upper moderate rain threshold;

[0095] Step 5: Qualitative forecast of subsequent rainfall and rainfall level:

[0096] Step 501: input the s-hour mean atmospheric precipitable water, s-hour mean temperature, s-hour mean air pressure, s-hour mean relative humidity and s-hour cumulative rainfall corresponding to the current annual accumulation time into N trained BPNN rainfall models, and output the predicted values ​​of the cumulative rainfall corresponding to N subsequent annual accumulation times;

[0097] Step 502: average the predicted values ​​of the cumulative rainfall corresponding to N subsequent years to obtain the average of the predicted values ​​of the cumulative rainfall during the subsequent years, and record the average of the predicted values ​​of the cumulative rainfall during the subsequent years as the surface rainfall in the area to be studied;

[0098] Step 503: Compare the surface rainfall in the area to be studied with the optimized value of the no-rainfall threshold, (the optimized value of the lower threshold of light rain to the upper threshold of light rain), and (the optimized value of the lower threshold of moderate rain to the upper threshold of moderate rain) to obtain a qualitative forecast of subsequent rainfall and rainfall level.

[0099] In this embodiment, step one, the specific process is as follows:

[0100] Step 101, using a computer to obtain the atmospheric precipitable water PWV at each measurement time of the Beidou inversion in the area to be studied according to the formula PWV=Π·ZWD; wherein ZWD represents the zenith wet delay, and Π represents the atmospheric water vapor conversion coefficient;

[0101] The atmospheric precipitable water PWV inverted by Beidou at each measurement time in the study area within one hour is averaged to obtain the 1-hour atmospheric precipitable water mean;

[0102] Step 102: Use a temperature and humidity sensor to monitor the area to be studied, obtain the temperature, air pressure, and relative humidity of the area to be studied at each sampling time within one hour, and perform mean processing on the temperature, air pressure, and relative humidity of the area to be studied at each sampling time within one hour to obtain a 1-hour temperature mean, a 1-hour air pressure mean, and a 1-hour relative humidity mean;

[0103] Step 103: Use a rainfall sensor to monitor the area to be studied, obtain the rainfall at each sampling time in the area to be studied within one hour, and accumulate and sum the rainfall at each sampling time in the area to be studied within one hour to obtain a 1-hour rainfall cumulative value;

[0104] Step 104, repeating steps 101 to 103 multiple times, obtaining the average value of atmospheric precipitable water for each hour, the average value of temperature for each hour, the average value of air pressure for each hour, the average value of relative humidity for each hour, and the cumulative value of rainfall for each hour;

[0105] Step 105: According to Get the annual cumulative time T corresponding to the jth time point j ; Where t(j) represents the cumulative day corresponding to the date at the jth time point; j is a natural number, and the value of j ranges from 0 to 23;

[0106] according to Get the annual cumulative time T corresponding to the j-(s-1)th time point j-(s-1) ;

[0107] The annual accumulation time T j-(s-1) To T j The corresponding 1-hour temperature averages are then processed to obtain the annual cumulative temperature T j The corresponding s hourly mean temperature; the annual accumulation time T j-(s-1) To T j The corresponding 1-hour pressure averages are then averaged to obtain the annual cumulative time T j The corresponding s-hour average pressure; the annual accumulation time T j-(s-1) To T j The corresponding 1-hour relative humidity averages are then averaged to obtain the annual cumulative time T j The corresponding s hour relative humidity average; the annual accumulation time T j-(s-1) To T j The corresponding 1-hour rainfall accumulation values ​​are then accumulated and summed to obtain the annual accumulation time T j The corresponding s-hour rainfall accumulation value; the annual accumulation time T j-(s-1) To T j The corresponding hourly atmospheric precipitation water averages are then processed to obtain the annual cumulative time T j The corresponding s hourly average of atmospheric precipitable water constitutes the historical data set.

[0108] In this embodiment, the specific process of obtaining the zenith wet delay ZWD and the atmospheric water vapor conversion coefficient Π in step 101 is as follows:

[0109] Step A, using GAMIT / GLOBK software to solve the Beidou station data by computer to obtain the zenith total delay ZTD;

[0110] Step B: Using the Saastamoinen model using a computer The zenith dry delay ZHD is obtained; where P represents the surface pressure at the Beidou station location, represents the latitude of the Beidou station, and H represents the geodetic height of the Beidou station;

[0111] Step C, using a computer to obtain the zenith wet delay ZWD according to the formula ZWD=ZTD-ZHD;

[0112] Step D: Use a computer to calculate the The atmospheric water vapor conversion coefficient Π is obtained; where ρ w represents the density of liquid water, R v is the water vapor gas constant, T m represents the weighted average temperature of the atmosphere, k2′ represents the first constant of the atmospheric refractive index, and k3 represents the second constant of the atmospheric refractive index.

[0113] In this embodiment, step 303, the specific process is as follows:

[0114] Step 3031, the rain state corresponding to the predicted value of the accumulated rainfall value for s hours in the subsequent year in step 301 is recorded as y; the rain state corresponding to the true value of the accumulated rainfall value for s hours in the subsequent year in step 302 is recorded as x;

[0115] The first no-rainfall threshold is set to l1 and is 0.1 mm. If the predicted value of the cumulative rainfall in the next s hours of the year is greater than or equal to l1, then y = 1, indicating rain; otherwise y = 0, indicating no rain; if the true value of the cumulative rainfall in the next s hours of the year is greater than or equal to l1, then x = 1, indicating rain; otherwise x = 0, indicating no rain;

[0116] Step 3032: If x=0 and y=0, the number of correct rainless forecasts Num TN Add 1; if x=0 and y=1, then the number of empty reports Num FP Add 1; if x=1 and y=0, then the number of missed reports Num FN Add 1; if x=1 and y=1, then the number of correct rainfall forecasts is Num TP Add 1; where Num TN 、Num FP 、Num FN 、Num TP The initial value of is zero;

[0117] Step 3033: Follow the method from step 3031 to step 3032 until the data in the test set is processed and the final correct number of rainless forecasts Num is obtained. TN 、Number of empty reports FP 、Num number of missed reports FN 、Num number of correct rainfall forecasts TP ;

[0118] Step 3034: According to Obtain the short-term rainfall forecast score TS(1) corresponding to the first trained BPNN rainfall model under the first no-rainfall threshold l1;

[0119] Step 3035, repeating steps 3031 to 3034 multiple times, inputting the test set into the nth trained BPNN rainfall model, and obtaining the short-term rainfall forecast score TS(n) corresponding to the nth trained BPNN rainfall model under the first no-rainfall threshold l1;

[0120] Step 3036: According to the formula The first short-term rainfall forecast score mean TS1 under the no rainfall threshold l1 is obtained; wherein n is a positive integer and 1≤n≤N;

[0121] Step 3037: Set the non - rainfall threshold for the \(i\) - th one as \(l\) i , and \(l\) i = 0.1+(i - 1)×0.1. According to the method from Step 3031 to Step 3036, obtain the average value \(TS\) of the short - term rainfall forecast score under the non - rainfall threshold \(l\) i for the \(i\) - th one; where \(i\) is a positive integer; i

[0122] Step 3038: Repeat Step 3037 multiple times, iterate the non - rainfall threshold with a step size of 0.1 mm until the non - rainfall threshold \(l\) I for the \(I\) - th one is 4.0 mm, then obtain the average values of the short - term rainfall forecast scores for \(I\) ones; where \(I\) is a positive integer, and \(1\leq i\leq I\);

[0123] Step 304 is specifically as follows:

[0124] Select the non - rainfall threshold corresponding to the maximum average value of the short - term rainfall forecast scores from the \(I\) average values of the short - term rainfall forecast scores and denote it as the optimized non - rainfall threshold value.

[0125] In this embodiment, Step 401 is specifically as follows:

[0126] Step 4011: Set the upper threshold \(H1\) for light rain of the first one as the optimized non - rainfall threshold value + 0.1 mm. Compare the predicted value of the cumulative rainfall of \(s\) hours in the subsequent annual accumulation time in Step 301 with the lower threshold \(H1\) for light rain of the first one, and obtain the rainfall level corresponding to the predicted value of the cumulative rainfall of \(s\) hours in the subsequent annual accumulation time and denote it as \(yd\);

[0127] Compare the true value of the cumulative rainfall in the subsequent annual accumulation time in Step 302 with the lower threshold \(H1\) for light rain of the first one, and obtain the rainfall level corresponding to the true value of the cumulative rainfall in the subsequent annual accumulation time and denote it as \(xd\); where the rainfall levels are 1, 2, 3, 1 is light rain, 2 is moderate rain, and 3 is heavy rain;

[0128] Step 4012: If \(xd\lt yd\), then the number of times \(Num\) OT of over - estimating the rainfall level is incremented by 1; if \(xd\gt yd\), then the number of times \(Num\) UT of under - estimating the rainfall level is incremented by 1; if \(xd = yd\), then the number of times \(Num\) TT of correctly estimating the rainfall level is incremented by 1; where the initial values of \(Num\) OT , \(Num\) UT , and \(Num\) TT are all zero;

[0129] Step 4013: According to the method from Step 4011 to Step 4012, until the data in the test set is processed completely, obtain the final number of times \(Num\) of over - estimating the rainfall levelOT 、Num number of underestimated rainfall levels UT The number of correct rainfall level forecasts TT ;

[0130] Step 4014: Obtain the short-term forecast rainfall grade score TS′(1) corresponding to the first trained BPNN rainfall model under the first light rain upper limit threshold H1;

[0131] Step 4015, repeating steps 4011 to 4014 multiple times, inputting the test set into the nth trained BPNN rainfall model, and obtaining the short-term forecast rainfall grade score TS′(n) corresponding to the nth trained BPNN rainfall model under the first light rain upper limit threshold H1;

[0132] Step 4016: According to the formula Get the mean score of the short-term forecast level of light rain under the first light rain upper limit threshold H1

[0133] Step 4017: Set the upper limit threshold of the e-th light rain to H e , and H e =H1+(e-1)×0.1, according to the method of step 4011 to step 4016, the upper limit threshold value H of the e-th light rain is obtained. e The average score of the short-term forecast level light rain under

[0134] Step 4018: Repeat step 4017 multiple times, and iterate the upper limit threshold of light rain with a step length of 0.1 mm until the upper limit threshold of light rain of the Eth time is H. E , then we get E short-term forecast level light rain score means; where e and E are both positive integers, and 1≤e≤E.

[0135] In this embodiment, step 402, the specific process is as follows:

[0136] Step 4021: Using a computer, take E light rain upper limit thresholds as abscissas, take E short-term forecast level light rain score means corresponding to the E light rain upper limit thresholds as ordinates, and draw a scatter plot of the short-term forecast level light rain score means;

[0137] Step 4022: Using a computer, record the centers of two adjacent light rain upper limit thresholds as exploration points on the horizontal axis to obtain E exploration points; among them, the e-th exploration point x e The upper limit threshold H of the light rain at the e-1th e-1 and the e-th light rain upper limit threshold H e the center of

[0138] Step 4023: Use a computer to obtain the e-th exploration point x e Perform linear regression on the first k light rain upper thresholds to obtain the front slope of the front linear regression equation Use a computer to take the e-th exploration point x e The k light rain upper thresholds are linearly regressed to obtain the posterior slope of the posterior linear regression equation. k is a positive integer, and the value of k is 2, 3 or 4; e is a positive integer greater than 4;

[0139] Step 4024: When k is 2, 3 or 4, the computer is used to calculate Get the e-th exploration point x e Weighted average slope when sliding forward window size is k and the e-th exploration point x e Weighted average slope when sliding backward with window size k

[0140] Step 4025: According to the method of step 4024, when e is greater than k+1, the e-1th exploration point x is obtained. e-1 Weighted average slope when sliding forward window size is k and the e-1th exploration point x e-1 Weighted average slope when sliding backward with window size k

[0141] Step 4026: Using a computer according to Get the e-th exploration point x e Second order difference value Δ 2 S(x e );

[0142] Step 4027, repeating steps 4021 to 4026 multiple times until the value of e is Ek, completing the second-order difference value before and after the Ek-th exploration point, and obtaining the second-order difference value before and after each exploration point;

[0143] Step 4028: Obtain the maximum second-order difference value of the exploration point from the second-order difference values ​​of each exploration point, and record the two exploration points corresponding to the maximum second-order difference value of the exploration point as the previous exploration point x e′-1 and the later exploration point x e′ ;

[0144] Step 4029: Using a computer according to Get the optimized value x of the upper threshold of light rain * ; Among them, Δ 2 S(x e′ ) represents the later explored point x e′ The corresponding second-order difference value before and after, Δ 2S(x e′-1 ) represents the previous exploration point x e′-1 The corresponding second-order difference value before and after, Δ 2 S(x e′+1 ) represents the later explored point x e′ A detection point x behind e′+1 The corresponding second-order difference values ​​before and after.

[0145] In this embodiment, the value of N is 100, taking into account the randomness of the BPNN model grid initialization.

[0146] In this embodiment, when it is implemented, the unit of the surface pressure P corresponding to the Beidou station position in step B is hPa, and the latitude of the Beidou station is The unit of is rad, and the unit of the Beidou station geodetic height H is km.

[0147] In this embodiment, during the specific implementation, the density of liquid water in step D is w , the value is 1.0×10 3 kg / m 3 , R v The value of T is 461.495 J / (kg·K), m The unit is K, and the value of k2′ is 17±10K / hPa, and the value of k3 is (3.776±0.04)×10 5 K 2 / hPa.

[0148] In this embodiment, during the specific implementation, the sampling time of the temperature, air pressure, and relative humidity in step 102 and the sampling time of the cumulative value of rainfall in step 103 are 5 minutes;

[0149] The measurement time for inverting the atmospheric precipitable water in step 101 is 30 seconds, so that the atmospheric precipitable water, temperature, air pressure, relative humidity and rainfall cumulative values ​​are all converted to the same 1-hour time.

[0150] In this embodiment, when s is 1, the upper limit threshold value setting value H for light rain is E is 4mm; when s is 2, the upper threshold value of light rain is set to H E When s is 3, the upper threshold value of light rain is set to H E When s is 12, the upper threshold value of light rain is set to H. E 15mm;

[0151] When s is 1, the upper threshold value of moderate rain is set to H′ E When s is 2, the upper threshold value of moderate rain is set to H′ E is 13 mm; when s is 3, the upper threshold value of moderate rain is set to H′E When s is 12, the upper threshold value of moderate rain is set to H′ E It is 30mm.

[0152] In this embodiment, by setting s differently, it is possible to adapt to forecasts at different subsequent times and meet different needs.

[0153] In this embodiment, it should be noted that there is no setting range of rainfall threshold, H E and H′ E It can be adjusted according to actual conditions.

[0154] In this embodiment, when obtaining the optimized value of the upper threshold of moderate rain in the area to be studied according to the method of step 401 to step 402 in 403, it should be noted that the predicted value of the cumulative value of rainfall in s hours in the subsequent year and the true value of the cumulative value of rainfall in s hours in the subsequent year need to be respectively compared with (light rain lower threshold ~ light rain upper threshold optimized value) and (moderate rain lower threshold ~ moderate rain upper threshold) to judge the rainfall level.

[0155] In this embodiment, if the subsequent annual cumulative rainfall cumulative value predicted value average is greater than the moderate rain upper limit threshold optimization value in step 503, the rainfall level is 3, that is, heavy rain.

[0156] In this embodiment, the weighted average temperature of the atmosphere T m It can be obtained by referring to conventional methods in the art.

[0157] In summary, the method of the present invention has simple steps and reasonable design. First, the atmospheric precipitable water inverted by Beidou and historical meteorological data are used to train the BPNN rainfall model, and the trained BPNN rainfall model is used to obtain the optimized value of the no-rainfall threshold and the optimized value of the rainfall level threshold, so as to realize the intelligentization of the subsequent qualitative forecast of surface rainfall and the qualitative forecast of future hourly rainfall, providing important data support for water disaster monitoring and water resources management.

[0158] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any way. Any simple modification, change and equivalent structural change made to the above embodiment based on the technical essence of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A spatial-temporal intelligent surface rainfall qualitative forecasting method based on Beidou water vapor, characterized in that: The method comprises the following steps: Step 1: Obtain atmospheric precipitable water and historical meteorological data inverted by Beidou: Obtain the s-hour mean atmospheric precipitable water, s-hour mean temperature, s-hour mean air pressure, s-hour mean relative humidity and s-hour cumulative rainfall corresponding to each annual accumulation time in the area to be studied to form a training set; where s is a positive integer; Step 2: Construction and training of BPNN rainfall model: Step 201, constructing N BPNN rainfall models; wherein N is a positive integer; Step 202: Take the average atmospheric precipitable water of an annual accumulation time and the corresponding s-hour average temperature, s-hour average air pressure, s-hour average relative humidity and s-hour cumulative rainfall value in the training set as input, and the s-hour cumulative rainfall value of the subsequent annual accumulation time as output, and input them into the BPNN rainfall model for training until N trained BPNN rainfall models are obtained; Step 3: Obtain the optimized value of the no-rainfall threshold based on the trained BPNN rainfall model: Step 301, input an annual accumulation time in the test set and the corresponding s-hour atmospheric precipitable water mean, s-hour temperature mean, s-hour air pressure mean, s-hour relative humidity mean and s-hour rainfall accumulation value into the first trained BPNN rainfall model, and output the predicted value of the s-hour rainfall accumulation value in the subsequent annual accumulation time; Step 302, obtaining the true value of the accumulated rainfall corresponding to the subsequent annual accumulation time in step 301 from the test set as the true value of the accumulated rainfall for s hours in the subsequent annual accumulation time; Step 303: Set the no-rainfall threshold value in the range of 0.1 mm to 4.0 mm, and compare the predicted value of the cumulative value of rainfall in s hours in the subsequent year and the true value of the cumulative value of rainfall in s hours in the subsequent year with the no-rainfall threshold value in accordance with a step length of 0.1 mm, and obtain the average value of the short-term rainfall forecast score corresponding to each no-rainfall threshold value under N trained BPNN rainfall models; Step 304, selecting the no-rainfall threshold corresponding to the maximum short-term rainfall forecast score mean value from step 303 as the no-rainfall threshold optimization value for the area to be studied; Step 4: Obtain the optimized value of rainfall level threshold based on the trained BPNN rainfall model: Step 401: The optimized value of the no-rainfall threshold is used as the lower limit threshold of light rain, and the upper limit threshold of light rain is between the optimized value of the no-rainfall threshold + 0.1 mm and H E mm interval, with a step length of 0.1 mm, the predicted value of the cumulative rainfall in the next s hours of the annual accumulation and the true value of the cumulative rainfall in the next s hours of the annual accumulation are respectively compared with the upper limit threshold of light rain to judge the rainfall level, and the average score of the short-term forecast rainfall level corresponding to each upper limit threshold of light rain under N trained BPNN rainfall models is obtained; where H E Set a value for the upper threshold of light rain; Step 402: Based on the mean values ​​of the short-term forecast rainfall grade scores corresponding to the light rain upper limit thresholds, a slope change point analysis method is used to obtain an optimized value of the light rain upper limit threshold for the area to be studied; Step 403: The optimized value of the upper threshold of light rain is used as the lower threshold of moderate rain, and the upper threshold of moderate rain is within the range of the optimized value of the upper threshold of light rain + 0.1 mm to H′ E In the interval, according to the method from step 401 to step 402, the optimized value of the upper threshold of moderate rain in the area to be studied is obtained; wherein, H′ E Set a value for the upper moderate rain threshold; Step 5: Qualitative forecast of subsequent rainfall and rainfall level: Step 501: input the s-hour mean atmospheric precipitable water, s-hour mean temperature, s-hour mean air pressure, s-hour mean relative humidity and s-hour cumulative rainfall corresponding to the current annual accumulation time into N trained BPNN rainfall models, and output the predicted values ​​of the cumulative rainfall corresponding to N subsequent annual accumulation times; Step 502: average the predicted values ​​of the cumulative rainfall corresponding to N subsequent years to obtain the average of the predicted values ​​of the cumulative rainfall during the subsequent years, and record the average of the predicted values ​​of the cumulative rainfall during the subsequent years as the surface rainfall in the area to be studied; Step 503: Compare the surface rainfall in the area to be studied with the optimized value of the no-rainfall threshold, (the optimized value of the lower threshold of light rain to the upper threshold of light rain), and (the optimized value of the lower threshold of moderate rain to the upper threshold of moderate rain) to obtain a qualitative forecast of subsequent rainfall and rainfall level.

2. A method for qualitative forecasting of spatial-temporal intelligent surface rainfall based on Beidou water vapor according to claim 1, characterized in that: Step 1: The specific process is as follows: Step 101, using a computer to obtain the atmospheric precipitable water PWV at each measurement time of the Beidou inversion in the area to be studied according to the formula PWV=Π·ZWD; wherein ZWD represents the zenith wet delay, and Π represents the atmospheric water vapor conversion coefficient; The atmospheric precipitable water PWV inverted by Beidou at each measurement time in the study area within one hour is averaged to obtain the 1-hour atmospheric precipitable water mean; Step 102: Use a temperature and humidity sensor to monitor the area to be studied, obtain the temperature, air pressure, and relative humidity of the area to be studied at each sampling time within one hour, and perform mean processing on the temperature, air pressure, and relative humidity of the area to be studied at each sampling time within one hour to obtain a 1-hour temperature mean, a 1-hour air pressure mean, and a 1-hour relative humidity mean; Step 103: Use a rainfall sensor to monitor the area to be studied, obtain the rainfall at each sampling time in the area to be studied within one hour, and accumulate and sum the rainfall at each sampling time in the area to be studied within one hour to obtain a 1-hour rainfall cumulative value; Step 104, repeating steps 101 to 103 multiple times, obtaining the average value of atmospheric precipitable water for each hour, the average value of temperature for each hour, the average value of air pressure for each hour, the average value of relative humidity for each hour, and the cumulative value of rainfall for each hour; Step 105: According to Get the annual cumulative time T corresponding to the jth time point j ; Where t(j) represents the cumulative day of the year corresponding to the date at the jth time point; j is a natural number, and the value of j ranges from 0 to 23; according to Get the annual cumulative time T corresponding to the j-(s-1)th time point j-(s-1) ; The annual accumulation time T j-(s-1) To T j The corresponding 1-hour temperature averages are then processed to obtain the annual cumulative temperature T j The corresponding s hourly mean temperature; the annual accumulation time T j-(s-1) To T j The corresponding 1-hour pressure averages are then averaged to obtain the annual cumulative time T j The corresponding s-hour average pressure; the annual accumulation time T j-(s-1) To T j The corresponding 1-hour relative humidity averages are then averaged to obtain the annual cumulative time T j The corresponding s hour relative humidity average; the annual accumulation time T j-(s-1) To T j The corresponding 1-hour rainfall accumulation values ​​are then accumulated and summed to obtain the annual accumulation time T j The corresponding s-hour rainfall accumulation value; the annual accumulation time T j-(s-1) To T j The corresponding hourly atmospheric precipitation water averages are then processed to obtain the annual cumulative time T j The corresponding s hourly average of atmospheric precipitable water constitutes the historical data set.

3. A method for qualitative forecasting of surface rainfall based on Beidou water vapor in time and space according to claim 2, characterized in that: The specific process of obtaining the zenith wet delay ZWD and the atmospheric water vapor conversion coefficient Π in step 101 is as follows: Step A, using GAMIT / GLOBK software to solve the Beidou station data by computer to obtain the zenith total delay ZTD; Step B: Using the Saastamoinen model using a computer The zenith dry delay ZHD is obtained; where P represents the surface pressure at the Beidou station location, represents the latitude of the Beidou station, and H represents the geodetic height of the Beidou station; Step C, using a computer to obtain the zenith wet delay ZWD according to the formula ZWD=ZTD-ZHD; Step D: Use a computer to calculate the The atmospheric water vapor conversion coefficient Π is obtained; where ρ w represents the density of liquid water, R v is the water vapor gas constant, T m represents the weighted average temperature of the atmosphere, k2′ represents the first constant of the atmospheric refractive index, and k3 represents the second constant of the atmospheric refractive index.

4. A method for qualitative forecasting of surface rainfall based on Beidou water vapor in time and space according to claim 1, characterized in that: Step 303, the specific process is as follows: Step 3031, the rain state corresponding to the predicted value of the accumulated rainfall value for s hours in the subsequent year in step 301 is recorded as y; the rain state corresponding to the true value of the accumulated rainfall value for s hours in the subsequent year in step 302 is recorded as x; The first no-rainfall threshold is set to l1 and is 0.1 mm. If the predicted value of the cumulative rainfall in the next s hours of the year is greater than or equal to l1, then y = 1, indicating rain; otherwise y = 0, indicating no rain; if the true value of the cumulative rainfall in the next s hours of the year is greater than or equal to l1, then x = 1, indicating rain; otherwise x = 0, indicating no rain; Step 3032: If x=0 and y=0, the number of correct rainless forecasts Num TN Add 1; if x=0 and y=1, then the number of empty reports Num FP Add 1; if x=1 and y=0, then the number of missed reports Num FN Add 1; if x=1 and y=1, then the number of correct rainfall forecasts is Num TP Add 1; where Num TN 、Num FP 、Num FN 、Num TP The initial value of is zero; Step 3033: Follow the method from step 3031 to step 3032 until the data in the test set is processed and the final correct number of rainless forecasts Num is obtained. TN 、Number of empty reports FP 、Num number of missed reports FN 、Num number of correct rainfall forecasts TP ; Step 3034: According to Obtain the short-term rainfall forecast score TS(1) corresponding to the first trained BPNN rainfall model under the first no-rainfall threshold l1; Step 3035, repeating steps 3031 to 3034 multiple times, inputting the test set into the nth trained BPNN rainfall model, and obtaining the short-term rainfall forecast score TS(n) corresponding to the nth trained BPNN rainfall model under the first no-rainfall threshold l1; Step 3036: According to the formula Get the mean score of the short-term rainfall forecast under the first no-rainfall threshold l1 Wherein, n is a positive integer, and 1≤n≤N; Step 3037: Set the no-rainfall threshold of the i-th node to l i , and l i =0.1+(i-1)×0.1, according to the method from step 3031 to step 3036, the i-th no-rainfall threshold l is obtained. i The average score of short-term rainfall forecast under Wherein, i is a positive integer; Step 3038: Repeat step 3037 multiple times, and iterate the no-rainfall threshold with a step length of 0.1 mm until the first no-rainfall threshold is l. I is 4.0 mm, then I short-term rainfall forecast score means are obtained; where I is a positive integer and 1≤i≤I; Step 304 is specifically as follows: The no-rainfall threshold corresponding to the largest short-term rainfall forecast score mean is selected from I short-term rainfall forecast score means and recorded as the optimized no-rainfall threshold value.

5. The method for qualitative forecasting of spatial-temporal intelligent surface rainfall based on Beidou water vapor according to claim 1, characterized in that: Step 401, the specific process is as follows: Step 4011, set the first light rain upper limit threshold H1 to the no-rainfall threshold optimization value + 0.1 mm, compare the predicted value of the cumulative value of rainfall in the next s hours in the next year in step 301 with the first light rain lower limit threshold H1, and obtain the rainfall level corresponding to the predicted value of the cumulative value of rainfall in the next s hours in the next year, recorded as yd; Compare the true value of the subsequent annual rainfall accumulation value in step 302 with the first light rain lower limit threshold H1, and obtain the rainfall level corresponding to the true value of the subsequent annual rainfall accumulation value, which is recorded as xd; wherein the rainfall levels are 1, 2, and 3, 1 is light rain, 2 is moderate rain, and 3 is heavy rain; Step 4012: If xd < yd, increment the number of times the predicted rainfall level is overestimated, Num OT by 1; if xd > yd, increment the number of times the predicted rainfall level is underestimated, Num UT by 1; if xd = yd, increment the number of times the predicted rainfall level is correct, Num TT by 1; where the initial values of Num OT , Num UT , and Num TT are all zero; Step 4013: Follow the method from step 4011 to step 4012 until the data in the test set is processed and the final number of overestimation of the forecast rainfall level Num is obtained. OT 、Num number of times the forecast rainfall level is underestimated UT The number of correct rainfall level forecasts TT ; Step 4014: Obtain the short-term forecast rainfall grade score TS′(1) corresponding to the first trained BPNN rainfall model under the first light rain upper limit threshold H1; Step 4015, repeating steps 4011 to 4014 multiple times, inputting the test set into the nth trained BPNN rainfall model, and obtaining the short-term forecast rainfall grade score TS′(n) corresponding to the nth trained BPNN rainfall model under the first light rain upper limit threshold H1; Step 4016: According to the formula Get the mean score of the short-term forecast level of light rain under the first light rain upper limit threshold H1 Step 4017: Set the upper limit threshold of the e-th light rain to H e , and H e =H1+(e-1)×0.1, according to the method of step 4011 to step 4016, the upper limit threshold value H of the e-th light rain is obtained. e The average score of the short-term forecast level light rain under Step 4018: Repeat step 4017 multiple times, and iterate the upper limit threshold of light rain with a step length of 0.1 mm until the upper limit threshold of light rain of the Eth time is H. E , then we get E short-term forecast level light rain score means; where e and E are both positive integers, and 1≤e≤E.

6. A method for qualitative forecasting of surface rainfall based on Beidou water vapor in time and space according to claim 5, characterized in that: Step 402, the specific process is as follows: Step 4021: Using a computer, take E light rain upper limit thresholds as abscissas, take E short-term forecast level light rain score means corresponding to the E light rain upper limit thresholds as ordinates, and draw a scatter plot of the short-term forecast level light rain score means; Step 4022: Using a computer, record the centers of two adjacent light rain upper limit thresholds as exploration points on the horizontal axis to obtain E exploration points; among them, the e-th exploration point x e The upper limit threshold H of the light rain at the e-1th e-1 and the e-th light rain upper limit threshold H e the center of; Step 4023: Use a computer to obtain the e-th exploration point x e Perform linear regression on the first k light rain upper thresholds to obtain the front slope of the front linear regression equation Use a computer to take the e-th exploration point x e The k light rain upper thresholds are linearly regressed to obtain the posterior slope of the posterior linear regression equation. k is a positive integer, and the value of k is 2, 3 or 4; e is a positive integer greater than 4; Step 4024: When k is 2, 3 or 4, the computer is used to calculate Get the e-th exploration point x e Weighted average slope when sliding forward window size is k and the e-th exploration point x e Weighted average slope when sliding backward with window size k Step 4025: According to the method of step 4024, when e is greater than k+1, the e-1th exploration point x is obtained. e-1 Weighted average slope when sliding forward window size is k and the e-1th exploration point x e-1 Weighted average slope when sliding backward with window size k Step 4026: Using a computer according to Get the e-th exploration point x e Second order difference value Δ 2 S(x e ); Step 4027, repeating steps 4021 to 4026 multiple times until the value of e is Ek, completing the second-order difference value before and after the Ek-th exploration point, and obtaining the second-order difference value before and after each exploration point; Step 4028: Obtain the maximum second-order difference value of the exploration point from the second-order difference values ​​of each exploration point, and record the two exploration points corresponding to the maximum second-order difference value of the exploration point as the previous exploration point x e′-1 and the later exploration point x e′ ; Step 4029: Using a computer according to Get the optimized value x of the upper threshold of light rain * ; Among them, Δ 2 S(x e′ ) represents the later explored point x e′ The corresponding second-order difference value before and after, Δ 2 S(x e′-1 ) represents the previous exploration point x e′-1 The corresponding second-order difference value before and after, Δ 2 S(x e′+1 ) represents the later explored point x e′ A detection point x behind e′+1 The corresponding second-order difference values ​​before and after.

Images